Phase 1 dose-escalation trial of sub-endometrial injection of human embryonic stem cells-derived immunity-and-matrix-regulatory cells to promote endometrial angiogenesis in refractory intrauterine adhesion

Abstract

Clinical application of mesenchymal stem cells for endometrial repair has been hampered by variability in cell quality, large-scale production, and uncertainty regarding the optimal delivery route. In this study, we investigated the therapeutic potential of clinical-grade human embryonic stem cell-derived immunity-and-matrix-regulatory cells (IMRCs) for treating refractory moderate-to-severe intrauterine adhesion (IUA). In a rabbit IUA model, sub-endometrial injection of IMRCs significantly reduced fibrosis and enhanced endometrial angiogenesis, outperforming uterine perfusion. Transcriptomic analysis revealed distinct pro-angiogenic gene expression profiles between the two delivery routes. In vitro, IMRCs co-cultured with endometrial stromal cells (ESCs) markedly enhanced angiogenic potential compared to either cell type alone. Protein array analysis of the co-culture supernatant showed elevated levels of angiogenic factors, with functional assays confirming that inhibition of ANGPTL4, a non-canonical pro-angiogenic mediator, impaired angiogenesis. In a first-in-human, single-center, phase 1 dose-escalation trial involving 18 patients with refractory IUA, high-dose sub-endometrial IMRC injection promoted angiogenesis, reduced uterine scarring, and improved pregnancy outcomes, with no safety concerns observed over 3 years of follow-up. These findings highlight the translational promise of IMRCs as a novel therapeutic strategy for endometrial regeneration in severe IUA.

Graphical abstract

Sub-endometrial injection of clinical-grade human endometrial stromal cell-derived immunity-and-matrix-regulatory cells (IMRCs) promotes angiogenesis and alleviates fibrosis in refractory intrauterine adhesion. This first-in-human trial demonstrates the safety and therapeutic potential of IMRCs, underscoring their promise as a regenerative strategy for endometrial repair.

Introduction

Intrauterine adhesion (IUA), also known as Asherman syndrome, is a fibrotic condition characterized by damage to the endometrial basal layer, followed by partial or complete adhesions within the uterine cavity.¹ Repeated intrauterine surgeries, endometrial tuberculosis, and chronic endometritis are the leading causes of IUA.^2,3,4 Severe damage to the endometrium can trigger a persistent inflammatory response, excessive deposition of extracellular matrix (ECM) proteins, and the formation of abnormal avascular fibrotic areas, all of which hinder endometrial regeneration.⁵ IUA is often linked to infertility, hypomenorrhea, pregnancy loss, and related complications, particularly in moderate-to-severe cases, posing a notable burden on affected couples.^2,6

Transcervical resection of adhesions using hysteroscopy (TCRA) is the standard treatment for IUA.⁷ To prevent re-adhesion and promote endometrial repair, physical barriers and hormone replacement therapy are often used as adjuncts.^7,8 While these strategies offer therapeutic benefits, they are limited by poor endometrial regeneration, high adhesion recurrence, and persistent refractory thin endometrium.^9,10 Thus, novel therapies are urgently needed to reconstruct the uterine cavity and restore endometrial function in moderate-to-severe IUA.

Mesenchymal stem cells (MSCs), a type of adult stem cell, possess self-renewal capacity and multipotent differentiation potential.¹¹ Emerging evidence supports the therapeutic potential of MSCs for IUA, with preclinical and clinical studies showing enhanced endometrial thickness (EMT) and improved pregnancy outcomes without major safety concerns.^12,13,14,15 However, there are some limitations in the clinical application of MSCs. Based on the different sources, MSCs can be classified into human bone marrow-derived MSCs, umbilical cord-derived MSCs (UCMSCs), and adipose tissue-derived MSCs. However, their therapeutic use is restricted by donor variability, inconsistent cell quality, and the lack of standardized, industrial-scale production protocols.¹⁶

These limitations may be addressed by using clinical-grade human embryonic stem cell-derived MSC-like cells, which we refer to as immunity-and-matrix-regulatory cells (IMRCs).^16,17 IMRCs are derived from clinical-grade human embryonic stem cells via stepwise, serum-free differentiation, yielding a homogeneous population with MSC surface markers and trilineage potential. Transcriptome profiling revealed close clustering with UCMSCs in unsupervised analyses.^16,17

IMRCs offer several advantages, including strict quality control, scalability for large-scale production, and ease of genetic modification.^16,18,19 Furthermore, a series of biosafety evaluations were performed according to the Guidelines for Human Somatic Cell Therapies and Quality Control of Cell-based Products issued by the China Food and Drug Administration to ensure both short-term and long-term safety.¹⁶ IMRCs exhibited a favorable safety profile in mice and monkeys, showing no long-term persistence or tumorigenic transformation. Their smaller size compared to UCMSCs may facilitate capillary transit and enhance therapeutic efficacy. Nonetheless, their effectiveness in repairing the endometrium in IUA remains to be determined.¹⁶

The therapeutic efficacy and underlying mechanisms of MSC treatment are influenced by the route of administration. For example, intravenous MSCs outperformed intraperitoneal delivery in colitis models, while intrathecal injection showed superior efficacy over intravenous infusion in progressive multiple sclerosis, suggesting that administration route critically influences therapeutic outcomes, potentially including IUA treatment.^20,21 These findings suggest that different administration routes may result in varying therapeutic effects in the treatment of IUA. Current MSC delivery methods for IUA include intrauterine perfusion (IUP), endometrial injection, intra-arterial injection, and intravenous infusion.^{12,14,15,22,23,24} Among these, IUP is the most commonly used method.²⁴ However, direct comparative studies assessing the therapeutic efficacy of different administration routes in IUA remain limited.

Angiogenesis is essential for the repair of damaged endometrium, as it activates endogenous mechanisms that support continuous regeneration and enhance endometrial receptivity, thereby improving pregnancy outcomes. MSCs promote this process mainly by modulating the local microenvironment through paracrine signaling rather than direct tissue integration.^14,25,26 Endometrial stromal cells (ESCs), which constitute the majority of resident endometrial cells, play a key role in tissue repair and angiogenesis.^27,28,29 Through sub-endometrial injection, IMRCs can directly interact with ESCs, potentially enhancing their pro-repair and pro-angiogenic functions.

In this study, we evaluated IMRC-mediated angiogenesis in a rabbit IUA model using sub-endometrial injection and uterine perfusion. To explore IMRC-ESC interactions, we performed single-cell RNA sequencing (scRNA-seq) and CellChat analysis, followed by a co-culture assay and protein array to identify pro-angiogenic factors. A clinical study in 18 patients with refractory moderate-to-severe IUA further assessed the safety and efficacy of sub-endometrial IMRC injection.

Results

Sub-endometrial injection of IMRCs exerted stronger pro-angiogenic and immunomodulatory effects

A schematic overview of the in vivo experimental workflow is shown in Figure 1A. One week after IUA modeling with 95% ethanol, IMRCs were administered via either sub-endometrial injection or uterine perfusion. To evaluate IMRC biodistribution and retention, we investigated the in vivo localization of GFP-labeled IMRCs. Immunofluorescence staining revealed clear colocalization of GFP⁺ IMRCs with ESCs (identified by the Collagen1a1 marker) at day 3 post-injection (Figures S1A and S1B), suggesting early retention and integration into the endometrial stroma. The GFP signal gradually declined over time and was nearly undetectable by day 14, indicating limited long-term engraftment (Figure S1A and S1B).

Figure 1.

The comparison of pro-angiogenic effect between two administration methods

(A) Schematic diagram of the animal experiment. (B and C) The fibrotic area of endometrial tissue among sham, model, IMRC injection, and IMRC perfusion groups was measured by Masson staining. Scale bar, 500 μm. (D) The ultrasonic color flow images. The uterine region is shown within the red dotted line. The red area is the blood flow toward the probe; the blue area is the blood flow away from the probe. (E and F) The microvessel density (MVD) of endometrial tissue was measured by counting the number of CD31⁺microvessels from five randomly selected fields per section. Scale bar, 40 μm. (G) The relative expression of VEGF was measured by quantitative real-time PCR. (H and I) The expression of VEGF was measured by the percentage of VEGF⁺ cells from five randomly selected fields per section. (J) The heatmap shows the angiogenesis-related pathways enriched among sham, model, IMRC injection, and IMRC perfusion groups. The red color in the legend indicates that the activity of the pathway was upregulated, whereas the blue color indicates that the activity of the pathway was downregulated. The red text shows the activated pathway in the IMRC injection group compared to that in the other three groups. (K) The heatmap shows the angiogenesis-associated differentially expressed genes (DEGs) among sham, model, IMRC injection, and IMRC perfusion groups. The red color in the legend indicates enhanced gene expression, whereas the blue color indicates reduced gene expression. The dots in the bar chart represent the number of biological replicates, which is five animals per group. ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Data shown as mean ± SD.

