Maternal obesity programs offspring metabolic dysfunction via small extracellular vesicle-mediated epigenetic remodeling

Abstract

Maternal obesity is linked to heightened metabolic disease risk in offspring, but the mediators of this intergenerational effect remain unclear. Using a diet-induced obesity (DIO) mouse model, we showed that maternal circulating small extracellular vesicles (sEVs) crossed the placenta and delivered obesity-associated miRNAs to the fetal liver, with lasting consequences for insulin sensitivity in male offspring. Among these miRNAs, miR-29a-3p was pathologically elevated and targeted both DNA methyltransferases and demethylases, thereby reshaping the DNA methylation landscape. This included hypomethylation of the Pgc-1α locus, a key regulator of gluconeogenesis, which resulted in premature activation of hepatic gluconeogenesis that contributed to the persistent metabolic dysfunction in adulthood in male offspring. These findings identify a transplacental sEV-miRNA-epigenetic axis that perturbs fetal metabolic programming and may represent a conserved mechanism underlying the developmental origins of metabolic disease.

The study identifies a transplacental sEV–miRNA–epigenetic axis as a mediator of maternal obesity. Maternal plasma sEVs transfer miR-29a-3p to the fetal liver, epigenetically reprogramming glucose metabolism and driving adult insulin resistance.

Introduction

Substantial evidence indicates that maternal obesity during pregnancy perturbs the intrauterine environment, reshaping fetal developmental trajectories and increasing offspring susceptibility to metabolic disorders, including overweight, insulin resistance, type 2 diabetes, and cardiometabolic disease^1–7. Beyond its immediate consequences for the neonate, maternal obesity confers enduring effects that compromise offspring health throughout childhood and persist into adulthood^8–12. The complex crosstalk between maternal metabolic status and fetal genetic programming has emerged as a key focus of investigation. Yet the mechanisms by which intrauterine environmental perturbations shape fetal genetic and epigenetic regulation remain to be fully defined.

Extracellular vesicles (EVs) are nanoscale membrane vesicles released by cells and have received much attention due to their intercellular communication in both physiological and pathological processes^13,14. Small EVs (sEVs), with a size range of ~50 to 150 nm in diameter, are a subset of EVs that can deliver bioactive cargo including proteins, lipids as well as small non-coding RNAs into recipient cells or organs^15–17. Such capability enables them to impact a wide range of biological processes including the maintenance of metabolic homeostasis^18–25. In addition to the interorgan crosstalk, our prior work indicates that circulating small extracellular vesicles (sEVs) can potentially mediate communication across individuals²⁶. Moreover, we have shown that maternal sEVs, enriched in miRNAs, are efficiently transferred to the fetus, where they modulate gene expression in recipient tissues²⁷. These findings demonstrate that the fetus can detect and respond to maternal signals via sEVs during gestation^28–31. Such interaction between the mother and the fetus via sEVs may have profound implications for the offspring’s metabolic homeostasis in case of maternal obesity.

In the present study, we investigate how maternal obesity imprints long-lasting metabolic disturbances in offspring. We demonstrate that circulating sEVs from DIO pregnant mice traverse the placenta and enter the fetal liver, leading to impaired glucose tolerance and insulin sensitivity in adult progeny. These maternal sEVs deliver pathologically elevated miR-29a-3p to the fetal liver, where they reshape the DNA methylation landscape by repressing key epigenetic regulators. This repression precipitates premature Pgc-1α expression and early activation of hepatic gluconeogenesis during fetal life. Selectively targeting sEVs cargo miRNAs during pregnancy markedly mitigates metabolic dysfunction in offspring. Taken together, our findings reveal that maternal sEVs exert a profound and enduring influence, shaping the metabolic characteristics of offspring during the prenatal period, with potential long-lasting effects on their future health trajectories.

Results

Maternal diet-induced obesity programs lasting metabolic dysfunction in male offspring

To investigate the potential impact of maternal diet-induced obesity (DIO) on the fetus during pregnancy, female mice were continuously fed a high-fat diet (HFD, DIO) or a standard chow diet (CD, Control) for 8 weeks starting at 3 weeks of age, before pairing with males (Fig. 1a). Body weight, glucose levels, insulin levels, glucose tolerance tests (GTT), and insulin tolerance tests (ITT) for both DIO and Control female mice before mating are presented in Supplementary Data 1 and Supplementary Fig. 1a, b. At 18.5 days of gestation, a comprehensive analysis was conducted on both the mother and the fetus (E18.5). DIO significantly increased maternal weight, insulin and cholesterol levels, whereas the blood glucose and triglyceride levels were not significantly changed (Supplementary Data 2). Offspring from the DIO mother (O-DIO) exhibited increased body weight and glucose levels compared to those from the control group (O-Con). Meanwhile, the litter size and placental weight remained comparable between the two groups (Fig. 1b–e). Notably, the difference in body weight trajectory observed in fetal mice persisted through the first postnatal week but was lost by the second week (Supplementary Fig. 1c, d). Sex-stratified analyses further showed that the changes in placental weight, offspring body weight, and blood glucose were comparable between males and females, indicating no apparent sex difference in these parameters at fetal stage (Supplementary Fig. 1e–j).

Fig. 1. Adverse effects of maternal diet-induced obesity on metabolic disorder in offspring.

a Experimental scheme diagram. Female C57BL/6J mice were fed a chow diet (CD) or a high-fat diet (HFD) for 8 weeks prior to mating with CD-fed males. The pregnant dams were maintained on their respective diets throughout gestation, and fetuses were collected at embryonic day 18.5 (E18.5) for downstream analyses. The litter size (b), placenta weight (c), body weight (d) and blood glucose (e) of O-Con and O-DIO mice at E18.5 (n = 5 biological replicates (b); O-Con: n = 41 biological replicates, O-DIO: n = 42 biological replicates (c–e), from 2 independent experiments). f Experimental scheme diagram. Maternal diets (CD or HFD) were maintained throughout gestation and lactation. The offspring were weaned at 3 weeks of age, placed on a standard chow diet, and subsequently evaluated at 8 weeks of age. g Body weight of mO-Con/fO-Con and mO-DIO/fO-DIO mice during 4–8 weeks (n = 12 biological replicates, from 2 independent experiments). h Food intake of mO-Con/fO-Con and mO-DIO/fO-DIO mice (n = 12 biological replicates, from 2 independent experiments). The O₂ consumption (i), CO₂ production (j), RER (k) activity (l) and energy expenditure (m) of mO-Con/fO-Con and mO-DIO/fO-DIO mice (n = 12 biological replicates, from 2 independent experiments). Plasma triglycerides levels (n), plasma cholesterol levels (o), glucose levels (p) and plasma insulin levels (q) of mO-Con/fO-Con and mO-DIO/fO-DIO mice (n = 12 biological replicates, from 2 independent experiments). Blood glucose excursions measured during an intraperitoneal glucose tolerance test (r) and insulin tolerance test (s) of mO-Con/fO-Con and mO-DIO/fO-DIO mice (n = 12 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed two-way ANOVA followed by Bonferroni’s multiple comparisons test (g, r, s), Mann–Whitney test (two-tailed) (b) and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a, f created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

To investigate potential long-term effects of maternal obesity on offspring, newborn mice were fed with standard chow diet after weaning and subjected to further analysis at the age of 8 weeks (Fig. 1f). Male offspring from the DIO (mO-DIO) and Control (mO-Con) pregnant mice exhibited comparable body weight and food consumption, along with similar performance in the metabolic cage analysis (Fig. 1g–m). The blood lipids or blood glucose levels between these two groups were also indistinguishable (Fig. 1n–p). Notably, mO-DIO group showed a tendency of higher plasma insulin levels compared to the mO-Con group (Fig. 1q). Glucose tolerance test and insulin tolerance test indicated that mO-DIO group had impaired glucose tolerance and insulin sensitivity by the age of 8 weeks (Fig. 1r, s). We further performed an insulin stimulation experiment (see Methods) and showed that insulin-stimulated p-AKT was reduced not only in the liver, but also in skeletal muscle and adipose tissue (Supplementary Fig. 1k), indicating the peripheral insulin resistance in these adult offspring. Unlike the male offspring, the female progeny born to DIO-fed mother (fO-DIO) seemed to be resistant to maternal obesity. The metabolic traits of both the fO-DIO and fO-Con groups were comparable in adulthood (Fig. 1g–s).

We further assessed the impact of maternal obesity on offspring challenged with metabolic stress. The offspring were subjected to HFD treatment at 3 weeks of age (Supplementary Fig. 2a). The male offspring from the DIO mother (mO-DIO HFD) exhibited increased weight gain, elevated levels of blood glucose and plasma insulin (Supplementary Fig. 2b–d), along with a more pronounced decline in glucose tolerance and insulin sensitivity compared to the mO-Con HFD group (Supplementary Fig. 2e, f). On the other hand, the metabolic profile of the female offspring seemed largely unaffected by the maternal obesity (Supplementary Fig. 2g–k).

These findings suggest that maternal obesity has long-lasting detrimental effects on the metabolic homeostasis of male offspring.

