Microfluidic chip-integrated vascularized endometrial complexes: Mitochondrial function and paracrine crosstalk enhance regenerative potential

Abstract

Endometrial injury is a prevalent gynecological condition that poses a significant threat to fertility and women's health. While the current reported endometrial organoids demonstrate potential in remodeling endometrial functions, they often lack the complexity and physiological relevance of in vivo tissue. Here, we introduce a vascularized triple-cellular endometrial complex integrating endometrial epithelial organoids, stromal cells, and endothelial cells within a microfluidic chip with a composite hydrogel comprising Matrigel and fibrin. This novel endometrial complex exhibits robust growth and endometrial repair capabilities in an immunodeficient mouse model of endometrial damage, significantly improving pregnancy rates. Single-cell RNA sequencing revealed bidirectional cellular paracrine crosstalk between epithelial, stromal, and endothelial cells in the vascularized endometrial complex. Endothelial cells secrete BMP6 and Galectin-9, which enhance mitochondrial function and promote epithelial cell proliferation. Conversely, epithelial and stromal cells secrete WNT7A and WNT5A, respectively, to stimulate angiogenesis and vascular network formation of endothelial cells. These findings reveal the paracrine interactions that underpin the superior regenerative properties of the vascularized triple-cellular endometrial complex, offering a potential therapeutic strategy for endometrial repair and a valuable in vitro model for endometrial pathophysiological studies.

Graphical abstract

Highlights

• •We developed a hormone-responsive vascularized endometrial complex in the composite hydrogel within a microfluidic chip.
• •Transplanted into the uterus, the complex promoted endometrial repair and restored fertility in mice with endometrial injury.
• •In the complex, endothelial cells boosted epithelial cell proliferation and mitochondrial function via paracrine interactions.

1. Introduction

The endometrium is the site for embryo implantation and development. Disruptions in normal endometrial processes, caused by factors such as uterine suction, curettage, or infections, can lead to various common health issues, including intrauterine adhesion, endometriosis, adenomyosis, embryo implantation failure, miscarriage, and endometrial carcinoma [[1], [2], [3], [4], [5]]. In recent years, endometrial epithelial organoids (EEOs) have emerged as a prominent research focus in reproductive biology [6,7], due to their value in studying endometrium regulation, disorders, and preclinical drug screening [8].

Various 3D models have been developed, such as assembloid models with both epithelial and stromal cells [9], endometrial epithelial cells cultured in high-purity collagen with stromal feeder cells [10], and models combining human embryonic stem cell-derived endometrial epithelial progenitor cells with endometrial stromal cells (ESCs) [11]. While these models have shown limited success in restoring fertility in rodent models [[12], [13], [14]], they still fall short of fully replicating the native tissue's complexity. Studies have highlighted the advantages of developing 3D organoid models with increased complexity and structural sophistication by incorporating multiple cell types [8]. Endothelial cells are crucial cellular components in the endometrium [15]. Incorporating endothelial cells into the construction of tissues and organs such as the liver [16], brain [17], kidneys [18], pancreatic islets [19], and cardiac system [20] aids in high-fidelity physiological function simulations and facilitates in vivo transplantation of the engineered tissues [21,22]. The endometrium is a complex tissue rich in vasculature, with spiral arteries playing pivotal roles in endometrial regeneration, menstrual shedding, maternal-fetal dialogue during embryo implantation, and placental oxygen supply [23,24]. Therefore, establishing a vascularized endometrium in vitro that includes endothelial cells, ESCs, and endometrial epithelial cells is of paramount importance for understanding the physiological and pathological regulatory roles of the endometrium.

The current endometrial organ models, which involve static cell culture in dishes or flasks, lack the dynamic fluidic environment crucial for in vivo cell functions – including fluid shear stress (which regulates vascular remodeling during gestation [25,26], facilitates embryo transport via uterine peristalsis [27,28], and contributes to maternal-fetal interface development [28]), substance transport, and cell-cell interactions. Significant progress has been made in the development of dynamic chip technologies, which integrate organoid methodologies with microfluidic systems to faithfully replicate physiological conditions [29]. These innovative approaches have successfully recreated organ architectures and functions, facilitating the study of cell-cell interactions, drug testing, and environmental toxicity assessment [29,30]. Thus, dynamic chip technologies hold great promise for advancing endometrial organ research by providing an accurate representation of in vivo physiological characteristics and a powerful tool to understand the complexities of the female reproduction system [31].

Our current research introduces an endometrium-on-a-chip (eCHIP) model that integrates EEOs, ESCs, and human umbilical vein endothelial cells (HUVECs) within a microfluidic chip embedded with a mixed hydrogel. The resulting triple-cell-type complex develops into a self-assembling microengineered vascularized endometrial organ-like complex. Importantly, this vascularized endometrial complex responds to reproductive hormones and effectively restores fertility potential in mouse models with endometrial damage. Through single-cell RNA sequencing (scRNA-seq), we unveiled the crucial role of the unique interactions between epithelial-stromal, epithelial-endothelial, and stromal-endothelial cells in mediating cell proliferation and extracellular matrix remodeling that consequently repairs the endometrium. Our study establishes a methodological framework for constructing robust vascularized endometrial complex, offering a potential therapeutic strategy for endometrial repair.

2. Results

2.1. Construction of the eCHIP in a microfluidic chip

To explore the culture conditions suitable for culturing EEOs, ESCs, and HUVECs, we conducted investigations into the composition ratios of the hydrogel and optimized the culture medium conditions (Fig. 1a). We used the positive area of vessel formation to gauge the network formation efficacy of HUVECs within the composite. We found that the optimal proportion of fibrin in the combination hydrogel of fibrin and Matrigel is 50 % (Fig. 1b and c). Increasing the concentration of fibrin beyond this point resulted in a significant level of adhesion of EEOs to the culture dish (Fig. 1b and c, P < 0.05). Furthermore, the hydrogel with 50 % fibrin demonstrated an intermediate stiffness and pore size between Matrigel and fibrin, providing an optimal physical environment for organoid growth and the establishment of vascular networks (Fig. 1d and e, Fig. S1). Different concentrations of fibrin did not lead to variation in the area of vessel positivity formed by HUVECs and EEO adhesion (Fig. 1f and g). We selected 5 mg/mL fibrin with a 1:1 ratio to Matrigel for subsequent experiments.

Fig. 1.

Optimization of the hydrogel composition for culturing the HEO complex. a. Experimental workflow diagram illustrating the cultivation of the HEO complex in hydrogels. b. Various ratios of fibrin and Matrigel, including 1:9, 3:7, 5:5, 7:3, and 9:1, were used to create composite hydrogels for static culturing the HEO complex (e.g., F1M9 refers to hydrogels composed of 10 % fibrin and 90 % Matrigel). Images were captured using fluorescent and bright-field microscopy at day 3. Green signals represent endometrial organoids and red signals represent endothelial cells. Scale bar: 200 μm. c. The left panel shows quantification of the vessel-positive area within the formed vascular network in various hydrogels using Image J, while the right panel shows the percentage of adherent organoids (day 3, n = 3–5). d. Scanning electron microscope images of different hydrogels, including Matrigel, fibrin, F1M9, F5M5, and F9M1. Scale bar: 2 μm. e. The left panel presents the pore size diameter of various hydrogels as determined by SEM results. The right panel displays the storage modulus values of the different hydrogels (n = 3). f. The static HEO complex was formed within F5M5 hydrogels containing varying concentrations of fibrin: 3 mg/mL, 5 mg/mL, and 10 mg/mL (day 3). Representative images are shown. Scale bar: 200 μm. g. Quantification of the vessel-positive area within the formed vascular network in various hydrogels and the percentage of adherent organoids are illustrated in histograms (n = 3–5). Statistical significance was determined using one-way ANOVA. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

We next developed a dynamic 3D culture system by embedding three distinct cell types into a microfluidic chip. Each chip featured six 30 μL wells interconnected by 1.0 mm-wide channels. The pre-solidified cell-laden hydrogels underwent perfusion using a flow controller programmed with a sinusoidal waveform (maximum flow rate: 10 μL/min; minimum flow rate: 1 μL/min; cycle period: 10 s) (Fig. 2a). We tested different ratios of EEO expansion medium to endothelial cell medium. A ratio of 5:5 led to a higher positive area of vessel formation while minimizing the adhesion of EEOs to the bottom of the microfluidic chip (Fig. 2b–d). Under dynamic flow conditions within the microfluidic chip, the triple-cell-type endometrial culture complex (comprising HUVECs, ESCs, and EEOs in hydrogels, termed the HEO complex) underwent self-assembly to form a highly interconnected network encompassing organoids. Live cell imaging revealed the self-organization of interconnected microvasculature on day 2, with substantial growth in the size of EEOs in the eCHIP over 6 days (Fig. 2e and Supplementary Video 1). Immunofluorescence staining using specific markers for endothelial cells (von Willebrand factor (vWF) and CD31), stromal cells (Vimentin), and epithelial cells (EPCAM) confirmed the presence of the three cell types within the HEO complex (Fig. 2f). Proliferative cells were observed in EEOs by Ki67 staining (Fig. 2f), which echoed the growth pattern of EEOs shown in Supplementary Video 1

Fig. 2.