Compared to the sham group, the rabbit IUA model group displayed a significant increase in fibrotic area, as demonstrated by Masson’s trichrome staining, along with a marked reduction in ultrasonic blood flow and microvessel density (MVD) (Figures 1B–1F). Both administration routes alleviated fibrosis and enhanced MVD and blood flow compared to the untreated model group, with the effects being more pronounced following sub-endometrial injection (Figures 1B–1F). Consistently, vascular endothelial growth factor (VEGF), a key pro-angiogenic factor, was significantly upregulated following IMRC treatment, particularly in the IMRC injection group when compared to the model group (Figures 1G–1I).

RNA-seq analysis was performed on rabbit endometrial tissues from the sham, model, IMRC injection, and IMRC perfusion groups. Pathways affected by IMRC were further evaluated through Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis (Figures S1C–S1H). Enrichment pathways in the perfusion group compared to the model group include “immune response” and “cell division” (Figures S1C–S1H). However, more enrichment pathways in the injection group were activated compared to the model group, which include “positive regulation of proliferation,” “integrin-mediated signaling pathway,” “T cell differentiation,” “cell chemotaxis,” “immune response,” “phosphatidylinositol 3-kinase (PI3K)-Akt signaling pathway,” and “cytokine-cytokine receptor interaction” (Figures S1C–S1H). Enrichment pathways in the injection group compared to the perfusion group include “immune response,” “cell division,” “cytokine-cytokine receptor interaction,” “T helper 1 (Th1) and Th2 differentiation,” and “Th17 cell differentiation” (Figures S1C–S1H). The quantitative real-time PCR analysis (Figures S1I–S1J) demonstrated increased tumor necrosis factor α (TNF-α) and decreased interleukin-10 (IL-10) in the model group compared to the sham group. Both IMRC-treated groups reversed this imbalance, with the injection group showing lower TNF-α and higher IL-10 levels than the perfusion group, indicating a stronger anti-inflammatory effect. These results suggest that both IMRC delivery routes promote angiogenesis, stimulate the growth of endometrial cells, and modulate immunity, with sub-endometrial injection exhibiting more pronounced effects.

Angiogenesis-associated expression patterns

Theoretically, sub-endometrial injection enables IMRCs to directly interact with ESCs and vascular endothelial cells, thereby facilitating localized pro-angiogenic responses. In contrast, IMRCs delivered via uterine perfusion are more likely to engage with fibroblasts and residual luminal epithelial cells in the uterine cavity. These differences in cellular targets suggest that the angiogenic mechanisms activated by the two delivery routes may be fundamentally distinct.

RNA-seq revealed significant suppression of angiogenic activity in the model group. Specifically, 16 out of 23 angiogenesis-associated signaling pathways were significantly downregulated compared to the sham group (Figure 1J), and approximately three-quarters of angiogenesis-related differentially expressed genes (DEGs) (63/86) were significantly reduced (Figure 1K).

Treatment with IMRCs—particularly via sub-endometrial injection—reversed this angiogenic suppression. Compared to the model group, 14 out of 23 angiogenesis-related pathways were significantly activated in the IMRC injection group, including critical pathways such as angiogenesis, blood vessel remodeling, VEGF receptor (VEGFR) signaling, and fibroblast growth factor receptor (FGFR) signaling (Figure 1J). Similarly, 71 out of 86 pro-angiogenic DEGs were upregulated in the injection group relative to the model group (Figure 1K).

When directly comparing the two administration routes, the sub-endometrial injection group exhibited a more robust angiogenic response than the uterine perfusion group, with 56 of 86 DEGs significantly upregulated (Figure 1K). Notably, key regulators involved in ECM remodeling and endothelial proliferation—such as matrix metalloproteinase-2 (MMP-2), FGF2, and epidermal growth factor (EGF)—were markedly elevated in the injection group (Figure 1K).

The distinct gene expression profiles between the injection and perfusion groups further suggest that the underlying pro-angiogenic mechanisms may differ depending on the route of administration.

IMRCs cooperated with ESCs to promote angiogenesis in vitro

The mechanism by which sub-endometrial injection of IMRCs promotes angiogenesis remains unclear. We previously performed scRNA-seq analysis of human endometrial tissue and demonstrated that ESCs constituted approximately 90% of endometrial cells and played a crucial role in shaping the local endometrial microenvironment and promoting angiogenesis.²⁷ Sub-endometrial injection, unlike uterine perfusion, facilitates direct contact between IMRCs and ESCs. Therefore, the interaction between IMRCs and ESCs may be a key mechanism contributing to the superior pro-angiogenic effects observed with sub-endometrial administration.

We analyzed the intercellular communication networks between IMRCs and endometrial cells using CellChat. The results identified IMRCs and ESCs as the primary signal senders (Figure 2A). The interactions among IMRCs, ESCs, and proliferating ESCs are shown in Figure 2B. Signaling pathways enriched in the communication between IMRCs and ESCs are shown in Figure S2. When IMRCs serve as signal senders, enriched pathways include platelet-derived growth factor (PDGF), FGF, angiopoietin-like 4 (ANGPTL4), midkine (MDK), and collagen signaling pathways (Figure S2A), suggesting that IMRCs may promote the proliferation of ESCs. Conversely, when ESCs and proliferating ESCs act as senders, MDK, fibronectin, and collagen signaling pathways were enriched, indicating that ESCs may regulate the proliferation and homing of IMRCs (Figure S2B).

Figure 2.

IMRCs cooperate with ESCs to promote angiogenesis

(A) Circle plots showing the intercellular communication network among IMRCs and nine other endometrial cells: stromal cells, a transcriptionally distinct proliferating stromal subpopulation, immune cells (T cells, natural killer cells, and macrophages), endothelia cells, smooth muscle cells, ciliated epithelia, and unciliated epithelia. IMRCs, immunity-and-matrix-regulatory cells. (Left) The number of significant ligand-receptor interactions. (Right) The cumulative interaction strength. The thickness of each connecting line reflects the magnitude of the corresponding interaction. (B) Subnetwork analysis revealed interaction among IMRCs, stromal cells, and proliferating stromal cells. The number and strength of signaling events indicate robust intercellular communication, suggesting active crosstalk within the local microenvironment. (C) The viability of HUVECs was measured by CCK-8 assays at 24 and 48 h after being treated by the conditional medium. CM, conditional medium; Ctrl, control, basal medium for IMRC; ESC-CM, ESCs-derived CM; HUVECs, human umbilical vein endothelial cells; IMRC-CM, IMRCs-derived CM; co-culture-CM, co-culture-derived CM. (D) Migration of HUVECs was observed under the inverted microscope. Scale bar, 200 μm. (E) The migration rates of HUVECs were measured at 4 h after being treated by CM. (F) Tube formation of HUVECs was observed under the inverted microscope. The black dot with a red circle outside is the branchpoint. The yellow lines are the capillary lengths. Scale bar, 100 μm. (G and H) The branchpoints and capillary lengths were measured at 6 h after treatment with CM. (I–K) Supernatant from co-culture of IMRCs and ESCs under hypoxia condition (1% O₂) promoted the tube formation of HUVECs compared to the Ctrl group. The dots in the bar chart represent the number of technical replicates. ∗p < 0.05; ∗∗p < 0.01; ns, not significant. Data shown as mean ± SD.

Furthermore, we established a transwell co-culture model of IMRCs and ESCs to investigate the pro-angiogenic effects of their interaction on human umbilical vein endothelial cells (HUVECs), including proliferation, migration, and tube formation. The co-culture model and representative morphologies of IMRCs and ESCs are shown in Figures S3A and S3B. The stability of IMRCs was confirmed by assessing the expression of MSC-specific markers CD90 and CD105 after 72 and 96 h of co-culture with ESCs (Figure S3C). Conditioned media (CM) were collected at 72 h from the ESC group (ESC-CM), the IMRC group (IMRC-CM), and the co-culture group (co-culture-CM), and subsequently used to culture HUVECs.

The Cell Counting Kit-8 (CCK-8) assay demonstrated that compared to the control (Ctrl) group, the co-culture-CM, ESC-CM, and IMRC-CM groups promoted the proliferation rate of HUVECs. Although co-culture-CM induced a higher proliferation rate than the other two groups, the difference was not statistically significant (Figure 2C). Moreover, the wound healing assay revealed that only co-culture-CM significantly promoted the migration of HUVECs compared to the Ctrl group (Figures 2D and 2E). Additionally, the migration rate in the co-culture group was significantly greater than that in the ESC-CM and IMRC-CM groups (Figures 2D and 2E).

