Fetal growth restriction (FGR) is a major cause of perinatal morbidity and mortality, yet its underlying mechanism remains elusive. Here, we show that growth differentiation factor 15 (GDF15) levels are significantly reduced in the placenta and fetal circulation in mouse and pig FGR models and correlate positively with fetal weight. Placenta-specific GDF15 deficiency causes FGR, while placenta-specific Gdf15 mRNA delivery through targeted lipid nanoparticle (LNP) effectively alleviates FGR. Mechanistically, placental GDF15 is a major contributor to the fetal circulating GDF15 pool and enters the fetal compartment, where it activates the hepatic STAT3 signaling pathway to stimulate IGF2 production during pregnancy. Placental GDF15 deficiency decreases hepatic STAT3 phosphorylation and IGF2 levels in fetuses. Inhibition of STAT3 in primary mouse fetal hepatocytes blocks GDF15-induced IGF2 upregulation. Placenta-specific Gdf15 mRNA-LNP administration rescues the decreased hepatic p-STAT3 and IGF2 levels in FGR fetuses. Collectively, our findings identify a placenta-fetal liver metabolic axis where placental GDF15 promotes fetal growth by activating the fetal hepatic STAT3-IGF2 pathway, revealing a potential therapeutic target for FGR.

Placental and fetal circulating GDF15 levels are reduced in mouse and pig FGR models

GDF15 is highly expressed in the placentas [18,36] during pregnancy. To investigate whether placental GDF15 is associated with FGR, we constructed two types of FGR mouse models [6] induced by protein restriction diet and Dexamethasone, respectively. Successful induction of FGR was confirmed by reduced fetal weight, short crown-rump length, decreased fetal liver weight, placental weight and placental efficiency at embryonic 18.5 days (E18.5) (Figs 1A and S1A–S1J), although no significant differences in the embryo number and resorption rate compared to controls (S1K and S1L Fig). Placental GDF15 expression and fetal plasma GDF15 levels were significantly decreased in FGR mice compared to normal controls (Figs 1B–1E, S1M, and S1N), and were positively correlated with fetal weight (Fig 1F and 1G). Maternal plasma GDF15 levels were also markedly decreased in FGR pregnant mice (S1O and S1P Fig). Consistent with these findings in mice, placental GDF15 expression and circulating GDF15 levels in FGR piglets were also significantly downregulated (Fig 1H–1L), and were positively correlated with birth weight (Fig 1M and 1N). Together, these results emphasize the potential role of GDF15 in FGR.

Placental GDF15 deficiency impairs fetal growth in utero

To determine the role of GDF15 in FGR, global Gdf15-knockout mice were generated and verified by PCR genotyping (S2A and S2B Fig). We first employed reciprocal crossing schemes in which wild-type (WT) (Gdf15^+/+) female mice × homozygous (Gdf15^−/−) male mice and the Gdf15^−/− female mice × WT male mice to investigate whether maternal GDF15 regulates fetal growth (S2C Fig). We found that plasma GDF15 levels in heterozygous fetuses from WT dams were comparable to those from Gdf15-null dams (S2D Fig), indicating that maternal GDF15 deficiency does not substantially reduce fetal circulating GDF15. Maternal GDF15 deficiency did not significantly alter fetal weight and crown-rump length at E18.5 (S2E–S2G Fig). Moreover, there is no differences in the embryo number and resorption rate between the two groups (S2H and S2I Fig). These data indicate that maternal-derived GDF15 may be not required to maintain fetal growth or fetal circulating GDF15 levels.

To further investigate whether the role of fetal-placental unit-derived GDF15 in FGR, we analyzed the growth phenotypes of fetuses with different Gdf15 genotypes (Gdf15^+/+, Gdf15^+/−, Gdf15^−/−) at E18.5, which had been produced by using a heterozygous × heterozygous breeding scheme (Fig 2A). Placental and fetal plasma GDF15 levels, along with the growth parameters including body weight and crown-rump length in Gdf15^−/− fetuses were significantly lower than those in wild-type (Gdf15^+/+) or heterozygous (Gdf15^+/−) fetuses (Fig 2B–2G). No significant differences in the placental weight among them (Fig 2H), whereas GDF15 deficiency significantly decreased placental efficiency (Fig 2I). In addition, there is no significant differences in fetal growth parameters and placental parameters were shown between Gdf15^+/+ and Gdf15^+/− fetuses (Fig 2F–2I). These data suggest that GDF15 deficiency impairs fetal growth. We further assessed the postnatal growth effect of GDF15 deficiency. Gdf15-deficient pups exhibited low birth weight compared to Gdf15^+/+ or Gdf15^+/− (Fig 2J), and the differences were also observed at weaning (Fig 2K). These findings demonstrate that GDF15 derived from fetal-placental unit increases fetal circulating GDF15 levels and regulates fetal and early postnatal growth.