Maternal plasma-derived sEVs transmit obesity-associated metabolic disturbances to offspring

Maternal EVs have the potential to mediate communication between the mother and fetus. We next asked whether such mechanism contributes to the intergenerational transfer of metabolic perturbations. Firstly, plasma-derived sEVs were isolated and characterized. Nanosight analysis revealed that 95% of purified EVs had a diameter range between 50-150 nm. Transmission electron microscopy and immunoblotting (CD9, CD63, CD81, TSG101, and Alix) assay confirmed that these sEVs were exosome-enriched sEVs (Supplementary Fig. 3a–c). To rule out potential contamination from lipoproteins, we measured the levels of ApoA1 and ApoB in the isolated sEVs. Neither protein was detected, confirming the high purity of our sEV samples (Supplementary Fig. 3d). Afterwards, to explore the impact of maternal sEVs in the intergenerational transmission of metabolic disturbance, plasma-derived sEVs isolated from DIO pregnant mice were intravenously administered to chow-diet fed pregnant mice for 6 consecutive days starting at G13 (Fig. 2a). Following sEVs administration, recipient dams showed no changes in body weight, food intake, blood lipids, blood glucose, or plasma insulin (Supplementary Fig. 3e–j). Litter size and placental weight at G18.5 were also unchanged (Fig. 2b, c), as was fetal (E18.5) body weight (Fig. 2d). Notably, fetuses from DIO-sEV–treated dams (O-DIO-sEV) displayed significantly higher blood glucose (Fig. 2e), similar to the phenotype observed in the DIO-treated group. This increase was observed in both male and female fetuses, indicating no sex-specific effect at fetal stage (Supplementary Fig. 3k–p). Given the comparable changes observed in male and female fetuses, subsequent analyses at the fetal stage were performed without sex stratification.

Fig. 2. Adverse effects of plasma-derived sEVs on metabolic disorder in offspring.

a Experimental scheme diagram. Chow diet fed pregnant mice were assigned at gestational day 13 (G13) to receive daily intravenous injections of age-matched plasma sEVs from either Con or DIO donor dams. The administration continued until G18, and the fetuses were collected for analysis at E18.5. The litter size (b), placenta weight (c), body weight (d) and blood glucose (e) of O-Con-sEV and O-DIO-sEV mice at E18.5 (n = 5 biological replicates (b); O-Con-sEV: n = 39 biological replicates, O-DIO-sEV: n = 40 biological replicates (c–e), from 2 independent experiments). f Experimental scheme diagram. g Body weight of mO-Con-sEV and mO-DIO-sEV mice during 4–8 weeks (n = 12 biological replicates, from 2 independent experiments). h Food intake of mO-Con-sEV and mO-DIO-sEV mice (n = 12 biological replicates, from 2 independent experiments). Plasma triglycerides level (i), plasma cholesterol levels (j), glucose levels (k) and plasma insulin levels (l) of mO-Con-sEV and mO-DIO-sEV mice (n = 12 biological replicates, from 2 independent experiments). The glucose tolerance test (m) and insulin tolerance test (n) of mO-Con-sEV and mO-DIO-sEV mice (n = 12 biological replicates, from 2 independent experiments). o Experimental scheme diagram. p Body weight of mO-Control-sEV and mO-Obesity-sEV mice during 4–8 weeks (n = 10 biological replicates, from 2 independent experiments). q Food intake of mO-Control-sEV and mO-Obesity-sEV mice (n = 10 biological replicates, from 2 independent experiments). Plasma triglycerides levels (r), plasma cholesterol levels (s), glucose levels (t) and plasma insulin levels (u) of mO-Control-sEV and mO-Obesity-sEV mice (n = 10 biological replicates, from 2 independent experiments). The glucose tolerance test (v) and insulin tolerance test (w) of mO-Control-sEV and mO-Obesity-sEV mice (n = 10 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed two-way ANOVA followed by Bonferroni’s multiple comparisons test (g, m, n, p, v, w), Mann–Whitney test (two-tailed) (d, k) and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a, f, o created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

To evaluate lasting effects of maternal sEVs, we analyzed adult offspring maintained on chow diet (Fig. 2f). Across groups, body weight, food intake, blood lipids, and blood glucose were comparable (Fig. 2g–k), but male offspring from DIO-sEV–treated dams (mO-DIO-sEV) showed elevated circulating insulin (Fig. 2l). Consistent with this, GTT and ITT revealed impaired glucose tolerance and reduced insulin sensitivity in mO-DIO-sEV males versus mO-Con-sEV controls (Fig. 2m, n). In females, metabolic parameters did not differ between groups (Supplementary Fig. 3q–x), so subsequent analyses focused on male progeny. To assess the relevance of these findings in a human context, we treated pregnant mice with plasma sEVs from pregnant women with obesity (Fig. 2o and Supplementary Fig. 3y). Their male offspring (mO-Obesity-sEV) showed higher plasma insulin and pronounced impairments in glucose tolerance and insulin sensitivity compared with offspring of mice given sEVs from pregnant women without obesity (mO-Control-sEV) (Fig. 2p–w), highlighting the cross-species relevance and translational potential of human-derived EVs.

Placental transfer and hepatic enrichment of maternal sEVs drive fetal liver metabolic programming

To elucidate how maternal sEVs influence fetal development, we traced their distribution in vivo by labeling plasma sEVs with PKH67 and administering them intravenously to pregnant mice (Fig. 3a). Fluorescence was detected on both maternal and fetal sides of the placenta, indicating placental transfer of maternal sEVs (Fig. 3b, c and Supplementary Fig. 4a, b). Labeled sEVs were detected in fetuses within 6 h and were preferentially enriched in the fetal liver, with no significant differences observed between male and female fetuses (Fig. 3d and Supplementary Fig. 4c). Co-localization of PKH67 with albumin confirmed delivery to hepatocytes (Fig. 3e), and flow cytometry showed that over 4% of fetal liver cells were PKH67-positive after a single injection (Fig. 3f). These results demonstrate that maternal plasma sEVs can cross the placental barrier and enter a substantial fraction of fetal hepatocytes.

Fig. 3. Maternal plasma-derived sEVs can be transplacentally delivered to fetal liver in mice.

a Experimental scheme diagram. PKH-labeled plasma sEVs derived from C57BL/6J mice were intravenously injected into recipient pregnant mice at G18.5. b Hematoxylin and eosin (H&E) stained images of the placenta. Scales bar: 1 mm (left) and 200 μm (right) (n = 3 each group, representative images were shown). c Confocal microscopy images. Scale bars: 75 μm (upper panel) and 25 μm (lower panel) (n = 5 each group, 5 random fields per sample, representative images were shown). d Confocal microscopy images and quantitative analysis. Scale bars: 25 μm (upper panel) and 5 μm (lower panel) (n = 5 each group, 5 random fields per sample, representative images were shown). e Confocal microscopy images. Scale bars: 25 μm (left) and 5 μm (right) (n = 5  each group, 5 random fields per sample, representative images were shown). f Flow cytometry images and quantification (n = 5 each group, representative images were shown). g Experimental scheme diagram. Heterozygous female CD63^FlagEGFP-mCherry^+/− mice were crossed with wild-type (WT) males. The resulting transgenic (CD63^FlagEGFP-mCherry^+/−) and WT (CD63^FlagEGFP-mCherry^−/−) fetuses were analysis. h Representative western blot (n = 5 each group, representative images were shown). i Flow cytometry images (n = 5 each group, representative images were shown). j Diagram of the experimental protocol and confocal microscopy images. Scale bars: 25 μm (n = 5 each group, 5 random fields per sample, representative images were shown). k H&E staining and Oil Red O staining (n = 3 each group, 5 random fields per sample, representative images were shown). l Western blotting and densitometry analysis (n = 6 each group, from 2 independent experiments). m H&E staining and Oil Red O staining (n = 3 each group, 5 random fields per sample, representative images were shown). n Volcano plot analysis of RNA-seq data of liver (n = 3 each group). o Gene Ontology (GO) enrichment analysis. Data are presented as mean ± SEM. Statistical analysis was performed one-way ANOVA followed by Bonferroni’s multiple comparisons test (d), and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a, g, j created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

To exclude dye leakage as a confounder, we used a genetic reporter model (Rosa26-CD63^FlagEGFP-mCherry) that produces EGFP-labeled EVs in the maternal circulation (Fig. 3g–i). In CD63^FlagEGFP-mCherry^+/− offspring, both CD63-EGFP and mCherry were detected in liver tissue, whereas non-transgenic offspring from transgenic dams exhibited EGFP without mCherry, confirming maternal sEVs transfer to the fetal liver (Fig. 3j).

We next assessed the hepatic consequences of exposure to maternal sEVs from mice with obesity. In adult mO-DIO-sEV offspring, H&E and Oil Red O staining were comparable to controls (Fig. 3k), but hepatic insulin signaling was blunted, evidenced by reduced p-AKT levels (Fig. 3l). At the embryonic stage, fetal livers likewise showed no overt histological changes (Fig. 3m). However, RNA-seq revealed 560 differentially expressed genes in sEV-exposed fetal livers (226 upregulated, 334 downregulated, Fig. 3n). GO analysis highlighted enrichment of pathways related to hormonal responses, particularly insulin signaling (Fig. 3o), aligning with the attenuated hepatic insulin sensitivity observed in adult offspring.