We found that, compared to the Ctrl group, ESC-CM, IMRC-CM, and co-culture-CM groups enhanced tube formation by increasing both the number of branchpoints and the total length of capillary-like structures (Figures 2F–2H). This enhancement was most pronounced in the co-culture group compared to the ESC and IMRC groups (Figures 2F–2H). Considering the hypoxic environment in the uterine cavity of patients with IUA, tube formation assays were repeated under low oxygen conditions (1% O₂). Similar pro-angiogenic effects were observed under both normoxic and hypoxic conditions, with no significant differences between the two environments (Figures 2I–2K). These results indicate that ESCs, IMRCs, and their co-culture promote angiogenesis, with the greatest effect seen in the co-culture group, suggesting that IMRCs may enhance angiogenesis through interactions with ESCs.

The co-culture-CM increased levels of angiogenic factors

To identify the active components in the co-culture-CM, we performed a protein array analyzing 60 angiogenesis-related secreted proteins in the ESC-CM, IMRC-CM, and co-culture-CM groups. The normalized expression data and raw images are provided in Table S1 and Figure S4, respectively. As shown in Figures 3A and 3B, 16 proteins were differentially expressed among the groups. Most of these proteins were significantly elevated in the co-culture-CM group compared to the IMRC-CM group (13/16) and the ESC-CM group (11/16). These findings strongly suggest that IMRCs synergize with ESCs to enhance angiogenesis.

Figure 3.

The expression levels of angiogenetic proteins in conditional medium

(A and B) Fluorescence intensity of differentially expressed proteins among three groups: ESC-CM, IMRC-CM, and co-culture-CM. ANG-1, angiopoietin-1;ANGPTL4, angiopoietin-like-4; CXCL16, chemokine ligand-16; ENA-78, epithelial neutrophil-activating protein-78; FGF, fibroblast growth factor; GRO, growth-related oncogene; HGF, hepatocyte growth factor; IL-8, interleukin-8; IP-10, interferon-gamma-inducible protein-10; MCP-1, monocyte chemoattractant protein-1; MMP-1, matrix metalloproteinases-1; TIMP-1, tissue inhibitor of metalloproteinase-1; TPO, thrombopoietin; uPAR, urokinase-type plasminogen activator receptor. The dots in the bar chart represent the number of technical replicates. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; ns, not significant. Data shown as mean ± SD.

The expression levels of several pro-angiogenic factors in the co-culture-CM group were higher than those in the ESC-CM and IMRC-CM groups, including ANGPTL4, monocyte chemotactic protein-1 (MCP-1, also known as CCL2), growth-related oncogene (GRO), thrombopoietin (TPO), chemokine ligand-16 (CXCL16), and IL-8 (Figure 3B). As shown in Figure 3A, the fluorescence intensity of ANGPTL4 was significantly greater than that of MCP-1, GRO, TPO, CXCL16, and IL-8, suggesting that ANGPTL4 may serve as an important mediator in promoting angiogenesis.

Additionally, proteins involved in the regulation of the ECM were highly expressed in the co-culture-CM, including MMP-1, tissue inhibitor of metalloproteinase-1 (TIMP-1), and TIMP-2. ECM remodeling is a critical process that governs angiogenic sprouting and the stabilization of the vascular network.^30,31,32 Although the fluorescence intensity of MMP-1 was higher than that of all other secreted proteins (Figure 3A), its expression levels did not differ significantly among the three groups (Figure 3B), indicating that MMP-1 may play an important role in the degradation of the ECM and the reduction of fibrotic deposition. Moreover, the expression of TIMP-1 and TIMP-2, inhibitors of MMP-1,³³ was significantly elevated in the co-culture-CM group compared to the ESC-CM and IMRC-CM groups (Figure 3B). This implies a balanced regulation of ECM remodeling in the co-culture group, preventing excessive collagen degradation while stabilizing newly formed blood vessels. Therefore, IMRCs and ESCs may promote angiogenesis and fibrotic tissue repair through coordinated regulation of the ECM.

Several proteins were highly expressed in both the IMRC-CM and co-culture-CM groups compared to the ESC-CM group, including basic FGF, urokinase-type plasminogen activator receptor (uPAR), epithelial neutrophil-activating protein-78 (ENA-78, also known as CXCL5), and interferon-γ-induced protein 10 (IP-10, also known as CXCL10) (Figure 3B). These findings suggest that IMRCs have a strong ability to promote the angiogenesis and chemotaxis of other cells. Conversely, several proteins such as angiogenin, angiopoietin-1 (ANG-1), ANG-2, MCP-3, and hepatocyte growth factor (HGF) were more highly expressed in ESC-CM and co-culture-CM groups than in the IMRC-CM group, indicating that ESCs also have a potent angiogenic capability (Figure 3B).

ANGPTL4 played a key role in co-culture-CM-induced angiogenesis

Based on the protein array results, ANGPTL4 was identified as a key pro-angiogenic mediator in the co-culture-CM group. To validate this finding, we performed an enzyme-linked immunosorbent assay (ELISA) to quantify ANGPTL4 levels. The concentration of ANGPTL4 was significantly higher in the co-culture-CM group than in the other two groups, consistent with the protein array results (Figure 4A). To determine the cellular source of ANGPTL4, we further analyzed its intracellular expression. Quantitative real-time PCR revealed elevated ANGPTL4 mRNA levels in co-cultured ESCs, but not in co-cultured IMRCs (Figures 4B and 4C). Furthermore, intracellular protein levels of ANGPTL4 were also increased in co-cultured ESCs (Figure 4D). These findings indicate that ESCs are the primary source of ANGPTL4 in the co-culture system.

Figure 4.

ANGPTL4 plays a key role in co-culture-CM-induced angiogenesis

(A) The protein concentration of ANGPTL4 was measured by ELISA among ESC-CM, IMRC-CM, and co-culture-CM groups, respectively. (B and C) The mRNA expression of ANGPTL4 in ESCs (B) and IMRCs (C) was measured by quantitative real-time PCR. Co-cultured ESCs: ESC after co-culturing with IMRC; co-cultured IMRCs: IMRCs after co-cultured with ESCs. (D) The protein expression of ANGPTL4 in co-cultured ESCs was measured by Western blotting. (E) The relative expression of ANGPTL4 after siRNA-mediated knockdown of ANGPTL4 in ESCs. (F–H) Co-culture-CM-induced angiogenesis was impaired after the knockdown of ANGPTL4 in ESCs. Ctrl, control, basal medium for IMRCs; si-NC, si-ANGPTL4, siRNA of ANGPTL4; siRNA of negative control. Scale bar, 100 μm. (I–K) Angiogenesis increased in the rHuANGPTL4 group compared to that in the Ctrl group. A similar result was observed in the co-culture-CM and rHuANGPTL4 combined groups in comparison with that in co-culture-CM alone. The concentration of rHuANGPTL4 was 1 μg/mL. Scale bar, 100 μm. The dots in the bar chart represent the number of technical replicates. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns, not significant. Data shown as mean ± SD.

To determine whether co-culture-induced angiogenesis depends on ANGPTL4, small interfering RNA (siRNA) was used to knock down ANGPTL4 expression in ESCs. After 48 h of transfection, ANGPTL4 expression was significantly reduced (Figure 4E). These ESCs were then co-cultured with IMRCs for 72 h, and the resulting CM was collected for tube formation assays using HUVECs. Silencing ANGPTL4 markedly reduced the number of branchpoints and total capillary length (Figures 4F–4H). To further validate the role of ANGPTL4, HUVECs were treated with 1 μg/mL recombinant human ANGPTL4 (rHuANGPTL4), which significantly increased both the number of branchpoints and capillary length compared to the Ctrl group (Figures 4I–4K). Tube formation was further enhanced when rHuANGPTL4 was combined with co-culture-CM, compared to the co-culture-CM group alone (Figures 4I–4K). These results indicate that ANGPTL4 plays a key role in co-culture-induced angiogenesis.

Clinical trial protocol and clinical characteristics of participants

Based on the above findings, it was hypothesized that IMRCs cooperate with ESCs to repair fibrotic endometrium and promote endometrial angiogenesis. To further evaluate the safety and therapeutic potential of IMRCs in treating refractory moderate-to-severe IUA, a phase 1 dose-escalation clinical trial was conducted at Tongji Hospital, employing sub-endometrial injection of IMRCs. Between December 2019 and March 2021, 176 patients were screened for eligibility (Figure 5). Of these, 58 were excluded, 68 did not meet the inclusion criteria, 10 declined to participate, 4 conceived naturally, and 81 were excluded for other reasons (e.g., difficulty attending follow-up due to living far from Wuhan or other personal factors) (Figure 5). Ultimately, 18 patients were enrolled and sequentially assigned to three dose groups, each receiving sub-endometrial injections of IMRCs: low-dose (LD, 3 × 10⁶ cells, n = 6), medium-dose (MD, 1 × 10⁷ cells, n = 6), and high-dose (HD, 3 × 10⁷ cells, n = 6) (Figure 5). All 18 patients completed the 12-week follow-up, and 17 completed the 3-year follow-up (Figure 5). The clinical characteristics of the 18 patients are summarized in Table S2. Their ages ranged from 30 to 38 years. All experienced hypomenorrhea, infertility, recurrent implantation failure (RIF) after high-quality embryo transfer (ET), and recurrent IUAs. Most had undergone at least three hysteroscopic operations, with IUA scores exceeding 6 points (Table S2).