Cultivation of the HEO complex in the endometrium-on-a-chip (eCHIP) model. h. Schematic of the eCHIP microfluidic system for dynamic HEO complex cultivation. The HEO complex was cultured within a μ-Slide III 3D Perfusion Uncoated plate (IBIDI, 80376) featuring six interconnected channels (width: 1.0 mm). Cell-laden hydrogels (30 μL/well) were solidified prior to perfusion. Culture medium was delivered via an OB1 MK3+ flow controller (Elveflow) programmed with sinusoidal flow (10 s cycle; 1–10 μL/min). i. Bright-field images of the dynamically cultured HEO complex in various media conditions examined to enhance the growth of vascularized complexes. The culture media consisted of different ratios of components: 90 % endometrial expansion medium (ExM) and 10 % endothelial cell medium (ECM), 50 % ExM and 50 % ECM, and 10 % ExM and 90 % ECM. Scale bar: 200 μm. j. The bar chart shows the quantification of the vessel-positive area within the established vascular network in different media compositions. The ratios on the X-axis correspond to the media formulations (n = 3). ∗P < 0.05, one-way ANOVA. k. The percentage of organoid adhesion (n = 3). ∗P < 0.05, one-way ANOVA. l. Bright-field images of the dynamically cultured HEO complex were captured on day 1 and day 3 using the CytoSmart imaging system. Scale bar: 200 μm. m. Representative immunofluorescent images of the dynamically cultured HEO complex expressing endothelial cell markers (vWF and CD31), a stromal cell marker (Vimentin), an epithelial cell marker (EPCAM), and a proliferation marker (Ki67). The experiment was replicated three times using cells from three individuals. DAPI staining was used to visualize nuclei. Scale bar: 100 μm.

2.2. Comparison of endometrial-derived endothelial cells and HUVECs in constructing vascularized multicellular endometrial complexes

To enhance the physiological relevance of the vascularized multicellular endometrial complexes, endothelial cells were isolated directly from human endometrial biopsy samples using CD31-targeted magnetic bead sorting. This approach efficiently enriched CD31⁺ endometrial-derived endothelial cells (eEndothelial cells) (Fig. S2a). Vascularized multicellular endometrial complexes incorporating these eEndothelial cells (eEndEO complexes) were constructed and compared to complexes incorporating HUVECs (HEO complexes) in dynamic culture condition. Phenotypic and functional analyses demonstrated comparable vascular network formation between eEndEO and HEO complexes (Fig. S2b). Furthermore, key physiological parameters - including cellular proliferation (EdU⁺ cells; Fig. S2c), apoptosis rates (Caspase-3⁺ cells; Fig. S2d), and mitochondrial membrane potential (JC-1 aggregates; Fig. S2d) - showed no statistically significant differences between the two groups (quantified in Fig. S2e–g). These results indicate that eEndothelial cells and HUVECs exhibit equivalent functionality in supporting the in vitro formation and basic physiology of vascularized multicellular endometrial complexes under these conditions.

However, serial passaging of isolated eEndothelial cells resulted in a progressive decline in the CD31⁺ cell population (Fig. S2h), highlighting a challenge for their long-term expansion in large-scale applications. Based on the comparable functional support demonstrated here and considering the accessibility and scalability of HUVECs, we utilized HUVECs as the endothelial source in subsequent studies.

2.3. Dynamic culture conditions significantly enhance epithelial cell proliferation in the HEO complex

Compared to static culture, dynamic conditions significantly enhanced the growth activity of endometrial epithelial cells in the HEO complexes, evidenced by a significantly faster increase in endometrial epithelial organoid diameter (Fig. 3a–b, P < 0.001), higher percentages of Ki67⁺ (Fig. 3c–d, P < 0.001) and EdU + cells (Fig. 3e–f, P < 0.0001), a lower percentage of apoptotic cells (Fig. 3g–h, P < 0.05), although no significant increase in mitochondrial activity (JC-1 aggregates, Fig. 3g and i).

Fig. 3.

Dynamic Culture Enhances Endometrial Epithelial Organoid Growth and Reduces Apoptosis Compared to Static Culture a-b. Organoid diameter under static vs. dynamic culture: (a) Bright-field images of endometrial epithelial organoids at Day 7 cultured under static (up) or dynamic (down) conditions; (b) Quantification of organoid diameter. Dynamic culture significantly increased organoid size (P < 0.001). Scale bar: 200 μm. c-d. Ki67 staining reveals enhanced proliferation in dynamic culture: (c) Confocal imaging of organoids stained with phalloidin (green, actin), Ki67 (red, proliferation marker), and DAPI (blue, nuclei); (d) Quantification of Ki67⁺ cells per organoid. Dynamic culture significantly increased Ki67 positivity (P < 0.001). Scale bar: 100 μm. e-f. EdU incorporation confirms enhanced proliferation: (e) Confocal imaging of EdU staining (green) and DAPI (blue) in organoids; (f) Quantification of EdU⁺ cells. Dynamic culture significantly increased EdU positivity (P < 0.0001). Scale bar: 100 μm. g-h. Reduced apoptosis in dynamic-cultured organoids: (g) Confocal imaging of JC-1 (red, mitochondrial membrane potential) and caspase-3 (green, apoptosis marker); (h) Quantification of caspase-3+ cells. Dynamic culture significantly decreased apoptosis (P < 0.05). Scale bar: 40 μm.Data presented as mean ± SD; ∗P < 0.05, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 (Student's t-test).

Calcein/PI staining of the dynamic 3D cultured HEO complex revealed that most cells were viable (Fig. 4a). Comparable cell viability, death, and mitochondrial ROS levels were observed between HEO and EO Complexes (Fig. S3). Immunohistochemistry results further confirmed the existence of CD31-positive micro-vessels proximal to EEOs (CK7⁺) and ESCs (CD10⁺, Fig. 4b). HUVECs displayed remarkable self-organization, forming a multilayered network of interconnected microvascular featuring a lumenized 3D structure (Fig. 4c, Supplementary Video 1–3). A capillary network-like structure was clearly seen in the dynamically cultured HEO complex, wherein some endothelial cells were tightly accompanied by ESCs (Fig. 4d, Supplementary Video 4). HUVECs and ESCs established cell-cell interactions with each other and formed direct intercellular connections with EEO cells, as shown by confocal microscopy and transmission electron microscopy (Fig. 4d, Fig. S4 and Supplementary Video 5).

Fig. 4.

The dynamically cultured HEO complex exhibits capillary network-like structures and intricate cellular interactions. n. Calcein/PI staining images of the dynamically cultured HEO complex. Calcein-labeled live cells are shown in green, and PI-labeled dead cells are shown in red. Scale bar: 400 μm. o. Immunohistochemistry images of the dynamically cultured HEO complex showing the proliferative cells (Ki67⁺), endothelial cells (CD31⁺), stromal cells (CD10⁺), and epithelial cells (Cytokeratin 7⁺, CK7⁺). Scale bar: 100 μm. p. The 3D reconstructed morphology of the dynamically cultured HEO complex showing a multilayered network of interconnected microvasculature featuring a lumenized 3D structure. The reconstructed 3D visual images were generated from confocal multi-layer continuous scanning images taken at intervals of 2 μm along the Z-axis. The images are displayed at rotations of 0, 45, 90, and 180°, respectively. HUVECs are labeled in red, while EEOs are labeled in green. Scale bar: 40 μm. q. Representative immunofluorescent images show endothelial cells forming the lumen tightly in conjunction with ESCs and cell-cell interactions among HUVECs, ESCs, and EEOs within the dynamically cultured HEO complex. HUVECs are labeled in red, ESCs in green, and cell nuclei stained with DAPI are shown in blue. The scale bars are 100 μm and 40 μm.

Bulk RNA-sequencing of endometrial tissues, the dynamically cultured HEO complex, EEOs, ESCs, and HUVECs revealed that the gene expression pattern of endometrial tissues was most similar to that of the HEO complex, while EEOs, ESCs, and HUVECs displayed unique gene expression profiles (Fig. S5). These results suggested that the in vitro culture model bore a high resemblance to the in vivo endometrium.

2.4. Responsiveness of the HEO complex to sex hormones

The endometrium exhibits notable responsiveness to ovarian hormones, specifically estrogen (E2) and progesterone (P4). To evaluate whether the dynamic cultured HEO complex retains this hormonal sensitivity, we established a sequential perfusion protocol in the eCHIP system (Fig. 5a). Following hormonal stimulation, secretory-phase markers PRL (produced by ESCs [32]), PAEP and SPP1 (expressed by endometrial epithelial cells [33]) were significantly upregulated (Fig. 5b, P < 0.001). Western blot and qRT-PCR analyses confirmed increased PAEP and SPP1 protein and mRNA expression (Fig. 5c–e, P < 0.01), accompanied by significantly elevated HSD17B2 transcript levels on days 2 and 6 (Fig. 5e, P < 0.01). After six days of hormonal treatment, immunofluorescence revealed decreased progesterone receptor (PR) expression alongside increased PAEP and SPP1 (Fig. 5f–h). Histological examination further demonstrated cytoplasmic secretory vacuoles in epithelial cells (Fig. 5i), with corresponding reductions in Ki67, ER, and PGR expression, and increased Mucin 1 (MUC1, Fig. 5i–j) and periodic acid-Schiff staining (Fig. 5i) in EEOs following E2 + P4 and cAMP stimulation. These collective changes indicate early secretory transformation [34]. Meanwhile, HEO complexes cultured under static conditions showed similar responsiveness to sex hormones (Fig. S6).

Fig. 5.

Hormonal responsiveness of dynamically cultured HEO complexes r. The schematic diagram illustrates the hormonal stimulation of the dynamically cultured HEO complex. The complexes were cultured in 5:5 ExM/ECM with E2 for 2 days, followed by E2, P4, and cyclic AMP (cAMP) for another 4 days. s. ELISA quantification of prolactin (PRL), SPP1, and PAEP secretion. t. Representative western blots of SPP1 and PAEP expression at days 2 and 6. u. Densitometric analysis of SPP1 and PAEP protein levels from (c). v. qRT-PCR analysis of PAEP and HSD17B2 mRNA expression. f–h. Immunofluorescence of progesterone receptor (PR, f), PAEP (g), and SPP1 (h) at indicated time points. Green: Target proteins; Blue: DAPI (nuclei); violet: cytoskeleton. Scale bars: 40 μm. i. H&E and immunohistochemical staining showing cellular morphology and Ki67/ER/PGR expression. White arrowheads indicate secretory vacuoles. Scale bars: 100 μm. j. Quantitative analysis of ER and PGR expression from (i). k. MUC1 expression in EEOs (day 6). Scale bars: 100 μm. l. Periodic acid-Schiff (PAS) staining of EEOs (day 6). Scale bars: 100 μm. Data presented as mean ± SD; ∗∗P < 0.01, ∗∗∗∗P < 0.0001 (Student's t-test or one-way ANOVA).