Together, these findings show that under maternal obesity, plasma-derived sEVs traverse the placenta, accumulate in the fetal liver, and reprogram hepatic gene expression, providing a mechanistic basis for reduced hepatic insulin sensitivity in adult offspring.

Maternal plasma sEVs deliver miR-29a-3p to the fetal liver under maternal obesity

Given the role of microRNAs (miRNAs) in EV-mediated communication, we investigated whether maternal plasma sEVs deliver miRNA cargo to fetal tissues. Profiling of plasma sEVs from control and DIO pregnant mice identified 27 miRNAs significantly upregulated in DIO sEVs (fold change >2, adjusted P < 0.05; Fig. 4a). In parallel, 20 miRNAs were elevated in livers of offspring from DIO dams (Fig. 4b), with three miRNAs overlapping between maternal sEVs and offspring liver (miR-24-3p, miR-22-3p, miR-29a-3p). Cross-referencing with sEVs miRNAs upregulated in pregnant women with obesity prioritized miR-29a-3p as a conserved candidate (Fig. 4c, d). qPCR and an RNase protection assay confirmed increased miR-29a-3p in maternal sEVs under obesity (Fig. 4e, f, and Supplementary Fig. 5a, b). Consistent with this finding, analysis of our human cohort (n = 22) revealed a positive correlation between maternal BMI and sEVs miR-29a-3p levels (Supplementary Fig. 5c), supporting the potential clinical relevance of our observations. Such elevation remained relatively stable across different gestational stages in mice (Supplementary Fig. 5d). Administration of these miR-29a-3p-containing sEVs elevated circulating miR-29a-3p levels in recipient mice in a dose-dependent manner (Supplementary Fig. 5e). Notably, fetal livers from DIO pregnancies showed elevated mature miR-29a-3p without changes in pri-miR-29a (Fig. 4g, h), consistent with exogenous transfer rather than endogenous upregulation. We also measured miR-29a-3p expression in major metabolic organs of DIO pregnant mice, including the liver, skeletal muscle, adipose tissue, and islets. The results showed that mature miR-29a-3p levels were increased in both the islets and the liver (Supplementary Fig. 5f), suggesting that these tissues may contribute to the production of miR-29a-3p-enriched circulating sEVs.

Fig. 4. Identification of the miRNAs responsible for offspring metabolic disorder in plasma-derived sEVs from mothers with obesity.

Heatmap displaying: Differential miRNA expression between Con-sEV and DIO-sEV (a), O-Con and O-DIO fetal livers (b), and mO-Control-sEV and mO-Obesity-sEV (c) (n = 3 each group). d Venn diagram summarizing the overlap of the miRNA sequencing dataset. e RT-qPCR quantification of miR-29a-3p levels in Con-sEV and DIO-sEV (n = 6 biological replicates, from 2 independent experiments). f RT-qPCR quantification of miR-29a-3p levels in Control-sEV and Obesity-sEV (n = 11 biological replicates, from 2 independent experiments). RT-qPCR quantification of miR-29a-3p (g) and pri-miR-29a (h) levels in the heart, liver, lung, kidney, and brain tissues of O-Con and O-DIO mice (n = 6 biological replicates, from 2 independent experiments). i Experimental scheme diagram. HEK-293T cells were transfected with negative control (NC) or mutant miR-29 oligonucleotides. The secreted sEVs were isolated from the conditioned medium, labeled with a PKH lipophilic dye, and intravenously administered to recipient pregnant mice at gestational day 18.5 (G18.5). Fetal tissues were harvested 6 h post-administration for fluorescence evaluation. j Confocal microscopy images showing PKH67 fluorescence in the fetal livers of control and PKH67-labeled sEVs-injected mice. Scale bars: 25 µm (left panel) and 5 µm (right panel) (n = 5 each group, 5 random fields per sample, representative images were shown). k RT-qPCR quantification of mutant-miR-29 levels in the fetal livers of O-NC and O-mutant-miR-29a mice (n = 6 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed Mann–Whitney test (two-tailed) (f, k) and others performed two-sided unpaired t-test. Source data are provided as a Source data file. i created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

To directly demonstrate delivery, we expressed a mutant miR-29a-3p (distinguishable from endogenous by specific probes) in 293T cells, isolated PKH67-labeled sEVs, and injected them into pregnant mice (Fig. 4i). Six hours later, robust PKH67 signals and accumulation of the mutant miR-29a-3p were detected in fetal livers (Fig. 4j, k).

Together, these data show that maternal plasma-derived sEVs traverse the placenta and deliver miRNA cargo—specifically miR-29a-3p—to the fetal liver. Under maternal obesity, dysregulated sEVs miRNAs likely contribute to adverse metabolic programming in offspring.

Pathogenically increased miR-29a-3p derived from mothers with obesity is sufficient to program metabolic dysfunction in offspring

To determine whether sEV-delivered miRNAs drive intergenerational metabolic dysregulation, we overexpressed miR-29a-3p in 293T cells and isolated sEVs enriched for this miRNA (Fig. 5a, b and Supplementary Fig. 6a, b). These sEVs were then intravenously administered to normal pregnant mice, leading to a dose-dependent increase in circulating miR-29a-3p levels (Supplementary Fig. 6c). In fetal livers, miR-29a-3p levels were markedly increased at E18.5, remained high at birth, and returned to control by 1 week postpartum (Fig. 5c), indicating a perinatal window of action. Sex-stratified qPCR analysis confirmed comparable miR-29a-3p levels in male and female fetal livers following injection of miR-29a-3p-enriched sEVs (Supplementary Fig. 6d), indicating no sex difference in sEV-mediated RNA uptake. Maternal metabolic parameters were unchanged following sEV/miR-29a-3p treatment (Supplementary Fig. 6e–j). In contrast, fetal offspring exposed to sEV/miR-29a-3p displayed markedly increased blood glucose (Fig. 5d–h). The effect persisted into adulthood: male offspring prenatally exposed to maternal sEV/miR-29a-3p (mO-sEV/miR-29a-3p) exhibited elevated fasting insulin and impaired glucose tolerance and insulin sensitivity by GTT and ITT (Fig. 5i–q). Liver histology showed no overt differences (Fig. 5r), yet insulin-stimulated p-AKT was reduced, indicating blunted hepatic insulin signaling (Fig. 5s). These results demonstrate that increased miR-29a-3p within maternal sEVs is sufficient to increase fetal blood glucose and to induce higher insulin levels accompanied by evidence of hepatic insulin resistance in adulthood. Together, they support a causal role for sEVs miR-29a-3p in driving intergenerational metabolic disturbance.

Fig. 5. sEVs with elevated miR-29a-3p are sufficient to induce metabolic disturbance in offspring.

a Experimental scheme diagram. HEK-293T cells were transfected with miR-29a-3p mimics or negative control (NC) oligonucleotides to generate engineered sEVs. Chow diet -fed recipient pregnant mice received daily intravenous injections of NC-sEVs or miR-29a-3p-sEVs from G13 to G18. Systematic analyses were conducted on fetuses (E18.5), neonates (New born), and offspring at 1, 2, 4, and 8 weeks of age to evaluate intergenerational effects. b RT-qPCR quantification of miR-29a-3p levels in engineered sEVs following transfection (n = 6 biological replicates, from 2 independent experiments). c RT-qPCR quantification of miR-29a-3p levels in the livers (n = 8 biological replicates, from 2 independent experiments). d Experimental scheme diagram. The litter size (e), placenta weight (f), body weight (g) and blood glucose (h) of O-sEV/NC and O-sEV/miR-29a-3p mice at E18.5 (n = 5 biological replicates (e); O-sEV/NC: n = 43 biological replicates, O-sEV/miR-29a-3p: n = 42 biological replicates (f–h), from 2 independent experiments. i Experimental scheme diagram. j Body weight of mO-sEV/NC and mO-sEV/miR-29a-3p mice during 4–8 weeks (n = 12 biological replicates, from 2 independent experiments). k Food intake of mO-sEV/NC and mO-sEV/miR-29a-3p mice (n = 12 biological replicates, from 2 independent experiments). Plasma triglycerides levels (l), plasma cholesterol levels (m), glucose levels (n) and plasma insulin levels (o) of mO-sEV/NC and mO-sEV/miR-29a-3p mice (n = 12 biological replicates, from 2 independent experiments). The glucose tolerance tests (p) and insulin tolerance tests (q) in mO-sEV/NC and mO-sEV/miR-29a-3p mice (n = 12 biological replicates, from 2 independent experiments). r HE staining and Oil Red O staining of livers (n = 3 each group, 5 random fields per sample, representative images were shown). s Western blotting and densitometry analysis (n = 6 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed two-way ANOVA followed by Bonferroni’s multiple comparisons test (j, p, q), Mann–Whitney test (two-tailed) (e, h, o), and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a, d, i created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