Figure 5.

The CONSORT diagram shows the process of subject participation

Assessment of safety

All participants tolerated the injection procedure well. Adverse events (AEs) and serious adverse events (SAEs) were monitored throughout the 3-year follow-up period to assess the safety of IMRC administration. Seven types of AEs were reported: sore throat, breast pain, liver angioma, vulvovaginal candidiasis, endometrial polyps, maxillary protuberance, and conjunctivitis. Most of these AEs were deemed unrelated to IMRC treatment by independent experienced clinicians (Table S3). For example, liver angioma was attributed to estrogen use following TCRA. Only one case of endometrial polyps (in patient P6 from the LD group) may be related to the cell injection. Notably, no endometrial polyps occurred in the HD group, indicating that polyp formation was likely due to individual variability rather than to cell dose. In addition, no progression of endometrial polyps was observed during the 3-year follow-up (Table S3). No SAEs or complications related to surgery or intervention were reported during this period (Table S3). Moreover, no abnormalities were detected in the uterus, liver, gallbladder, pancreas, spleen, or kidneys after the IMRC injection (Table S3). Key clinical biochemical parameters, including white blood cell count, alanine aminotransferase, estimated glomerular filtration rate, C-reactive protein, α-fetoprotein, human epididymal protein 4, and thyroid-stimulating hormone, remained within normal ranges in all patients (Table S4).

IMRCs promoted the regression of scars and angiogenesis in patients with IUA

Hysteroscopy was performed monthly for three sessions (V1, V2, and V3) after IMRC injection to assess uterine cavity changes in patients with IUA (Figure 6A). The clinical efficacy outcomes for the three IMRC treatment groups and the TCRA-only control group are summarized in Tables 1 and S5.

Figure 6.

IMRCs promote the regression of scarring and angiogenesis in IUA patients

(A) The schematic diagram of the clinical trial. V0 means pre-treatment (before cell injection). V1, V2, and V3 mean the first follow-up (1 month), the second follow-up (2 months), and the third follow-up (3 months) after cell injection, respectively. (B) Hysteroscopic images of 18 patients in the LD, MD, and HD groups before and after IMRC injection. The black boxes mark regions of endometrial angiogenesis. (C and D) The area of re-adhesion (C) and type of re-adhesion (D) were checked in 18 patients at each visit. (E) The menstrual volume was measured in 18 patients before and after cell injection. (F) The endometrial thickness was measured in 18 patients at each visit. (G) Hysteroscopic images of eight patients in the TCRA-alone group. The images were obtained 1 month post-TCRA treatment. Minor angiogenic regions are highlighted by black squares. (H–J) The re-adhesion area (H), adhesion type (I), and endometrial thickness (J) were evaluated in eight patients before and 1 month after TCRA treatment. The dots in the bar chart represent the number of biological replicates. ∗p < 0.05; ∗∗p < 0.01; ns, not significant. Data shown in (F) and (J) as median ± quartile.

Table 1.

Efficacy outcomes of sub-endometrial injection of IMRCs

Group and patient no.	Type of adhesion				Area of adhesion				Endometrial thickness, mma				Menstrual volume scoreb				Pregnancy outcomes	Neonatal outcomes
	V0	V1	V2	V3	V0	V1	V2	V3	V0	V1	V2	V3	V0	V1	V2	V3
LD
01	dense	dense	filmy-dense	filmy-dense	>2/3	1/3–2/3	1/3–2/3	1/3–2/3	3.5	5.0	NA	NA	6	NA	NA	NA	live birth	preterm delivery (34 weeks)
02	dense	filmy	filmy	NA	<1/3	<1/3	<1/3	NA	6.0	4.9	7.1	6.0	9	NA	199	336	implantation failure	–
03	dense	dense	dense	filmy-dense	>2/3	1/3–2/3	1/3–2/3	1/3–2/3	2.2	3.0	2.6	3.3	9	10	8	9	miscarriage	–
04	dense	filmy-dense	filmy-dense	filmy-dense	>2/3	<1/3	<1/3	<1/3	4.0	5.0	4	5.4	26	NA	115	12	implantation failure	–
05	dense	dense	dense	filmy-dense	>2/3	1/3–2/3	1/3–2/3	1/3–2/3	3.0	2.4	4.1	4.6	16	6	NA	NA	implantation failure	–
06	dense	filmy-dense	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	5.0	5.6	7.3	6.0	17	NA	NA	NA	live birth	healthy
MD
07	dense	filmy-dense	filmy	filmy	1/3–2/3	<1/3	1/3–2/3	<1/3	4.0	3.7	4.0	4.0	26	NA	NA	NA	implantation failure	–
08	dense	dense	filmy-dense	filmy-dense	1/3–2/3	<1/3	<1/3	<1/3	4.0	6.2	5.0	8.0	50	NA	NA	NA	implantation failure	–
09	dense	dense	dense	dense	>2/3	>2/3	>2/3	>2/3	3.0	3.4	3.4	4.0	21	16	NA	NA	miscarriage	–
10	dense	filmy	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	3.4	4.6	9.0	7.0	190	NA	NA	NA	live birth	healthy
11	filmy-dense	filmy-dense	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	3.4	3.0	2.6	3.0	18	NA	18	82	implantation failure	–
12	dense	dense	dense	filmy	1/3–2/3	1/3–2/3	1/3–2/3	1/3–2/3	5.0	5.3	5.2	5.0	45	NA	308	NA	biochemical pregnancy	–
HD
13	dense	filmy	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	6.0	5.0	5.0	6.0	24	210	29	NA	live birth	healthy
14	dense	filmy	filmy	filmy	>2/3	<1/3	<1/3	<1/3	4.0	2.0	3.0	4.0	16	78	111	NA	miscarriage	–
15	dense	filmy	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	4.0	6.0	6.0	4.0	17	189	78	NA	live birth	healthy
16	dense	filmy-dense	filmy-dense	filmy-dense	>2/3	1/3–2/3	<1/3	<1/3	3.0	4.0	4.4	2.5	9	NA	NA	17	implantation failure	–
17	dense	filmy	filmy	filmy	1/3–2/3	<1/3	<1/3	<1/3	7.0	7.0	8.0	5	9	NA	NA	106	live birth	preterm delivery (36+3 weeks)
18	dense	filmy-dense	filmy-dense	filmy-dense	>2/3	1/3–2/3	1/3–2/3	<1/3	3.0	2.0	2.6	2	8	NA	NA	NA	implantation failure	–

NA, not applicable; V0–V3, follow-up.

^aData expressed as mean ± SD.

^bThe detail of menstrual volume scoring is included in Data S1.

Prior to IMRC injection, all patients exhibited abnormal uterine cavity anatomy characterized by extensive fibrous or muscular adhesions, predominantly at the uterine fundus, which is the primary site of embryo implantation (Figure 6B). However, re-adhesion occurred in both groups. The differences in adhesion area (LD and MD groups) and adhesion type (MD group) were not statistically significant (Figures 6C and 6D). Only the LD group showed significant improvement in adhesion type at V2 and V3 compared to that at V0 (Figure 6D). A slight increase in menstrual volume was observed in both the LD and MD groups; however, the changes did not reach statistical significance (Figure 6E). In contrast, the HD group exhibited more pronounced improvements, with extensive scar tissue replaced by thin or near-normal endometrium as early as V1 (Figure 6B). The adhesion area was significantly reduced in the HD group (Figure 6C), accompanied by notable regression of scar tissue at the uterine fundus. Post-injection, re-adhesions were predominantly membranous and filmy, with no dense adhesions observed (Figure 6D). Six patients in the HD group exhibited a marked increase in menstrual volume (Figure 6E). Although EMT increased slightly in the LD and MD groups following IMRC injection, the changes were not statistically significant (Figure 6F). In the HD group, the average EMT increased from 4.50 ± 1.64 mm to 4.83 ± 2.00 mm, and the maximum EMT rose from 6.23 ± 1.37 mm to 6.92 ± 1.87 mm, without reaching statistical significance (Figure 6F).

Angiogenesis or revascularization was clearly observed in the uterine cavities of 18 patients following IMRC treatment (Figures 6B and S5A). The uterine cavity color changed from pale red to bright red in all patients, accompanied by the appearance of crisscrossing microvascular networks and numerous glands, especially at the previously scarred fundus (Figures 6B and S5A). Serial hysteroscopic recordings from six patients in the HD group (V0–V3) further confirmed extensive neovascularization, scar tissue regression, and progressive endometrial regeneration (Videos S1, S2, S3, S4, S5, and S6). To evaluate the role of IMRCs in angiogenesis, a control group of eight patients with moderate to severe IUA received only conventional TCRA without cell therapy (Figure 6G). One month postoperatively, no visible angiogenesis was observed in controls 1, 5, 6, and 8, and only minimal vascular changes were noted in controls 2–4 and 7. Although TCRA led to a statistically significant reduction in adhesion area, there were no significant improvements in adhesion severity or EMT (Figures 6H–6J).

