Atosiban-conjugated 3WJ-pRNA nanoparticles delivering GAS1-enhanced extracellular vesicles: targeting the decidua to combat recurrent miscarriage.

Primary mouse endometrial stromal cells (MESCs) were harvested according to previously established protocols [ 22 ]. Human endometrial stromal cells (HESCs, ATCC Cat# CRL-4003, RRID: CVCL_C464) and primary MESCs were cultured in Dulbecco’s modified Eagle’s medium (DMEM)/F-12 medium (Boster, China) supplemented with 10% charcoal-stripped fetal bovine serum (CS-FBS, Vivacell, China), 1% insulin-transferrin-selenium (Gibco, US), 1.5 g/L sodium bicarbonate (Sigma, US), and 1 mM Sodium Pyruvate (Sigma, US). Decidualization was induced by treating HESCs/MESCs with medroxyprogesterone acetate (MPA, 1 µM, HY-B0469; MCE, China) and dibutyl cyclophosphamide (db-cAMP, 0.5 mM, HY-B0764; MCE, China) in 2% CS-FBS for varying durations. Immortalized HUVECs (ZQXZbio Cat# ZQ1099, RRID: CVCL_F0BB) were cultured in a special medium (ZM1099; ZQXZbio, China). HUVECs were used for the endothelial tube formation assay [ 23 ]. HTR8 (Servicebio Cat# STCC12001P, RRID: CVCL_7162) were cultured in Roswell Park Memorial Institute 1640 medium (RPMI 1640; Boster) supplemented with 10% FBS (CellMax, China). JEG3 cells (Servicebio Cat# STCC12003P, RRID: CVCL_0363) were cultured in DMEM-high glucose supplemented with 10% FBS. HTR8 and JEG3 cells were used for the wood-healing assay. All cells were cultivated in a humidified incubator at 37 °C with 5% CO 2 . Cells were not tested for cross-contamination of other cell lines or misidentification. Endometrial samples were collected from 10 patients with RM and 10 healthy fertile participants at the First Affiliated Hospital of Xi’an Jiaotong University, Xi’an, China. All participants provided written informed consent, and the study protocol was approved by the Ethics Committee of the First Affiliated Hospital of Xi’an Jiaotong University (XJTU1AF2024LSYY-430) in accordance with the Declaration of Helsinki. The samples were anonymized and processed according to the BRISQ guidelines (Level 2 compliance) for reporting human biospecimens. Single-cell transcriptomes of the early maternal–fetal interface in humans [ 24 ] and mice [ 2 ] were analyzed and visualized using the UCSC Cell Browser [ 25 ] and SCP package in R ( https://github.com/zhanghao-njmu/SCP ). A spatial multiomics map of human implantation sites, placenta, and decidua [ 26 ] was analyzed and visualized using CROST [ 27 ]. All microarray data were systematically extracted from the Gene Expression Omnibus (GEO) database ( https://www.ncbi.nlm.nih.gov/geo/ ) and The Cancer Genome Atlas (TCGA) ( https://portal.gdc.com/ ), preprocessed, and normalized using R statistical software (version 4.2.2, https://www.r-project.org/ ). Total RNA was isolated from the cultured cells using the SteadyPure Quick RNA Extraction Kit (AG, China). First-strand complementary DNA (cDNA) was synthesized using an EVO M-MLV Reverse Transcription Mix Kit (AG). The qRT-PCR reaction was performed in a 20-µL mixture with SYBR Green Premix Pro Taq HS qPCR Kit (AG). All PCR primers used are listed in Additional file 1 Table S1. For qRT-PCR, the reaction was conducted in a 20-µL mixture with 10 µL of SYBR Green Premix, cDNA templates (< 100 ng), and primers (10 pM each) for 40 cycles. Relative mRNA levels were calculated using Glyceraldehyde-3-phosphate dehydrogenase ( GAPDH) as a reference. All assays were performed at least thrice. Proteins were separated by 12.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes. The membranes were blocked with protein-free rapid blocking solution for 15 min at room temperature (RT) and incubated overnight at 4 °C with primary antibodies: GAS1 (Proteintech 17903-1-AP, 1:1000), GAPDH (Proteintech 60004-1-Ig, 1:50,000), RAB39B (Proteintech 12162-1-AP, 1:500), Alix (Proteintech 12422-1-AP, 1:5,000), calnexin (Proteintech 10427-2-AP, 1:5,000), CD63 (Abcam, ab193349, 1:500), and TSG101 (Proteintech 28283-1-AP, 1:2,000). After incubation with primary antibodies, the membranes were incubated with specific secondary antibodies (Thermo Fisher Scientific, 1:5,000) for 1 h at RT. Bands were visualized using SuperSignal West Pico (Thermo Scientific), according to the manufacturer’s instructions. Uncropped and unprocessed scans of blots are shown in Additional file 2. Fluorescence in situ hybridization assays were performed using a Hybridization Kit (GenePharma, Shanghai, China) according to the manufacturer’s instructions. Cy3-labeled riboprobe sequences for GAS1 were 5’-CGTCATATTCT + TCGTCGTAGTAGTCG-3’, 5’-CTGGAAAAGT + TTGTCCGAGT + T-3’, 5’-GCGTGTGGT + TGAGCTGGATAA-3’. Briefly, uterus paraffin Sect. (4 μm) were treated with proteinase K (37 °C, 20 min), followed by blocking (37 °C, 30 min), denaturing (78 °C, 2 min), and then hybridized with probes (biotin-probe: SA-Cy3: PBS = 1:1:8) in hybridization buffer overnight at 37 °C. For immunofluorescence, slices were incubated overnight at 4 °C with primary antibodies for GAS1 (Proteintech 17903-1-AP, 1:200), CD45/Ptprc (Proteintech 60287-1-Ig, 1:200), RAB39B (Proteintech 12162-1-AP, 1:50), CD63 (Abcam, ab193349, 1:50), OXTR (Proteintech 23045-1-AP, 1:50), and CD31 (Proteintech, 28083-1-AP), GJB3 (Proteintech, 12880-1-AP), and then were incubated with specific secondary antibodies Cy3 AffiniPure Goat Anti-Mouse IgG (H + L), Alexa Fluor 488 AffiniPure Goat Anti-Rabbit IgG (H + L) (Beyotime Biotechnology, 1:200), or with a multiplex fluorescence kit (ImmunoWay, RS0035) at 37 °C for 1 h. Nuclear staining was performed using Antifade Mounting Medium with DAPI (Beyotime Biotechnology, China). Images were visualized using a Leica fluorescence microscope. For immunohistochemistry, slices were incubated overnight at 4 °C with primary antibody for GAS1 (Proteintech 17903-1-AP, 1:500), and then incubated with goat anti-mouse/rabbit IgG polymer (Zsbio) at 37 °C for 30 min. The co-IP assay was performed using a GAS1 antibody (R&D AF2636-SP) and Pierce Co-IP kit (Thermo Scientific, US) according to the manufacturer’s instructions. Mass spectrometry was performed by OE Biotech (Shanghai, China). Peptides and proteins were identified using the UniProt FASTA database of Homo sapiens (Human), UP000005640_9606. siRNA transfection for GAS1 or RAB39B genession (Additional file 1 Table S2) was designed and synthesized by Qingke Biotechnology Company (Beijing, China) and transfected into HESCs using Lipofectamine™ RNAiMAX. GAS1 overexpression and GAS1 knockout lentivirus vectors were purchased from Qingke. The conditioned medium from each group of cells was collected. The cells were centrifuged at 3,000 × g for 10 min to remove dead cells. The supernatant was centrifuged at 4 °C, 16,500 × g for 20 min to remove cell debris and apoptotic bodies. The supernatant was centrifuged at 3,500 × g for 15 min in a 100-kD ultrafiltration tube. The filtrate was passed through a 0.22-µm filter membrane, and the filtrate was collected. The filtrate was transferred to an ultracentrifuge tube and centrifuged at 120,000 × g for 90 min at 4 °C. The supernatant was removed, and the pellet was resuspended in an appropriate volume of pre-cooled 1× PBS to obtain EVs, which were stored at −20 °C. Transmission electron microscopy (TEM) was used to visualize isolated EVs. The samples were resuspended in copper grids for 5–10 min, dried under incandescent light at 65 °C, stained with 3% uranyl acetate, and visualized using a Hitachi HT7700 transmission electron microscope. Nanoparticle tracking analysis (NTA) was used to quantify EV size and number using a Particle Metrix PMX-120 instrument (Germany). NTA software was used to measure nanoparticle concentration (particles/mL). Each sample was analyzed using the batch process included in the NTA software. Western blotting was used to detect the expression of EV markers Alix, CD63, and TSG101 and the absence of calnexin. The in vitro mouse blastocyst implantation model