Maternal sEVs miR-29a-3p reshapes fetal liver epigenetics to prematurely activate hepatic gluconeogenesis

Given the persistent effects of maternal sEV-miR-29a-3p on offspring metabolic disorders, we aimed to investigate the alterations in the epigenetic profile of fetal liver. After the administration of miR-29a-3p enriched sEVs to pregnant mice (Fig. 6a, b), we evaluated the expression of several key epigenetic modulators including those known as miR-29a-3p targets. In fetal liver (E18.5), miR-29a-3p enriched sEVs notably decreased the protein levels of key DNA methylation regulators, including DNMT3A, DNMT3B, TDG, and TET3 (Fig. 6c). Similarly, hepatocytes treated with plasma sEVs from DIO pregnant mice exhibited comparable reductions in these epigenetic modifiers (Supplementary Fig. 7a–c). To further confirm that miR-29a-3p directly suppresses these enzymes at the post-transcriptional level, we performed luciferase reporter assays. miR-29a-3p mimic significantly reduced the luciferase activity of all four wild-type reporters, whereas this effect was largely lost when the seed sequences were mutated (Supplementary Fig. 7d–g). We also performed AGO2-RIP assays and found significant enrichment of miR-29a-3p together with the transcripts of Dnmt3a, Dnmt3b, Tdg, and Tet3 in the AGO2 complex compared with the IgG control (Supplementary Fig. 7h). These results indicate that miR-29a-3p directly binds to and represses these epigenetic regulators at the post-transcriptional level.

Fig. 6. Elevated miR-29a-3p in sEVs modifies the fetal liver epigenome and promotes premature hepatic gluconeogenesis.

a Experimental scheme diagram. Chow diet-fed recipient pregnant mice received daily intravenous injections of NC-sEVs or miR-29a-3p-sEVs from G13 to G18. b RT-qPCR quantification of miR-29a-3p levels in the fetal livers (n = 7 biological replicates, from 2 independent experiments). c Western blotting and densitometry analysis (n = 6 biological replicates, from 2 independent experiments). d DNA methylation levels across various genomic elements in the fetal livers of O-sEV/NC and O-sEV/miR-29a-3p mice (E18.5) (n = 3 each group). e DMRs identified via whole-genome bisulfite sequencing comparing O-sEV/NC and O-sEV/miR-29a-3p fetal livers (n = 3 each group). f Distribution of DMRs present in O-sEV/miR-29a-3p mice livers across various genomic features. g KEGG pathway enrichment analysis of genes associated with promoter hyper-DMRs and hypo-DMRs. h A list of genes with hyper-DMRs and hypo-DMRs in the insulin signaling pathway. i Bisulfite pyrosequencing quantification of CpG methylation levels within the validated DMR of the Pgc-1α promoter in fetal livers (n = 8 biological replicates, from 2 independent experiments). j FAIRE-qPCR quantification of chromatin accessibility at the Pgc-1α promoter in the fetal livers of O-sEV/NC and O-sEV/miR-29a-3p mice (E18.5). (n = 12 per group biological replicates, from 2 independent experiments). k Relative luciferase activity in cells transfected with a CpG-free reporter vector harboring the Pgc-1α DMR. Prior to transfection, the reporter constructs were either mock-treated (Unmethylated) or artificially methylated in vitro utilizing M.SssI CpG methyltransferase (Methylated) (n = 6 biological replicates, from 2 independent experiments). l RT-qPCR quantification of Pgc-1α, Pepck, and G6Pase mRNA levels in the fetal livers. (n = 6 biological replicates, from 2 independent experiments). m Western blotting and densitometry analysis (n = 6 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed two-way ANOVA followed by Bonferroni’s multiple comparisons test (i), and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

Whole-genome bisulfite sequencing (WGBS) revealed unchanged global methylation but widespread remodeling of the methylome, with numerous hypo- (n = 7073) and hyper-methylated (n = 7750) DMRs, unveiled that miR-29a-3p has the capacity to perturb the DNA methylation landscape within the fetal offspring’s liver (Fig. 6d, e). Promoter-associated DMRs were enriched in genes linked to insulin signaling, consistent with impaired insulin sensitivity in offspring of DIO dams (Fig. 6f, g). Among those differentially methylated genes, we identified Pgc-1α, which acts as an essential component of the glucose metabolism in liver, with a hypo-DMR located in promoter (Fig. 6h). To further validate the findings from the WGBS, we conducted targeted bisulfite sequencing of Pgc-1α, as well as two additional genes (Phka2 and Mapk9) that also showed decreased methylation levels. Consistent with the WGBS data, all three genes exhibited significantly reduced methylation at multiple CpG sites within their regulatory regions in the miR-29a group (Fig. 6i and Supplementary Fig. 7i, j).

To determine whether reduced methylation at the Pgc-1α promoter was associated with altered chromatin status, we performed FAIRE-qPCR in fetal liver tissues. The miR-29a-3p group showed significantly higher FAIRE enrichment at the Pgc-1α promoter than controls, indicating a more open and accessible chromatin configuration at this locus (Fig. 6j). This increase was similar in male and female fetuses (Supplementary Fig. 7k, l). To further test whether DNA methylation at this region directly affects transcriptional activity, we cloned the Pgc-1α DMR into a CpG-free luciferase reporter vector and compared unmethylated and M.SssI-methylated constructs. The unmethylated construct showed robust luciferase activity, whereas methylation markedly suppressed reporter expression (Fig. 6k), supporting a functional role for DNA methylation at this regulatory region in restraining Pgc-1α transcription.

Consistent with these epigenetic changes, qPCR analysis disclosed an upregulation of Pgc-1α transcription in the livers of fetal mice from sEV/miR-29a-3p group (Fig. 6l), suggesting a premature Pgc-1α expression. In conjunction with the elevated expression of Pgc-1α, the mRNA levels of its downstream gluconeogenic genes, Pepck and G6Pase, were significantly upregulated in the fetal mouse liver (Fig. 6l). Consistently, their protein levels also increased accordingly (Fig. 6m), indicating a premature activation of hepatic gluconeogenesis. This premature activation was further supported by pyruvate tolerance tests conducted in adulthood, which demonstrated enhanced hepatic gluconeogenic capacity in the offspring of DIO/sEV-miR-29a-3p mother (Supplementary Fig. 7m, n).

To examine the potential involvement of other hormonal regulators or the sympathetic nervous system at this developmental stage, we measured plasma corticosterone and glucagon concentrations and examined hepatic sympathetic innervation in E18.5 fetuses. No significant differences in circulating corticosterone or glucagon levels were observed between the two groups (Supplementary Fig. 7o, p). In contrast to adult liver, sympathetic nerve signals were barely detectable in fetal liver in either group (Supplementary Fig. 7q). Together, these data suggest that circulating corticosterone, glucagon, and hepatic sympathetic innervation are unlikely to explain the phenotype observed at this stage.

Together, these data suggest that elevated maternal sEV-miR-29a-3p disrupts DNA methylation homeostasis in the fetal liver, leading to aberrant Pgc-1α expression and premature activation of hepatic gluconeogenesis.

Neutralizing maternal sEVs miR-29a-3p prevents intergenerational metabolic dysfunction

To establish miR-29a-3p as a key driver of intergenerational metabolic dysfunction, we neutralized miR-29a-3p in plasma sEVs from DIO pregnant mice by electroporating anti-miR-29a. RNase degradation assays confirmed that anti-miR-29a was encapsulated within sEVs rather than surface-bound (Supplementary Fig. 8a). Moreover, luciferase analysis showed that anti-miR-29a-loaded DIO-sEVs abolished the suppressive effect of DIO-sEVs on Dnmt3a reporter activity, supporting effective neutralization of endogenous sEVs miR-29a (Supplementary Fig. 8b).

In vitro, anti-miR-29a–modified sEVs co-cultured with Hepa1-6 cells significantly increased protein levels of the epigenetic regulators TET3, DNMT3A, DNMT3B, and TDG compared with controls (Fig. 7a, b). In vivo, administering these anti-miR-29a–modified sEVs (anti-miR-29a) to healthy pregnant dams for 6 consecutive days did not alter maternal metabolic parameters (Fig. 7c, Supplementary Fig. 8c–h), yet markedly mitigated offspring metabolic abnormalities. The intervention normalized fetal high glucose level (Fig. 7d–g) and was associated with reduced hepatic expression of Pgc-1α and its gluconeogenic targets Pepck and G6Pase in fetuses (Fig. 7h). Furthermore, in adulthood, male offspring exposed to anti-miR-29a in utero (mO-anti-miR-29a) were protected from glucose intolerance and insulin resistance (Fig. 7i–q). Elevated hepatic p-AKT levels indicated restored insulin signaling in male offspring from DIO pregnancies (Fig. 7r). Together, these results identify sEVs miR-29a-3p neutralization as a potential therapeutic strategy to disrupt the intergenerational transmission of metabolic dysfunction.