Video S1. The hysteroscopic surgery videos in subject no. 13

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#13 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second followup (two months), and the third follow-up (three months) after cell injection, respectively.

Video S2. The hysteroscopic surgery videos in subject no. 14

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#14 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second followup (two months), and the third follow-up (three months) after cell injection, respectively.

Video S3. The hysteroscopic surgery videos in subject no. 15

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#15 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second followup (two months), and the third follow-up (three months) after cell injection, respectively.

Video S4. The hysteroscopic surgery videos in subject no. 16

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#16 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second followup (two months), and the third follow-up (three months) after cell injection, respectively.

Video S5. The hysteroscopic surgery videos in subject no. 17

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#17 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second follow-up (two months), and the third follow-up (three months) after cell injection, respectively.

Video S6. The hysteroscopic surgery videos in subject no. 18

The hysteroscopic surgery videos demonstrate the intrauterine cavity changes before and after IMRCs injection in subject#18 of HD group. V0 means pre-treatment (before cell injection). V1, V2, V3 means the first follow-up (one month), the second followup (two months), and the third follow-up (three months) after cell injection, respectively.

Given the appearance of a bright red uterine cavity in many patients following IMRC treatment, endometrial biopsies were performed to exclude endometritis. Histopathological analysis showed no evidence of plasma cell infiltration, indicating the absence of endometritis. Furthermore, the proportion of CD45⁺ immune cells in the endometrium was assessed following IMRC administration. A significant reduction in CD45⁺cell infiltration was observed in IMRC-treated IUA patients compared to untreated controls, indicating that IMRCs exert immunomodulatory effects (Figures S5B and S5C). To further evaluate angiogenesis, hematoxylin and eosin (H&E) and Masson’s trichrome staining were performed. Compared with the proliferative-phase endometrium from healthy women, IUA patients showed disrupted tissue architecture and reduced glandular structures by H&E staining, while Masson staining revealed extensive collagen deposition (Figure S5D). In IMRC-treated patients (e.g., patient 16 at V3), collagen accumulation was markedly reduced, and vessel-like structures emerged within scar tissue (Figures 7A–7C). Immunofluorescence for CD31 and α-smooth muscle actin (α-SMA) confirmed that most of these structures were CD31⁺/α-SMA⁺, indicative of mature vasculature (Figures 7D and 7E). Immunohistochemistry showed ANGPTL4 expression predominantly in epithelial cells and weakly in stromal cells in normal proliferative endometrium and absent in fibroblasts. Following IMRC injection, ANGPTL4 expression was markedly upregulated in stromal cells (Figures 7F and 7G). Together, hysteroscopy and immunofluorescence showed that uterine scars regressed and both vasculature and endometrial tissue regenerated following cell therapy. These neovessels may facilitate long-term endometrial repair by activating endogenous stem cells.

Figure 7.

IMRCs significantly promote endometrial angiogenesis in IUA patients

(A–C) The hysteroscopic image (A), H&E staining (B), and Masson staining (C) of the endometrium of patient 16 at V3 after cell injection. Scale bar, 100 µm. The hysteroscopic image (A) of patient 16 at visit V3 is also presented in Figure S5A, as both figures correspond to the same patient at the same follow-up time point. (D and E) The representative images (D) and quantitative analysis (E) of immunofluorescence of CD31 and α-SMA in the normal proliferative endometrium, IUA patient endometrium, and the endometrium of patient 16 at V3. Scale bar, 50 μm. (F and G) The representative images (F) and quantitative analysis (G) of immunohistochemistry of ANGPTL4 in normal proliferative endometrium, endometrium of patient 16 at V0, and endometrium of patient 16 at V3. Scale bar, 100 µm. The dots in the bar chart represent the number of biological replicates, which is three. ∗p < 0.05; ∗∗p < 0.01. Data shown as mean ± SD.

Pregnancy outcomes after cell transplantation

The uterine cavities of all patients showed marked improvement after injection of IMRCs, and ET was subsequently planned. By July 2025, five patients (P1, P3, P6, P9, and P10) in the LD and MD groups achieved clinical pregnancy. Among them, three patients (P1, P6, and P10) successfully achieved live births (Table 1). Patients in the HD group demonstrated better pregnancy outcomes (Table 1): four participants (P13, P14, P15, and P17) became pregnant, and three (P13, P15, and P17) delivered live births (Table 1). Unfortunately, P14 experienced a miscarriage due to trisomy 21. All newborns were healthy, although those from P1 and P17 were born prematurely (Table 1). In the TCRA control group, eight patients received treatment, resulting in three clinical pregnancies—two live births and one miscarriage (Table S5). Based on these findings, the dose used in the HD group is recommended for future trials, as it led to substantial uterine cavity recovery, enhanced angiogenesis, and favorable pregnancy outcomes and demonstrated a good safety profile.

Discussion

Recently, MSC transplantation has emerged as a promising treatment for moderate to severe IUA.¹⁴ In this study, we report for the first time the use of IMRCs, derived from human embryonic stem cells, for the treatment of IUA. IMRCs possess enhanced immunomodulatory properties, as evidenced by the elevated expression of CD24, CD274, and PGE2.¹⁶ To comprehensively assess IMRC safety, a series of in vitro and in vivo assays were conducted.^16,34,35 IMRCs maintained stable diploid karyotypes and lacked chromosomal abnormalities or pluripotency markers, as confirmed by karyotyping, scRNA-seq, soft agar, and whole-genome analyses. No mutations were detected in coding or non-coding exons. In cynomolgus monkeys, both short- and long-term toxicity evaluations showed no adverse effects. Biosafety testing—including pathogen screening, tumorigenicity, and biopreparation quality—met regulatory standards set by the National Institutes for Food and Drug Control in China. These findings collectively support the clinical safety of IMRCs.

Our animal studies demonstrated that both sub-endometrial injection and uterine perfusion of IMRCs enhanced endometrial angiogenesis, with sub-endometrial injection exhibiting a markedly superior pro-angiogenic effect compared to perfusion. However, the underlying mechanism by which sub-endometrial IMRC injection enhances angiogenesis remains unclear. RNA-seq analysis revealed distinct pro-angiogenic profiles between the injection and perfusion groups. Several signaling pathways were significantly upregulated in the injection group, including angiogenesis, endothelial cell proliferation, and VEGFR signaling. These findings indicate that sub-endometrial injection of IMRCs more effectively enhances angiogenesis by activating these key pathways.

To further elucidate cellular interactions in IMRC-induced endometrial angiogenesis, scRNA-seq and CellChat analyses were performed to examine crosstalk between IMRCs and local endometrial cells. IMRCs and ESCs were identified as the main sources of angiogenesis-related signaling. ESCs are abundant in the endometrium and are recognized for their potent roles in tissue repair, microenvironment regulation, and promotion of angiogenesis.^28,36,37 Given that sub-endometrial injection enables direct IMRC-ESC contact, we hypothesized that their interaction enhances angiogenesis. CellChat analysis confirmed strong IMRC-ESC crosstalk via FGF, PDGF, ANGPTL, MDK, and collagen signaling. To validate this, transwell co-culture assays showed that co-culture supernatants significantly promoted HUVEC tube formation compared to IMRCs or ESCs alone, supporting a synergistic pro-angiogenic effect.

To explore the underlying mechanism, a protein array revealed that co-culture-CM contained higher levels of angiogenic factors than CM from IMRCs or ESCs alone. Notably, ANGPTL4, along with TPO, GRO, MCP-1, CXCL16, and IL-8, was significantly elevated, with ANGPTL4 showing the highest expression. Further analysis showed that IMRCs specifically induced ANGPTL4 expression in ESCs, but not in themselves. ANGPTL4 has diverse, tissue-dependent functions.^38,39,40 For example, it promotes angiogenesis in the endometrium during the peri-implantation period and in tendons.³⁸ ANGPTL4 regulates multiple downstream pathways involved in vascular and metabolic function. In HUVECs, ANGPTL4 knockdown altered over 5,000 genes, with increased lipid metabolism and activation of FoxO and Hippo signaling, while reducing glycolysis and angiogenesis-associated processes, including cell adhesion and growth factor signaling.⁴¹ ANGPTL4 promotes angiogenesis via MAPK signaling, as its silencing decreases the phosphorylation of p38, ERK, and JNK.⁴² It also activates integrin-mediated JAK2/STAT3 signaling⁴³ and modulates vascular cell behavior through the PI3K/AKT pathway,⁴⁴ contributing to smooth muscle cell phenotypic regulation. In our study, ANGPTL4 knockdown in ESCs impaired the pro-angiogenic effect of co-culture-CM, while adding recombinant ANGPTL4 further enhanced HUVEC tube formation. These findings indicate that ANGPTL4 cooperates with other factors to promote angiogenesis and may serve as a potential therapeutic target for restoring vascularization in scarred endometrium.