was modified from a previously published protocol [ 28 , 29 ]. Female mice aged 6–8 weeks were selected and injected intraperitoneally with 10–12 IU pregnant mare serum gonadotropin (PMSG; M2520, Nanjing Aibei Biotechnology Company, China). After 48 h, 10–12 IU human chorionic gonadotropin (HCG; M2620, Nanjing Aibei Biotechnology Company, China) was injected. Female mice were mated with male mice and sacrificed on the following day. Oviducts were dissected and placed in sterile culture dishes. The fertilized oocytes were obtained by cutting the enlarged ampulla of the oviduct using a syringe tip. The fertilized oocytes were transferred to KSOM (M1435, Nanjing Aibei Biotechnology Company, China) culture dishes after digestion with hyaluronidase and washing. The zona pellucida of the blastocysts was removed using acidic Tyrode’s solution (Sigma, T1788), and the blastocysts were washed again in KSOM. The blastocysts were then transferred to decidualized endometrial stromal cells. Blastocyst adhesion and migration were observed and counted over 24–48 h. pRNA-a3WJ, pRNA-b3WJ, and pRNA-c3WJ were synthesized by Shanghai Jierui Bioengineering Co., Ltd. The sequences of RNA strands (lowercase letters indicate 2′-deoxy-2′-Fluoro modified nucleotides) are as follows: a3WJ -TEG: 5′‑uuGccAuGuGuAuGuGGG‑ TEG-3′ b3WJ: 5′‑cccAcAuAcuuuGuuGAuccc‑3′ Atosiban-c3WJ-Alexa647: 5′‑(Atosiban) GGAucAAucAuGGcAA(C6‑NH) (Alexa647)-3′ a3WJ -TEG: 5′‑uuGccAuGuGuAuGuGGG‑ TEG-3′ b3WJ: 5′‑cccAcAuAcuuuGuuGAuccc‑3′ Atosiban-c3WJ-Alexa647: 5′‑(Atosiban) GGAucAAucAuGGcAA(C6‑NH) (Alexa647)-3′ The three RNA strands were mixed in TMS buffer (89 mM Tris, 5 mM MgCl 2 , pH 7.6) at an equal molar ratio and then heated to 80 °C for 5 min, followed by slow cooling to 37 °C for 1 h. The synthesized 3WJ-pRNA nanoparticles were characterized by 15% PAGE in TBE buffer for 2 h at 100 V. After staining with SYBR Green I (HY-K1004, MCE), the gel was visualized and analyzed using a UV illuminator. The morphology of the 3WJ-pRNA nanoparticles was recorded using atomic force microscopy (AFM) as previously described [ 19 ]. The thermal stability of the 3WJ-pRNA nanoparticles was assessed by monitoring fluorescence during continuous heating at 37 °C for 99 cycles, each lasting 10 min. To assess cytotoxic effects, a CCK8 assay was conducted to measure cell viability following treatments with Atosiban-3WJ-pRNA. The binding ratio was optimized by monitoring the time-dependent (10 min to 4 h) reduction in A647 fluorescence in the supernatants after incubation with Alexa Fluor 647-labeled 3WJ-pRNA with EVs at varying ratios (100:1 to 10,000:1) and subsequent cellular uptake. Biocompatibility was evaluated using standard hemolysis assays as previously described [ 30 ]. All animal experiments were approved by the Laboratory Animal Ethics Committee of Xi’an Jiaotong University (XJTUAE2024-14) and were conducted in compliance with the ARRIVE 2.0 guidelines. The mice were purchased from Shaanxi Pharmaceutical and Medical Biotechnology and housed under specific pathogen-free conditions (22 ± 2 °C, 55 ± 5% humidity, 12-h light/dark cycle) with free access to food and water. To establish the RM model, CBA/J females were mated with DBA/2 males, and BALb/c mice served as normal pregnancy controls. Tail vein injections of GAS1-EVs (10 8 particles) or Atosiban-3WJ-pRNA-GAS1-EVs (10 8 EVs + 10 11 nanoparticles) were administered daily from gestational day (GD) 5 to 9. Decidual microenvironments were analyzed on GD10, and fetal resorption rates were quantified on GD14. Data are presented as mean ± standard deviation (SD) for normally distributed variables. Two-group comparisons were performed using unpaired Student’s t-tests (normal distribution) or Wilcoxon rank-sum tests (non-parametric), whereas multi-group comparisons were performed using one-way analysis of variance (ANOVA) with Tukey’s post-hoc test (parametric) or the Kruskal–Wallis test with Dunn’s correction (non-parametric). Statistical significance was set at P  < 0.05 (two-tailed). Sample sizes (n) represent biologically independent replicates as specified in the figure legends. All analyses were performed using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA).

To establish a model of in vitro decidualization (IVD), we treated HESCs with 1 µM MPA and 0.5 mM db-cAMP over a 96-h time course. Decidualization was confirmed by time-dependent upregulation of canonical markers: insulin-like growth factor-binding protein 1 ( IGFBP1 ) plateaued after 48 h, whereas prolactin ( PRL ) showed sustained induction, particularly at 72–96 h (Fig.  1 A). RNA sequencing of HESCs at 0 and 96 h in IVD revealed significant enrichment of the Hallmark Hedgehog Signaling pathway by Gene Set Enrichment Analysis (GSEA) (Fig.  1 B). Differential expression analysis within this pathway identified GAS1 as a top upregulated gene (Fig.  1 C), prompting further investigation of its role. GAS1 expression increased progressively during IVD, with mRNA and protein levels peaking at 96 h (Fig.  1 D–E). Immunofluorescence of HESCs and primary MESCs (Additional file1 Figure S1) revealed that the cytoplasmic/membrane localization of GAS1 was markedly elevated in decidualized cells (enlarged and rounded morphology; Fig.  1 F). Single-cell RNA-seq analysis of ~ 70,000 cells from the maternal–fetal interface (UCSC Cell Browser) demonstrated enriched GAS1 expression in stromal cells of the decidual tissue, underscoring its tissue-specific relevance (Fig.  1 G). Fig. 1 Dynamic expression of GAS1 during decidualization. ( A ) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of decidual markers IGFBP1 (top) and PRL (bottom) in HESCs during IVD (mean ± standard deviation [SD]; n  = 3; analysis of variance [ANOVA] with Tukey’s post-hoc test). ( B ) GSEA plot of Hallmark Hedgehog Signaling pathway enrichment in IVD (96 h vs. 0 h). ( C ) Volcano plot and heatmap of Hedgehog pathway genes differentially expressed during IVD (96 h S1-S3 vs. 0 h C1-C3; RNA-seq). ( D-E ) Time course of GAS1 mRNA (qRT-PCR) and protein (western blotting) expression during IVD (mean ± SD; n  = 3; ANOVA with Tukey’s post-hoc test). ( F ) Immunofluorescence of GAS1 (Cy3, red), F-actin (FITC-phalloidin, green), and nuclei (DAPI, blue) in HESCs/MESCs pre- and post-IVD. Scale bars: 100 μm. ( G ) UMAP of single-cell transcriptomes from the maternal–fetal interface, with GAS1 expression feature plots across 32 cell clusters of blood, placenta, and decidua. IVD, in vitro decidualization Dynamic expression of GAS1 during decidualization. ( A ) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of decidual markers IGFBP1 (top) and PRL (bottom) in HESCs during IVD (mean ± standard deviation [SD]; n  = 3; analysis of variance [ANOVA] with Tukey’s post-hoc test). ( B ) GSEA plot of Hallmark Hedgehog Signaling pathway enrichment in IVD (96 h vs. 0 h). ( C ) Volcano plot and heatmap of Hedgehog pathway genes differentially expressed during IVD (96 h S1-S3 vs. 0 h C1-C3; RNA-seq). ( D-E ) Time course of GAS1 mRNA (qRT-PCR) and protein (western blotting) expression during IVD (mean ± SD; n  = 3; ANOVA with Tukey’s post-hoc test). ( F ) Immunofluorescence of GAS1 (Cy3, red), F-actin (FITC-phalloidin, green), and nuclei (DAPI, blue) in HESCs/MESCs pre- and post-IVD. Scale bars: 100 μm. ( G ) UMAP of single-cell transcriptomes from the maternal–fetal interface, with GAS1 expression feature plots across 32 cell clusters of blood, placenta, and decidua. IVD, in vitro decidualization To define the spatiotemporal expression of GAS1 during early pregnancy, we analyzed mouse uteri from GD1 to GD10. The RNA and protein expression patterns were largely concordant, although protein localization was more widespread (Fig.  2 A and B). On GD1, there was significant expansion of the uterine cavity, with nearly undetectable GAS1 expression in the uterine epithelium and stromal cells. On GD4, which corresponds to the preimplantation period