Notably, qPCR and immunohistochemical staining showed that GDF15 was highly expressed in the placenta, predominantly in the labyrinth layer, whereas remained low in various tissues (including heart, liver, lung, kidney, brain, intestine, muscle) of fetal mice during pregnancy (S3A–S3C Fig), suggesting that placenta-derived GDF15 may play a key role in fetal growth. To further investigate whether these effects are due to placental GDF15, placenta-specific Gdf15-deficient (Gdf15^PKO, Gdf15^fl/fl; Cyp19a1^cre/+) mice were generated by using a Cyp19a1-Cre/Gdf15-loxP system and were identified by PCR (Fig 3A and 3B). qPCR and western blot analysis showed that CYP19A1 is highly expressed in placenta compared to fetal and maternal tissues in mice (S4A-S4B Fig). Recombinase activity was confirmed specifically in placenta but not in fetal tissues and maternal liver in mice (S4C Fig), as evidenced by crossing Cyp19a1-Cre with R26-LSL-EGFP reporter mice. We intercrossed Gdf15^fl/fl female mice with Gdf15^PKO male mice to generate both Gdf15^PKO and Gdf15^fl/fl conceptuses in the same pregnant mice (Fig 3C). GDF15 levels were significantly decreased in the placentas and plasma of Gdf15^PKO conceptuses compared to Gdf15^fl/fl littermates (Fig 3D–3F). However, placental GDF15 deficiency did not alter Gdf15 mRNA expression in fetal various tissues (S5A–S5F Fig). These data indicate that placenta contributes to the fetal circulating GDF15 pool. Further analysis showed that placenta-specific GDF15 deficiency resulted in reducing fetal weight and crown-rump length at E18.5 (Fig 3G–3J), suggesting that placental GDF15 deficiency is sufficient to cause defective FGR. In addition, placental weight was lower in Gdf15^PKO conceptuses compared to Gdf15^fl/fl controls, although the difference was not statistically significant (Fig 3K). Placenta-specific GDF15 deficiency also reduced placental efficiency (Fig 3L). Together, these findings demonstrate that fetal circulating GDF15 is derived predominantly from the placenta and that placental GDF15 is essential for fetal growth in mice.

Placental Gdf15 overexpression ameliorates FGR

Based on the finding that placental GDF15 deficiency causes FGR in mice, we next evaluated the therapeutic potential of GDF15 for FGR. We employed a placenta-tropic ionizable lipid nanoparticle (LNP) system, LNP55, to generate placenta-targeted Cy5-labeled mouse Gdf15 mRNA nanoparticles (Cy5-Gdf15-LNPs) as previously described [37]. Dynamic light scattering (DLS) characterization of the Cy5-Gdf15-LNP revealed a mean hydrodynamic diameter of 114.4 ± 4 nm and a polydispersity index (PDI) of 0.136 ± 0.013 (S6A and S6B Fig), indicating consistent formulation quality, with a high encapsulation efficiency (S6C Fig). Following intravenous (i.v.) administration of Cy5-Gdf15-LNPs to pregnant mice (S6D Fig), in vivo imaging revealed robust placenta-specific delivery and accumulation of Cy5-Gdf15-LNP (S6E and S6F Fig). Six hours after i.v. administration, Gdf15-LNPs (0.5 mg kg⁻¹) treatment significantly elevated placental GDF15 levels (~10-fold) in pregnant mice compared with those of controls (S6G Fig), confirming efficient Gdf15 overexpression in placenta.

We further evaluated the pharmacodynamic characteristics of different doses of Gdf15 mRNA-LNPs in healthy pregnant mice. At 6 hours after intravenous administration, plasma GDF15 levels in fetal mice receiving doses of 0.1, 0.25, 0.5, 1, 2, and 5 mg kg⁻¹ of Gdf15 mRNA-LNP were approximately 1.2-fold, 2-fold, 5.3-fold, 11.5-fold, 27.4-fold, and 71.8-fold, respectively, compared with the control group (S6H Fig). These data demonstrate that Gdf15 mRNA-LNP achieves dose-dependent GDF15 protein expression in the fetal circulation. Additionally, no obvious alterations in liver injury or inflammatory markers were shown in healthy pregnant mice treated with different dose (S6I–S6L Fig). Given that excessively high GDF15 during pregnancy may cause adverse effects such as nausea and vomiting [17], we assessed the safety of Gdf15 mRNA-LNPs by measuring food intake 24 h after a single injection in FGR pregnant mice. Doses of 0–1 mg kg⁻¹ did not affect food intake, whereas 2 and 5 mg kg⁻¹ significantly reduced it compared with controls (S6M Fig). Accordingly, 0.5 mg kg⁻¹ Gdf15 mRNA-LNPs were selected as the therapeutic dose for subsequent experiments.

To evaluate its therapeutic effect, FGR pregnant mice received Gdf15 mRNA-LNP treatment at E10 (Fig 4A). Compared with healthy pregnant mice (control group), FGR mice exhibited significantly decreased fetal weight and crown-rump length, as well as reduced placental and fetal plasma GDF15 levels (Fig 4B–4H). Gdf15 mRNA-LNPs (0.5 mg kg⁻¹) treatment significantly elevated GDF15 levels in the placenta and fetal plasma relative to the FGR group, but did not alter maternal plasma GDF15 levels (Figs 4B–4D and S6N). Moreover, this treatment ameliorated the FGR phenotype with increased fetal weight, crown-rump length, fetal liver weight, placental weight and placental efficiency (Fig 4E–4K), without affecting embryo number or resorption rate (Fig 4L and 4M). These results demonstrate that placenta-targeted GDF15 administration is a promising approach for treating FGR.