Fig. 7. Inhibiting sEVs-miR-29a-3p improves metabolic dysfunction in offspring of dams with obesity.

a Experimental scheme diagram. Plasma sEVs isolated from pregnant mice with obesity were loaded with anti-NC or anti-miR-29a oligonucleotides via electroporation. Hepa1-6 cells were subsequently incubated with the engineered sEVs for 48 h, followed by immunoblotting analysis. b Western blotting and densitometry analysis (n = 6 biological replicates, from 2 independent experiments). c Experimental scheme diagram. Chow diet -fed recipient pregnant mice received daily intravenous injections of the modified sEVs from G13 to G18. Fetuses were subsequently harvested at E18.5 for downstream analysis. The litter size (d), placenta weight (e), body weight (f) and blood glucose (g) of O-anti-NC and O-anti-miR-29a mice at E18.5. (n = 5 biological replicates (d); O-anti-NC: n = 42 biological replicates, O-anti-miR-29a: n = 41 biological replicates e–g From 2 independent experiments). h RT-qPCR quantification of Pgc-1α, Pepck, and G6pase mRNA levels in the fetal livers of O-anti-NC and O-anti-miR-29a mice (n = 6 biological replicates, from 2 independent experiments). i Experimental scheme. j Body weight of mO-anti-NC and mO-anti-miR-29a mice from 4 to 8 weeks of age (n = 12 biological replicates, from 2 independent experiments). k Food intake of mO-anti-NC and mO-anti-miR-29a mice (n = 12 biological replicates, from 2 independent experiments). Plasma triglycerides levels (l), plasma cholesterol levels (m), glucose levels (n) and plasma insulin levels (o) of mO-anti-NC and mO-anti-miR-29a mice (n = 12 biological replicates, from 2 independent experiments). The glucose tolerance tests (p) and insulin tolerance tests (q) in mO-anti-NC and mO-anti-miR-29a mice (n = 12 biological replicates, from 2 independent experiments). r Western blotting and densitometry analysis of insulin-stimulated AKT phosphorylation in the livers of mO-anti-NC and mO-anti-miR-29a mice (n = 6 biological replicates, from 2 independent experiments). Data are presented as mean ± SEM. Statistical analysis was performed two-way ANOVA followed by Bonferroni’s multiple comparisons test (j, p, q), and others performed two-sided unpaired t-test. Source data are provided as a Source data file. a, c, i created in BioRender. Tian, L. (2026) https://BioRender.com/i9gb3bw.

Discussion

Maternal obesity increases offspring risk for metabolic disorders and is recognized as a developmental origin predisposing to later-life metabolic complications^32–35. Obesity during pregnancy perturbs the intrauterine milieu and alters fetal developmental trajectories, leading to long-term metabolic abnormalities. A subset of these persistent effects is thought to arise from epigenetic reprogramming^2,36–38, with evidence suggesting that maternal obesity can alter the fetal epigenome through modifications in DNA methylation and histone marks^39–41. However, the mechanisms by which fetal tissues sense maternal cues and translate them into epigenomic alterations remain incompletely understood.

While most studies emphasize transferred nutrients (for example, glucose and FFA) or hormones (such as adiponectin)^42–49, our work highlights maternal small extracellular vesicles (sEVs) as important mediators shaping offspring metabolic phenotype. In prenatal obesity, circulating maternal sEVs can traverse to the fetal liver and are capable of inducing long-term metabolic disorders in offspring. Pathologically elevated miR-29a-3p within these sEVs targets key epigenetic regulators, reshaping the hepatic epigenetic landscape and influencing gluconeogenesis and long-term insulin sensitivity. We also observe sexual dimorphism, with male offspring displaying greater susceptibility than females, consistent with the notion that estrogen may confer partial protection^46,50–52. This sex bias became evident only in the postnatal metabolic phenotype, whereas fetal-stage alterations were broadly comparable between males and females. These findings suggest that maternal sEV-mediated fetal programming is likely established in both sexes during development, and that the male-specific metabolic phenotypes observed in adulthood are unlikely to result from major sex differences in prenatal sEVs exposure or in the initial fetal hepatic response. Instead, the long-term phenotypic consequences may be unmasked later in life within a sex-dependent physiological context. One possible explanation is that estrogen-related hepatometabolic protection in females partially buffers this developmental insult, thereby reducing susceptibility to adverse metabolic outcomes in adulthood⁴⁰.

sEVs serve as interorgan messengers that reprogram recipient cells through targeted miRNA delivery, thereby linking distal physiological or pathological signals to gene regulatory networks^53–57. Our data indicate that maternal sEVs can cross the maternal–fetal interface during pregnancy, enabling intergenerational communication with potential consequences for offspring health. In our model, given that sEVs likely need to cross the placenta and enter the fetal circulatory system to exert their effects, we believe the earliest time for sEVs to influence the fetus is around E10.5, when maternal–fetal circulatory exchange becomes functional⁵⁸. However, our current data cannot exclude the possibility that maternal sEVs may also influence fetal development through circulation-independent mechanisms, potentially at earlier developmental stages. The elevated sEVs miR-29a-3p may represent a molecular signature of maternal metabolic stress that is imprinted onto the fetal genome via epigenetic modulation, predisposing offspring to metabolic vulnerability during juvenile and adult life. Notably, sEVs from pregnant women with obesity exhibit a similar miRNA signature, suggesting a potentially conserved mechanism across species.

The tissue origin of the altered maternal sEVs remains unclear. We found increased mature miR-29a-3p levels in the liver and pancreatic islets of DIO pregnant mice (Supplementary Fig. 5f). Previous studies have reported that both organs can secrete EVs to mediate intercellular communication^59–63. It is suggested that the pancreas can secrete miR-29 to the liver in obesity models⁶⁴. Combined with our current data, the liver and islets are the likely sources organs for those sEVs. However, this remains speculative, and contributions from other tissues cannot be excluded. The mechanisms by which maternal DIO reshapes the miRNA cargo of circulating sEVs also remain to be fully elucidated. We speculate that metabolic stress may both reshape cellular miRNA expression and affect the selective sorting of miRNAs into sEVs through RNA-binding proteins^22,65,66. These processes may together account for the altered miRNA profile of maternal sEVs during obesity in pregnancy.

miR-29a-3p exerts profound regulatory effects on DNA methylation dynamics^67–70. It targets DNMT3A and DNMT3B, enzymes critical for de novo DNA methylation, thereby modulating gene silencing^71,72. Meanwhile, miR-29a’s action extends to TETs and TDG, enzymes involved in the active demethylation process, leading to a downregulation of DNA methylation^73,74. Interestingly, the impact of miR-29a-3p on DNA methylation appears context-dependent, with opposing effects observed in different cellular environments⁷⁵. In our study, miR-29a-3p suppresses several enzymes (such as DNMT3A, DNMT3B, TDG, and TET3) involved in both methylation and demethylation processes after entry into the fetal liver via maternal sEVs. It perturbs DNA methylation landscape without altering the global methylation levels. These findings suggest the intricate regulatory impact of miR-29a-3p on the developing epigenome. Importantly, neutralizing elevated miR-29a-3p in maternal sEVs ameliorated metabolic dysfunction in offspring of dams with obesity, suggesting that this pathway may represent a candidate target for limiting obesity-associated intergenerational metabolic programming.

Among those aberrantly methylated genes in the fetal mice liver, we have identified Pgc-1α, which is known as a key regulator of hepatic gluconeogenesis, with a hypo-DMR located in promoter⁷⁶. This was associated with premature Pgc-1α expression, accompanied by upregulation of its gluconeogenic targets Pepck and G6Pase, indicating precocious activation of hepatic gluconeogenesis, a program normally restricted to the postnatal period^77–80. Under normal circumstances, fetal glucose homeostasis is maintained through a steady maternal supply, with negligible endogenous gluconeogenesis by the fetus itself until the neonatal stage⁷⁸. The early activation of gluconeogenesis aligns with the notion that heightened neonatal gluconeogenic drive may foreshadow later glucose intolerance^81–83, concordant with our observations.

While our findings establish a mechanistic link between maternal sEVs cargo and fetal epigenomic remodeling, important questions remain. The specific placental transport routes and tissue tropism of maternal sEVs need to be defined more precisely. Moreover, although integrated miRNA profiling implicates miR-29a-3p as a key effector, our target prediction analyses indicate that the broader miRNA repertoire in sEVs from DIO dams likely acts cooperatively to trigger premature hepatic gluconeogenesis in the fetal liver (Supplementary Figs. 9 and 10). This network-level effect highlights the importance of moving beyond single–miRNA frameworks in future studies.

In summary, we provide evidence that maternal sEVs and their miRNA cargo orchestrate epigenetic reprogramming of the fetal liver, offering a mechanistic framework for understanding how maternal metabolic stress may influence the developmental origins of metabolic disease. By linking maternal metabolic stress to fetal locus-specific epigenomic remodeling and premature activation of gluconeogenesis, our study expands current understanding of sEVs-mediated maternal–fetal communication and identifies sEVs miR-29a-3p as a potential intervention target that warrants further investigation.