of mouse embryos, GAS1 expression remained low. However, by GD5, coinciding with the embryonic implantation period, embryonic adhesion induces changes in the endometrium, leading to the formation of numerous glands around the implantation site. At this stage, prominent GAS1 expression was observed in the luminal/glandular epithelium and scattered stromal cells. On GD6, uterine stromal cells initiate decidualization, resulting in the formation of a primary decidualization zone (PDZ) characterized by robust expression of GAS1 throughout the PDZ. By GD8, these stromal cells undergo further decidualization, leading to the establishment of a secondary decidualization zone (SDZ) where GAS1 expression is even more pronounced. By GD10, as embryonic development progresses, decidual cells organize into a substantial, rounded mesometrial decidua (MD) at the maternal–fetal interface, with strong GAS1 expression persisting in the MD. We further analyzed GAS1 expression patterns in human early pregnancy uteri using spatial multiomics data from the CROST database, focusing on 6- and 8-week gestational specimens, encompassing the critical window of embryo implantation and decidualization. Spatial transcriptomic analysis revealed that GAS1 was predominantly localized in DSCs populations, exhibiting a distribution pattern highly concordant with that of the classical decidual markers IGFBP1 and PRL (Fig.  2 C). This conserved spatiotemporal pattern between humans and murine models suggests that GAS1 plays an important role in decidualization and early pregnancy maintenance. To investigate the potential association between GAS1 and RM, we conducted immunohistochemical assessments of decidual tissue samples obtained from 10 healthy pregnant controls and 10 patients with RM. Our findings indicated that GAS1 is ubiquitously expressed in decidual tissues. Notably, compared to that of healthy controls, the decidual tissue of patients with RM exhibited a deficiency and significantly lower expression levels of GAS1 (Fig.  2 D, P  = 0.0040). Given the substantial alterations in GAS1 expression during decidualization, we examined gene expression profiles from public databases, including GEO and TCGA, for endometrial tissues associated with recurrent implantation failure (RIF) ( GSE111974 ), endometriosis (EMs; GSE120103 ), and endometrial cancer (TCGA-uterine corpus endometrial carcinoma [UCEC]). Figure  2 E and F, and 2 G revealed a significant reduction in GAS1 expression among patients with RIF, infertile women with EMs, and patients with UCEC, respectively. These results suggested that GAS1 plays a significant role in female infertility and UCEC. Fig. 2 GAS1 is expressed in a spatiotemporally specific manner during early pregnancy and reproductive disorders. (A) Fluorescent in situ hybridization localization of GAS1 (red) in WT uteri on days 1, 4, 5, 6, 8, and 10 of pregnancy. (B) Immunofluorescence of GAS1 protein (green) in WT uteri on days 1, 4, 5, 6, 8, and 10 of pregnancy. CD45 staining outlines uterine leukocytes. Asterisks indicate the location of the blastocysts. Scale bars: 500 μm. (C) Spatial multiomics maps of the implantation site, placenta, and decidua of human females at 6 and 8 weeks of gestation, and localization of GAS1 and decidual cell markers IGFBP1 and PRL. (D) Representative images of GAS1 antigen staining (top) and GAS1 antigen immunostaining scores (bottom) in decidual tissue from healthy pregnant women and RM patients (n = 10, Mann-Whitney U test). (E) Expression landscape of GAS1 in the endometrium of women with RIF following in vitro fertilization treatment (n = 24) and fertile control women (n = 24) ( GSE111974 ). (F) Expression landscape of GAS1 in the endometrium of fertile and infertile women with or without stage IV EMs in dataset GSE120103 (n = 9, respectively). (G) The clinical stage expression of GAS1 in TCGA-UCEC patients (Control, n = 35; I, n = 340; II, n = 52; III, n = 124; IV, n = 29). The statistical difference of two groups was compared through the Wilcoxon rank-sum test, and that of the four groups was tested with Kruskal–Wallis test. EMs, endometriosis; IVD, in vivo decidualization; GD, gestation day; ge, glandular epithelium; le, luminal epithelium; M, mesometrial pole; MD, mesometrial decidua; NES, normalized enrichment score; PDZ, primary decidual zone; RIF, recurrent implantation failure; RM, recurrent miscarriage; SDZ, secondary decidual zone; st, stroma; TCGA, The Cancer Genome Atlas; UCEC, uterine corpus endometrial carcinoma; WT, wildtype GAS1 is expressed in a spatiotemporally specific manner during early pregnancy and reproductive disorders. (A) Fluorescent in situ hybridization localization of GAS1 (red) in WT uteri on days 1, 4, 5, 6, 8, and 10 of pregnancy. (B) Immunofluorescence of GAS1 protein (green) in WT uteri on days 1, 4, 5, 6, 8, and 10 of pregnancy. CD45 staining outlines uterine leukocytes. Asterisks indicate the location of the blastocysts. Scale bars: 500 μm. (C) Spatial multiomics maps of the implantation site, placenta, and decidua of human females at 6 and 8 weeks of gestation, and localization of GAS1 and decidual cell markers IGFBP1 and PRL. (D) Representative images of GAS1 antigen staining (top) and GAS1 antigen immunostaining scores (bottom) in decidual tissue from healthy pregnant women and RM patients (n = 10, Mann-Whitney U test). (E) Expression landscape of GAS1 in the endometrium of women with RIF following in vitro fertilization treatment (n = 24) and fertile control women (n = 24) ( GSE111974 ). (F) Expression landscape of GAS1 in the endometrium of fertile and infertile women with or without stage IV EMs in dataset GSE120103 (n = 9, respectively). (G) The clinical stage expression of GAS1 in TCGA-UCEC patients (Control, n = 35; I, n = 340; II, n = 52; III, n = 124; IV, n = 29). The statistical difference of two groups was compared through the Wilcoxon rank-sum test, and that of the four groups was tested with Kruskal–Wallis test. EMs, endometriosis; IVD, in vivo decidualization; GD, gestation day; ge, glandular epithelium; le, luminal epithelium; M, mesometrial pole; MD, mesometrial decidua; NES, normalized enrichment score; PDZ, primary decidual zone; RIF, recurrent implantation failure; RM, recurrent miscarriage; SDZ, secondary decidual zone; st, stroma; TCGA, The Cancer Genome Atlas; UCEC, uterine corpus endometrial carcinoma; WT, wildtype The structure of a protein determines its functions. As a glycosylphosphatidylinositol-anchored cell membrane protein (GPI-AP), GAS1 is embedded within sphingolipid/cholesterol-rich lipid rafts and dynamic membrane microdomains that orchestrate polarized trafficking, endocytic sorting, and signalosome assembly [ 31 ]. To map the interactome of GAS1 during decidualization, we performed co-immunoprecipitation coupled with mass spectrometry (Co-IP/MS) in HESCs at 0 h and 96 h post-IVD. This revealed 100 IVD-specific interactors (Fig. 3 A), with Gene Ontology (GO) functional enrichment analysis showing profound enrichment for the term “extracellular exosome” ( P = 4.13 −14 ; 40/100 proteins; Fig. 3 B). These data suggested that GAS1 is a potential regulator of EV biogenesis in decidual cells. CD63, a tetraspanin located on the surface of EVs, plays a critical role in EVs biosynthesis and secretion, and is a common EV tracer [ 32 , 33 ]. Immunofluorescence assays were conducted on decidual tissues obtained from healthy pregnant women and patients with RM. In the decidual tissues of healthy pregnant individuals, GAS1 and CD63 colocalize within the cytoplasm, with CD63 + EVs widely distributed in and around DSCs, glandular epithelial cells, and epithelial cells. Conversely, patients with RM exhibit compromised decidual development, characterized by abnormal size and morphology of DSCs and the absence of distinct glandular and spiral artery structures. The expression of GAS1 was lower, and CD63 + EVs were observed surrounding GAS1-positive DSCs, with both exhibiting reduced expression levels in the RM decidua (Fig. 3 C). To establish causality, artificially synthesized siRNA-GAS1 fragments were transiently transfected into HESCs to generate GAS1 knockdown (KD) HESCs (Additional file1 Figure S2A–B). siRNA-mediated GAS1 KD in IVD HESCs reduced the EV diameter (Additional file1 Figure S2C) and CD63 + EVs counts, and induced perinuclear clustering (Fig. 3 D). Proteomic analysis also revealed that the cargo of GAS1 KD EVs was functionally enriched in pathways related to pregnancy, EV function, and cellular interactions in the decidual microenvironment (Additional file1 Figure S2E). These findings demonstrated that GAS1 governs the quantity, spatial distribution, and function of EVs in decidual cells. To investigate how GAS1 regulates EVs, we conducted RNA-seq on si-NC and GAS1 KD HESCs at 96 h after IVD induction. The volcano plot and heatmap both showed that RAB39B levels were notably reduced in GAS1 KD HESCs (log2FoldChange = −5.45; Fig.  3 E). RAB39B, a member of the Rab GTPase family, is involved in plasma membrane budding, EVs formation, cytoskeletal transport, and absorption by receptor cells. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses revealed that the differentially expressed genes were notably enriched in pathways linked to the decidual microenvironment, such as cell adhesion molecules, ECM-receptor interactions, and glycerophospholipid metabolism (Fig.  3 F). Genetic manipulation experiments established a linear GAS1-RAB39B-EVs axis: (i) GAS1 knockout reduced RAB39B protein, whereas overexpression (OE) increased it (Fig.  3 G and I); (ii) IVD induction increased RAB39B expression, paralleling the CD63 + EVs surge (Fig.  3 H); and (iii) GAS1 OE boosted EV production, an effect abolished by RAB39B siRNA (Fig.  3 I, Additional file1 Figure S3). Fig. 3 GAS1 regulates decidual EVs secretion and delivery through RAB39B. ( A ) Venn diagram depicting shared and unique protein hits identified in IVD 96 h and 0 h HESCs of anti-GAS1 Co-IP. ( B ) Chord diagram of unique protein functional enrichment analysis of IVD 96 h with anti-GAS1 Co-IP. The “extracellular exosome” term contained 40/100 proteins. ( C ) Subcellular localization of CD63 + EVs and GAS1 in the pregnancy decidual tissue of healthy pregnant women and RM patients. ( D ) Subcellular localization of CD63 + EVs and GAS1 in si-NC and GAS1 KD HESCs. Immunofluorescence shows GAS1 protein (green), CD63-labeled EVs (red), and DAPI-labeled nuclear staining (blue). le, epithelial cells; st, stromal cells; ge, glandular epithelial cells. White arrows point to CD63 + EV-enriched regions. ( E ) Volcano plot and heatmap showing differentially expressed genes in RNA-seq of si-NC and GAS1 KD HESCs. ( F ) KEGG functional enrichment analysis of si-NC and GAS1 KD HESCs. ( G ) RAB39B protein levels in GAS1 KO and OE HESCs were detected by western blot. ( H ) Representative immunofluorescence images of RAB39B and CD63 + EVs in IVD 96 h vs. 0 h. ( I ) Representative immunofluorescence images of RAB39B and CD63 + EVs in GAS1 OE HESCs with/without si-Rab39b GAS1 regulates decidual EVs secretion and delivery through RAB39B. ( A ) Venn diagram depicting shared and unique protein hits identified in IVD 96 h and 0 h HESCs of anti-GAS1 Co-IP. ( B ) Chord diagram of unique protein functional enrichment analysis of IVD 96 h with anti-GAS1 Co-IP. The “extracellular exosome” term contained 40/100 proteins. ( C ) Subcellular localization of CD63 + EVs and GAS1 in the pregnancy decidual tissue of healthy pregnant women and RM patients. ( D ) Subcellular localization of CD63 + EVs and GAS1 in si-NC and GAS1 KD HESCs. Immunofluorescence shows GAS1 protein (green), CD63-labeled EVs (red), and DAPI-labeled nuclear staining (blue). le, epithelial cells; st, stromal cells; ge, glandular epithelial cells. White arrows point to CD63 + EV-enriched regions. ( E ) Volcano plot and heatmap showing differentially expressed genes in RNA-seq of si-NC and GAS1 KD HESCs. ( F ) KEGG functional enrichment analysis of si-NC and GAS1 KD HESCs. ( G ) RAB39B protein levels in GAS1 KO and OE HESCs were detected by western blot. ( H ) Representative immunofluorescence images of RAB39B and CD63 + EVs in IVD 96 h vs. 0 h. ( I ) Representative immunofluorescence images of RAB39B and CD63 + EVs in GAS1 OE HESCs with/without si-Rab39b EVs secreted by OE-GAS1 (GAS1-EVs) and NC HESCs were isolated and purified using ultracentrifugation. TEM and NTA analyses revealed that these EVs were 30–200 nm in size, round or oval, with a clear double-layer membrane, matching typical EV characteristics (Fig.  4 A). Western blotting confirmed the presence of the EV markers Alix, CD63, and TSG101, and the absence of calnexin (Fig.  4 B). These findings indicate the successful separation and high purity of EVs from the HESCs supernatant. Notably, GAS1-EVs displayed an 18% larger mean diameter than NC-EVs (151.7 vs. 136.8 nm, P  = 0.0008; Fig.  4 C), suggesting an altered cargo-loading capacity. Deep proteomic profiling of the EVs identified 2,336 proteins with an abundance spanning eight orders of magnitude (Fig.  4 D). Among the 370 differentially expressed proteins (235 upregulated and 135 downregulated), GAS1 emerged as the most significantly enriched cargo in GAS1-EVs (Log2Foldchange = 8.11, P  = 0.00020), underscoring its selective packaging into EVs (Fig.  4 E). Additional file 1 Table S3 lists the top 20 upregulated and downregulated proteins. Twenty-four EV markers were found in the two groups, including 11 conventional EV markers (FLOT1, HSP90AA1, HSP90AB1, CD63, HSPA8, CD9, HSPA4, CD81, FLOT2, PDCD6IP, and TSG101) and 13 newly EV markers (RAP1BL, A2M, JCHAIN, FN1, GSN, FLNA, MSN, STOM, PRDX2, ACTB, B2M, HBB, and LGALS3BP), confirming the reliability of the proteomics data (Fig.  4 F). GO analysis linked these proteins to key biological processes, including embryonic development, cell function, and angiogenesis (Fig.  4 G). Further functional validation experiments demonstrated the following. (i) In the blastocyst adhesion model, mouse blastocysts co-cultured with OE-GAS1 HESCs post-IVD showed increased outgrowth areas (2.3-fold larger outgrowth area vs. NC HESCs post-IVD, P  = 0.0159; Fig.  4 H). (ii) GAS1-EVs notably enhanced wound healing in trophoblast cell lines (HTR8: 24.9% faster wound closure at 24 h, P  < 0.0001; JEG3: 22.0% faster, P  = 0.0045; Fig.  4 I). (iii) GAS1-EVs significantly improved tube formation in HUVEC, with more nodes, junctions, and total vessel lengths than NC-EVs ( P  < 0.0001; Fig.  4 J). Our in vitro experimental results collectively indicate that GAS1-EVs undergo cargo reprogramming and potentially establish a self-reinforcing mechanism, which appears to contribute to enhanced blastocyst adhesion, trophoblast invasion, and angiogenesis in the tested model systems. Fig. 4 EVs secreted by overexpressing GAS1 promote blastocyst adhesion, trophoblast cell migration, and angiogenesis. ( A ) Identification of EVs secreted by NC and OE-GAS1 HESCs. Representative TEM images of EVs, showing typical round or oval structures with a clear double-layer membrane. Scale bars: 100 nm. ( B ) Representative western blotting images of EVs markers Alix, clanexin, CD63, and TSG101 secreted by NC and OE-GAS1 IVD 96 h. ( C ) NTA (particles/mL) of EVs secreted by NC and OE-GAS1 IVD 96 h. Data expressed as mean ± SD ( n  = 3, unpaired t-test). ( D ) Protein abundance plots comparing the proteomic profiles of EVs from NC and OE-GAS1 HESCs. ( E ) Volcano plot showing differentially expressed proteins (P-value < 0.05 and fold change ≤ 0.5 or ≥ 2). ( F ) Heatmap showing levels of 11 conventional EV protein markers and 13 new EV protein markers detected in samples. ( G ) GO analysis of differentially expressed proteins, including development, cell function, and angiogenesis highlighted in a bar plot. ( H ) Mouse blastocyst outgrowth analysis after co-cultured