2.5. Restoration of reproductive capacity by the HEO complex

To evaluate the regenerative potential of the multicellular endometrial complexes, we established a mouse model of endometrial damage. All transplanted complexes underwent 2-day pretreatment with E2 prior to intrauterine injection. The experimental groups received PBS control, Matrigel-only control, E2-pretreated dynamically cultured EO complexes, and E2-pretreated dynamically cultured HEO complexes. After seven days, the damaged endometrium treated with PBS (termed Nature Repair group, NR), exhibited a significantly reduced thickness and a lower number of endometrial glands compared with the endometrium in the normal control group (NC) (Fig. 6a–c, P < 0.05), confirming successful endometrial damage modeling. The EO complex showed superior endometrial repair capacity compared with Matrigel, but was inferior to the HEO complex which resulted in a significantly thicker endometrium and more developed endometrial glands (Fig. 6a–c, P < 0.05). Notably, the delta number of embryos per uterine horn, calculated as the embryo number in the left (damaged uterus) minus the embryo number in the right (normal uterus), was significantly lower in the HEO group compared with that in the NR group (Fig. 6d and f, P < 0.05). Additionally, the pregnancy rate was highest in the HEO group among all groups (Fig. 6e and f, P < 0.05). These findings suggest that HEO treatment resulted in the most favorable reproductive outcomes among the four groups.

Fig. 6.

The dynamically cultured HEO complex facilitates endometrial regeneration and improves reproductive capacity in mice. w. Histological assessment of endometrial repair. Hematoxylin and eosin staining of uterine sections from: Natural repair control (NR): PBS-injected injured uteri, Matrigel control (MG): Injured uteri repaired with Matrigel™, E2-pretreated EO complexes, E2-pretreated HEO complexes, Normal control (NC): Uninjured uteri. All transplanted cellular complexes (EO/HEO) received 2-day E2 pretreatment prior to injection. Red arrows indicate glands and blue dashed lines denote the endometrium–myometrium boundary in the 4 × images. Scale bar: 200 μm and 100 μm. x. Bar chart displaying the average number of endometrial glands per slide in each group. n = 12; ∗P < 0.05 versus NC; #P < 0.05 versus NR; $P < 0.05 versus MG; ^P < 0.05 versus EO. Analysis was performed using one-way ANOVA. y. Bar chart illustrating the average endometrial thickness of the uteri in each group. n = 12; ∗P < 0.05 versus NC; #P < 0.05 versus NR; $P < 0.05 versus MG; ^P < 0.05 versus EO. One-way ANOVA was conducted for statistical analysis. z. Bar chart displaying the difference (△) in the number of embryos per uterine horn (left horn count minus right horn count) in each group. n = 10; ∗P < 0.05 versus NR. Analysis was carried out using one-way ANOVA. aa. Pregnancy rate of each group represented by the ratio of pregnant right uterine horns to pregnant left uterine horns. n = 10; ∗P < 0.05. Fisher's exact test was conducted for statistical analysis.bb. Images of uterine, embryo, and placental morphologies under different treatments. The right uterus was damaged by ethanol infusion (as indicated by the red arrow).cc. Immunohistochemical staining images of CK7, ER, and PGR in EO and HEO groups. ER and PGR antibodies specifically target human ER and PR, respectively. Scale bar: 200 μm dd. Microscopic observation of the formed graft from the subcutaneously injected HEO complex (top panel, scale bar: 400 μm). Visualization of human (HUVECs, green) and host (mouse endothelial cells, mEC, red) vessels and their connections (arrow) within the graft (middle panel, scale bar: 100 μm). Dextran infusion (red) demonstrates the functional formation of human vessels (HUVECs, green; bottom panel, scale bar: 100 μm).

To investigate the mechanisms underlying the beneficial effects of the HEO complex, we tracked transplanted cell dynamics and assessed human-specific markers. Fluorescent signals from CMTPX-labeled EEOs and CMFDA-labeled HUVECs persisted robustly at graft sites through days 3 and 7 post-transplantation, with progressive signal attenuation by day 14 (Fig. S7a–b). CMFDA⁺cells organized into distinct annular formations indicative of nascent vascular structures (Fig. S7c, arrowheads), while CMTPX ⁺ epithelial cells stably integrated within endometrial tissue layers. Immunostaining confirmed the presence of human CK7⁺, Vimentin⁺, and CD31⁺ cells in HEO grafts by day 3 (Fig. S8). By day 7, human-derived epithelial cells expressing ERα and PGR were specifically localized to the endometrial layer of the mouse uterus (Fig. 6g), demonstrating successful engraftment of human tissue. Notably, EO grafts lacked detectable human CD31 expression at this stage.

Furthermore, we subcutaneously injected the dynamically cultured HEO complex into mice and retrieved the grafts after 14 days. Stereomicroscopy revealed well-developed blood vessels in the grafts (Fig. 6h). By infusing human-specific fluorescent substances (UEA-1) and mouse-specific fluorescent substances (Lectins), we confirmed the connection between the human and mouse blood vessel networks in the grafts (Fig. 6h, Supplementary Video 6). Additionally, perfusion of fluorescein-conjugated dextran and UEA-1 revealed that the green-labeled human blood vessels exhibited a red fluorescent signal (Fig. 6h, Supplementary Video 7), indicating perfusion of mouse blood within the human blood vessel network in the grafts.

2.6. Induced epithelial and stromal cell proliferation by HUVECs

To decipher the intercellular interactions between EEOs, ESCs, and HUVECs, we conducted scRNA-seq analysis on the dynamically cultured HEO complex and EO complex under organoid expansion medium (ExM)/endothelial cell medium (ECM) + estrogen (e for short) and ExM/ECM + estrogen + progesterone + cAMP (p for short) treatment, therefore consisting of four groups (HEOe, HEOp, EOe, and EOp). A comprehensive map of 99315 cells from 12 samples was generated, which were divided into 11 clusters. Tissue/cell-specific markers assigned the cells into four distinct categories: endometrial epithelial cells (Epi), endometrial stromal cells (Str), endothelial cells (Endo), and macrophages (Mf) (Fig. 7a).

Fig. 7.

Single-cell RNA sequencing (scRNA-seq) analysis of the dynamically cultured EO and HEO complexes. a. UMAP visualization of 12 samples across four culture conditions: HEOe (HEO + estrogen), HEOp (HEO + estrogen + progesterone + cAMP), EOe (EO + estrogen), EOp (EO + estrogen + progesterone + cAMP). Endometrial epithelial cell clusters were annotated as main Epi (mEpi), ribosome-related Epi (rEpi), proliferative Epi (pEpi), and ciliated Epi (cEpi). Endometrial stromal cell clusters were annotated as main Str (mStr), ribosome-related Str (rStr), proliferative Str (pStr), and a second proliferative Str cluster (pStr2); the endothelial cell clusters included main endothelial cells (mEndo) and ribosome-related endothelial cells (rEndo). b–c. Relative abundance of epithelial (b) and stromal (c) clusters within their respective compartments across groups.]d. Comparison of the percentage of proliferative epithelial cells (pEpi, left panel) and proliferative stromal cells (pStr, right panel) between the dynamically cultured EO and HEO complexes under estrogen culture.e. Validation of enhanced proliferation in HEO complexes by flow cytometry. Significantly higher proportions of EdU + cells in epithelial cells of HEO vs. EO complexes under estrogen culture.f. Ligand-receptor pairs exhibiting significant differences in communication probability between HEOe and EOe complexes. Pathways of biological interest include WNT7A, WNT5A, BMP6, and LGALS9. Dot size represents the maximum communication probability across compared groups; dot color indicates statistical significance (P < 0.05). Cluster abbreviations: Epi (endometrial epithelial cells), Str (endometrial stromal cells) and Endo (endothelial cells).g. Circle plot showing the BMP6, LGALS9, WNT7A, WNT5A signaling networks among different cell types in EOe and HEOe complexes.

Based on the expression of distinct marker genes and gene ontologies (Fig. S9), the four clusters belonging to endometrial epithelial cells were designated as main Epi (mEpi), ribosome-related Epi (rEpi), proliferative Epi (pEpi), and ciliated Epi (cEpi) (Fig. S10a–b and Supplementary Table 1). Endometrial stromal clusters were designated as main Str (mStr), ribosome-related Str (rStr), proliferative Str (pStr), and a second proliferative Str cluster (pStr2) (Fig. S10a and c, Supplementary Table 2); the endothelial cells included main endothelial cells (mEndo) and ribosome-related endothelial cells (rEndo) (Fig. S10a and d). All rEpi, rStr, and rEndo cells exhibited a high expression of ribosome-related genes. When compared to other non-ribosome-related clusters, their functions were enriched in Gene Ontology (GO) terms such as “Ribosome, cytoplasmic” and “40S ribosomal subunit, cytoplasmic” (Fig. S11a-b, Fig. S12 and Supplementary Table 3–10). This expression proximity therefore attributed to the close positioning of rEpi, rStr, and rEndo in the UMAP plot (Fig. 7a).