Methods

Description of the cohort study population

The study cohort included 11 pregnant women with overweight (24 ≤ pre-pregnancy BMI < 28 kg/m²) or obesity (pre-pregnancy BMI ≥ 28 kg/m²) and 11 pregnant women without obesity as controls. A detailed summary of the clinical characteristics was provided in Supplementary Data 3. Participants with cardiovascular disease, depression, smoking, or alcohol use were excluded. Blood samples were collected from all participants. The plasma sEVs isolated from human subjects were pooled within their respective clinical groups before administration.

This research was approved by the Ethics Review Committee at Nanjing Drum Tower Hospital, which is affiliated with Nanjing University (Approval numbers: 2016-030-01). Informed consent was obtained from all eligible patients and control subjects upon their enrollment.

Mice

All animal procedures were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Nanjing University (IACUC-2109005). C57BL/6J mice were obtained from the Model Animal Research Centre of Nanjing University (Nanjing, China). Mice were housed in a specific pathogen-free (SPF) facility at Nanjing University under a 12-h light/12-h dark cycle, with ambient temperature maintained at 21 ± 2 °C and relative humidity at 55 ± 10%. Animals had ad libitum access to water and either a standard chow diet (Cat# XTI01ZJ-009, Xietong Shengwu; 22.7% protein, 62.4% carbohydrate, and 14.9% fat by calories; 4.02 kcal/g) or a high-fat diet (HFD; Cat# D12492i, Research Diets; 20% protein, 20% carbohydrate, and 60% fat by calories; 5.24 kcal/g). Pregnancy was confirmed by the presence of a vaginal plug, which was designated embryonic day 0.5 (E0.5). Regarding sex as a biological variable, both male and female offspring were initially evaluated at the fetal and adult stages. While similar changes were observed in both sexes at the fetal stage, significant metabolic abnormalities were exclusively observed in adult male offspring—in both the maternal obesity model and the maternal obesity-derived plasma sEV injection model—whereas adult female offspring did not exhibit these persistent metabolic phenotypes. Consequently, subsequent mechanistic studies were focused on adult male offspring.

Rosa26-CD63^FlagEGFP-mCherry mice (C57BL/6 J background) were also sourced from the Model Animal Research Centre of Nanjing University (Nanjing, China) and under pathogen-free conditions at Nanjing University. The transgenic mice harbored a CD63^FlagEGFP-mCherry reporter system inserted at the Rosa26 locus, in which the modified tetraspanin CD63 contains an extracellular domain fused to a FLAG tag and a C-terminus fused to EGFP for sEV tracking, while co-expressing mCherry for tissue localization. A loxP-floxed stop codon was inserted upstream of CD63, and mating with CAG-Cre mice generated CD63^FlagEGFP-mCherry^+/− heterozygous offspring, enabling the expression of CD63^FlagEGFP-mCherry. Consequently, plasma-derived sEVs from these mice exhibited CD63 with both FLAG tagging and EGFP fluorescence labeling. Female CD63^FlagEGFP-mCherry^+/− heterozygous mice were then crossed with wild-type male mice, producing offspring that were either transgenic (CD63^FlagEGFP-mCherry^+/−) or wild-type (CD63^FlagEGFP-mCherry^−/−).

Cell cultures

HEK-293T (SCSP-502) and Hepa1-6 (SCSP-512) cells were obtained from the Shanghai Cell Bank, Chinese Academy of Sciences, and were confirmed to be free of mycoplasma contamination. Cells were maintained in DMEM medium (Cat#10564011, Gibco) containing 10% fetal bovine serum (FBS) (Cat#1099141 C, Gibco) under standard culture conditions (37 °C, 5% CO₂).

Isolation of fetal mouse liver cells

Fetal mouse liver tissue was isolated and placed in a sterile centrifuge tube. The tissue was minced and digested with collagenase (Cat# 17104019, Gibco) at 37 °C with shaking for 15 min. Digestion was terminated by adding DMEM supplemented with 10% fetal bovine serum (FBS). The cell suspension was filtered through a 70 μm cell strainer, treated with red blood cell lysis buffer (Cat#420301, BioLegend), centrifuged at 1200 rpm for 3 min, and resuspended in PBS for further use.

Metabolic measurement

The experiment involved 10-week-old mice. Parameters such as heat production (HP), activity, carbon dioxide production (VCO₂), respiratory exchange ratio (RER), oxygen consumption (VO₂) were measured using a Comprehensive LabMaster home cage system (TSE Systems, Thuringia, Germany). Before data collection, the mice were given 2 days to acclimate to the system. The data collection process then proceeded continuously over 48 h.

Glucose and insulin tolerance tests

For the glucose tolerance test (GTT), mice were fasted for 12 h, then received an intraperitoneal injection of glucose at 1 g/kg body weight. Blood glucose levels were measured using a glucometer (Cat# 590, Yuwell, PRC) at 15, 30, 60, and 90 min post-injection.

For the insulin tolerance test (ITT), mice were fasted for 6 h, followed by intraperitoneal injection of insulin at 0.5 U/kg body weight. Blood glucose levels were measured at the same time points (15, 30, 60, and 90 min after administration).

Pyruvate tolerance test

To evaluate hepatic gluconeogenic capacity in vivo, mice were subjected to a pyruvate tolerance test (PTT). Following a 16-h overnight fast with ad libitum access to water, baseline blood glucose levels (0 min) were determined from tail vein blood. Subsequently, the mice received an intraperitoneal (i.p.) injection of sodium pyruvate dissolved in sterile saline at a dose of 1.5 g/kg body weight. Blood glucose concentrations were serially monitored from the tail vein at 15, 30, 60, and 120 min post-administration.

Biochemical analyses

For the assessment of systemic metabolic parameters, circulating hormones and lipid profiles were evaluated using commercially available assay kits according to the manufacturers’ protocols. Specifically, the plasma concentrations of insulin (Cat#10-1247-01, Mercodia), glucagon (Cat#10-1281-01, Mercodia), and corticosterone (Cat# E-OSEL-M0001, Elabscience) were quantified via enzyme-linked immunosorbent assay (ELISA). Concurrently, the circulating lipid panel, including triglycerides (Cat# A110-1, Nanjing Jiancheng) and total cholesterol (Cat# A111-1, Nanjing Jiancheng), was determined using standard enzymatic colorimetric methods.

Insulin stimulation experiment

To assess in vivo insulin sensitivity, 8-week-old male offspring mice were fasted for 6 h. Following the fast, mice were randomly assigned to receive an intraperitoneal (i.p.) injection of either insulin (0.5 U/kg body weight) or an equal volume of sterile saline (as a vehicle control). Thirty minutes post-injection, the mice were euthanized. Liver tissues were rapidly harvested, snap-frozen in liquid nitrogen, and stored at −80 °C. Total protein was extracted, and Western blot analysis was performed following standard laboratory protocols to evaluate the expression levels of phosphorylated AKT (p-AKT, Ser473) and total AKT.

Extracellular vesicles isolation

Small extracellular vesicles (sEVs) were isolated from plasma using differential ultracentrifugation combined with a commercial isolation kit. Briefly, plasma samples were diluted with an equal volume of PBS and subjected to sequential centrifugation steps at 4 °C: 500 × g for 5 min, 3000 × g for 25 min, and 10,000 × g for 60 min. The resulting supernatant was then ultracentrifuged at 120,000 × g for 90 min. The sEV pellet was collected. The remaining supernatant was further processed using a plasma sEV isolation kit (Cat# 4484450, Invitrogen) according to the manufacturer’s instructions. Specifically, the supernatant was treated with proteinase K for 10 min, followed by the addition of 0.2 volumes of exosome precipitation reagent. The mixture was thoroughly vortexed and incubated at 4 °C for 30 min. Finally, the sample was centrifuged at 10,000 × g for 10 min at 4 °C. The supernatant was discarded, and the sEV pellet was resuspended in an appropriate volume of PBS for downstream applications.

For the cell culture medium, initial centrifugation was performed at 300 × g for 10 min, followed by additional centrifugations at 3000 × g for 30 min and 10,000 × g for 60 min. The remaining supernatant was subjected to ultracentrifugation at 120,000 × g for 120 min to isolate sEVs. The resulting supernatant was further processed using a total sEV isolation kit for cell culture medium. The final sEV pellet was resuspended in PBS.

Density gradient purification

All steps were performed at 4 °C. Plasma was sequentially centrifuged at 500 × g for 5 min, 3000 × g for 25 min, and 10,000 × g for 60 min. The supernatant was then ultracentrifuged at 120,000 × g for 90 min. OptiPrep™ density medium (Cat#D1556, Sigma-Aldrich) was diluted with ice-cold PBS to prepare a discontinuous gradient consisting of 18%, 24%, 30%, and 36% (w/v) iodixanol solutions. The sEV pellet was resuspended directly in the 36% iodixanol solution, and this mixture was placed at the bottom of an ultracentrifuge tube. The lighter gradient layers (30%, 24%, and 18%) were then carefully overlaid sequentially. The tube was ultracentrifuged at 120,000 × g for 16 h to allow sEVs to float upward. The top 6 mL, containing high-purity sEVs, were collected, diluted with sterile PBS, and centrifuged again at 120,000 × g for 4 h to remove residual iodixanol. The final sEV pellet was resuspended in sterile PBS for immediate use.