with NC or OE-GAS1 HESCs post-IVD ( n  = 6, unpaired t-test). ( I ) Wound healing assay of HTR8/JEG3 treated with NC-EVs or GAS1-EVs ( n  = 6, unpaired t-test). ( J ) Evaluation of angiogenesis function of HUVEC treated with NC-EVs or GAS1-EVs. The number of nodes, number of junctions, and total length were quantitatively analyzed. Data are expressed as mean ± SD ( n  = 9, unpaired t-test) EVs secreted by overexpressing GAS1 promote blastocyst adhesion, trophoblast cell migration, and angiogenesis. ( A ) Identification of EVs secreted by NC and OE-GAS1 HESCs. Representative TEM images of EVs, showing typical round or oval structures with a clear double-layer membrane. Scale bars: 100 nm. ( B ) Representative western blotting images of EVs markers Alix, clanexin, CD63, and TSG101 secreted by NC and OE-GAS1 IVD 96 h. ( C ) NTA (particles/mL) of EVs secreted by NC and OE-GAS1 IVD 96 h. Data expressed as mean ± SD ( n  = 3, unpaired t-test). ( D ) Protein abundance plots comparing the proteomic profiles of EVs from NC and OE-GAS1 HESCs. ( E ) Volcano plot showing differentially expressed proteins (P-value < 0.05 and fold change ≤ 0.5 or ≥ 2). ( F ) Heatmap showing levels of 11 conventional EV protein markers and 13 new EV protein markers detected in samples. ( G ) GO analysis of differentially expressed proteins, including development, cell function, and angiogenesis highlighted in a bar plot. ( H ) Mouse blastocyst outgrowth analysis after co-cultured with NC or OE-GAS1 HESCs post-IVD ( n  = 6, unpaired t-test). ( I ) Wound healing assay of HTR8/JEG3 treated with NC-EVs or GAS1-EVs ( n  = 6, unpaired t-test). ( J ) Evaluation of angiogenesis function of HUVEC treated with NC-EVs or GAS1-EVs. The number of nodes, number of junctions, and total length were quantitatively analyzed. Data are expressed as mean ± SD ( n  = 9, unpaired t-test) To specifically target the decidual microenvironment, we focused on the Atosiban–OXTR–ligand–receptor system because of its established benefits for uterine function, particularly in managing uterine contractions and improving implantation and pregnancy rates [ 34 ]. Molecular docking analysis revealed that Atosiban, a clinically used oxytocin antagonist, binds to the extracellular domain of OXTR (Fig. 5 A), identifying potential targetable sites for the pharmacological modulation of uterine activity. On day 2 of IVD in HESCs, we introduced 20 µM estradiol (E2) to simulate conditions of impaired endometrial receptivity characteristic of RM, which mimics the aberrant hormonal milieu observed in patients with RM. Quantitative analysis revealed that IVD alone reduced OXTR levels by 41% ( P < 0.0001), consistent with the contractile-inhibitory role of OXTR during normal implantation. However, E2 restored OXTR expression to 79.8% of baseline ( P = 0.0038; Fig. 5 B). This paradoxical estrogen-mediated OXTR upregulation mirrors the elevated OXTR levels clinically observed in patients with pregnancy failure [ 35 ]. Spatial analysis in murine pregnancy showed dynamic OXTR localization: on GD6-8, OXTR was expressed in cells of both the PDZ and SDZ, and its expression scope expanded along with the development of the decidua. On GD10, OXTR was expressed in the early placenta, particularly in the MD group (Fig. 5 C). In normal mid- to late-term pregnancy in mice, OXTR is mainly expressed in the MD and placental labyrinths. The RM mouse model demonstrated pathological OXTR overexpression at the early embryo absorption site and in some abnormal placental lobules at the late embryo absorption site (Fig. 5 D), revealing precise localized targets for pharmacological intervention. Fig. 5 OXTR expression dynamics in decidual microenvironment during normal pregnancy and RM. ( A ) Binding mode of Atosiban (red; PubChem CID: 5311010) to OXTR (green; PDB ID: 6TPK) by molecular docking, with key interaction sites (dashed lines). ( B ) Representative western blot images and relative quantification of OXTR protein levels in IVD interventions, with or without the addition of 20 µM E2 ( n  = 3, one-way ANOVA with Tukey’s test). ( C ) Immunofluorescence of OXTR (green) in WT murine uteri on days 6, 8, and 10 of pregnancy. Asterisks indicate the location of the blastocysts. ( D ) Hematoxylin and eosin (HE) staining and immunofluorescence of OXTR in a normal pregnant and RM mouse (early and late resorption site). CD45 staining (red) outlines uterine leukocytes. ERS, early resorption site; GD, gestation day; JZ, junctional zone; L, labyrinth; LRS, late resorption site; MD, mesometrial decidua; PDZ, primary decidual zone; SDZ, secondary decidual zone; std, stroma OXTR expression dynamics in decidual microenvironment during normal pregnancy and RM. ( A ) Binding mode of Atosiban (red; PubChem CID: 5311010) to OXTR (green; PDB ID: 6TPK) by molecular docking, with key interaction sites (dashed lines). ( B ) Representative western blot images and relative quantification of OXTR protein levels in IVD interventions, with or without the addition of 20 µM E2 ( n  = 3, one-way ANOVA with Tukey’s test). ( C ) Immunofluorescence of OXTR (green) in WT murine uteri on days 6, 8, and 10 of pregnancy. Asterisks indicate the location of the blastocysts. ( D ) Hematoxylin and eosin (HE) staining and immunofluorescence of OXTR in a normal pregnant and RM mouse (early and late resorption site). CD45 staining (red) outlines uterine leukocytes. ERS, early resorption site; GD, gestation day; JZ, junctional zone; L, labyrinth; LRS, late resorption site; MD, mesometrial decidua; PDZ, primary decidual zone; SDZ, secondary decidual zone; std, stroma We developed an innovative, targeted therapeutic platform using a multistep bioengineering process (Fig.  6 A). First, GAS1 OE DSCs were cultured to harvest EVs. We synthesized Atosiban-conjugated 3WJ-pRNA nanoparticles. GAS1-EVs were subsequently functionalized with these nanoparticles via cholesterol-mediated membrane anchoring, creating the final Atosiban-GAS1-EV complexes. This engineered system was formulated for intravenous delivery by leveraging both the uterine-targeting capability and decidual-supportive properties of GAS1-enhanced EVs. The 3WJ-pRNA nanostructure was engineered using three functionalized strands (Fig.  6 B). The cholesterol-anchored a3WJ-TEG strand enabled membrane integration into GAS1-EVs. The structural b3WJ strand maintained a three-way junction architecture, while the Atosiban-c3WJ-Alexa647 strand delivered both targeting specificity (via OXTR-binding Atosiban) and fluorescent tracking capability. All uracil and cytosine residues (lowercase) were 2’-deoxy-2’-fluoro modified to enhance nuclease resistance, with mass spectrometry confirming the molecular weights of these modified strands: a3WJ-TEG, 6,539.8 Da; b3WJ, 6,558.9 Da; Atosiban-c3WJ-Alexa647, 7,127.2 Da. Molecular characterization confirmed successful assembly, with 150–200 bp Atosiban-3WJ-pRNA nanoparticles showing distinct, sharp bands on 15% native polyacrylamide gel electrophoresis (Fig.  6 C). The AFM image illustrates the three-branched triangular structure of self-assembled trivalent Atosiban-3WJ-pRNA nanoparticles (Fig.  6 D). The constructs demonstrated excellent pharmaceutical properties, including 12-h thermal stability at body temperature (Fig.  6 E) and the absence of cytotoxicity up to 10,000 pM (Fig.  6 F). To facilitate the binding of 3WJ-pRNA nanoparticles to engineered GAS1-EVs, the two components were mixed at a ratio of 1,000:1 (determined by fluorescence-based binding assays, which showed complete EV saturation with less free 3WJ-pRNA in the supernatant; Additional file1 Figure S4) and incubated at 37 °C for 1 h. This incubation period was selected based on the rapid binding kinetics (observed within 10 min) and sustained complex stability (no significant degradation over 4 h; Additional file1 Figure S4). After incubation, the mixture was placed on ice for 1 h to stabilize the complex and yield the