Placenta-derived GDF15 promotes fetal growth by stimulating fetal hepatic IGF2 production

Based on the contribution of placental GDF15 to fetal circulating GDF15 and fetal growth, we speculated that it controls fetal growth in an endocrine manner. To test this, we examined whether placental GDF15 enters into the fetal circulation during pregnancy. We first employed a BeWo trophoblast monolayers as an in vitro placental barrier model [38] and the integrity of the monolayers was confirmed by the absence of fluorescein isothiocyanate (FITC)-dextran translocation from the apical to the basolateral chamber (S7A and S7B Fig). We then added recombinant mouse GDF15 (rmGDF15) to the apical chamber and found detectable rmGDF15 in the basolateral supernatant (S7C Fig), demonstrating that GDF15 is able to cross the trophoblast layer. We next validated this finding in vivo by injecting the fluorescently labeled recombinant GDF15 (Cy5.5-rGDF15) into the placenta of pregnant mice (S8A Fig). Cy5.5-rGDF15 was detectable within the placentas and fetuses (S8B Fig), confirming that Cy5.5-rGDF15 in the placenta is able to enter the fetal circulation. Consistent with this, a single administration of Gdf15-LNP for 6 hours significantly increased fetal plasma GDF15 levels by more than 5-fold (S6H Fig), indicating that placenta-derived GDF15 successfully secreted into fetal circulation. Notably, Cy5.5-rGDF15 preferentially accumulated in the fetal liver (S8C Fig). These data suggest that placental GDF15 can be actively transferred into the fetal circulation and reveal a potential endocrine mechanism by which placental GDF15 acts on fetal liver to promote fetal growth.

Given the specific accumulation of Cy5.5-rGDF15 in the fetal liver, we further performed proteomic analyses on liver samples from placenta-specific Gdf15^PKO fetuses and Gdf15^fl/fl control fetuses to identify the potential downstream mediator of placental GDF15 (Fig 5A). The analysis revealed 142 differentially expressed proteins, including 38 secretory proteins, significantly enriched in Gdf15^fl/fl fetal livers (fold change ≥ 2; P < 0.05) compared to Gdf15^PKO fetal livers (Fig 5B). Among these differentially secretory proteins, IGF2 was identified to rank top1 (Fig 5C). IGF2 has been reported as a key regulator of fetal growth [39,40] and was highly expressed in the fetal liver (Figs 5D and S9A). Further analysis showed that IGF2 protein levels were significantly downregulated in the liver and plasma of Gdf15^PKO fetuses compared to controls (Fig 5E–5H). qPCR analysis confirmed that placenta-specific GDF15 deficiency markedly decreased Igf2 mRNA expression in fetal liver (Fig 5I). We also detected the alterations of IGF2 expression in other tissues of fetal mice and found that placental GDF15 deficiency did not significantly alter Igf2 mRNA or protein expression in other fetal tissues, including muscle, lung and heart (S9B–S9G Fig). These results suggest that placental GDF15 deficiency specifically suppresses IGF2 expression in fetal liver.

We further analyzed the correlation between GDF15 and IGF2 during pregnancy and showed that fetal plasma IGF2 levels positively correlated with fetal plasma and placental GDF15 levels, and fetal weight at E18.5 (S10A–S10C Fig). Besides these in vivo evidences, treatment of primary mouse fetal hepatocytes with rGDF15 protein markedly upregulated Igf2 mRNA and protein expression in a dose-dependent manner (Fig 6A and 6B). Furthermore, incubation of primary mouse fetal hepatocytes with conditioned medium from GDF15-overexpressing trophoblast cells also significantly increased IGF2 expression (Fig 6C–6H), demonstrating that trophoblast-derived GDF15 stimulates IGF2 expression in fetal hepatocytes. Together, these findings reveal a causal link between placental GDF15 and fetal hepatic IGF2 production.

Placental GDF15 promotes fetal hepatic IGF2 expression by activating STAT3 signaling during fetal growth

To elucidate the underlying mechanism by which GDF15 improves fetal hepatic IGF2 expression to promote fetal growth, we further analyzed the liver proteomic data from Gdf15^PKO fetuses and Gdf15^fl/fl control fetuses. Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis showed that differentially expressed proteins are markedly enriched in the JAK-STAT signaling pathway (Fig 7A). Signal Transducer and Activator of Transcription 3 (STAT3) is a major component in this pathway and has been shown to play a critical role in regulating IGF2 expression in human tumor cells [41]. We then investigated whether STAT3 is involved in placental GDF15-induced fetal hepatic IGF2 expression. Western blot analysis showed that placental GDF15 deficiency significantly inhibits STAT3 phosphorylation (p-STAT3) in fetal liver but not in fetal muscle, lung, heart, or kidney (Figs 7B and S11A–S11D), indicating that placenta-derived GDF15 specifically regulates the STAT3 signaling in the fetal liver. Additionally, rGDF15 treatment significantly increased STAT3 phosphorylation (p-STAT3) levels in primary mouse fetal hepatocytes (Fig 7C). To further establish the role of STAT3 activation in this process, STAT3 was inhibited with Stattic (2 μM), which effectively abolished rGDF15-induced STAT3 phosphorylation (Fig 7C) and blocked the increases of Igf2 mRNA and protein expression in primary mouse fetal hepatocytes (Fig 7C and 7D). These data demonstrate that STAT3 signaling is essential for placental GDF15 to drive fetal hepatic IGF2 expression.