Both pEpi and pStr were characterized by a high expression of genes including CKS1B, CKS2, MKI67, PTTG1, STMN1, TUBA1B, and UBE2S genes, which involved cellular proliferation–related GO terms such as “cell division,” “cell cycle, Mitotic,” and “cell cycle” (Fig. S11a–b, Fig. S12, and Supplementary Table 3–10). Fig. 7b and c presents the proportions of epithelial and stromal clusters within their respective total cell populations. Critically, a significant increase in pEpi and pStr proportions was observed in the HEO group versus the EO group (Fig. 7d; P < 0.05). This finding was independently validated by flow cytometry, which confirmed significantly higher proportions of EdU⁺ cells in epithelial of HEO complexes compared to EO complexes under estrogen culture (Fig. 7e, P < 0.01; Fig. S13), demonstrating enhanced proliferative activity in HEO-derived endometrium. Proportions of all other clusters remained consistent between groups (Fig. S14).

To investigate cell-to-cell communication, we performed CellChat analysis under estrogen culture conditions. Cellular communication occurred not only between stromal and epithelial cells, but also extensively between endothelial and epithelial cells, as well as between endothelial and stromal cells (Fig. S15). In estrogen-cultured HEO complexes, specific ligand-receptor pairs were identified, including WNT7A-FZD4-LRP5/6 in epithelium-to-endothelium signaling and WNT5A-MCAM-FZD4/6 in stroma-to-endothelium signaling (Fig. 7f–g). Furthermore, a distinct endothelium-to-stroma signaling axis mediated by VEGFA/B-VEGFR1/2 was identified under progesterone culture (Fig. S15).

2.7. Impact of HUVECs on epithelial cells

The differentially expressed genes (DEGs) between the dynamically cultured HEO and EO complexes are shown in Supplementary Table 11–18. For both mEpi and pEpi clusters of HEO complexes compared to their counterparts in EO complexes, their DEGs were enriched in GO terms such as “ovulation cycle,” “cellular response to hormone stimulus,” and “G2/M transition of mitotic cell cycle” (Fig. S16 and Supplementary Table 19–22). These findings, combined with experimental validation of increased EdU⁺ cells in HEO complexes (Fig. 7e), support that epithelial cells in HEO undergo enhanced proliferation. To further investigate the functional consequences of these proliferative changes, we examined the epithelial barrier integrity formed by HEO-derived epithelial cells. Key components of intercellular junctions—such as CDH5, CTNNB1, CLDN5, JAM3, TJP1, and GJA1 (Fig. S17). This gene expression evidence of enhanced cell-cell adhesion suggested a potential for improved barrier function. To test whether this molecular signature translated into a functional phenotype, we performed Transepithelial Electrical Resistance (TEER) measurements on confluent monolayers, as TEER is the gold-standard assay for quantifying epithelial barrier integrity. TEER measurements using the EVOM2 voltohmmeter on Transwell inserts revealed that the HEO complex exhibited a significantly faster increase in electrical resistance over time compared to the EO complex (Fig. S18a, P < 0.05). This accelerated resistance growth implied a greater abundance of functional cell-cell connections in the HEO complex, consistent with the upregulation of junctional genes. Together, these results demonstrate that the co-culture with HUVECs not only promotes epithelial cell proliferation but also enhances the structural and functional integrity of the epithelial barrier, likely through the reinforcement of tight and adhesion junctions.

We also observed higher expression of mitochondrial genes MT-RNR1, MT-RNR2, and MT-ND4L in the dynamically cultured HEO complex relative to the dynamically cultured EO complex; these emerged as the top upregulated DEGs by scRNA-seq (Fig. 8a) and were validated by qRT-PCR (Fig. 8b and Supplementary Table 31). JC-1 staining and flow cytometry revealed that the number of JC-1 aggregates of epithelial cells in the dynamically cultured HEO complex was higher than those in the dynamically cultured EO complex, indicating a stronger mitochondrial transmembrane potential in epithelial cells of the HEO complex compared with that in the EO complex (Fig. 8c and Fig. S18b). This elevated mitochondrial membrane potential is a key indicator of enhanced mitochondrial function and bioenergetic capacity [35,36]. Critically, robust mitochondrial function, characterized by high membrane potential, is essential for supporting cellular ATP production required for active processes like epithelial cell growth and proliferation [37]. Transmission electron microscopy revealed a significantly increased number of mitochondria in epithelial cells within the dynamically cultured HEO complex compared with that in the dynamically cultured EO complex (Fig. 8d and Fig. S18c, P < 0.05). Collectively, these findings demonstrate that HUVECs promote the growth of epithelial cells in the HEO complex and are associated with enhanced mitochondrial function (as evidenced by increased gene expression, membrane potential, and mitochodrial abundance) in these epithelial cells.

Fig. 8.

Endothelial cells promote mitochondrial function and extracellular matrix gene expression of endometrial cells within the dynamically cultured HEO complex.a. Comparison of expression levels of MT-RNR1, MT-RNR2, and MT-ND4L genes between the dynamically cultured EO and HEO complexes in different clusters.b. EEOs were isolated from the dynamically cultured EO and HEO complexes using a mouth pipette. Comparison of MTRNR1, MTRNR2, and MT-ND4L genes in EO and HEO complexes. GAPDH was used as the internal control. ∗∗P < 0.01, ∗∗∗P < 0.001, Student's t-test.c. Flow cytometry analysis showing JC-1 aggregates of mitochondria in the dynamically cultured EO and HEO complexes.d. Ultrastructure of mitochondrion in the dynamically cultured EO and HEO complexes by transmission electron microscopy. Triangle, nucleus; arrow, mitochondrion. Scale bar: 5 μm.e. Flow cytometry analysis of mitochondrial JC-1 aggregates in the dynamically cultured EO and HEO complexes. In the HEO complex, inhibitors of BMP6 (dorsomorphin, 10 μM) and CD44 (HA-CD44, 10 μM) were used to inhibit the respective receptors of EEOs. In the EO complex, recombinant proteins BMP6 (50 ng/mL) or Galectin-9 (10 ng/mL) were introduced as ligands.f. Quantitative results of flow cytometry assays in panel e. ∗P < 0.05, ∗∗∗P < 0.001, one-way ANOVA.g. The electrical resistance, measured using the EVOM2 voltohmmeter in Transwell inserts, reflects the growth of cells in HEO and EO complexes treated with inhibitors and recombinant proteins, respectively. ∗∗∗P < 0.001, two-way ANOVA.h. Comparison of the gene expression level of MMPs (MMP1 and MMP3) and collagens (COL1A1, COL4A1 and COL5A3) between the dynamically cultured EO and HEO complexes in different clusters.i. Gene expression levels of MMP1 and MMP3 in EEOs isolated from the dynamically cultured HEO complex relative to that from the dynamically cultured EO complex. 18S was used as internal control. ∗∗∗P < 0.001, Student's t-test.j. MMP and collagen expression in EO and HEO complexes under estrogen treatment. Left: Representative western blots of MMP1, MMP3, COL1A1, and COL4A1. β-ACTIN was used as loading control. Right: Quantitative analysis of protein expression levels (normalized to β-ACTIN), measured using ImageJ software.Data presented as mean ± SD; ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001 (Student's t-test or one-way ANOVA).

In the intercellular communication findings, BMP6 and Galectin-9 were identified as distinct and specific to the communication from endothelial cells to epithelial cells (Fig. 7f–g). To validate the roles of BMP6 and Galectin-9 in epithelial cells, cell proliferation activity was assessed using EdU staining. Inhibition of BMP6 and Galectin-9 receptors (by dorsomorphin and HA-CD44) resulted in reduced proliferation of epithelial cells in the dynamically cultured HEO complex, while the addition of BMP6 and Galectin-9 in the dynamically cultured EO complex led to increased proliferation of epithelial cells (Fig. S18d-e, P < 0.001). Organoid formation was hindered by inhibitors of BMP6 or Galectin-9 receptors in the HEO complex and enhanced by the presence of BMP6 and Galectin-9 in EO cultures (Fig. S18f-g, P < 0.001). Consistent with the growth of organoids, inhibition of BMP6 and Galectin-9 reduced the percentage of JC-1 aggregates in the HEO complex, while BMP6 and Galectin-9 increased the percentage of JC-1 aggregates in the EO complex (Fig. 8e and f, P < 0.05; Fig. S18h). Similar trends were observed in electrical resistance testing using the EVOM2 voltohmmeter in Transwell inserts; resistance significantly decreased with dorsomorphin and HA-CD44 and significantly increased with BMP6 and Galectin-9 (Fig. 8g, P < 0.001). These findings provide robust evidence that endothelial cells promote the growth of endometrial epithelial cells in the HEO complex through BMP6 and Galectin-9 signaling.

2.8. Impact of HUVECs on the formation of extracellular matrix

Compared to the dynamically cultured EO controls, MMP1 and MMP3 expression was significantly reduced across all four stromal cell clusters in the HEO complexes (Fig. 8h and Supplementary Table 11–18). Conversely, COL1A1, COL4A1 and COL5A3 expression increased specifically in the mStr and pStr subpopulations (Fig. 8h and Supplementary Table 11–18). Significantly lower MMP1 and MMP3, and higher COL1A1and COL4A1 expression levels were also confirmed in the dynamically cultured EO complex compared with the levels in the HEO complex by WB and qRT-PCR (Fig. 8i and j, P < 0.05). Additionally, the DEGs that were more prominently identified in the mStr and pStr clusters of the dynamically cultured HEO complex compared with those of the EO complex demonstrated a heightened correlation with GO terms such as “ECM proteoglycans,” “blood vessel development,” “cell-substrate junction assembly,” and “response to wounding” (Fig. S16b and Supplementary Table 23–27). This result suggested that endothelial cells might have other effects on ESCs in the dynamically cultured HEO complex.

