Evaluation of optimal trophoblastic models to mimic uterine vessel remodeling in human pregnancy

Abstract

During human pregnancy, a certain proportion of endothelial cells in uterine spiral arteries are replaced by placental extravillous trophoblasts (EVTs). This process is a pivotal step in spiral artery remodeling (SAR), ensuring the adequate supply of oxygen and nutrients to the fetus. Given the ethical constraints and the lack of suitable in vivo animal models, developing ideal in vitro models is crucial for investigating this cellular event. Therefore, it is imperative to evaluate key trophoblastic properties to optimize the in vitro model and replicate the in vivo context of SAR. In this study, we refined a three-dimensional co-culture system involving human trophoblast cells and tubular structure formed by human umbilical vein endothelial cells, allowing for dynamical monitoring of cell behaviors. Using this model, we conducted a comparative analysis of vascular remodeling capabilities among primary EVTs and various trophoblast cell lines. Meanwhile, we examined the expression profiles of multiple SAR-associated genes in trophoblast cells. The correlation between molecular characteristics and vascular remodeling performance was statistically analyzed using a comprehensive scoring system. Our findings highlight the critical roles of EVT-derived NCAM1, ITGB3, ITGAV, and JAG1 in vascular remodeling. Furthermore, JEG-3 and human trophoblast stem cell-derived EVT cells demonstrate significant advantages as optimal models for mimicking primary EVTs in vitro, thereby facilitating investigations into SAR. Consequently, we propose an evaluation framework to assess key attributes of trophoblast cell lines that enable them to accurately represent EVT behaviors in SAR. This study establishes a robust foundation for future exploration of the molecular mechanisms underlying human SAR.

Summary Sentence 

We demonstrate an evaluation framework to assess key attributes of trophoblast cells that enable them to accurately represent extravillous trophoblast behaviors in spiral artery remodeling, which is a robust foundation for future exploration of the molecular mechanisms underlying human spiral artery remodeling.

Graphical Abstract

Graphical Abstract.

Introduction

In utero development of semi-allogeneic fetus relies on precisely regulated interactions between fetus-derived placental trophoblast cells and multiple uterine cell types [1]. Following embryonic implantation, human trophoblasts differentiate into two primary categories: villous trophoblasts (VTs) and extravillous trophoblasts (EVTs) [2, 3]. Villous trophoblasts primarily consist of mononucleated cytotrophoblasts (CTBs) and multinucleated syncytiotrophoblasts (STBs). STBs, formed through the fusion of CTBs, serve as the principle maternal-fetal exchange surface and are responsible for hormone production, including human chorionic gonadotrophin subunit beta (CGB) and progesterone, etc. EVTs exhibit highly invasiveness, infiltrating the uterine decidua and interacting with endothelial cells within uterine spiral arteries. This process is accompanied by endothelial cell apoptosis, dedifferentiation of vessel smooth muscle cells, and degradation of extracellular matrix with deposition of fibrinoid materials [4, 5]. Collectively, these changes are referred to as spiral artery remodeling (SAR), which transforms narrow, high-resistance vessels into high-capacity, low-resistance vessels that ensure optimal uteroplacental perfusion and thereby guarantee adequate oxygen and nutrient supply to the developing fetus [5]. SAR is a critical placentation event essential for safeguarding maternal and fetal health, as impaired SAR is strongly associated with severe obstetric complications such as early pregnancy loss, preeclampsia, and fetal growth restriction [6].

Despite extensive studies on the regulation of EVT properties, our understanding of the process and mechanisms of human SAR remains limited. Ethic considerations restrict in vivo investigation of this process in humans. Moreover, the significant species-specific differences of SAR among viviparous mammals limit the availability of suitable animal models to study this event [7, 8]. Therefore, it is crucial to develop ideal in vitro models that can mimic the complex cell interactions during human SAR. Considerable efforts have been made to establish models that simulate the interactions of trophoblasts with vessel endothelium. A practical three-dimensional (3D) system was developed by Aldo et al., which involves the co-culture of human trophoblast cells with tubular structures formed by human umbilical vein endothelial cells (HUVECs) on Matrigel, where trophoblast cells migrate toward and incorporate into these tubular structures [9]. We further refined this model and conducted dynamic monitoring of trophoblast behaviors in response to various growth factors and cytokines [10]. Other studies have employed this model to show trophoblast cell-induced endothelial cell apoptosis through diverse signaling pathways [11, 12]. However, different lines of trophoblast cells have been utilized across various studies, and it remains unclear which specific properties of trophoblasts are essential for the physiological interactions between EVTs and endothelial cells. Therefore, it is crucial to evaluate the key trophoblastic properties to optimize the in vitro model, thereby more accurately replicating the in vivo context of SAR.