We further analyzed how STAT3 regulates Igf2 transcription in primary mouse fetal hepatocytes. Mouse Igf2 regulation is highly promoter-specific, and its transcripts from promoters 2 (P2) and 3 (P3) are present in developing embryo, with dominant expression from the P3 [42–44]. A putative STAT3 binding site was shown in the P3 of mouse Igf2 gene by using the JASPAR website (S12A Fig). Approximately 1 kb of the murine Igf2 gene P3 region was cloned into a PGL3 luciferase reporter vector (mIgf2-Luc) (S12B Fig). The luciferase reporter assay showed that mutation of the putative STAT3 response element significantly decreased the luciferase activity of Igf2 P3 (S12C Fig), whereas Stat3 overexpression increased its luciferase activity (S12D Fig), demonstrating STAT3-driven Igf2 gene transcription. ChIP-PCR confirmed that STAT3 directly binds to the endogenous Igf2 promoter in primary mouse fetal hepatocytes (S12E Fig). Together, these findings establish STAT3-mediated transcriptional activation of Igf2 in mouse fetal hepatocytes.

We further determined whether placental GDF15 regulates the fetal hepatic STAT3-IGF2 signaling in FGR mice. Compared to normal controls, FGR fetuses exhibited significantly reduced placental and fetal GDF15 levels (Fig 4B–4D), accompanied by decreased hepatic p-STAT3 and IGF2 expression (Fig 7E–7H). Administration of Gdf15 mRNA-LNP to FGR pregnant mice markedly elevated the placental and fetal plasma GDF15 levels (Fig 4B–4D), and restored the reduced hepatic p-STAT3 and IGF2 protein levels and circulating IGF2 levels in FGR fetuses (Fig 7E–7I), and ultimately ameliorated FGR (Fig 4E–4H). Collectively, these results confirm that placenta-derived GDF15 activates fetal hepatic STAT3-IGF2 signaling axis, thereby rescuing FGR.

Animals

All animal procedures followed the Guide for the Care and Use of Laboratory Animals (Eighth Edition) and were approved by the Northwest A&F University Animal Care and Use Ethics Committee (approval ID: IACUC2025-0517). Male and female C57BL/6J mice were obtained from Xi’an Jiao Tong University and housed under standard conditions (12-h light/dark cycle, 23 ± 2 °C, 50 ± 10% humidity), with ad libitum food and water. Mice were randomly assigned to experimental groups. Newborn piglets were obtained from pregnant sows (Yorkshire × Landrace) at the Shaanxi Jingyang Xingao Laozhuang Breeding Co. (Xianyang, China). All pregnant sows (2–4 parities) were individually housed in individual gestation stalls (2.20 × 0.64 m) under controlled environmental conditions (mean temperature, 20 °C; relative humidity, 60%−70%) maintained by an automatic ventilation system. The room temperature of the lactating barn was kept at 28 °C ± 2 °C, and the baby house under a heating lamp was kept at 32 °C ± 2 °C, both automatically regulated. After parturition, piglets were suckled by the sows.

Experimental design

FGR mouse models were established as previously described [6]. Briefly, C57BL/6J pregnant mice were fed with low-protein diet (8% protein, normal diet for 20% protein) to construct the protein restriction diet-induced FGR (PR-FGR) mouse model. The Dex-FGR model was induced by subcutaneous injection with dexamethasone (HY-14648, MedChemExpress, USA) (2 mg kg⁻¹, every other day) from E11 to E17. Pregnancy was confirmed by vaginal plug (E0.5). At E18.5, pregnant mice were euthanized by isoflurane inhalation. The growth parameters (fetal weight and crown-rump length) and embryo absorption rate were recorded. Placental efficiency was calculated as the ratio of fetal weight to placental weight (g/g). Placentas were weighed after removal of membranes and umbilical cord and blotting of surface fluid, and each fetus was paired with its corresponding placenta. Blood was centrifuged (1,000g for 15 min at 4 °C) to obtain plasma. Placentas and fetuses were collected via cesarean section, snap-frozen in liquid nitrogen, or fixed in 4% paraformaldehyde for subsequent analyses. For fetal biochemical measurements (fetal plasma GDF15 and IGF2), we randomly sampled one male and one female fetus per litter, maintaining a 1:1 ratio to minimize potential confounding by sex. For growth phenotypes, all fetuses per litter were included. For FGR piglet model, mean birth weight of piglet was 1.42 ± 0.201 kg (mean ± SD). Piglets with birth weight 2 SD below the mean were defined as FGR, and those between 1.3 and 1.7 kg as normal. From these cohorts, 18 pairs of normal and FGR piglets were selected, ensuring one male and one female from each litter were chosen. Placentas located 3–4 cm from the umbilical cord insertion and umbilical vein blood samples for piglets were collected for subsequent analyses.

Generation of Gdf15 gene-edited mice

Gdf15-knockout (Gdf15^−/−) mice (S-KO-07014) and Cyp19a1-IRES-Cre mice (C001382) were obtained from Guangzhou Cyagen Biosciences. The IRES-Cre is inserted into the termination codon downstream of endogenous Cyp19a1 gene. Gdf15^fl/fl mice were donated by Meng Zhou’s lab from Anhui Medical University. Eight to ten weeks female Gdf15^+/− mice were obtained by heterozygote breeding. Gdf15^fl/fl mice and Cyp19a1-Cre mice were used to generate placenta-specific Gdf15-deficient (Gdf15^PKO, Gdf15^fl/fl; Cyp19a^Cre/+) mice. Gdf15^PKO fetuses and littermate controls (Gdf15^fl/fl) were acquired by crossing Gdf15^fl/fl female mice with Gdf15^PKO male mice. R26-CAG-LSL-Luc-2A-EGFP mice (NM-KI-190095) were purchased from Shanghai Model Organisms Center Recombinase activity was measured by crossing Cyp19-Cre with R26R-EYFP reporter mice. Genotyping was performed by PCR (primers listed in S1 Table).