2.9. Impact of epithelial and stromal cells on HUVECs

We observed that the primary intercellular communications were from the epithelial and stromal cells targeting endothelial cells, specifically the communications from mEpi and pEpi to mEndo and from mStr and pStr to mEndo (Fig. S15). We thus explored whether epithelial and stromal cells have a potential impact on endothelial cells. In the ligand-receptor interaction results, WNT7A and WNT5A garnered our attention, with the former exclusively acting from epithelial cells to endothelial cells, while the latter operating solely from stromal cells to endothelial cells (Fig. 7g and Fig. S15).

In the HUVEC only culture system, both WNT7A and WNT5A promoted sprouting within the endothelial cells in the gel and formation of vascular networks on the gel surface (Fig. 9a–d). Conversely, the WNT pathway inhibitor DKK1 (targeting LRP5/6 receptors required for WNT7A signaling [38,39]) and the specific WNT5A antagonist BOX5 [40,41] significantly inhibited sprouting and vascular network formation (Fig. 9a–d, P < 0.05). Notably, neutralizing antibodies against WNT7A (α-WNT7A) and WNT5A (α-WNT5A) produced comparable inhibitory effects (Fig. S19), confirming the specificity of observed phenotypes. Additional experiments in a multicellular dynamic culture system validated that the introduction of WNT7A in the hydrogel containing HUVECs and ESCs (HE complex) or WNT5A in the hydrogel with HUVECs and EEOs (HO complex) significantly enhanced the vascular network development in dynamically cultured complexes (Fig. 9e and f, P < 0.05). In contrast, the addition of DKK1 and BOX5 separately in the dynamically cultured HEO complex led to a significant reduction in the formation of vascular networks within the complex (Fig. 9e and f, P < 0.05).

Fig. 9.

Endometrial epithelial and stromal cells enhance the formation of vascular networks in HUVECs within the HEO complex through WNT7A and WNT5A signaling pathways. a. Representative images of sprouting assays using HUVEC spheroids treated as indicated. HUVEC spheroids cultured with basic ECM medium are shown as the blank group (Control). VEGFA added to the basic ECM medium served as the positive control. Scale bar: 400 μm. b. Quantification of sprouting number was determined by Image J software. ∗P < 0.05 versus Control; #P < 0.05 versus WNT7A; $P < 0.05 versus DKK1; ^P < 0.05 versus WNT5A; &P < 0.05 versus BOX5; one-way ANOVA was conducted for statistical analysis. c. Representative images of the tube formation of HUVECs on Matrigel surface treated as indicated. HUVECs cultured with basic ECM medium are shown as the blank (Control). VEGFA added to the basic ECM medium served as the positive control. Scale bar: 400 μm. d. Quantification of branch number using Image J software. ∗P < 0.05 versus Control; $P < 0.05 versus DKK1; &P < 0.05 versus BOX5; one-way ANOVA was conducted for statistical analysis.e. Vascular networks of complex with HUVECs and ESCs (HE), complex with HUVECs and EEOs (HO), and HEO treated with WNT7A (or its inhibitor DKK1) and WNT5A (or its inhibitor BOX5) as indicated. Vascular networks are depicted in green. Scale bar: 100 μm.f. Quantification of the vessel-positive area within the formed vascular network across different groups. ∗P < 0.05, ∗∗P < 0.01, Student's t-test.g. The electrical resistance of HEO and HE with DKK1 and WNT7A treatments, respectively. ∗∗∗P < 0.001, two-way ANOVA.h. The electrical resistance of HEO and HO with BOX5 and WNT5A treatments, respectively. ∗∗∗P < 0.001, two-way ANOVA.

The formation of vascular networks was further confirmed by measuring electrical resistance using the EVOM2 voltohmmeter in Transwell inserts. The HEO complex exhibited a significantly faster increase in electrical resistance over time compared with the EO complex (Fig. 9g, P < 0.001), indicating a higher abundance of cell connections in the HEO complex. The HE complex, which lacked epithelial cells, demonstrated a notably reduced resistance compared with the HEO complex that contained a higher proportion of integral epithelial cells. DKK1 significantly reduced the resistance properties of the HEO complex (Fig. 9g, P < 0.001), suggesting that DKK1 inhibits the capacity of epithelial cells within the HEO complex to facilitate the assembly of vascular networks. Conversely, WNT7A significantly increased the resistance within the HE complex, partially restoring the compromised potential for vascular network formation in the absence of epithelial cells (Fig. 9g, P < 0.001). This finding underscores the critical role of WNT7A, a protein secreted by epithelial cells, in enhancing the formation of vascular networks within the HEO complex. The TEER assay also highlighted the importance of WNT5A, secreted by stromal cells, in promoting the formation of HUVEC vascular networks in the HEO complex (Fig. 9h, P < 0.001).

3. Discussion

Vascularized organoids, referred to as the next-generation organ model [42], can be constructed using various methods, including genetic engineering [19,43], multidirectional differentiation [20,44,45], microfluidics and 3D bioprinting [21,46], in vivo organoid transplantation [47], and coculture with endothelial cells [17,48]. Notably, the use of induced pluripotent stem cells or embryonic stem cells for organoid construction for application in tissue regeneration and repair raises ethical and safety concerns. While animal models have been successfully used to construct complex vascularized organoids for studying physiological and pathological processes [49], they are unsuitable for direct application in tissue and organ repair for humans. Hence, using human endothelial cells along with microfluidics or 3D bioprinting techniques has emerged as the main approach for constructing vascularized organoids aimed at regenerating damaged tissues. In our study, we established a microfluidic-based culture system to generate vascularized endometrial complexes consisting of endometrial epithelial cells, stromal cells, and endothelial cells, with the aim of facilitating tissue regeneration. Incorporating fibrin into our hydrogel formulation facilitated the robust growth of endothelial cells and the formation of a well-developed vascular network. The presence of fibrin fibers in the hydrogel creates an optimal microenvironment for endothelial cells, enabling the formation of robust vascular microcapillaries [50,51]. The vascularized endometrial complex exhibits a morphological architecture akin to the in vivo human endometrium, accompanied by noteworthy endometrial cell proliferation and substantial hormone responsiveness. Furthermore, the comprehensive gene expression of the entire cell population demonstrates a striking similarity to that of the female endometrium in vivo, a feat unmatched in prior research endeavor [9].

Previous studies have consistently demonstrated the regenerative potential of organoids in various tissue systems, as evidenced by their successful application in liver rejuvenation [52,53], kidney repair [44,54], and cerebellum rebuilding [55]. While in vitro studies have reported the co-formation of endometrial complexes using both endometrial stromal cells and epithelial cells [[56], [57], [58], [59]], referred to as “endometrial assembloids” [9,59], few studies using in vivo animal experiments to validate the reparative potential in damaged endometrium have been conducted [11]. In our study, the dynamically cultured vascularized multicellular endometrial complexes demonstrated significant recovery effects on injured endometrium in a mouse model, accompanied by pronounced angiogenesis and robust regeneration of the endometrial tissue. These vascularized endometrial complexes not only exhibited remarkable capacity for restoring damaged endometrium when transplanted into a mouse model, but also significantly enhanced embryo implantation rates. This phenomenon, which has not been previously reported in endometrial assembloids, highlights the crucial role of neovascularization in graft survival after in vivo transplantation [60]. Furthermore, our subcutaneous transplantation experiments in mice demonstrated the involvement of vascular endothelial cells within the dynamically cultured vascularized endometrial complexes in the formation of new blood vessels within the host. Importantly, the endothelial cells from the complexes seamlessly integrated with the newly formed blood vessels in mice, contributing to the functional perfusion of the transplanted tissue. This finding provides insight into one of the significant advantages of the dynamically cultured vascularized endometrial complexes in repairing damaged endometrium. These complexes exhibit a high abundance of vascular generation following transplantation into the uterus. This observation aligns with the increased presence of CD31⁺ endothelial cells within the endometrium of the transplanted mice.

Through single-cell sequencing analysis, we identified a notable increase in the proportion of proliferating endometrial epithelial cells and stromal cells when co-cultured with HUVECs, potentially explaining the enhanced capacity for endometrial repair. Furthermore, these cells exhibited enrichment in crucial cell cycle regulatory pathways such as G1/S transition, spindle checkpoint control, and mitotic cytokinesis. Notably, we observed elevated expression of mitochondrial genes and enhanced mitochondrial function, which play key roles in regulating cellular energy production during proliferation. Crucially, the enhanced mitochondrial function observed in the HEO complexes, evidenced by increased gene expression, mitochondrial number, membrane potential, is strongly implicated in supporting the heightened proliferative state. Cellular proliferation is an energetically demanding process, requiring substantial ATP to fuel essential biosynthetic activities such as DNA replication and protein synthesis [61]. The augmented mitochondrial capacity likely provides the necessary bioenergetic support for these processes, thereby contributing directly to the increased proliferation seen in the vascularized complexes and their enhanced reparative potential. This mechanistic link between improved mitochondrial energetics and cellular proliferation is well-supported in regenerative contexts [[62], [63], [64], [65]]. Blood vessels supply nutrients and oxygen, which may explain these outcomes. Additionally, the paracrine activities of endothelial cells should not be overlooked. Increasing evidence suggests that blood vessels influence the functional status of neighboring tissues [66], as seen in their involvement in regulating liver lipid metabolism [67], promoting neurogenesis and alleviating cellular stress [17], and supporting pancreatic islet cell survival and proliferation [68]. Therefore, deeper understanding of the mechanisms underlying endometrial injury repair and functional regulation necessitates research into the direct influence of endothelial cells on endometrial cells. This approach will enhance the comprehension of the underlying physiological and pathological processes.