The primary culture of EVTs represents the most physiologically relevant cell model for investigating SAR. However, the limited availability and individual variability of fresh clinical samples to isolate EVTs pose significant challenges to their practical application. Various trophoblastic cell lines, including JEG-3 [13], BeWo [14], HTR-8/SVneo [15], as well as human trophoblast stem cells (hTSCs) [16], offer stable and accessible experimental models. Nevertheless, their ability to accurately represent EVT behaviors in SAR requires further systematic evaluation. Traditional histological studies and recent single-cell transcriptomic analysis on human maternal–fetal interface have suggested genes potentially associated with human EVT interaction with endothelium, such as human leukocyte antigen G (HLA-G), neural cell adhesion molecule 1 (NCAM1), integrin subunit beta 3 (ITGB3), angiopoietin 4 (ANGPT4), gamma-glutamyltransferase 1 (GGT1), jagged canonical Notch ligand 1 (JAG1), integrin subunit alpha V (ITGAV), matrix metallopeptidase 12 (MMP12), C-X-C motif chemokine ligand 12 (CXCL12), EPH receptor A1 (EPHA1), ADAM metallopeptidase with thrombospondin type 1 motif 4 (ADAMTS4), C-C motif chemokine receptor 3 (CCR3), integrin subunit beta 1 (ITGB1), and fms related receptor tyrosine kinase 4 (FLT4) [17–21]. The extent to which these genes are implicated in the vessel remodeling by EVTs remains incompletely understood.

In this study, we conducted a comparative analysis of the expression profiles of the aforementioned SAR-associated molecules in primary EVTs and multiple trophoblast cell lines. The capability of the various lines of trophoblasts to participate in vascular remodeling was evaluated using the refined 3D co-culture model incorporating human trophoblasts and HUVECs. We analyzed the correlation between molecular characteristics and vascular remodeling performance using a comprehensive scoring system. Consequently, we propose an evaluation framework to assess key attributes of trophoblast cell lines that enable them to accurately represent EVT behaviors in SAR. The findings provide critical insights for developing optimal in vitro models to explore the molecular mechanisms underlying human SAR.

Materials and methods

Human clinical sample collection

Human placental villi and decidual tissues at gestational weeks 6–9 were collected from healthy pregnant women upon elective surgical abortion, and the umbilical cord tissues were collected from healthy term pregnant women at Peking University Third Hospital (Beijing, China). The tissues were immersed in iced phosphate-buffer solution (PBS) and subjected to cell isolation within 1 h of surgery. Women with fetal chromosomal or congenital abnormalities, renal and cardiovascular diseases, history of gestational complications such as recurrent spontaneous miscarriage, gestational diabetes, gestational hypertensive disorders, and intrauterine fetal death, or pregnancies conceived with assisted reproductive technologies were excluded from this study. The study for human sample collection was approved by the Ethics Committees of the Institute of Zoology, Chinese Academy of Sciences and the Peking University Third Hospital.

Isolation of human umbilical vein endothelial cells

Human umbilical vein endothelial cells were isolated from the umbilical vein using a previously developed protocol [10]. Briefly, umbilical vein was cannulated with a blunt-tipped, handleless 18-gauge needle, ensuring minimal damage to the vessel. The umbilical cord was securely fastened over the needle using a hemostat to maintain stability during the process. To eliminate any residual blood, the umbilical vein was thoroughly perfused with 100 mL of sterile PBS. Following perfusion, the opposite end of the umbilical cord was clamped with a hemostat to create a closed system. A solution of 0.25% trypsin (Sigma, T4799) and 2% EDTA was then injected into the vein to enzymatically dissociate the endothelial cells from the vessel wall. Following digestion, the cell-containing solution was gently flushed out using PBS supplemented with 2% fetal bovine serum (FBS; Gibco, 10099141), and the cells were collected by centrifugation and resuspended in fresh M199 medium enriched with Low Serum Growth Supplement (LSGS, Gibco, S00310) and 1% penicillin–streptomycin (PS, Gibco, 15140122). The cells were cultured under standard conditions at 37°C, 20% O₂, and 5% CO₂.

Isolation of primary EVTs and primary decidual NK cells

The isolation of Pr-EVTs and primary decidual natural killer cells (Pr-dNK) cells was conducted following a previously established protocol [22]. Briefly, fresh human decidual tissues were mechanically dissociated using the GentelMACS Dissociator (Miltenyi Biotec, 130-093-235) to ensure tissue homogenization. The dissociated tissues were digested with a solution containing 1.0 mg/mL type IV collagenase (Gibco, 9001121) and 10 U/mL type I DNase (Sigma Aldrich, DN25) to achieve single-cell suspension. The resulting cell suspensions were collected and incubated overnight in RPMI-1640 medium supplemented with 10% FBS and 1% PS.