Building on the above findings, IMRCs likely enhance angiogenesis and reverse fibrosis through cooperation with ESCs via sub-endometrial injection. A clinical study was conducted to assess their safety and efficacy in IUA patients. No IMRC-related SAEs were observed over a 3-year follow-up. Therapeutic benefits, including endometrial regeneration and anti-fibrotic effects, were most notable in the HD group. To date, six patients have achieved live births, most of whom had refractory IUA with prior RIF, multiple hysteroscopies, and IUA scores >6. The sub-endometrial IMRC injection appears to be safe and effective for treating refractory moderate to severe IUA.

Notably, abundant neovessels were observed in scarred endometrium after IMRC treatment, while minimal angiogenesis was seen with traditional TCRA alone, aligning with animal data. As angiogenesis is essential for tissue repair and endometrial receptivity, this difference highlights the therapeutic potential of IMRCs.^45,46 Prior hysteroscopic studies rarely reported visible angiogenesis, possibly due to the superior proliferative and angiogenic properties of IMRCs, their direct interaction with ESCs via sub-endometrial injection, and improved survival through local vascular support.^12,13,14

This study has several key strengths. First, hESC-derived IMRCs offer advantages such as stringent quality control, large-scale production, and gene-editing compatibility, making them promising for IUA therapy. Second, we demonstrate that sub-endometrial injection more effectively enhances angiogenesis and reduces fibrosis in an IUA model, supporting this delivery route. Third, bioinformatic and in vitro analyses revealed that IMRCs and ESCs synergize to promote angiogenesis. Finally, our clinical application of IMRCs in moderate-to-severe IUA yielded encouraging pregnancy outcomes and neovascularization, highlighting their translational potential.

Several limitations warrant consideration. First, although ESCs were the focus of our in vitro studies, IMRCs likely interact with other endometrial cell types. Further studies are needed to elucidate these broader cellular interactions and underlying mechanisms. Second, etiological heterogeneity existed among the participants. In cases of tuberculosis-related IUA, basal layer atrophy and reduced sub-endometrial blood flow may delay recovery and result in poorer reproductive outcomes.^45,47 Finally, our single-arm study lacked both direct comparison of administration routes and multicenter validation, limiting generalizability. Future randomized trials are needed to confirm the efficacy of sub-endometrial IMRC therapy for refractory IUA.

Conclusion

In summary, this study establishes the safety profile of IMRCs in treating moderate-to-severe IUA and highlights their therapeutic potential in enhancing endometrial angiogenesis via sub-endometrial injection. These findings underscore the necessity for future randomized, double-blind, controlled clinical trials to validate efficacy and inform clinical translation.

Materials and methods

Experimental design

We first performed an animal experiment to study the effect of IMRCs on treating IUA and compare the effect of uterine perfusion and endometrial injection of IMRCs. Afterward, we investigated the interaction between IMRCs and ESCs using scRNA-seq and CellChat analyses. Then, we constructed a transwell co-culture model of IMRCs and ESCs to evaluate the angiogenic effect of a co-culture-conditioned medium. After that, we explored the molecular mechanism of IMRCs’ and ESCs’ collaboration to promote angiogenesis. Finally, a clinical investigation was conducted on 18 patients with refractory moderate to severe IUA by sub-endometrial injection with IMRCs. All subjects were divided into three groups: LD, MD, and HD. The injected cell concentration was 3 × 10⁶ cells/mL, 1 × 10⁷ cells/mL, or 3 × 10⁷ cells/mL, respectively. Safety and efficacy on endometrial repair, endometrial angiogenesis, and pregnancy outcomes were assessed.

Animal experiments

All animal experiments were approved by the institutional review board of Tongji Hospital (no. 2021S326). The normal sexually mature adult New Zealand female rabbits weighing 2.5–3.0 kg were purchased from the Experimental Animal Center of Tongji Hospital Affiliated to Tongji Medical College, Huazhong University of Science and Technology (animal license no. SCXK (E) 2021-0020). Twenty rabbits were randomly and evenly divided into sham, model, injection, and perfusion groups, in five rabbits per group. After 1 week of adaptive feeding of the New Zealand female rabbits, the vaginal orifice of the female rabbits was observed every day. If there was significant swelling, a bright red color, and secretions flowing out, then it was considered to be the estrous period. The model was conducted during the estrous period. Rabbits were treated with 3% pentobarbital sodium for ear margin intravenous anesthesia. The anesthetized New Zealand rabbits were fixed in the supine position on the operating table, and the lower abdomen was shaved and sterilized with iodine. A median incision in the lower abdomen was performed to cut the skin and muscle of New Zealand rabbits successively, expose the uterus, and then provide different treatments according to different groups. In the sham group, rabbits only received laparotomy. In the IUA model group, a 1-mL sterile insulin syringe (29G, BD Biosciences, USA) was used to penetrate through the wall of the uterus and then retracted. If air appeared, it indicated that the needle had entered the uterine cavity. Then, 200 μL of 95% ethanol was perfused and remained for 1 min. Ethanol was then withdrawn and saline was irrigated. In the injection group, 1 week after modeling, a total of 200 μL 4 × 10⁶ IMRC suspension was injected with the syringe at four sites around the longitudinal axis of the uterine wall. We injected 50 μL IMRC suspension at each site. In the perfusion group, rabbits were perfused with 200 μL 4 × 10⁶ IMRC suspension, similar to the model group. One week after IMRC treatment, the animals were killed, and specimens were collected for subsequent experiments.

Masson staining

The rabbit uteri collected in each group (n = 5) were fixed with 4% paraformaldehyde (PFA) and made into serial paraffin sections (4 μm). Three sections were selected for Masson staining with a standardized procedure (Solarbio, China). Sections are observed with a microscope and photographed. Each Masson-stained section was randomly selected to observe five high-power fields of view. The proportion of fibrotic area was quantitatively calculated using Image Pro Plus 6.0 software. The area of fibrosis (%) is equal to the fibrosis area/endometrial area.

Ultrasound scanning

The rabbits in each group were anesthetized using the above methods. The skin and muscles of an anesthetized rabbit was cut through a median incision in the lower abdomen, and the rabbit uterus was fully exposed. After the coupling agent was evenly applied to the uterus, ultrasound measurement (Esaote VET X5, Italy) of color flow mapping was performed to obtain the blood flow signal diagram inside the uterus.

Immunohistochemistry

In brief, the endometrium tissue was dehydrated, embedded in paraffin, and cut into serial sections (4 μm). Immunohistochemistry labeling was performed with primary antibodies, including mouse monoclonal anti-VEGF (1:100, Abcam, UK), mouse monoclonal anti-CD31 (1:100, Novus, USA), rabbit polyclonal anti-CD31 (1:200, ServiceBio, China), mouse monoclonal anti-α-SMA (1:200, ServiceBio), and rabbit polyclonal anti-ANGPTL4 (1:200, ABclonal, USA) according to the manufacturer’s manual. For MVD count, three sections were selected from each sample (n = 5). The MVD was measured by counting the number of CD31⁺ microvessels under five randomly high-power lenses (400×) per section.

RNA-seq

RNA-seq was analyzed using five rabbits per group. Total RNA was extracted by the Trizol method. RNA-seq was performed in the BGISEQ platform (BGI Genomics, China). In short, the first step is library construction for RNA-seq. The procedure of library construction was based on the manufacturer’s protocol. The quality of the constructed libraries was tested, and the qualified libraries were replicated through a rolling circle to form DNA nanoballs (DNBs). Finally, the BGISEQ platform combines the DNBs-based nanoarrays and stepwise sequencing using the combinational probe-anchor synthesis sequencing method. As for analyzing the RNA-seq data, reads with low-quality adapter contamination and an excessively high content of unknown base N were filtered out to obtain clean reads. The clean reads were mapped to the rabbit genome (GCF_000003625.3_OryCun2.0), and the read counts were calculated. R 4.0.2 was used for downstream analysis. The DESeq2 package was used to analyze DEGs between the two groups. Significant DEGs were identified as q < 0.05 and log2 |fold-change| >1. The pheatmap package was used to draw clustered heatmaps. The GSVA package was used for gene set variation analysis.

Patient recruitment for scRNA-seq

The selection process of participants for scRNA-seq was described in our previous study.²⁷ In short, the human endometrial tissue samples in proliferative phase are derived from three participants who performed in vitro fertilization due to male factors (e.g., oligo-/asthenozoospermia or azoospermia) and signed informed consent. The inclusion criteria include ages ranging from 18 to 35 years, body mass index between 20 and 24 kg/m², no history of pathology of the endometrium or myometrium. This study has been approved by the medical ethics committee of Tongji Hospital affiliated with Tongji Medical College of Huazhong University of Science and Technology (no. TJ-IRB20210909).