Nanoparticle tracking analysis (NTA)

Extracellular vesicles (sEVs) extracted from plasma or cell culture medium were diluted to 5–10 μg protein per mL, followed by further dilution for subsequent analysis. Particle diameter data of the sEVs were obtained by analyzing them using the NanoSight instrument (Malvern Instruments), with data analyzed via NanoSight NTA 3.2 software (Malvern Instruments).

Transmission electron microscopy (TEM)

Purified EV suspensions were deposited onto carbon-film copper grids for 5 min, after which excess fluid was wicked away using filter paper. The samples were subsequently negatively stained with 2% phosphotungstic acid (Cat# G1102, Servicebio) for 2 min and blotted dry. Following air-drying, the morphological features of the EVs were visualized at 120 kV utilizing a Hitachi HT7800 transmission electron microscope.

PKH67 labeling and sEVs trafficking

To visualize the transplacental transfer of vesicles, purified plasma sEVs derived from C57BL/6 J mice were structurally tagged using the green-fluorescent lipophilic membrane dye PKH67 (Cat#MIDI67-1KT, Sigma). Briefly, the sEVs were uniformly suspended in PBS and incubated with the PKH67 dye for 5 min. The labeling reaction was subsequently quenched by the introduction of 1% bovine serum albumin (BSA). To remove free, unbound fluorescent dyes, the quenched reaction mixture was substantially diluted in ice-cold PBS and subsequently washed via ultracentrifugation at 100,000 × g for 90 min at 4 °C. Following centrifugation, the supernatant containing the unreacted dye was carefully discarded, and the resulting labeled sEVs pellet was resuspended in PBS. These 200 μg fluorescently tagged sEVs were systemically administered to pregnant dams. At 6 h post-injection, maternal organs and fetal tissues were harvested for downstream spatial assessments.

Immunofluorescence staining and images quantification

Mice were perfused with PBS followed by 4% paraformaldehyde. Tissues were collected, fixed in 4% paraformaldehyde for 10 h, and sectioned at 10 or 20 μm using a cryostat (Leica CM1950). Sections were washed with PBS for 15 min and then incubated in blocking buffer (0.1% Triton X-100 and 5% BSA in PBS) for 60 min. Sections were incubated with anti-albumin antibody (Cat# ab207327, 1:200, Abcam) for at least 12 h. After washing, incubate with the secondary fluorescent antibodies Alexa Fluor™ 594 (Cat# A-21207, 1:1000, Invitrogen) for at least 60 min. Counterstain nuclei with DAPI (Cat#C1002, Beyotime) for 15 min, followed by an additional 15-min wash with PBS. Detect the fluorescent signals using a Zeiss LSM 880 confocal microscope.

Fluorescence intensity was quantified using ImageJ. Images were converted to 8-bit format, and the total grayscale value within the region of interest was measured.

Flow cytometry

Plasma-derived small extracellular vesicles (sEVs) from WT or CD63^FlagEGFP-mCherry^+/− mice were isolated as detailed above. The purified sEVs were subjected to high-resolution flow analysis on a DxFLEX flow cytometer (Beckman Coulter, USA).

For cellular assessments, primary fetal liver cells were isolated from the offspring of control or PKH67-sEV-administered pregnant mice using established protocols. PKH67 fluorescence within the single-cell suspensions was subsequently detected via an Attune NxT Flow Cytometer (Thermo Fisher Scientific, USA). Data analysis was consistently performed utilizing FlowJo software (version 10; BD Biosciences). The gating strategy was shown in Supplementary Fig. 4d.

MicroRNA sequencing and analysis

Small RNA libraries were prepared using 3 μg RNA from liver tissue or 5 ng RNA from sEVs. The NEBNext Multiplex Small RNA Library Prep Set Illumina was utilized to generate sequencing libraries. Quality assessment of the libraries was conducted with the DNA High Sensitivity Chip on the Agilent Bioanalyzer 2100 system, followed by sequencing on the Illumina NovaSeq 6000 platform. Sequencing and analysis were carried out by BGI (Shenzhen, China). Analysis of the small RNA sequencing data was performed using DESeq2 (version 1.40.0). To identify highly abundant and biologically relevant miRNA candidates in the liver tissues, stringent filtering criteria were applied: only miRNAs with baseline expression >800 reads per million (RPM) were retained. To control the false discovery rate (FDR), P-values were adjusted using the Benjamini–Hochberg (BH) multiple-testing correction method. An FDR-adjusted P-value < 0.05 was predefined as the threshold for statistical significance. Furthermore, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of the targeted genes was conducted using KOBAS software (version 3.0) to elucidate the significantly altered metabolic pathways.

In vitro extracellular vesicle treatment

Approximately 50 μg of plasma-derived sEVs isolated from control mice or mice with obesity was added to 1 × 10⁵ cells and incubated for 24 h. Afterward, cells were harvested for further analysis.

In vivo extracellular vesicle treatment

Physiological transfer of plasma-derived sEVs: To investigate the pathophysiological role of plasma-derived sEVs from pregnant mice with obesity, chow diet (Con)-fed recipient pregnant mice were randomly allocated into two groups at gestational day 13 (G13). These mice received daily intravenous injections of plasma-derived sEVs isolated from either Con or high-fat diet (DIO) donor dams. The administration was performed for 6 consecutive days from G13 to G18. Each recipient was injected with a daily dose of 200 μg of sEVs, which is equivalent to the physiological yield from approximately 500 μL of donor plasma. Importantly, to accurately recapitulate the dynamic gestational environment, a strict age-matched administration protocol was adopted: recipient mice only received sEVs isolated from donor dams at the exact corresponding gestational stage (e.g., G13 donor sEVs were injected into G13 recipients).

Delivery of engineered sEVs enriched with miR-29a-3p: To determine whether sEVs-delivered miR-29a-3p drives intergenerational metabolic dysregulation, engineered sEVs were generated utilizing HEK-293T cells. Briefly, HEK-293T cells were transfected with either synthesized miR-29a-3p mimics or negative control (NC) oligonucleotides using Lipofectamine 2000, according to the manufacturer’s instructions. Chow diet fed pregnant mice at G13 were randomly assigned to two groups to receive daily intravenous injections of either sEV/NC or sEV/ miR-29a-3p. The treatment was administered for 6 consecutive days (G13 to G18). Each mouse received a daily dose of 100 μg of sEVs, representing the yield collected from the conditioned medium of approximately 2.5 × 10⁷ cultured HEK-293T cells.

RNA Extraction and qRT-PCR Analysis

The total RNA extraction from cells, tissues, sEVs, and plasma was performed using TRIzol reagent (Cat#15596018CN, Invitrogen). The reverse transcription of mRNA was conducted with the PrimeScript Reagent Kit (Cat#RR047A, Takara), incorporating gDNA Eraser for genomic DNA removal. For miRNA analysis, cDNA was generated using the HiScript III RT SuperMix (cat. no. R323, Vazyme). qRT-PCR was performed using TaqMan miRNA probes with LightCycler ®480 Real-time PCR System (Roche Diagnostics). Relative quantification of miRNA and mRNA expression was normalized against U6 snRNA and β-actin. Absolute quantification of miRNAs was achieved through the use of synthetic miRNA oligonucleotides at defined concentrations. Primer sequences are shown in Key resources table.

Western blot analysis

Liver tissue or cell lysates used for western blot analysis were extracted using RIPA buffer (Cat#P0013B, Beyotime) containing a protease inhibitor (Cat#P1048, Beyotime), and the protein concentrations were determined using a BCA protein assay kit (Cat#23227, Thermo Scientific). These lysates were subsequently loaded onto SDS-PAGE gels (10% or 8%) and then transferred to polyvinylidene fluoride membranes. The membrane was treated with the blocking solution (5% BSA) at room temperature for 60 min, and then the membrane was incubated with different primary antibodies at 4 °C overnight. Membranes were incubated with the secondary antibody at room temperature for 60 min before washed three times with TBST. Afterward, protein expression levels were determined using an ECL Kit and analyzed with ImageJ software. The following antibodies were used for western blot: CD81 antibody (Abcam, cat#ab109201, 1:1000), ALIX antibody (Cell Signaling Technology, cat#2171, 1:1000), CD9 antibody (Cell Signaling Technology, cat#98327, 1:1000), CD63 antibody (Santa Cruz Biotechnology, cat# sc5275, 1:1000), ALBUMIN antibody (Abcam, cat#ab207327, 1:2000), GFP antibody (Cell Signaling Technology, cat#2956S, 1:1000), p-AKT antibody (Cell Signaling Technology, cat#4060, 1:1000), AKT antibody (Cell Signaling Technology, cat#9272, 1:1000), DNMT3A antibody (Abcam, cat#ab2850, 1:1000), DNMT3B antibody (Abcam, cat# ab2851, 1:1000), DNMT1 antibody (Abcam, cat#ab188453, 1:1000), TDG antibody (Abcam, cat#ab154192, 1:1000), TET1 antibody (Abcam, cat#ab191698, 1:1000), TET2 antibody (Abcam, cat#ab124297, 1:1000), TET3 antibody (Novus, cat#NBP2-20602, 1:1000), β-ACTIN antibody (Cell Signaling Technology, cat#4968S, 1:1000), PGC-1a (Abcam, ab54481, 1:1000), PEPCK antibody (Santa Cruz Biotechnology, cat# sc-271029, 1:500), G6PC antibody (Proteintech, cat# 29084-1-AP, 1:1000), Horse anti-mouse IgG (Cell Signaling Technology, cat#7076, 1:1000), Goat anti-rabbit IgG (Cell Signaling Technology, cat#7074, 1:1000).