For the isolation of Pr-EVTs, the adherent cells were harvested and labeled with an anti-HLA-G antibody (Sigma Aldrich, SAB4700317), a specific marker for EVTs. The labeled cells were subjected to fluorescence activated cell sorting (FACS) using the MoFlo XDP system (Beckman Coulter, Inc., CA, USA) to isolate HLA-G⁺ EVT cells. The purified EVT cells were cultured in RPMI-1640 medium (Gibco, 31800-022) supplemented with 10% FBS and 1% PS.

For the isolation of Pr-dNK cells, the non-adherent cells were collected and incubated with magnetic beads conjugated to an anti-CD56 antibody (Miltenyi Biotec, 130050401), a marker specific for dNK cells. The labeled cells were then separated and collected using an LS column (Miltenyi Biotec, 130042401) according to the manufacturer’s instructions. The purified dNK cells were cultured in RPMI-1640 medium supplemented with 10% FBS and 1% PS.

Isolation of primary CTB cells

Primary CTBs (Pr-CTBs) were isolated from fresh placental villi according to previous report [23]. In brief, the placental villi were delicately trimmed into diminutive fragments, and subjected to enzymatic digestion in PBS solution containing 0.25% trypsin (Sigma, T4799) and 2% EDTA. The resulting cell suspension was sieved through a 75-μm mesh and harvested via centrifugation. The cells were resuspended and carefully layered onto a bovine serum albumin (BSA; Sigma, A9418) gradient at concentrations of 1.25% and 2.5%. Primary CTBs were collected at the bottom of the gradient and cultured in DMEM/F12 medium (Gibco, 11320033) containing 1 ng/mL EGF (Wako, 053-07871) and 1 μg/mL Insulin (Sigma, 11061-68-0).

Human trophoblast stem cell culture and differentiation induction

Human trophoblast stem cells were isolated from human early pregnant placental villi according to previous report [23]. Human trophoblast stem cells were seeded into 6-well plates precoated with 5 mg/mL type IV collagen (Corning, 354233) and maintained in DMEM/F12 medium supplemented with a specialized cocktail of growth factors and inhibitors, including 0.1 mM β-mercaptoethanol (Gibco, 21985023), 0.2% FBS, 0.5% PS, 0.3% BSA, 1% insulin-transferrin-selenium-ethanolamine (ITS-X; Gibco, 51500056), 0.5 μM A83-01 (MedChemExpress (MCE); HY-10432), 2 μM CHIR99021 (MCE; HY-10182), 1 μM SB431542 (MCE; HY-10431), 5 μM Y27632 (MCE; HY-10071), and 0.8 mM valproic acid (Wako, 227-01071).

To induce hTSCs differentiation into STBs, the cells were cultured in differentiation medium consisted of DMEM/F12 supplemented with 0.1 mM β-mercaptoethanol, 0.5% PS, 0.3% BSA, 1% ITS-X, 4% Knockout Serum Replacement (KSR), 2.5 μM Y27632, and 0.2 μM Forskolin (FSK, Selleck, S2449) for 12 h. The formation of multinucleated STBs was characterized with the expression of syncytial markers, CGB.

For the differentiation of hTSCs into EVTs, the cells were cultured in DMEM/F12 medium supplemented with 0.1 mM β-mercaptoethanol, 0.5% PS, 0.3% BSA, 1% ITS-X, 100 ng/mL Neuregulin-1 (NRG1; Cell Signaling, 5218SC), 7.5 μM A83–01, 2.5 μM Y27632, 4% KSR, and 2% Matrigel for 24 h. The formation of EVT cells were determined with the expression of HLA-G, an EVT biomarker.

Culture of human trophoblast cell lines

JEG-3 cells (ATCC, HTB-36) and HTR-8/SVneo cells (ATCC, CRL-3271) were separately cultured in DMEM (Gibco, 11965092) and RPMI-1640 medium supplemented with 10% FBS (Gibco, 10270-106) and 1% PS. BeWo cells (ATCC, CCL-98) were maintained in F-12K (Gibco, 21127022):DMEM medium supplemented with 1% sodium pyruvate, 1% glutamine, 10% FBS, and 1% PS. To induce the differentiation of BeWo cells into STBs, the cells were treated with 2 μM FSK in complete medium for 24 h. The formation of multinucleated STBs was characterized with the expression of syncytial markers, human chorionic gonadotropin (hCG).