The scRNA-seq procedure

The details of scRNA-seq were described in our previous study.²⁷ Briefly, three human endometrial tissue samples and IMRCs (prior to sub-endometrial injection) were independently processed for droplet-based scRNA-seq using the 10x Genomics Chromium platform, following the manufacturer’s protocol. Approximately 8,000 cells per sample were loaded per channel to generate gel bead-in-emulsions. Libraries were constructed with the Chromium Single Cell 3′ Reagent Kit version 3 (10x Genomics, USA), converted using the MGIEasy Universal Library Conversion Kit (BGI Genomics), and sequenced on the MGISEQ-2000 system (BGI Genomics). Raw sequencing data were processed using Cell Ranger (10x Genomics) with default settings and aligned to the GRCh38 reference genome. The resulting expression matrices were analyzed with Seurat (version 3.1.5) in R (version 4.0.2). Cells expressing fewer than 600 genes or containing >15% mitochondrial transcripts were excluded. Data were normalized to transcripts per 10,000 and log transformed. Dimensionality reduction and clustering were performed using uniform manifold approximation and projection and the shared nearest neighbor algorithm, with clustering defined using Seurat’s FindClusters function (resolution = 0.2). To integrate the endometrium and IMRC datasets and correct for batch effects, Harmony was applied. Cell-cell communication networks were inferred using CellChat (version 1.0.0), based on averaged expression of ligand-receptor pairs across clusters. All data visualization was performed in R (version 4.0.2).

Cell culture

The IMRCs were provided after quality control by the Chinese National Stem Cell Resource Center. Generation, identification, and maintenance of the growth of IMRCs were reported in the previous study.¹⁶ The Good Manufacturing Practices (GMP)-compliant laboratory was used to generate IMRCs, and all the processes strictly followed GMP standard operating procedures. Cells with diploid karyotypes at passage 5 were used in the clinical trial. Human endometrial stromal cell lines (THESCs) and HUVECs were purchased from the American Type Culture Collection. These cells were cultured and maintained in the non-phenol red DMEM medium supplemented with 10% fetal bovine serum. We used 5% trypsin to passage these two types of cells. All cells were cultured at 37°C in 5% CO₂. As for cell cryopreservation, they were slowly frozen in freeze medium and stored at −80°C.

Co-culture of IMRCs and ESCs

The non-contact co-culture model of IMRCs and ESCs was constructed on a 24-well plate-sized transwell system containing 6 inserts with microporous membrane (Corning, USA). We plated 100 μL IMRCs (passage 5–7, 4 · 5 × 10⁵ cells/mL) on transwell inserts as upper components, and 600 μL THESCs (1 · 65 × 10⁵ cells/mL) were loaded on 24-well plates as lower components. Then, cells were cultured at 37°C with 5% CO₂ to allow the cells to be firmly attached to the membrane. After cell adherence, the transwell inserts were put back into the well plate to enable the co-culture of IMRCs and ESCs.

Cell immunofluorescence

IMRCs were loaded on the glass slides. After cell attachment and co-culture, cells were fixed with 4% PFA for 25 min and incubated with a blocking buffer (1% BSA) for 1 h at room temperature. Cells were stained with primary antibodies overnight at 4°C and then incubated with secondary antibodies for 1 h. As primary antibodies, we used mouse monoclonal anti-CD90 (1:100; catalog no. 66766-1-lg, Proteintech, USA) and rabbit polyclonal anti-CD105 (1:100; catalog no. 10862-1-AP, Proteintech). As secondary antibodies, donkey anti-mouse immunoglobulin G (IgG)-Cy3 (1:100; catalog no. abs20016, Absin Bioscience, China) and donkey anti-rabbit IgG-Alexa Fluor 488 (1:100; catalog no. abs20020, Absin Bioscience) were used. Hoechst 33258 (Thermo Fisher Scientific, USA) was used to stain the nuclear DNA.

CCK-8 assay

CCK-8 was performed according to the manufacturer’s instructions (catalog no. 40203ES76, Yeasen, USA). HUVECs with 2,500 cells/well were seeded into 96-well plates and cultured. After cell attachment, they were treated by ESC-CM, IMRC-CM, and co-culture-CM. Supernatants of the above groups were collected after culturing for 72 h. ESC-CM: 50% DMEM medium + 50% ESC supernatant; IMRC-CM: 50% DMEM medium + 50% IMRC supernatant; co-culture-CM: 50% DMEM medium + 50% co-culture supernatant. At 24 and 48 h after treatment, 10 μL CCK-8 solution mixed with 90 μL serum-free DMEM medium was added into every well with cells for incubating for 2 h. Then, a microplate reader (BioTek, USA) was used to detect the absorbency at a test wavelength of 450 nm. This experiment was repeated biologically at least three times.

Migration assay

We used culture inserts (Ibidi, Germany) to perform the migration assay. Specifically, the culture inserts were transferred in 24-well plates. Each compartment of the insert was filled with a 70-μL suspension of HUVECs (3 × 10⁵ cells/mL). After cell adherence, the inserts were gently removed. Then, these cells were washed by PBS twice and treated by ESC-CM, IMRC-CM, and co-culture-CM. ESC-CM, IMRC-CM, and co-culture-CM were consistent with that used in the CCK-8 assay. Images of HUVECs were captured at 0 and 4 h after treatment. Five visual fields were randomly selected in each sample to take photographs, after which the average area of intercellular space of the five visual fields was calculated. This experiment was repeated biologically at least three times.

Tube formation assay

We added 50 μL cold reduced growth factor basement membrane extract (catalog no. 3445-010-01, Trevigen, USA) to each well of pre-cooling 96-well plates. Then, the plates were placed in the incubator at 37°C with 5% CO₂ for 30 min A 100-μL suspension of HUVECs (1.5 × 10⁵ cells/mL) was seeded in the basement membrane extract (BME) and incubated for another 30 min. After cell attachment, HUVECs were treated by ESC-CM, IMRC-CM, co-culture-CM, si-NC-CM, si-ANGPTL4-CM, or rHuANGPTL4 (catalog no. 4487-AN-050, R&D Systems, USA). Finally, pictures of tube formation were taken at 6 h after treatment. Five visual fields were randomly selected in each sample to take photographs, after which the average number of branches or lengths of the five visual fields was calculated. This experiment was repeated biologically at least three times.

In this assay, we generated conditioned media from IMRC-ESC co-cultures under hypoxic conditions (1% O₂), referred to as hypoxia co-culture-CM. The term “co-culture-CM” denotes media collected from the same co-culture system maintained under normoxic conditions. As a control, we used unconditioned basal medium. HUVECs were then treated with these three types of media to assess tube formation capacity.

Protein array

The Human Angiogenesis Array 1000 (catalog no. GSH-ANG-1000-1, RayBiotech, USA) was used to screen for the differential expression of 60 proteins associated with angiogenesis. Supernatants of ESC-CM, IMRC-CM, and co-culture-CM were collected after centrifuging at 2,000 × g for 15 min. Then, the array was performed according to the manufacturer’s instructions. We quantified the number of viable cells under each culture condition at the time of sample collection. Protein concentrations were subsequently normalized to the corresponding total cell counts, ensuring valid cross-group comparisons irrespective of cellular density. This experiment was repeated biologically at least three times.

ELISA

The concentrations of ANGPTL4 (catalog no.70-EK1292-96, Liankebio, China) in ESC-CM, IMRC-CM, and co-culture-CM were measured using human ELISA kits. Supernatants of the above three groups were collected after culturing for 72 h and centrifuged at 2,000 × g for 15 min. Then, each ELISA assay was performed according to protocols provided by the kit manufacturers. This experiment was repeated biologically at least three times.

Quantitative real-time PCR

We used an RNA-easy isolation reagent (Vazyme, China) to extract the total RNA in cells according to the manufacturer’s protocol. After that, complementary DNA (cDNA) was synthesized by using Hifair Ⅲ 1st Strand cDNA Synthesis SuperMix for qPCR (gDNA Digester Plus) (catalog no. 11141ES60, Yeasen). Quantitative real-time-polymerase chain reaction was performed by using Hieff qPCR SYBR Green Master Mix (No Rox) (catalog no. 11201ES03, Yeasen). The relative mRNA levels of each gene were calculated by the 2^–ΔΔCT method and normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The sequences of all used primers are listed as VEGF sense 5′-ATGGCAGAAGAAGGAGAC-3′, anti-sense 5′-GCAGGAAGGCTTGAATAT-3′; ANGPTL4 sense 5′-TCCTGGACCACAAGCACCTA-3′, anti-sense 5′-CTGGAACAGCTCCTGGCAAT-3′; GAPDH sense 5′-TGTGGGCATCAATGGATTTGG-3′, anti-sense 5′-ACACCATGTATTCCGGGTCAAT-3′. This experiment was repeated biologically at least three times.