final product, termed Atosiban-GAS1-EVs. To evaluate biocompatibility, we first assessed hemocompatibility and observed that the Atosiban-GAS1-EVs group exhibited a clear and transparent supernatant, comparable to that of the NS control (Fig.  6 G). Next, we investigated the biological effects of GAS1-EVs and Atosiban-GAS1-EVs on HESCs decidualization. Confocal imaging confirmed the enhanced uptake of CD63 + EVs in the Atosiban-GAS1-EVs group, with clear co-localization of Atosiban-3WJ-pRNA nanoparticles and CD63 + EVs (Fig.  6 H). Functionally, both GAS1-EVs and Atosiban-GAS1-EVs significantly upregulated PRL mRNA levels in IVD 2D HESCs, with Atosiban-GAS1-EVs showing superior efficacy (110.4-fold vs. 71.5-fold; P  = 0.029; Fig.  6 I). Consistent with this, HESCs treated with GAS1-EVs or Atosiban-GAS1-EVs exhibited cytoskeletal remodeling, appearing more rounded and plumper than the IVD 2D controls (Fig.  6 J). Collectively, these results indicate that Atosiban-GAS1-EVs possess enhanced targeting abilities while simultaneously promoting decidual development. Fig. 6 Bioengineering and functional characterization of Atosiban-GAS1-EVs therapeutic platform. ( A ) Construction workflow: Isolation of GAS1-EVs from GAS1 OE DSCs, 3WJ-pRNA self-assembly with Atosiban conjugation, cholesterol-mediated EV-nanoparticle fusion, and intravenous therapeutic delivery. ( B ) Sequences and molecular weights of Atosiban-3WJ-pRNA nanoparticles oligos with 2’-fluoro modifications (lowercase bases). ( C ) Naive PAGE verifying Atosiban-3WJ-pRNA nanoparticle assembly (150–200 bp). ( D ) AFM revealing the triangular three-branched structure of Atosiban-3WJ-pRNA nanoparticles. ( E ) Thermal stability profile showing maintained fluorescence intensity over 12 h at 37 °C. ( F ) Cytotoxicity assessment of Atosiban-3WJ-pRNA was conducted using the CCK8 assay with HESCs. No toxicity observed ( n  = 4, P  > 0.05). ( G ) hemolysis rates upon various treatments ( n  = 3). ( H ) Representative immunofluorescence images illustrate the presence of CD63 + EVs (green) and Alexa647 nanoparticles (purple) following incubation with HESCs and the respective GAS1-EVs or Atosiban-GAS1-EVs. ( I ) The relative quantitative changes in PRL mRNA levels were measured following the addition of GAS1-EVs or Atosiban-GAS1-EVs based on the IVD day 2 HESCs ( n  = 3, one-way ANOVA with Tukey’s test). ( J ) Images of the cytoskeleton after incubation with additional GAS1-EVs or Atosiban-GAS1-EVs, based on the IVD day 2 model, with merged data displaying the fluorescence channels for FITC (phalloidin) and DAPI Bioengineering and functional characterization of Atosiban-GAS1-EVs therapeutic platform. ( A ) Construction workflow: Isolation of GAS1-EVs from GAS1 OE DSCs, 3WJ-pRNA self-assembly with Atosiban conjugation, cholesterol-mediated EV-nanoparticle fusion, and intravenous therapeutic delivery. ( B ) Sequences and molecular weights of Atosiban-3WJ-pRNA nanoparticles oligos with 2’-fluoro modifications (lowercase bases). ( C ) Naive PAGE verifying Atosiban-3WJ-pRNA nanoparticle assembly (150–200 bp). ( D ) AFM revealing the triangular three-branched structure of Atosiban-3WJ-pRNA nanoparticles. ( E ) Thermal stability profile showing maintained fluorescence intensity over 12 h at 37 °C. ( F ) Cytotoxicity assessment of Atosiban-3WJ-pRNA was conducted using the CCK8 assay with HESCs. No toxicity observed ( n  = 4, P  > 0.05). ( G ) hemolysis rates upon various treatments ( n  = 3). ( H ) Representative immunofluorescence images illustrate the presence of CD63 + EVs (green) and Alexa647 nanoparticles (purple) following incubation with HESCs and the respective GAS1-EVs or Atosiban-GAS1-EVs. ( I ) The relative quantitative changes in PRL mRNA levels were measured following the addition of GAS1-EVs or Atosiban-GAS1-EVs based on the IVD day 2 HESCs ( n  = 3, one-way ANOVA with Tukey’s test). ( J ) Images of the cytoskeleton after incubation with additional GAS1-EVs or Atosiban-GAS1-EVs, based on the IVD day 2 model, with merged data displaying the fluorescence channels for FITC (phalloidin) and DAPI NP, RM, and RM mice treated with GAS1-EVs or Atosiban-GAS1-EVs (GD5 to 9) were sacrificed on GD10 for decidual microenvironment analysis or GD14 for fetal resorption rate assessment (Fig. 7 A). We first analyzed Gas1 expression in RM and NP decidua by reanalyzing two publicly available single-cell sequencing datasets comprising cells from GD9/GD10 decidual tissues [ 2 ]. Major maternal-derived cell types, including uterine endothelial cells (uECs), DSCs, immune cells, smooth muscle cells, and epithelial cells, were identified in both RM and NP mice (Additional file1 Figure S5). The RM mice exhibited significantly lower Gas1 levels than those in the NP mice, supporting their functional relevance to our therapeutic model (Fig. 7 B). Using small-animal live imaging technology to assess the retention effects in the uterus, we observed that Atosiban-GAS1-EVs exhibited an extended retention time within the gravid uterine tissue (Fig. 7 C). HE staining revealed the absence of an inflammatory response and tissue abnormalities in the major organs of the mice (Fig. 7 D). Gross morphological examination on GD10 revealed distinct pregnancy phenotypes across the groups (Fig.  7 E). Histological evaluation revealed significant differences in decidual maturation between the experimental groups. Mesenchymal–epithelial transition (MET) hub refers to the histological areas within the decidua that contain persistent, undifferentiated endometrial stromal fibroblasts (eSFs) and proliferative DSCs that have not fully undergone epithelial-like transformation. Therefore, an increased MET hub reflects impaired or delayed decidual maturation. GD10 RM mice exhibited an expanded MET hub and delayed decidual maturation, whereas NP mice showed typical MET progression, with mature DSCs exhibiting a characteristic vacuolated cytoplasm and fused nuclei (Fig.  7 F). GAS1-EVs and Atosiban-GAS1-EVs treatment significantly rescued these MET defects, reducing the MET hub by 21.8% and 25.2%, respectively, compared to untreated RM mice ( P  = 0.0026, P  = 0.001; Figs.  7 G). Immunofluorescence analysis revealed key therapeutic mechanisms (Fig.  7 H). Compared with normal pregnancy controls, RM mice exhibited characteristic pathological features in decidual tissues, including markedly delayed decidualization (reduced Gas1 expression in DSCs), aberrant vascular formation (fragile and narrow CD31-positive vessels), and impaired trophoblast invasion (decreased Gjb3-positive cell infiltration). Both treatment groups demonstrated significant restorative effects, with Atosiban-GAS1-EVs showing superior therapeutic outcomes. The treatments promoted the formation of larger, more mature blood vessels in the decidua, while improving the overall vascular architecture. Additionally, they substantially enhanced trophoblast invasion capacity. Notably, the interventions increased both the number and size of DSCs, with the cellular morphology shifting toward the mature phenotype observed in normal pregnancy. These cellular improvements translated into significant pre-clinical outcomes. While RM mice showed a 32.3% fetal resorption rate at GD14, treatment with GAS1-EVs and Atosiban-GAS1-EVs reduced the resorptions to 21.5% and 6.2% ( P  = 0.0153; Figs.  7 I). Fig. 7 Atosiban-GAS1-EVs contribute to the decidual microenvironment and prevented pregnancy loss in abortion mice. ( A ) Schematic diagram of the experimental process for abortion mice. Pregnant mice were injected in the tail vein with normal saline (control), GAS1-EVs or Atosiban-GAS1-EVs on GD5 to 9 daily, with endpoints at GD10 (decidual microenvironment) and GD14 (fetal resorption rates). ( B ) UMAP projection of single-cell sequencing data from GD9–10 decidua in RM and NP mice. Left: UMAP showing the major cell types. Right: Gas1-positive cells (red gradient). ( C ) Enhanced uterine accumulation