Our findings also revealed that BMP6 and Galectin-9 secreted by vascular endothelial cells significantly promoted the growth of endometrial epithelial cells. This may partly explain why including endothelial cells in our dynamic 3D culture system contributed to the proliferation of endometrial epithelial cells and stromal cells. It may also elucidate why the dynamically cultured HEO complexes containing endothelial cells exhibited a greater capacity to promote endometrial repair and enhance fertility in animal models with damaged endometrium compared with the dynamically cultured EO complexes without endothelial cells. However, the specific roles of BMP6 and Galectin-9 in endometrial cells need to be explored in depth. BMP6 belongs to the BMPs, a class of highly conserved members of the transforming growth factor β (TGFβ) family with crucial functions during development, morphogenesis, and reproduction [69]. BMPs signal through a heterotetrameric cell surface receptor complex comprising two BMP type 1 receptors (ALK2/ALK3/ALK6) and two BMP type 2 receptors (ACVR2A/ACVR2B/BMPR2) that activate intracellular signaling via the SMAD1/5 transcription factors [70]. Conditional deletion of BMPR2 in the mouse endometrium leads to female fertility defects [71], proving the vital role of BMP signaling in endometrial function. Given that both HEO and EO complexes utilized endometrial cells from identical donors, inherent differences in basal BMP receptor expression are unlikely to account for enhanced signaling in HEO cultures. Rather, our scRNA-seq data suggest that paracrine BMP6 secretion from incorporated HUVECs may drive this effect [72,73]. Alternatively, endothelial crosstalk might upregulate BMP receptors in endometrial cells, enhancing ligand sensitivity - a potential positive feedback mechanism meriting future investigation.

Dorsomorphin selectively inhibits BMP type I receptors ALK2, ALK3, and ALK6, blocking downstream BMP signaling pathways [74]. In our study, dorsomorphin significantly inhibited the growth of organoids with vascular endothelial cells in the dynamically cultured HEO complex, while supplementation of BMP6 in the dynamically cultured EO complex without endothelial cells promoted organoid growth. The functions of BMP2, BMP7, and other BMPs in the endometrium are related to the classical SMAD pathway and non-classical pathways such as MAPK, AKT, and RhoA signaling [75]. The regulatory mechanisms of BMP6 in endometrial function require further exploration.

Galectin-9, a β-galactoside lectin protein and an important member of the galectin family, is involved in cell development, adhesion processes, innate and adaptive immune regulation, inflammatory responses, and immune suppression activity [[76], [77], [78]]. Galectin-9 is expressed in all uterine cell types, including endometrial epithelial and endothelial cells [79,80]. Recent research in a preeclampsia mouse model showed that trophoblast-derived Galectin-9 from the placenta activates decidual macrophages through CD44 binding to inhibit spiral artery remodeling [81]. Administration of Galectin-9 in mice induces preeclampsia-like phenotypes, suggesting that Galectin-9 may serve as a biomarker for predicting and intervening in preeclampsia [81]. Studies on decidualization during pregnancy demonstrated that Galectin-9 expressed by macrophages and fibroblasts may work with fetal extravillous trophoblasts to reduce immune cytotoxicity following natural killer cell-induced disruption of arterial smooth muscle [82,83]. We found that Galectin-9 derived from vascular endothelial cells promoted the growth of EEOs in the dynamically cultured HEO complex through CD44. The specific mechanisms of Galectin-9/CD44 in the endometrium require further investigation.

Single-cell analysis revealed distinct expression patterns of genes involved in extracellular matrix (ECM) regulation between the dynamically cultured HEO and EO complex. We observed marked elevation of COL1A1, COL4A1, and COL5A3, which encode collagen types I, IV, and V respectively, key components of the endometrial ECM, in the HEO complex. Conversely, MMP1 and MMP3, enzymes that degrade ECM components like collagen, fibrin, laminin, and fibronectin, were notably downregulated in the HEO complex. The enhanced accumulation of collagen and reduced MMP activity in the HEO complex may result in a more stable ECM structure. This balanced maintenance of the ECM is crucial for supporting sustained cell growth and proliferation [84], potentially explaining the heightened proliferative potential and improved reparative capacity of the HEO complex. Further research is required to elucidate the precise mechanisms through which HUVECs in the HEO complex directly influence collagen and MMP expression and ECM remodeling in endometrial cells. This influence may be attributed to the paracrine function of HUVECs, involving the secretion of various growth factors and cytokines [85].

Previous research has primarily focused on the effects of blood vessels on other cells within organoids, overlooking the reciprocal influence of other cells on blood vessels. In our comprehensive analysis of intercellular communication within the dynamically cultured HEO complex, we discovered that the secretion of WNT7A by endometrial epithelial cells and WNT5A by endometrial stromal cells significantly enhanced vascular network formation and branching proliferation of HUVECs. The WNT family, a highly conserved group of secreted glycoproteins, plays vital roles in various biological processes [86], regulating angiogenesis and impacting the growth, migration, and lumen formation of vascular endothelial cells during organoid formation [87,88]. This suggests that increasing levels of WNT7A or WNT5A may stimulate endometrial vascular neogenesis and facilitate endometrial repair processes. Moreover, in vitro manipulation through the addition of WNT may also regulate the cultivation of vascularized complexes. However, the impact of WNT7A and WNT5A on angiogenesis in the endometrial complex needs to be further investigated.

Our in vitro model, which incorporates endometrial cells and vascular endothelial cells, is highly effective for studying cellular interactions within the endometrium. The single-cell sequencing results revealed the presence of a small number of macrophages in both the dynamically cultured HEO and EO complexes. We speculate that these macrophages may originate from trace amounts of macrophages present in the low-passage HUVECs or ESCs used to establish our culture system. Given their minimal presence, we believe these macrophages do not significantly impact our study's conclusions. This observation also provides an interesting insight. Macrophages and other immune cells, such as NK cells, play crucial roles in endometrial physiology and pathology [89]. Building on this, we propose that incorporating patient-specific immune cells into the HEO or EO culture system could enhance the model. This modification may create a more accurate immune microenvironment, offering a powerful tool for investigating the interactions between immune cells and endometrial cells. Such research could significantly advance our understanding of endometrial immune responses and related pathologies. On the other hand, we acknowledge an important limitation in our in vivo model: the use of NOD-SCID mice inherently excludes functional immune components. While this approach was necessary to prevent xenotransplant rejection, it precludes investigation of immune-mediated repair mechanisms, a critical aspect of endometrial regeneration. Collectively, these considerations underscore the imperative to prioritize immune components in future investigations.

In summary, we successfully constructed and characterized an in vitro vascularized endometrial complexes and demonstrated its significant effects in promoting endometrial repair using animal experiments. We believe that the positive effects of vascularization may involve various aspects such as improving mitochondrial function in endometrial cells and modulating ECM structure. Additionally, we revealed the promoting effect of WNTs derived from endometrial cells on endothelial cell angiogenesis. These findings not only provide a novel in vitro model for exploring the physiological and pathological mechanisms of the endometrium but also provide potential new strategies for clinical treatment of endometrial injuries. Given the inadequate stability in the culture system from continuous consumption of the hydrogel, future research should focus on achieving continuous culture mimicking complete menstrual cycles or multiple menstrual cycles and establishing a stable culture environment to accommodate the requirements of embryo implantation. Furthermore, using publicly available single-cell RNA-sequencing datasets of primary tissue to benchmark the cell-cell interactions described in this study will likely further help elucidate the molecular mechanisms underlying endothelial-endometrial interactions.

4. Methods

4.1. Study approval and sample acquisition

This study was approved by the Ethics Committee of Run Run Shaw Hospital, Zhejiang University. Prior to sample collection, informed written consent was obtained from all patients. Uterine endometrial samples were obtained from reproductive-age women in the proliferative phase undergoing routine hysteroscopy and seeking assisted reproductive technologies because of male or embryonic factors causing infertility. The endometrial functional layer was collected by gently scraping the endometrium with a curette under hysteroscopy-guided.

4.2. Patient inclusion and exclusion criteria

The patient inclusion criteria were as follows: 1. regular menstrual cycle and on days 2–5 of the menstrual cycle (proliferative phase); 2. age ≤40 years; 3. 18 kg/m² < BMI <25 kg/m²; 4. no history of uterine fibroids or adenomyosis; 5. no history of tubal hydrosalpinx, tubal tuberculosis, or pelvic inflammatory disease; and 6. no history of hepatitis B, AIDS, or syphilis.

The exclusion criteria were the following: 1. irregular menstruation or menstrual cycles in the non-proliferative phase; 2. hormonal medication, anti-endometriosis medication, and placement of intrauterine devices in the 3 months prior to the hysteroscopy; 3. history of pregnancy and breastfeeding in the 6 months prior to the hysteroscopy; 4. history of pelvic inflammatory disease or combination of tubal hydrosalpinx; and 5. history of malignant and juxtapapositional tumors of the ovary.

4.3. Primary culture of EEOs, ESCs and HUVECs

Human ESCs and epithelial cells were isolated from tissue biopsies following a previously described method [6,90] with slight modifications. Endometrial biopsies were washed three times with wash buffer, which consisted of RPMI 1640 (MA0215, Meilunbio, Dalian, China) with penicillin-streptomycin (P7630, Solarbio, Beijing, China), and minced into small fragments of approximately 1 mm³ using scissors. The fragments were subjected to a 20-min digestion in a solution of Collagenase IV (C9263, Sigma-Aldrich, MO, USA), Dispase II (D4693, Sigma), and DNase I (MB3069, Meilunbio). After centrifugation, the supernatant was removed, and the resulting pellet was washed three times. The cell pellet was then suspended in wash buffer and filtered through a 100-μm filter (352360, BD Biosciences, NJ, USA). The suspension was further filtered through a 40-μm filter (352340, BD Biosciences). The endometrial epithelial cell clusters on the top of the filter were collected for culture [6]. The cells in the filtered supernatant were spun down and cultured in DMEM/F12 medium (MA0214, Meilunbio) supplemented with 10 % fetal bovine serum (FBS, 10437028, Gibco, MA, USA). Following a 24-h incubation, the adherent cells were detached using trypsin (25200, Genom Bio, Hangzhou, China) and reseeded in a new culture dish to obtain ESCs. HUVECs were isolated from human umbilical cord veins [91] and cultured in Endothelial Cell Medium (ECM, 1001, Sciencell, CA, USA). EEOs were employed at passage 6 or higher to ensure phenotypic stability. ESCs and HUVECs at passage 3–6 were used for experiments.