Whole-genome bisulfite sequencing (WGBS)

WGBS was performed by Novogene (Beijing, China). Briefly, high-quality genomic DNA was treated with bisulfite, followed by library preparation and sequencing on an Illumina platform. The average read depth was 30× per sample, covering about 90% of CpG sites across the genome. At the single-base level, methylated cytosines were identified using a binomial test based on the number of methylated reads, total reads, and the bisulfite non-conversion rate. Sites with an FDR-adjusted P-value < 0.05 were considered significantly methylated. At the regional level, differentially methylated regions (DMRs) between groups were identified using DSS (version 2.12.0). A DMR was required to contain at least five CpG sites, span at least 50 bp, and show an absolute methylation difference of at least 20% between groups. A strict P-value threshold of 1 × 10⁻⁵ was used to reduce false positives, as DSS accounts for the spatial correlation between nearby CpG sites.

Lucia luciferase activity assay for DNA methylation

To test whether CpG hypomethylation directly regulates Ppargc1a transcription, an in vitro methylation reporter system was used⁸⁴. A DNA fragment covering the DMR of Ppargc1a was cloned into the pCpGfree-promoter-Lucia plasmid (InvivoGen, San Diego, CA, USA), which contains a basal promoter for studying cis-regulatory elements. The DMR reporter plasmids were fully methylated using M.SssI CpG methyltransferase (Cat# M0226S, NEB) following the manufacturer’s instructions. After methylation, all plasmid DNA was purified using a column-based kit (Tiangen, Beijing, China). Hepa 1-6 cells were seeded in 24-well plates and transfected with either unmethylated or methylated DMR reporter plasmids. A firefly luciferase reporter was co-transfected as an internal control. After 48 h, secreted Lucia luciferase activity in the culture medium was measured using QUANTI-Luc reagent (InvivoGen). Cells were then lysed, and firefly luciferase activity was measured using the Bio-Lite Luciferase Assay System. The relative transcriptional activity of the Ppargc1a DMR was calculated as the ratio of Lucia luminescence to firefly luminescence.

Luciferase reporter assay for miRNA target validation

To validate the direct interaction between miR-29a-3p and its predicted targets, luciferase reporter assays were performed. The wild-type (WT) 3′-UTRs of DNMT3A, DNMT3B, TDG, and TET3 containing the putative miR-29a-3p binding sites were cloned into the pMIR-REPORT vector (Genscript) downstream of the firefly luciferase gene. Mutant (MUT) reporters were also generated by mutating the seed-pairing sequences. HEK-293T cells were seeded in multi-well plates and transfected using Lipofectamine 2000 (Invitrogen) with either WT or MUT reporters along with miR-29a-3p mimics or negative control (NC) oligonucleotides (GenePharma). A β-galactosidase (β-gal) expression plasmid was co-transfected as an internal control. After 24 h, cells were lysed. Firefly luciferase activity was measured using a kit (Cat# DD1101, Vazyme), and β-gal activity was measured by absorbance at 420 nm. Firefly luciferase activity was normalized to β-gal activity for each sample.

RNA immunoprecipitation (RIP) assay

To test whether miR-29a-3p binds to its target mRNAs via the RISC complex, an AGO2-based RIP assay was performed. Hepa 1-6 cells were transfected with miR-29a-3p mimics or NC using Lipofectamine 2000. After 36 h, cells were collected and lysed in RIP lysis buffer. The lysates were incubated with an anti-AGO2 antibody (Cat# ab186733, Abcam) or normal IgG (Cat# 2729S, CST) overnight at 4 °C with rotation. Pre-blocked magnetic beads (Cat# PB201, Vazyme) were then added and incubated for another 4 h at 4 °C. The beads were washed three times, and RNA was extracted from the immunoprecipitates using TRIzol. The enrichment of miR-29a-3p and target mRNAs (Dnmt3a, Dnmt3b, Tdg, Tet3) was measured by RT-qPCR. Gapdh mRNA was used as a non-target control.

Formaldehyde-assisted isolation of regulatory elements (FAIRE)

Chromatin accessibility in liver tissues was assessed using FAIRE⁸⁵. Briefly, frozen liver samples were ground into fine powder and immediately suspended in 1% formaldehyde. The suspension was rotated at room temperature for 10 min to cross-link DNA and proteins. Glycine was added to a final concentration of 125 mM to stop the reaction, followed by another 5 min of rotation. The homogenates were centrifuged at 300–500 × g for 5 min at 4 °C. The cell pellet was washed three times with cold PBS before lysis. Chromatin was sheared by sonication into small fragments. Ten percent of the lysate was kept as total input control. The input sample was treated with RNase A for 30 min at 37 °C, followed by Proteinase K and overnight incubation at 65 °C to reverse cross-links. The remaining 90% of the lysate was extracted with phenol:chloroform:isoamyl alcohol. The aqueous layer was collected and extracted again with chloroform. DNA was precipitated with ethanol, then treated with RNase A and Proteinase K as described above. Finally, both FAIRE-enriched DNA and input DNA were purified using a spin-column kit for downstream analysis.

sEVs loading

The Gene Pulser X Cell Electroporation System (BioRad) was used to load anti-NC and anti-miR-29a into plasma sEVs of DIO mice by electroporation⁸⁵. Approximately 200 μg sEVs were electroporated with 0.5 OD of anti-NC and anti-miR-29a in 400 μL of electroporation buffer (25 mM potassium chloride; 1.15 mM potassium phosphate, pH 7.2; 21% Optiprep), using 400 V and 125 μF in 4 mm wide electroporation cuvettes. The solution was then incubated on ice. Afterwards, the sEVs were isolated and diluted in 200 μL of PBS for subsequent tail vein injections.

Encapsulation of anti-miR-29a into sEVs was assessed by an RNase protection assay. The loading efficiency was calculated as the proportion of anti-miR-29a remaining detectable in RNase A-treated sEVs relative to the total amount of anti-miR-29a used for electroporation, yielding an estimated efficiency of approximately 20%. Functional neutralization of endogenous miR-29a within sEVs by the loaded anti-miR-29a was further verified using a luciferase reporter assay containing the Dnmt3a 3′UTR.

Statistical

Data are presented as mean ± standard error of the mean (SEM). Statistical analyses were conducted using GraphPad Prism 8. Data normality was assessed via the D’Agostino-Pearson (n ≥ 8) or Shapiro-Wilk (n < 8) tests. Normally distributed variables were analyzed using a two-tailed unpaired Student’s t-test, or a one-way/two-way ANOVA analysis followed by Bonferroni’s multiple comparisons test. When normality of samples failed, the Mann–Whitney U test was applied. We also conducted post hoc power analyses to further assess sample-size adequacy (statistical power >0.8), and the results supported the overall reliability of the statistical analyses presented.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.L., C.Y.Z., Y.Z., and Z.G. conceived and designed the work. H.S. performed the experiments. H.S., D.L., L.M., and N.Z. analyzed the data. H.S., D.L., L.M., J.W., C.Y., K.Z., X.C., Q.Z., X.J., J.L., Z.G., Y.Z., C.Y.Z., and L.L. contributed materials and analysis tools. L.L. and H.S. wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Hiroshi Itoh, who co-reviewed with Kenichiro Kinouchi; Cong-Yi Wang, who co-reviewed with Abdelilah Arredouani; and the other anonymous reviewer for their contribution to the peer review of this work. A peer review file is available.

Funding

L.L. discloses support for the research of this work from the National Natural Science Foundation of China Major Research Plan (92357306), National Science and Technology Major Project (2023ZD0507200), National Natural Science Foundation of China (32270998, 32471037). C.Y.Z. discloses support for the research of this work from the National Natural Science Foundation of China (32341007), the CAMS Innovation Fund for Medical Sciences (No. CIFMS-2021-I2M-5-015), and the National Center of Technology Innovation for Biopharmaceuticals Technology Breakthrough Projects (NCTIB2022HS02008). H.S. discloses support for the research of this work from the China Postdoctoral Science Foundation (Grant No. 2024M761421). J.W. discloses support for the research of this work from the National Science and Technology Major Project (2023ZD0507903), National Natural Science Foundation of China (82470865), and Natural Science Foundation of Jiangsu Province (BK20240116).

Data availability

The sequencing data have been deposited in the NCBI Gene Expression Omnibus under accession number GEO: GSE295756 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE295756), GSE295983 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE295983), GSE297729 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE297729), GSE298821 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE298821), GSE299405 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE299405). Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77161-4.