Placenta-tropic Cy5-Gdf15 mRNA-LNP55 production and characterization

Placenta-targeting ionizable lipids (LNP55 formulation) were obtained from APEx Biological Co., LTD in accordance with the LNP55 formulation as previously described [37]. Mouse Gdf15 mRNA was synthesized and capped by in vitro transcription from a double-stranded DNA template (Qingke Biological Co.) using T7 Co-transcription Kit with CAG Trimer (DD4203, Vazyme, China, Nanjing). Gdf15 mRNA was labeled with Cy5 fluorescent dye (APEx Biological, USA). Cy5-labeled Gdf15 mRNA was formulated into LNP55 at a 10:1 molar ratio of ionizable lipid to mRNA by mixing ethanol and aqueous phases using pipette, dialyzed against 1× PBS for 4 h using Pierce microdialysis plates. Gdf15 mRNA-LNPs were dialyzed in the Pierce microdialysis plates (A50472, Thermo Fisher Scientific, USA) using 1× PBS solution for 4 h. LNPs were sterilized by 0.22 µm filters and stored at 4 °C for further experiment. RNA encapsulation efficiencies of LNPs were measured by using a Quant-iT-RiboGreen RNA Quantitative kit (R32700, ThermoFisher Scientific, USA) as the manufacturer’s instructions. The size and PDI of LNPs were determined by using a Malvern Zetasizer Nano ZSP (Malvern, UK).

Treatment of pregnant mice with Gdf15 mRNA-LNP

For tracking in vivo, pregnant mice at E15.5 were intravenously injected with fluorescent Gdf15 mRNA-LNP (0.5 mg kg⁻¹ body weight). Six hours later, mice were euthanized; organs, placentas, and fetuses were imaged using 3D optical Living Imaging system (IVIS Spectrum, USA). To evaluate the pharmacodynamic characteristics of Gdf15 mRNA-LNP, healthy pregnant mice were intravenously injected with different doses of Gdf15 mRNA-LNP (0.1, 0.25, 0.5, 1, 2, and 5 mg/kg) or LNP-Luc mRNA (5 mg/kg). At 6 hours after injection, fetal blood was collected and GDF15 levels were measured. To assess the role of placental Gdf15 overexpression, FGR pregnant mice were injected intravenously with Gdf15 mRNA-LNP (0.5 mg kg⁻¹) or LNP-Luc mRNA every other day from E10. Control group in which healthy pregnant mice receiving normal diet and the same volume of PBS injection as the FGR + Gdf15-LNP group. At E18.5, mice were euthanized, and samples were collected and growth parameters were analyzed.

Intraplacental injection of fluorescent recombinant GDF15

Intraplacental injection of E17 pregnant mice was performed as described previously [66]. Cy5.5-labeled recombinant GDF15 (5 μg/10 μl, 8944-GD, R&D Systems, USA) or free Cy5.5 (produced by Qiangyao Biotechnology, Shanghai, China) was injected into the placenta. The incision was sutured and mice were euthanized with isoflurane after 3 h; fetuses and placentas were collected and imaged using Living Image software.

HTR8 cell culture and treatment

HTR8/SVneo trophoblast cells (gifted by Xifeng Xiao’s lab, Air Force Medical University, Shaanxi, China) were cultured in DMEM/F-12 medium (supplemented with 10% FBS (Z7186FBS-500, ZETA LIFE, USA) and 1% penicillin/streptomycin in 37 °C incubator with 5% CO₂. Cells were transfected with His-tagged human GDF15 plasmids or vehicle using Lipofectamine 3000 (L3000001, Invitrogen, California, USA) for 48 h.

BeWo cell culture and treatment

BeWo cells (CCL-98, ATCC, USA) were maintained in Ham’s F-12K medium (21127022, Gibco, USA) with 10% FBS at 37 °C and 5% CO₂. For placental barrier assay in vitro, FITC-dextran (46944, Sigma, USA) or rmGDF15 (8944-GD, R&D Systems, USA) was added to the apical chamber and were incubated for 6 h. Supernatants from the basolateral chambers were collected for further analysis.

Primary mouse fetal hepatocytes culture and treatment

Primary mouse fetal hepatocytes were isolated as previously described [66] and cultured in high-glucose DMEM (H30022.01, HyClone, USA) with 10% FBS and 1% penicillin/streptomycin. Cells were treated with different dose rmGDF15 (0–200 ng/ml, TMPK-01244, TargetMol, USA) for 24 h. Hepatocytes were cultured with supernatant from hGDF15-OE plasmid-treated HTR8 cells for 24 h. STAT3 phosphorylation was inhibited using 2 μM Stattic (HY-13818, MedChemExpress, USA).