4.4. Culture of endometrial triple-cell-type complexes

EEOs and ESCs were separately collected and then combined either with or without HUVECs in a hydrogel matrix. For the HEO complex, 100 EEOs, 5 × 10⁴ ESCs, and 1.5 × 10⁵ HUVECs were incorporated into every 30 μL of hydrogel. For the EO complex, 100 EEOs and 2 × 10⁵ ESCs were used for every 30 μL of hydrogel. To ensure precise dispensing of 100 endometrial epithelial organoids (EEOs), organoids were harvested during the optimal 3–5 day post-passage window (typically day 4). The counting protocol proceeded as follows: Following Triple-digestion (10 min) and centrifugation (500×g), the EEO pellet was resuspended in Advanced DMEM using a standardized volume ratio based on initial Matrigel® quantity. Specifically, 200 μL of Advanced DMEM was added per 10 droplets of original Matrigel® (25 μL/droplet). The suspension underwent gentle mixing via 10 repeated aspirations through a 1 mL pipette tip to ensure homogeneous dispersion. A 10 μL aliquot of the homogenized suspension was loaded onto a hemocytometer, where exclusively intact 3D structures were enumerated. Organoid density (EEOs/μL) was calculated to determine the exact volume required for dispensing 100 EEOs.

We developed eCHIP, a dynamic 3D culture system, by encapsulating three distinct cell types within a microfluidic chip. Specifically, 30 μL of cell-laden hydrogel was loaded into each chamber of a μ-Slide III 3D Perfusion Uncoated plate (80376, IBIDI, WI, USA), followed by 20-min incubation at 37 °C for matrix polymerization. Physiological perfusion was achieved using an OB1 MK3+ microfluidic flow controller (Elveflow, Paris, France) programmed with a sinusoidal waveform (maximum flow rate: 10 μL/min; minimum: 1 μL/min; 10-sec cycle period), selected to closely mimic physiological hemodynamic variations while enabling real-time system verification through waveform monitoring.

For optimizing vascular network formation and endometrial complex culture in our culture system, fibrinogen (F4883, Sigma) was introduced together with Matrigel® (356231, Corning, MA, USA) to create hydrogel matrix [16,48]. For experiments exploring the proportions of the constituent parts of hydrogels, different ratios of fibrinogen and Matrigel® (1:9, 5:5, and 9:1 v/v ratio, referred to as F1M9, F5M5, and F9M1, respectively) were used. For experiments exploring the stiffness of hydrogels, fibrinogen at different final concentrations (3 mg/mL, 5 mg/mL, and 10 mg/mL) was mixed with Matrigel® in a 5:5 ratio. For experiments optimizing the ideal medium for co-culturing multiple cells, culture medium was prepared with different ratios of ExM and ECM (9:1, 1:1 or 1:9 v/v ratio) and tested. The thrombin concentration was 1.5 U/mL, following a previous study [28]. For static culture, hydrogel with cells was placed in a 48-well plate (3548, Corning) containing ExM or ExM/ECM medium.

To assess organoid adhesion quantification in the different culture conditions of HEO complexes, organoids were pre-stained with a green fluorescent dye prior to experimentation. Adhesion was evaluated using a combination of brightfield microscopy and fluorescence imaging. Non-adherent organoids were identified under brightfield microscopy by their characteristic round morphology with distinct, sharply contrasting edges. Adherent organoids exhibited obscured edges within the brightfield image, but their adherent regions were confirmed by the presence of associated fluorescent signal spreading along the culture surface. The proportion of adherent organoids was calculated by analyzing eight randomly selected microscopic fields per sample group. Image J was used to quantify the vessel-positive area in different hydrogels and media conditions.

4.5. scRNA-seq analysis

This study included endometrial biopsy specimens from three women seeking in vitro fertilization assistance because of male infertility and who were undergoing routine hysteroscopy. EEOs and ESCs derived from the same patient were used to construct the dynamically cultured HEO and EO complexes, respectively. The complexes were dynamically cultured under a microfluidic flow controller by sequential culture conditions with estrogen and progesterone supplementation. After 6 days of culture, HEO and EO complexes were enzymatically digested with TrypLE™ Express (12604013, Gibco) for 5–8 min at 37 °C. Following digestion, the cells were passed through a 40 μm cell strainer (352340, BD Falcon, NJ, USA) to obtain dissociated single cells. The cells were then stained with acridine orange (C0233, Beyotime, Shanghai, China) and propidium iodide (ST511, Beyotime) for viability assessment using the Countstar Fluorescence Cell Analyzer (RuiYu Biotech, Shanghai, China). scRNA-Seq libraries were generated using the 10X Genomics Chromium Controller Instrument and Chromium Single Cell 3′ V3.1 Reagent Kits (10X Genomics, Pleasanton, CA, USA). Finally, all libraries were sequenced on the HiSeq Xten (Illumina, San Diego, CA, USA) using a 150 bp paired-end run. The hormone-stimulated HEO and EO complexes were denoted as HEOe, HEOp, EOe, and EOp (n = 3, biological replicates).

The R package Seurat (v. 4.3.0) was used to perform dimensionality reduction, clustering, and visualization for the scRNA-Seq data [92]. Cells were subjected to rigorous quality control filtering, excluding those with fewer than 200 detected genes (indicating low RNA content/capture) or greater than 5 % of reads mapping to mitochondrial genes (suggesting potential cell stress/damage). Clustering was performed on log-normalized expression data using principal component analysis (PCA) via RunPCA. The top 10 statistically significant principal components, identified using the JackStraw method, were selected for graph-based clustering (FindNeighbors and FindClusters, resolution = 0.5). Cluster identities were determined by identifying significantly DEGs for each cluster compared to all others (Wilcoxon rank-sum test; adjusted p-value <0.05, avg_log2FC > 0.25) and assigned based on the expression of canonical endometrial cell type markers (e.g., EPCAM/CDH1 for epithelial, COL1A1/DCN for stromal, PECAM1/VWF for endothelial, CD68 for macrophages). Cell clusters were visualized in two dimensions using UMAP (RunUMAP). For each annotated cluster, the FindMarkers function was used to identify DEGs between treatment conditions. Functional enrichment analysis of these DEGs was conducted using Metascape (https://metascape.org/gp/index.html), and cell-cell communication networks were inferred using the CellChat package (v.1.6.1) [93].

4.6. Endometrial injury mouse model experiments

Animal breeding followed the guidelines outlined in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Female NOD-SCID mice (QizhenLab, Hangzhou, PR China), 6–8 weeks old, were used in experiments. The mice were anesthetized and a laparotomy was performed to fully expose the left uterus. The uterine horns near the oviductal end and cervical end were ligated. Next, 95 % ethanol was injected into the left uterine cavity from the cervical end, followed by a 15-sec pause, and then the uterine cavity was flushed with PBS (MA0015, Meilunbio) injected from the oviductal end. The dynamically cultured EO and HEO complexes, cultured in eCHIP with estradiol for 2 days, were collected. Using a 1 mL syringe, 75 μl of the complexes were injected into the damaged uterine cavity from the oviductal end. An equal volume of Matrigel or PBS was injected into other endometrial injury model mice to establish the Matrigel group and the NR group, respectively. Three minutes after injection, the ligatures at both ends were removed, and the uterus was repositioned in the abdomen and sutured. At 3 and 7 days post-surgery, the mice were euthanized, and the uteri were collected for pathological staining and immunohistochemical staining. At 14 days post-surgery, the mice were cohabitated with male mice at reproductive age to observe the fertility of the model mice.

4.7. Subcutaneous implantation mouse model experiments

To investigate the vascular formation ability of the dynamically cultured HEO complex in vivo, a subcutaneous implantation model was established in female, 8-week-old NOD/SCID mice. The mice were anesthetized with 4 % chloral hydrate by intraperitoneal injection and their backs were shaved with clippers. After the skin was sterilized with povidone-iodine, small incisions were made on both sides of the back. Next, 30 μl of HEO complexes were implanted under the subcutaneous pockets bilaterally created by closed blunt forceps. At 14 days after transplantation, two types of lectins were administered via the tail vein. The FITC-conjugated UEA-I lectin (L32476, ThermoFisher, CA, USA) exhibits specific binding affinity to human endothelial cells, while the DyLight 594-labeled GSL-I isolectin B4 (DL-1207, VectorLabs, CA, USA) exhibits specific binding affinity to mouse endothelial cells. To visualize the vasculature with functional blood perfusion, tetramethylrhodamine-conjugated dextran (D1818, ThermoFisher) was administered via the tail vein. After 5 min of injection, the transplants were retrieved under a stereomicroscope for confocal microscopy fluorescence imaging and IHC.