Western blotting

Radioimmunoprecipitation assay lysis buffer (ServiceBio, China) mixed with 1% protease and phosphatase inhibitor (ServiceBio) was used to extract total intracellular protein. Proteins with different molecular weights were separated on sodium dodecyl sulfate-polyacrylamide gel and then transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA). The PVDF membranes were blocked by protein-free rapid blocking buffer for 1 h and then incubated with specific primary antibodies at 4°C overnight. As primary antibody, rabbit polyclonal anti-ANGPTL4 (1:1,000; catalog no. A2011, ABclonal) was used. After incubation with corresponding horseradish peroxidase-conjugated secondary antibodies for 1 h, proteins were visualized by Super ECL Detection Reagent (Vazyme) in the Bio-Imaging System. Mouse polyclonal anti-GAPDH (1:5,000, Proteintech) was used as internal controls. This experiment was repeated biologically at least three times.

Transfection of siRNA

RiboFECT CP Transfection Kit (catalog no. C10511-05, Ribo, China) was used for the transfection of ANGPTL4 siRNA or control (Negative control, NC) siRNA (Ribo). 50 nM siRNA (ANGPTL4 siRNA or control siRNA), 30 μL RiboFECT CP Buffer (1×), 3 μL RiboFECT CP Reagent, and 465 · 75 μL DMEM complete medium were mixed and transfected into ESCs in 24-well plates. The controls consisted of scrambled siRNAs. The medium was replaced after 48 h of transfection, and ESCs were co-cultured with IMRCs for another 72 h. The co-culture-CMs (si-NC-CM and si-ANGPTL4-CM) were collected. The sequence of ANGPTL4 siRNA is 5′-CCACAAGCACCTAGACCAT-3′. The efficiency of knockdown was confirmed by quantitative real-time PCR.

Ethics statement

This clinical trial complied with the Declaration of Helsinki. All procedures were approved by the ethics committee and academic committee of Tongji Hospital, affiliated with Tongji Medical College, Huazhong University of Science and Technology (no. TJ-IRB20180804) in August 2018 and filed at the National Health Commission of China in January 2019. All patients signed an informed consent form and were enrolled in this clinical trial from December 2019 to March 2021 (http://clinicaltrials.gov/ct2/show/NCT82004232592?term=Qi+Zhou&draw=2&rank=1, NCT04232592). Before the start of this clinical trial, the study protocol form was submitted to the ethics committee of Tongji Hospital for approval, and the written approval of the ethics committee was obtained before the experiment. Therefore, this trial was conducted in strict accordance with the requirements of the Criterions for the Quality Control of Clinical Trials of Drugs and relevant laws and regulations.

Patient recruitment for IMRC transplantation

This was an open-label, single-arm, single-center, phase 1 dose-escalation trial funded by the Institute for Stem Cell and Regeneration, Chinese Academy of Sciences. The inclusion criteria were as follows: 18–38 years old, moderate to severe IUA according to American Fertility Society (AFS) criteria (1988), having one or more of the following clinical symptoms: hypomenorrhea or amenorrhea, periodic abdominal pain, infertility, recurrent abortion, RIF, recurrent IUA, and premature delivery or other adverse pregnancy outcomes. The patients gave informed consent and voluntarily underwent the operation. Patients would be excluded if they had any of the following abnormalities: having hysteroscopy or estradiol-supplementation contraindications, endometriosis, pelvic inflammation, contraindications of surgery or pregnancy, history of malignant tumors, and difficulty in cooperating with follow-up. Whether the patients fit the requirements was determined by an experienced gynecologist based on their history. This study included 18 subjects, and all subjects were divided into LD, MD, and HD groups in order of enrollment, with 6 people in each group.

Hysteroscopic operations

The operation was performed following 12 days of 4 mg/day estradiol valerate from the second day of menstruation. The shape of the uterine cavity was observed under hysteroscopy to evaluate the AFS score. Then, micropliers and scissors were carefully used to separate adhesion until a normal uterus cavity could be seen, which was characterized by two uterine horns and proper womb volume. The same dosage of estrogen was applied in all subjects for 3 months after operation, while a COOK balloon without water was placed into the uterus for 2 weeks after the operation. All surgeries were completed by the same gynecologist who was skilled at hysteroscopy using ultrasound guidance.

Cell delivery

The cell concentration of IMRCs was 5 × 10⁷ cells/mL in a cryopreservation tube, which would be counted again and diluted to 3 × 10⁶, 1 × 10⁷, and 3 × 10⁷ cells/mL by 0.9% saline before injection. Patients lay on the examination bed in a lithotomy position, and cells were delivered to the junction area between the endometrium and myometrium (4 points, each with 500 μL) by a 17G needle that punctured through the posterior fornix of the vagina under B-mode ultrasound guidance.

Follow-up of clinical research

AEs refer to any unexpected medical events, including abnormal and clinically significant symptoms or laboratory results that occurred after the intervention (cell injection). SAEs refer to any unexpected medical events, including death, permanent disability, and other harmful events, that could damage the patient’s health seriously and require hospitalization or surgery. During the 3-year follow-up, all AEs, SAEs, surgery complications, and blood biochemical parameters were recorded and evaluated by two independent experienced clinicians to be irrelevant to IMRCs interventions; they judged the correlation between AEs or SAEs and cell injection. Hysteroscopy and ultrasound were performed once per month three times after cell injection to observe the changes in the uterine cavity (re-adhesion and area covering endometrium). Menstrual volume was recorded using a chart,⁴⁶ and EMT was measured by transvaginal three-dimensional ultrasound (maximum thickness in midsagittal plane) the day before hysteroscopic operations. If the anatomical structure of the uterus was normal 3 months after the operation, then patients were advised to prepare for spontaneous pregnancy or receive ET. The pregnant patients were followed until delivery, and their pregnancy-related complications were recorded. Clinical pregnancy refers to the presence of one or more gestational sacs on ultrasonographic visualization after positive human chorionic gonadotrophin (hCG) tests. Ongoing pregnancy means the presence of a living intrauterine fetus on ultrasonographic visualization at least after the 12th week of gestation. Live birth is defined as at least one live-born baby being delivered. Biochemical pregnancy refers to a transient increase in blood hCG that drops to normal levels quickly, and ultrasound finds no morphological evidence of pregnancy, suggesting the implantation failure of embryos. Miscarriage is defined as any pregnancy loss after clinical pregnancy.^48,49

Statistical analysis

In animal and in vitro experiments, ImageJ was used to analyze the results of the Masson staining, immunofluorescence, immunohistochemistry, wound healing assay, tube formation assay, and western blotting. Calculations and figures were performed using GraphPad Prism 8 (GraphPad, USA). An unpaired t test was used to assess the comparison in two groups, and a one-way analysis of variance (ANOVA) was used for the comparison in multiple groups.

In the clinical trial, the Statistical Program for Social Sciences version 24 (SPSS, USA) was used for the statistical analysis. Outcomes measures were analyzed based on per-protocol population. Continuous variables were analyzed by unpaired t test or ANOVA, and categorical variables were analyzed by chi-squared test or Fisher’s exact test. All data are represented as bar graphs, with error bars representing mean ± SD. A two-tailed p < 0.05 was considered statistically significant. The experiments described in this study were based on the analysis of at least three biological replicates or three technical replicates.

Data availability

The data and protocol are available upon request by contacting a corresponding author.

Acknowledgments

We would like to thank the professors from the Institute of Zoology of the Chinese Academy of Sciences, and the Reproductive Center of RTongji Hospital strongly supported this project. We thank SMART (https://smart.servier.com/) and BioRender (https://biorender.com/) for abundant image materials for schematic drawing. This study has been funded by the National Key Research and Development Program of China (grant nos. 2018YFA0108401 and 2021YFA1101604) and the Clinical Key Project of the Institute of Stem Cell and Regenerative Medicine Innovation in the Chinese Academy of Sciences (GXBLCYJ02).

Author contributions

B.H., J.W., L.J., Q.Z., W.L., and L.W. supervised the experiments. K.Q., C.S., and J.H. designed the study, with assistance from J.F. and F.N. Y.T. and Z.F. analyzed the scRNA-seq data, with help from Y.G., K.Q., and G.F. Z.L. and X.H. performed experiments and data analysis and created illustrations, with suggestions from K.Q., C.S., J.W., and H.Z. Q.L. and K.Q. participated in the hysteroscopic surgical procedures and cell injection, with help from B.H. and C.S. Q.L., and Y.L. followed up with the patients. Z.L., C.S., Y.L., and K.Q. wrote and revised the manuscript, with suggestions from J.L., T.G., X.H., W.L., W.W., Y.D., Z.L., J.Z., and Y.Z. All authors contributed to this manuscript and gave approval to the submitted version.

Declaration of interests

The authors declare no competing interests.

Supplemental information

Data S2. Statistical analysis

plan