Enzyme-linked immunosorbent assay (ELISA)

GDF15 levels were measured using mouse GDF15 ELISA (MGD150, R&D Systems, USA), pig GDF15 ELISA (SEKP-0034, Solarbio, China), and human GDF15 ELISA (DY957, R&D Systems, USA). IGF2 levels were determined using a mouse IGF-2 ELISA kit (SEKM-0285, Solarbio, Beijing, China) according to the manufacturer’s instructions.

Plasmid construction

The coding sequences of human GDF15 or mouse Stat3 were cloned into pcDNA3.1 for overexpression. The mouse Igf2 promoter was inserted into pGL3-Basic (firefly luciferase, Promega), with pRLTK (Renilla luciferase) as internal control. A mutated STAT3-binding site within the Igf2 promoter was synthesized by Qingke Biotechnology (Beijing, China).

Luciferase reporter assay and ChIP-qPCR

Luciferase reporter assay and ChIP-qPCR were performed as previously described [67]. Primary mouse fetal hepatocytes were transfected with mouse Igf2 promoter luciferase reporters (WT or mutant), pcDNA3.1-Stat3, and pRLTK (50 ng) using Lipofectamine 3000. After 48 h, luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega). ChIP-qPCR was conducted with STAT3 or IgG antibody according to manufacturer’s protocol (ab500, Abcam, UK). Primer sequences are listed in S2 Table.

Real-time quantitative reverse transcription PCR (RT-qPCR) and western blotting analyses

RT-qPCR and western blotting were performed as previously described [68]. The relative gene expression level was calculated using the comparative Ct method (2^−ΔΔCt) and quantified by normalization to GAPDH. Primer sequences for all genes are listed in S2 Table. The target protein was normalized to GAPDH. Antibody information for all proteins is listed in S3 Table.

Mass spectrometry (MS) proteomics analysis

E18.5 fetal liver samples from Gdf15^PKO and Gdf15^fl/fl fetuses were collected for proteomics analysis in Novogene Bioinformatics Co. (Beijing, China). Protein lysates were centrifuged at 13,000g using YM-10 filters, reduced with 10 mM DTT (56 °C, 30 min), alkylated with 20 mM iodoacetamide (1 h, dark), and digested overnight with trypsin (50:1, 37 °C). Tryptic peptides were desalted on C18 resin column and analyzed by UHPLC coupled with DIA-MS. Differentially expressed proteins were defined as | log₂ (fold change) | ≥ 1 and P < 0.05. The false discovery rate was set to <1%. The MS proteomics data have been submitted to the ProteomeXchange consortium. MS proteomics data have been uploaded to iProX (https://www.iprox.cn/page/home.html) under the dataset identifier PXD084012.

Hematoxylin and eosin, immunofluorescence, and immunohistochemical staining

Hematoxylin and eosin and immunofluorescence staining were performed as previously described [68] according to standardized procedures. The whole-slide digital images were collected with a Pannoramic DESK Scanner (P-MIDI, P250, 3D HISTECH, Hungary). The fetal tissues samples were labeled with GDF15 or IGF2 antibodies. Nuclei were stained with DAPI (ID22502, dilution 1:1000, Solarbio, Beijing, China). Immunohistochemical staining was performed as previously described [69]. The placental samples were labeled with anti-mouse GDF15 antibodies (sc-515675, Santa Cruz, USA). Images were collected with BioTEK gen5 Software (Agilent Technologies, USA).

Statistical analysis

Data are presented as mean ± SEM. Statistical comparisons between two groups were performed using two-tailed unpaired Student t test. For multiple groups, one-way ANOVA with Tukey’s post hoc test or two-way ANOVA with Sidak’s post hoc test was used. Correlation was assessed by two-tailed Pearson correlation (95% CI). All analyses were conducted using GraphPad Prism 9 software (La Jolla, CA, USA). Linear mixed-effects model (LMM) analyses were performed in R (version 4.3.1) using the lme4 and lmerTest packages and were applied to analysis fetal body weight and crown-rump length across all groups, adjusting for fetal sex. Dam ID was included as a random intercept to account for clustering of fetuses within litters, while treatment, and fetal sex were included as fixed effects. Significance was defined as *P < 0.05; **P < 0.01. The n number and statistical methods used for each analysis were indicated in figure legends.

S1 Fig. The GDF15 levels of placenta and fetal plasma and fetal growth phenotype in FGR pregnant mice.

(A) Quantification of crown-rump length at E18.5 in normal (n = 42) and PR-induced FGR fetuses (n = 39). (B and C) Body weight (B) and crown-rump length (C) at E18.5 in normal (n = 25) and Dex-induced FGR fetuses (n = 21). (D and E) Representative images (D) and H&E staining of fetal sagittal histological sections (E) in normal and PR-induced FGR groups at E18.5. Scale bar, 2 mm. (F) Fetal liver weight in normal and PR-induced FGR groups (n = 15 fetuses from 5 litters). (G and H) Placental weight (G) and placental efficiency (H) in normal (n = 42) and PR-induced FGR groups (n = 39). (I and J) Placental weight (I) and placental efficiency (J) in normal (n = 25) and Dex-induced FGR groups (n = 21). (K and L) Embryo number (K) and resorption rate (L) in normal and PR-induced FGR dams (N = 5 dams). (M) ELISA analysis to quantify placental GDF15 levels in normal and Dex-induced FGR groups (n = 9 placentas from 3 litters). (N) ELISA analysis to quantify fetal GDF15 levels in normal and Dex-induced FGR groups (n = 6 from 3 litters). (O) GDF15 concentrations of maternal plasma in normal and PR-induced FGR groups (N = 5 dams). (P) GDF15 concentrations of maternal plasma in normal and Dex-induced FGR groups (N = 3 dams). Data are presented as the mean ± SEM. P values were calculated by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01, and ns (not significant). Dex, dexamethasone. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s001

(TIF)

S2 Fig. Maternal GDF15 deficiency has no effect on fetal growth.