4.8. Immunohistochemistry and immunofluorescence staining

The EO and HEO complexes were fixed in a 4 % (w/v) solution of paraformaldehyde. For immunohistochemistry, the complexes were embedded in paraffin and sliced according to the standard protocol. The paraffin sections then underwent immunohistochemical staining following standard procedures. For immunofluorescence, the fixed complexes were permeabilized with a 0.1 % solution of Triton X in PBS for 1 day, followed by incubation with primary antibodies for 2 days and secondary antibodies for 1 day at 4 °C. The nuclei were stained with DAPI overnight. After each incubation step, the samples were washed with PBS three times for 30 min per wash. Immunofluorescence images were obtained using a confocal laser-scanning microscope (ZEISS 800, ZEISS, Oberkochen, Germany). Movie files of 3D rendered images were created using Zeiss ZEN Black 2010 software (Carl Zeiss Canada). Antibodies for immunohistochemistry and immunofluorescence are listed in Supplementary Table 30. Periodic acid Schiff staining was performed on paraffin sections using the Periodic Acid-Schiff Staining Kit (C0142S, Beyotime) and standard protocols.

4.9. Morphology of vessel formation

The CytoSMART (Axion BioSystems, Netherlands) incubator system was used to capture time-lapse imaging of the dynamically cultured EO and HEO complexes self-organizing into a network-like structure (Supplementary Video 1), with 15-min intervals captured between each image.

To ascertain the impact of epithelial and stromal cells on the tube formation of HUVECs in multicellular complexes, CellTracker™ Green CMFDA (C7025, ThermoFisher) was used to pre-label the HUVECs. Following a 72-h dynamic culture in different treatments, images were acquired under a confocal microscope. Five fields per group were randomly selected to quantify the vessel-positive area using Image J software.

Confocal image stacks were acquired with a ZEISS LSM 800 Inverted Microscope (ZEISS, Germany) for the visualization of capillary network-like structures and implanted vessels. Z-stack image reconstructions were generated in Imaris 9.0.1 (Bitplane, Swiss) using the Z-projection function with the maximum pixel intensity setting. Additionally, 3D image reconstructions and rotating videos were generated with the Imaris software.

4.10. Determination of mitochondrial membrane potential

Mitochondrial membrane potential was assessed using the JC-1 kit following the manufacturer's protocol (C2003S, Beyotime) together with confocal microscopy and flow cytometry.

Briefly, the dynamically cultured HEO and EO complexes were cultured in eCHIP for 3 days with treatments including dorsomorphin (10 μM, HY-13418, MCE), HA-CD44 (10 μM, HY-149897, MCE), BMP-6 (50 ng/mL, HY-P700029AF, MCE), Galectin-9 (10 ng/mL, C808, Novoprotein, Jiangsu, China), or DMSO (PWL064, Meilunbio) as a negative control. The cells were washed with cold PBS and then incubated with JC-1 dye (C2006, Beyotime) for 30 min in the dark at room temperature. The cells were then washed and examined using a ZEISS LSM800 confocal microscope and ZEN microscope software. Loss of mitochondrial membrane potential was indicated by increased green fluorescence of JC-1 monomers and decreased red fluorescence of JC-1 aggregates.

For flow cytometric analysis, cells were first harvested using TrypLE, washed with cold PBS and incubated with JC-1 dye for 30 min in the dark at room temperature. Epithelial cells were gated with an antibody against EPCAM using an Alexa Fluor™ 647 secondary antibody. JC-1 aggregates were measured in the FL-2 channel and green fluorescence (both JC-1 monomer and/or GFP) was assessed in the FL-1 channel using a flow cytometer (DxFLEX, Beckman Coulter, CA, USA). The red-to-green fluorescence intensity ratio was calculated as an indicator of mitochondrial membrane potential using FlowJo v10.1 (BD Biosciences).

4.11. TEER measurement

TEER was measured on intact HEO/EO complexes within Matrigel/fibrin-filled insert. Resistance reflects ion flow impedance through the cell-laden hydrogel matrix, proportional to cell density and network integrity. For the HEO complex, 400 EEOs, 2 × 10⁵ ESCs, and 6 × 10⁵ HUVECs were incorporated into each 120 μL of hydrogel. For the EO complex, 400 EEOs and 8 × 10⁵ ESCs were used per 120 μl of hydrogel. A total of 120 μL hydrogel containing cells was seeded on the apical side of insert membranes. The culture medium was a mixture of ExM and ECM with different treatments in a 1:1 ratio and changed daily. After cell seeding on 24-well Transwell inserts (3470, Corning), TEER was measured daily for one week, as previously described [94]. TEER values were obtained by subtracting blank values from the filters and medium. The TEER was measured every 24 h using an EVOM2 voltohmmeter with “chopstick” probes (World Precision Instruments, FL, USA). All TEER experiments were performed with three duplicate wells and at least three independent differentiations.

4.12. Transmission electron microscopy

Samples were fixed with 2.5 % glutaraldehyde followed by 1 % osmic acid and then dehydrated using a series of ethanol solutions with increasing concentrations (50 %, 70 %, 90 %, 100 %) and 100 % acetone solution. The samples were embedded in resin and then sectioned using an ultramicrotome (EM UC7, Leica, Hessian, Germany). The sections were stained with uranyl acetate and lead citrate. The ultra-thin sections were analyzed using a FEI Tecnai G2 spirit TEM at 120 kV. Fifteen regions were randomly selected in the HEO or EO complex to count the number of mitochondria in the EEOs at low-magnification level. Values are expressed in number of mitochondria per 121 μm² of the epithelial organoid area.

4.13. Spheroid sprouting assay

The assay was performed as previously described [95] with the following modifications. Briefly, HUVECs were trypsinized, counted, and adjusted to a density of 2 × 10⁴ cells/mL in ECM containing 20 % methylcellulose. Drops of a 25 μL cell suspension were pipetted into a 6 cm Petri dish. The plates were then inverted and incubated in a humidified cell culture incubator at 37 °C for 24 h. Spheroids were collected, centrifuged at 200 g for 5 min, and resuspended in methylcellulose supplemented with 20 % FBS. The collagen mixture was prepared on ice using Collagen I (CB-40236, Corning), Medium 199 (M0650, Sigma) and NaOH (1 mol−1) in a ratio of 8:1:1. The collagen solution was mixed with the spheroid solution in a 1:1 ratio, and 0.5 mL of the mixture was transferred to a 48-well plate. The plate was incubated at 37 °C for 30 min to allow polymerization of the collagen matrix. The spheroids were stimulated with 100 μL of ECM with different treatments, including WNT5A (200 ng/mL, CSB-EP026138HU, CUSABIO, Wuhan, China), WNT7A (100 ng/mL, P06680, Solarbio), DKK1 (200 ng/mL, C12B, Novoprotein), BOX5 (HY-123071A, MCE), anti-WNT5A neutralizing antibody (2 μg/mL, MAB645, R&D), anti-WNT7A neutralizing antibody (10 μg/mL, sc-365665, Santa Cruz), or VEGFA (10 ng/mL, C744, Novoprotein) as a positive control, by adding dropwise onto the collagen matrix. After 24 h of incubation at 37 °C, the assay was terminated by adding 1 mL of 4 % PFA per well. Images of 10 spheroids per gel were captured using an inverted microscope (ECLIPSE Ts2R, Nikon, Tokyo, Japan). To evaluate the sprouting capacity of spheroids from different groups, each sprout from five spheroids per group was examined using Fiji software (National Institutes of Health). From these data, the average cumulative sprouting number per spheroid was calculated.

4.14. Tube formation assay

Tube formation assays were performed using μ-slide 15-well 3D ibiTreat plates (81506, IBIDI). Matrigel was thawed on ice overnight. Next, 10 μL Matrigel was added per well and allowed to polymerize for 30 min at 37 °C. HUVECs were trypsinized and counted, and 5000 HUVECs per well were seeded on the Matrigel in 50 μL of ECM with various treatments including WNT5A, WNT7A, DKK1, BOX5, anti-WNT5A neutralizing antibody, anti-WNT7A neutralizing antibody, or VEGFA as a positive control. After 6 h incubation at 37 °C, tube formation capacity was assessed by quantifying the number of branches using Fiji software (National Institutes of Health).

CRediT authorship contribution statement

Yongdong Dai: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization. Fanxuan Zhao: Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Qiuli Chen: Visualization, Software, Methodology, Formal analysis. Biya Zeng: Validation, Resources, Investigation, Formal analysis. Weijia Gu: Validation, Resources, Investigation, Formal analysis. Yi Zhang: Resources, Methodology, Investigation, Formal analysis. Fangying Sun: Validation, Resources, Investigation. Xinyu Wang: Validation, Resources, Methodology. Xiang Lin: Writing – review & editing, Funding acquisition, Data curation. Na Liu: Resources, Methodology. Yulu Wang: Validation, Resources, Investigation. Feng Zhou: Writing – review & editing, Methodology, Funding acquisition. Jianhua Yang: Writing – review & editing, Resources, Methodology. Shangjing Xin: Writing – review & editing, Validation, Methodology, Formal analysis. Ye Feng: Writing – review & editing, Writing – original draft, Visualization, Supervision, Software, Methodology, Investigation. Songying Zhang: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Funding

This work was supported by the National Natural Science Foundation of China [grant numbers 82371637 to Y.D.]; "Pioneer" and "Leading Goose" R&D Program of Zhejiang [grant numbers 2025C02119 to Y.D. and 2023C03033 to S.Z.]; Science and Technology Program Project of Zhejiang Provincial Administration of Traditional Chinese Medicine [grant numbers GZY-ZJ-KJ-23028 to S.Z.]; National Natural Science Foundation of China [grant numbers 82061160494 to S.Z., 82101722 to X.L.]; Natural Science Foundation of Zhejiang Province Grants [grant numbers LY22H040005 to Y.D., LTGY23H040010 to F.Z., Q22H048670 to X.L.]; and the Medical Science and Technology Project Foundation of Zhejiang Province [grant number 2020KY606 to Y.D.].

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Run Run Shaw Hospital, Zhejiang University (approval number: SRRSH202302348 and Research 20230211-286). Prior to sample collection, informed written consent was obtained from all patients.

Declaration of competing interest

None.

Appendix A. Supplementary data

The following are the Supplementary data to this article.