of Atosiban-GAS1-EVs. Left: In vivo images. Right: Ex vivo images of the major organs (excitation: 608 nm, emission: 680 nm). ( D ) In vivo toxicity of Atosiban-GAS1-EVs. HE-stained slice images of major organs: heart, brain, lung, liver, kidney, and spleen. ( E ) Representative pictures of GD10 uteri. ( F ) HE staining showing the decidual morphology in the antimesometrial region: (a) endometrial stromal fibroblasts (eSF), (b) proliferative DSCs, (c) mature DSCs with vacuolated cytoplasm, and (d) mature DSCs with fused nuclei. ( G ) Quantitative analysis of MET hub area ( n  = 3, one-way ANOVA). ( H ) Immunofluorescence images of decidual vascularization, trophoblast invasion, and Gas1 expression patterns. Representative images from each experimental group show low-magnification (6×) and high-resolution (40×) overviews of the decidual architecture, high-resolution (40×) views of CD31 (vascular endothelial cells, red), Gjb3 (trophoblasts, green, white arrows indicate invasive fronts penetrating the decidua), and Gas1 (DSCs, yellow) co-staining with Dapi (nuclei, blue). Scale bars: 500 μm (6× images), 100 μm (40× images). ( I ) Representative pictures of GD14 uterine, placenta, and embryo tissues. ( J ) Embryo resorption rates in NP ( n  = 3) and RM models with various treatments ( n  = 5, one-way ANOVA). Data are presented as mean ± SD. EP, early placenta; MD, mesometrial decidua Atosiban-GAS1-EVs contribute to the decidual microenvironment and prevented pregnancy loss in abortion mice. ( A ) Schematic diagram of the experimental process for abortion mice. Pregnant mice were injected in the tail vein with normal saline (control), GAS1-EVs or Atosiban-GAS1-EVs on GD5 to 9 daily, with endpoints at GD10 (decidual microenvironment) and GD14 (fetal resorption rates). ( B ) UMAP projection of single-cell sequencing data from GD9–10 decidua in RM and NP mice. Left: UMAP showing the major cell types. Right: Gas1-positive cells (red gradient). ( C ) Enhanced uterine accumulation of Atosiban-GAS1-EVs. Left: In vivo images. Right: Ex vivo images of the major organs (excitation: 608 nm, emission: 680 nm). ( D ) In vivo toxicity of Atosiban-GAS1-EVs. HE-stained slice images of major organs: heart, brain, lung, liver, kidney, and spleen. ( E ) Representative pictures of GD10 uteri. ( F ) HE staining showing the decidual morphology in the antimesometrial region: (a) endometrial stromal fibroblasts (eSF), (b) proliferative DSCs, (c) mature DSCs with vacuolated cytoplasm, and (d) mature DSCs with fused nuclei. ( G ) Quantitative analysis of MET hub area ( n  = 3, one-way ANOVA). ( H ) Immunofluorescence images of decidual vascularization, trophoblast invasion, and Gas1 expression patterns. Representative images from each experimental group show low-magnification (6×) and high-resolution (40×) overviews of the decidual architecture, high-resolution (40×) views of CD31 (vascular endothelial cells, red), Gjb3 (trophoblasts, green, white arrows indicate invasive fronts penetrating the decidua), and Gas1 (DSCs, yellow) co-staining with Dapi (nuclei, blue). Scale bars: 500 μm (6× images), 100 μm (40× images). ( I ) Representative pictures of GD14 uterine, placenta, and embryo tissues. ( J ) Embryo resorption rates in NP ( n  = 3) and RM models with various treatments ( n  = 5, one-way ANOVA). Data are presented as mean ± SD. EP, early placenta; MD, mesometrial decidua

To achieve uterine-targeted therapy, a novel nanoplatform was developed by conjugating the oxytocin receptor-targeting peptide Atosiban to three-way junction pRNA (3WJ-pRNA) nanoparticles, enabling the precise delivery of GAS1-EVs. This targeted approach restored decidual microenvironment organization and reduced fetal resorption rates in murine RM models. Our results unveil the GAS1-RAB39B axis as a master regulator of EV-mediated decidual communication and pioneer a non-invasive, RNA nanotechnology-driven strategy for RM treatment. As a cell cycle inhibitor and tumor suppressor, GAS1 inhibits tumor growth in various cancers by affecting the cell cycle and apoptosis pathways [ 6 ]. GAS1 plays diverse roles depending on the context; it inhibits colorectal cancer by reducing aerobic glycolysis [ 36 ] triggers apoptosis in neuroblastoma via caspase-3 [ 7 ] and induces glioblastoma cell death through autocrine and paracrine actions [ 8 ]. In contrast, during embryonic development, GAS1 is highly expressed in proliferative areas without G0 phase arrest [ 9 ], indicating a positive regulatory role. Additionally, in endothelial cells, GAS1 prevents apoptosis and supports vascular integrity [ 10 ], showcasing its varied regulatory functions. This indicates that GAS1 functions as a pleiotropic regulatory protein, exhibiting distinct roles contingent on the cellular environment [ 37 ]. In reproductive medicine, GAS1 is crucial in ovarian granulosa cells, aiding ovulation and luteinization and directing primordial germ cell migration during embryogenesis [ 11 , 12 ]. GAS1 mutants show reduced fertility due to fewer germ cells [ 12 ]. Recent studies have revealed increased GAS1 levels during implantation and lower levels in patients with RIF. GAS1, regulated by miR-34c-5p in small EVs, supports early embryonic implantation [ 13 ]. Our multiomics analyses showed that GAS1 is prevalent in mature decidual cells but is reduced in RIF, endometriosis, and UCEC. Immunohistochemistry confirmed low GAS1 expression in patients with RM, highlighting its positive role in pregnancy. GAS1, a membrane receptor, enhances the Hedgehog signaling pathway alongside Patched1, Cdo, and BOC, influencing cell growth and embryonic development [ 38 – 41 ]; however, it exhibits non-canonical functions in the decidual tissue. Even without the GPI anchor sites, GAS1 mutants show inhibited growth, suggesting that its function may not depend on membrane localization [ 42 ]. This paradox prompted us to investigate the alternative mechanisms of action. Co-IP/MS and GO analyses identified a significant link between GAS1 and EV localization. In tissues from patients with RM, low GAS1 levels were associated with abnormal EV secretion, supporting previous findings that decidual dysfunction affects EV communication and placental function [ 43 ]. Gene editing confirmed the role of GAS1 in the regulation of EV secretion and localization to maintain decidual homeostasis. RAB39B is a pivotal molecule mediating the influence of GAS1 on EV regulation. As a Rab GTPase located in the endoplasmic reticulum/cis-Golgi network, RAB39B has been implicated in studies reporting that mutations associated with Parkinson’s disease affect basolateral exosome release [ 44 , 45 ]. Furthermore, proteomic analysis of GAS1-enhanced EVs revealed two critical features reverse decidual dysfunction: autologous enrichment (GAS1 itself, Log2FC = 8.11) and reproductive cargo involved in embryonic differentiation, trophoblast invasion, and placental vascular formation (such as MOB1B, ADAMTS-4, and WNT2) [ 46 – 48 ]. The therapeutic effect of GAS1-EVs may be explained by an engineered self-amplifying cycle; the delivered GAS1 protein activates RAB39B-mediated EV biogenesis in recipient cells, which in turn produces new GAS1-enriched EVs. These experiments demonstrate that GAS1-EVs enhance trophoblast migration, blastocyst adhesion, and endothelial tube formation. Our research complements yet extends Ma et al.