(A) Schematic diagram for the generation of Gdf15-deficient mice using CRISPR/Cas9 technology. (B) Genotype identification of Gdf15^+/+, Gdf15^+/−, Gdf15^−/− mouse by PCR. (C) Schematic diagram of breeding strategy in which WT (Gdf15^+/+) female mice × Gdf15^−/− male mice and the Gdf15^−/− female mice × WT male mice (N = dams, n = fetuses). (D) Fetal plasma GDF15 levels in WT and Gdf15^−/− groups at E18.5 (WT, n = 18; Gdf15^−/−, n = 20). (E) Representative images of fetuses at E18.5 in each group. Scale bar, 2 mm. (F and G) Fetal weight (F) and crown-rump length (G) at E18.5 in each group (WT, n = 63; Gdf15^−/−, n = 69). (H and I) Embryo number (H) and resorption rate (I) of pregnant mice at E18.5 in each group (WT, N = 9; Gdf15^−/−, N = 10). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s002

(TIF)

S3 Fig. Tissue distribution of GDF15 in mice during pregnancy.

(A) Relative Gdf15 mRNA expression measured by qRT-PCR in the uterus, placenta, and fetal tissues in pregnant female at E18.5 (n = 6). (B and C) Representative GDF15 immunohistochemistry staining in placentas (B) and fetal mice (C) at E18.5. Scale bar, 2 mm and 500 μm. Data are presented as the mean ± SEM. P values were calculated by one-way ANOVA with Tukey’s post hoc test (A). *P < 0.05, **P < 0.01. JZ, Junction layer; Lab, Labyrinth layer. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s003

(TIF)

S4 Fig. Tissue distribution of CYP19A1 in mice during pregnancy.

(A) Relative mRNA expression of Cyp19a1 in various tissues of fetuses and pregnant mice at E18.5 (n = 6). (B) western blot analysis of CYP19A1 protein abundance in various tissues of fetuses and pregnant mice at E18.5. (C) Representative immunofluorescence images of Cyp19a1-Cre: R26-EGFP in fetal tissues, placenta, and adult liver. Scale bar, 2 mm and 500 μm. P values were analyzed by one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01. The underlying data for this figure can be found in S1 Data. The Original blot for this figure can be found in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004061.s004

(TIF)

S5 Fig. Placental GDF15 deficiency does not alter Gdf15 mRNA expression in fetal tissues.

(A–F) Relative Gdf15 mRNA expression in fetal liver (A), muscle (B), lung (C), heart (D), brain (E), and intestine (F) from Gdf15^fl/fl and Gdf15^PKO fetuses (n = 6). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s005

(TIF)

S6 Fig. Cy5-Gdf15 mRNA-LNP specifically targets to the placenta.

(A–C) Mean hydrodynamic diameter (A), polydispersity index (B) and encapsulation efficiency (C) of Gdf15 mRNA-LNPs (n = 5). (D) Schematic diagram of pregnant mice treatment. 8−10 weeks pregnant dams were administrated by i.v. injection with Cy5-Gdf15-LNP or PBS. (E) Fluorescence imaging of dissected tissues in pregnant mice 6 h after i.v. injection of Cy5-Gdf15-LNP or PBS in vivo. (F) Fluorescence imaging of placentas and fetuses 6 h after i.v. injection of Cy5-Gdf15-LNP or PBS in vivo. Radiance efficiency scales reflect photon emission intensity (photons s⁻¹ cm⁻² sr⁻¹ μW⁻¹ cm²). (G) Placental GDF15 levels of pregnant mice 6 h after i.v. injection of Cy5-Gdf15-LNP or LNP-Luc mRNA in vivo (n = 6). (H) Fetal plasma GDF15 levels, 6h following i.v. administration of various doses of Gdf15-LNP or LNP-Luc mRNA to healthy pregnant mice (n = 6). (I–L) Plasma levels of alanine transaminase (ALT) (I), aspartate aminotransferase (AST) (J), tumor necrosis factor alpha (TNF-α) (K), and interleukin-6 (IL-6) (L), 6h following i.v. administration of various doses of Gdf15-LNP to pregnant mice (n = 6). (M) Cumulative 24-h food intake after i.v. administration of various doses of Gdf15-LNP or LNP-Luc mRNA to pregnant mice (n = 6). (N) Maternal GDF15 levels in pregnant mice at E18.5 in each group (n = 6). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test (G) or one-way ANOVA with Tukey’s post hoc test (H-N). *P < 0.05, **P < 0.01. Cy5-Gdf15-LNP, Cy5-labeled Gdf15 mRNA lipid nanoparticle; i.v., intravenous injection; F, fetus; P, placenta; H, heart; Li, liver; S, spleen; Lu, lung; K, kidney. Panel D created in BioRender. liu, X. (2026) https://BioRender.com/ax7lt7r. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s006

(TIF)

S7 Fig. GDF15 could cross the placental barrier in vitro.

(A) Schematic of an in vitro BeWo cells transwell assay mimicking the placental barrier. rmGDF15 or PBS was added to the apical chamber of Transwell inserts with monolayer BeWo cells. FITC-dextran was then added, and the chambers were incubated for 6 h. Supernatants from the basolateral chambers were collected for further analysis. (B) Fluorescence intensity of FITC-dextran in the supernatant of the basolateral chamber of the Transwell with or without monolayer BeWo trophoblast cells (n = 6). (C) GDF15 levels in the supernatant of the basolateral chamber of the Transwell were measured by ELISA (n = 3). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01. rmGDF15, recombinant mouse GDF15. Panel A created in BioRender. liu, X. (2026) https://BioRender.com/ax7lt7r. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s007

(TIF)

S8 Fig. Placental GDF15 could circulate into fetal liver.

(A) Schematic diagram of mice treatment. 8-week-old pregnant dams were intraplacentally injected with Cy5.5-rGDF15 or free Cy5.5 at E17. (B) Fluorescence imaging of placentas (left) and fetuses (right) 3 h after intraplacental injection of Cy5.5-rGDF15 or free Cy5.5 in vivo. (C) Fluorescence imaging of dissected fetal organs following Cy5.5-rGDF15 administration in pregnant dam. The left panel displays fluorescence imaging of fetal organs, and the right panel displays the images of fetal organ. Radiance efficiency scales reflect photon emission intensity ((photons s⁻¹ cm⁻² sr⁻¹ μW⁻¹ cm²). Cy5.5-rGDF15, Cy5.5-labeled recombinant GDF15. Panel A created in BioRender. liu, X. (2026) https://BioRender.com/ax7lt7r.

https://doi.org/10.1371/journal.pbio.3004061.s008

(TIF)

S9 Fig. Placenta-specific GDF15 deficiency does not alter IGF2 expression in other fetal tissues.

(A) qPCR analysis of Igf2 mRNA expression in various tissues of fetuses at E18.5 (n = 6). (B–D) Relative Igf2 mRNA expression in fetal muscle (B), lung (C), and heart (D) from Gdf15^fl/fl and Gdf15^PKO fetuses (n = 6). (E) Representative immunofluorescence staining of IGF2 (Green) in muscle, lung, heart, pancreas, intestine, and brain from Gdf15^fl/fl and Gdf15^PKO fetuses. Scale bar, 200 μm. (F and G) western blot analysis of IGF2 protein expression in muscle (F) and lung (G) from Gdf15^fl/fl and Gdf15^PKO fetuses (n = 3). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01. The underlying data for this figure can be found in S1 Data. The Original blot for this figure can be found in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004061.s009

(TIF)

S10 Fig. The correlation between fetal IGF2 and fetal GDF15, placental GDF15, and fetal weight.

(A) Pearson’s correlation coefficient analysis of plasma IGF2 and GDF15 levels in fetal mice at E18.5 (n = 20). (B) Pearson’s correlation coefficient analysis of fetal plasma IGF2 and placental GDF15 levels in pregnant mice at E18.5 (n = 20). (C) Pearson’s correlation coefficient analysis of fetal plasma IGF2 and body weight in pregnant mice at E18.5 (n = 20). The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s010

(TIF)

S11 Fig. Placental GDF15 deficiency has no effect on the STAT3 signaling pathway in fetal other tissues.

(A–D) Relative protein expression of phosphorylated STAT3 (Tyr705) and total STAT3 in muscle (A), lung (B), heart (C), and kidney (D) from Gdf15^fl/fl and Gdf15^PKO fetuses (n = 3). The Original blot for this figure can be found in S1 Raw Images.

https://doi.org/10.1371/journal.pbio.3004061.s011

(TIF)

S12 Fig. STAT3 binds with the promoter of murine Igf2.

(A) Schematic images of structure and putative STAT3-binding motif in the P3 of murine Igf2 gene predicted by JASPAR. (B) Wild-type and mutant Igf2 P3 promoter luciferase reporter. (C) Luciferase activity of wild-type or mutant Igf2 reporters in the presence of STAT3 (n = 6). (D) Luciferase reporter analysis in WT or Stat3-OE primary mouse fetal hepatocytes transfected with Igf2 P3 constructs (n = 6). (E) ChIP analysis of STAT3 occupancy in the Igf2 P3 in primary mouse fetal hepatocytes (n = 3). Data are presented as the mean ± SEM. P values were analyzed by two-tailed unpaired Student t test. *P < 0.05, **P < 0.01. E, exon; P, promoter. The underlying data for this figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3004061.s012

(TIF)

S1 Data. Contains underlying data for Figs 1A, 1B, 1D–1I, 1K–1N, 2B, 2C, 2F–2K, 3E, 3F, 3I–3L, 4B, 4C, 4G–4M, 5B, 5C, 5E, 5F, 5I, 6A, 6D–6F, 6H, 7A, 7D, 7G–7I, S1A–S1C, S1F–S1P, S2D, S2F–S2I, S3A, S4A, S5A–S5F, S6A–S6C, S6G–S6N, S7B, S7C, S9A–S9D, S10A–S10C, S12C–S12E.

https://doi.org/10.1371/journal.pbio.3004061.s017

(ZIP)