‘s findings on the HIF2α-RAB27B pathway in PNAS [ 3 ], which regulates EV basal secretion and metabolic functions. Our data demonstrate that GAS1-RAB39B primarily governs EV biogenesis and reprograms reproductive-functional proteins in EV cargo, establishing EV-based communication as the primary mode of maternal–fetal crosstalk. In recent years, reproductive endocrinology has seen increased interdisciplinary research combining medicine and engineering, such as the use of hydrogel-encapsulated EVs [ 49 , 50 ] and collagen scaffolds with exosomes [ 51 ] for uterine treatments. However, studies on gestational treatments are limited. One study found that injecting hyaluronic acid gel-encapsulated trophoblast-derived exosomes into the uterus could treat RM [ 52 ]; however, invasive methods during pregnancy are challenging. For RM due to decidual dysfunction, treatment must avoid altering the uterine structure. Traditional materials, such as hydrogels and collagen scaffolds, are not suitable; therefore, a new absorbable, targetable, and structureless biomaterial is required. Our study used 3WJ-pRNA nanotechnology to create an Atosiban-GAS1-EVs delivery system that precisely regulates the decidual microenvironment. The delivery system offers two main advantages over traditional implantable materials: enhanced stability, targeting, and tracing due to the Atosiban-3WJ-pRNA nanostructure; and non-invasive administration, which avoids mechanical interference with the pregnant uterus. This approach shows potential for intervention in the uterine microenvironment during pregnancy; however, further validation is needed. Previous scStereo-seq data revealed dysfunctional decidual hubs in the E8.5/E9.5 implantation sites of DBA/2-mated CBA/J mice, characterized by spatial disorganization and altered cell composition. This indicates that the decidualization process was significantly delayed in abortion-prone mice [ 2 ]. A recent study further confirmed the profound decidual dysfunction of RM mice, characterized by a reduction in the Str_0 stromal cell subpopulation (critical for decidual support) and dysregulated intercellular signaling [ 53 ]. These findings suggest that the RM model is a suitable framework for the investigation into the therapeutic effects of EVs on the decidual microenvironment. Experiments demonstrated that Atosiban-GAS1-EVs significantly improved the decidual microenvironment, re-establishing decidual MET hubs and functional markers. While this study presents a promising nanoplatform for uterine-targeted therapy, some limitations should be acknowledged. First, although our stable GAS1-overexpressing stromal cell system yields relatively homogeneous EVs and circumvents the low and inconsistent loading efficiency reported in active-loading strategies [ 54 , 55 ], the regulatory mechanisms underlying the self-amplifying GAS1-EVs cycle remain incompletely defined. Particularly, potential crosstalk between GAS1, RAB39B, and other Rab GTPases may introduce variability in EV biogenesis, highlighting the need for mechanistic studies to achieve more predictable EV production. Second, while our murine model provides valuable insights into decidual dysfunction and demonstrates therapeutic efficacy, species-specific differences in placental development and uterine physiology limit its ability to fully recapitulate human reproductive biology. Although our findings are supported by re-analysis of human single-cell datasets, future studies should incorporate humanized models [ 56 ] or advanced in vitro systems such as ex vivo decidual tissues [ 57 ] and organoids [ 58 – 60 ] to better mimic human pathophysiology. Third, a comprehensive safety assessment remains crucial for clinical translation. While our current data demonstrate favorable short-term biosafety through cytotoxicity assays, hemolysis tests, histological evaluation, and normal embryonic development, the long-term biosafety and potential off-target effects of the Atosiban-GAS1-EVs system still require thorough evaluation. Importantly, because the envisioned clinical use would involve only short-term administration during early pregnancy rather than chronic treatment, some long-term risks may be mitigated. Nevertheless, extended toxicology studies, detailed biodistribution analyses, immune response profiling, and multi-generational offspring follow-up remain necessary to fully establish safety. Several technical challenges need to be addressed from a translational perspective. The long-term stability of GAS1 overexpression requires further optimization to ensure consistent EV production. Additionally, standardization of decidual induction protocols is needed to minimize batch variability in EV cargo composition. From a manufacturing perspective, while traditional methods, such as ultracentrifugation, yield EVs of sufficient purity for research purposes, more scalable and reproducible techniques are essential for clinical-grade production [ 61 ]. Consequently, addressing these aspects will be crucial for advancing this technology toward therapeutic applications.

The Atosiban-GAS1-EV platform demonstrates therapeutic efficacy through three synergistic mechanisms as follows: self-amplifying GAS1-EV production, regulatory cargo delivery via the GAS1-RAB39B axis, and uterine-targeted delivery using a nanoparticle-EV complex. This approach shifts recurrent miscarriage treatment from transient supplementation to sustained microenvironment reprogramming. Our findings suggest the potential significance of the GAS1-RAB39B-EV axis in decidual homeostasis and establish a targeted 3WJ-pRNA nanotechnology delivery system for uterine intervention, providing a promising non-invasive strategy for RM that warrants further investigation.

Recurrent miscarriage (RM), which affects 1% of couples trying to conceive, remains a clinical challenge with a poorly understood pathophysiology [ 1 ]. Decidual microenvironment dysfunction is a key contributor to RM, particularly through impaired intercellular communication mediated by decidual stromal cell (DSC)-derived extracellular vesicles (EVs) [ 2 – 5 ]. While EVs have emerged as crucial regulators of maternal–fetal crosstalk, fundamental questions remain regarding the molecular triggers governing their release and functional specificity within the decidual microenvironment. The pleiotropic growth arrest-specific gene 1 (GAS1) has recently garnered attention in the field of reproductive biology. Originally characterized as a cell cycle regulator [ 6 ], GAS1 exhibits context-dependent functions, including tumor suppression [ 7 , 8 ], embryogenesis promotion [ 9 ], and vascular maintenance [ 10 ]. In reproduction, GAS1 participates in ovulation [ 11 ], germ cell migration [ 12 ], and implantation window dynamics [ 13 ], yet its role in EV-mediated decidual microenvironment remodeling remains unclear. EV-based therapies offer compelling advantages in the treatment of reproductive disorders, including innate biocompatibility and modifiable cargo, with no ethical concerns [ 14 ]. These natural EVs demonstrate exceptional potential to restore endometrial function, enhancing implantation receptivity, and stimulating uterine angiogenesis [ 15 ]. However, their clinical translation faces a critical challenge: their inability to selectively accumulate in target tissues, leading to suboptimal efficacy and potential systemic exposure [ 16 ]. RNA nanotechnology has been utilized to overcome this limitation. Bacteriophage phi29-derived three-way junction pRNA (3WJ-pRNA) nanoparticles provide an ideal solution, exhibiting excellent attributes such as [ 1 ] an ultrastable scaffold structure resistant to RNase degradation [ 2 ], precise modular assembly, and [ 3 ] programmable targeting capacity [ 17 – 20 ]. Mechanistically, 3WJ-pRNA can be precisely functionalized with ligands that specifically bind to target cell receptors. By integrating functionalized 3WJ-pRNA nanoparticles into EVs, it is feasible to display aptamers on the membrane surface of the EVs, thereby achieving selective targeting and specific delivery of EVs [ 21 ]. Therefore, we aimed to determine how GAS1 affects the release of EVs and recipient cells into the decidual microenvironment. We explored the use of Atosiban-conjugated 3WJ-pRNA for the targeted delivery of therapeutic GAS1-enhanced EVs to the uterus in early pregnancy, offering a promising non-invasive targeted treatment for women with RM.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.