Plasmid construction and transfection

We designed a short hairpin RNA (shRNA) targeting NCAM1 (5′-AATTCAGAGGATGGAA-ACTCTATTAATTCAAGAGATTAATAGAGTTTCCATCCTCTTTTTTTG-3′), and cloned it into the lentiviral vector LV011-pHBLV-CMV-MCS-3flag-EF1-ZsGreen-T2A-Puro vector. The lentiviral transfer plasmid and packaging plasmid were transfected into HEK293T cells for production of infectious lentivirus. Supernatants were collected at 48- and 72-h post-transfection, filtered through a 0.22-μm filter membrane, and prepared for further use. JEG-3 cells were cultured in 6-well plates and then infected with lentivirus for 6–12 h. The infected JEG-3 cells were subsequently treated with puromycin at a concentration of 1.5 μg/mL for 48 h to select stably transfected cells. The survived cells were utilized for subsequent experiments. The silencing efficiency of shRNA on NCAM1 was assessed by analyzing the mRNA expression levels via quantitative real-time PCR.

Flow cytometry analysis

To assess the expression of HLA-G and NCAM1 in various cell types, the cells were incubated with antibody against HLA-G (Sigma, SAB4700317) or NCAM1 (BioLegend, 362546) at 4°C for 30 min, followed by flow cytometry analysis using the CytoFLEX system (Beckman Coulter). Data acquisition and analysis were conducted using CytExpert and FlowJo Version 10.9 software. The relative mean fluorescence intensities of HLA-G and NCAM1 were quantified and normalized against IgG isotype controls.

RNA preparation and quantitative real-time PCR

Total RNA was extracted from cultured cells using TRIzol reagent (Invitrogen, 15596018) following the manufacturer's instructions. Reverse transcription was conducted with Oligo Primer (TIANGEN, CD106), and the resultant cDNA was used as the template for quantitative real-time PCR employing the SYBR Premix Ex Taq reagent kit (Takara, DRR820A). Primer sequences are provided in Supplementary Table S1. Following the PCR reaction, the relative expression level of the target gene was normalized to H2A using the 2^−ΔΔ CT method.

Three-dimensional (3D) co-culture of trophoblast cells and HUVECs and live-cell monitoring

Co-culture of trophoblast cells and HUVECs were modified from the previously established methods [10, 24]. In brief, HUVECs cells at 40 000 were labeled with CellTracker (Thermo Fisher, C34552) and seeded into an 8-well cell culture plate (ibidi, 808024) pre-coated with Matrigel (Corning, 354230) at thickness of 100 µm. The HUVECs were incubated at 37°C for 4 h to facilitate the formation of a 3D tubular structure in Matrigel matrix. Trophoblast cells, including Pr-EVT, Pr-CTB, Pr-dNK, JEG-3, HTR-8/SVneo, BeWo, BeWo-STB, hTSC, hTSC-STB, and hTSC-EVT cells, at 20 000 were labeled with F-actin live cell dye (Spirochrome, SC001) and added to culture well with HUVECs tubular structure. Cell behaviors in co-culture system were monitored in real time using the Andor Dragfly 200 instrument, under the following conditions: a total imaging duration of 24 h, with time intervals set at 20 min, and Z-axis layer scan thickness of 50 μm. The acquired data were analyzed using Imaris 10.0 software to show the 3D imagine.

Cell apoptosis assay

Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) fluorescence staining procedure was performed in accordance with the manufacture’s instruction (Roche, 12156792910). Briefly, cells were fixed with 4% paraformaldehyde solution for 15 min, permeabilized using 0.1% Triton X-100, and incubated with TUNEL reaction mixture in dark for 60 min. The cells were washed with PBS, and subjected to fluorescent images with a Leica Stellaris microscope, and the images were processed using LAS X software for further analysis.

Analysis of integration rate and apoptosis index

The rate of trophoblast cell integration into tubular structures was assessed using the Andor Dragonfly 200 instrument in combination with ImageJ software. Briefly, fluorescence images of the cells were captured using the Andor Dragonfly 200, and ImageJ software was employed to analyze and quantify the pixel values. Specifically, the pixel values corresponding to the tubular structures formed by CellTracker labeled HUVECs and F-actin labeled trophoblast cells adhering to these structures were measured. The integration rate was calculated as the ratio of the pixel value of the fluorescent trophoblast cells adhering to the tubular structures to the total pixel values of the tubular structures, expressed as a percentage.

The apoptotic index of HUVECs was determined using LAS X and ImageJ software. Fluorescence images of the cells were acquired with LAS X software, and ImageJ software was used to quantify the pixel values based on the TUNEL staining pattern and the tubular structures formed by HUVECs. The apoptotic index was calculated as the ratio of the pixel value of apoptotic cells to the total pixel value of all HUVECs, expressed as a percentage.

Correlation analysis and weighting analysis

GraphPad Prism (version 8.0) was utilized to assess the correlations between gene expression levels, and the integration rate of trophoblast cells or the apoptotic index of HUVECs. Specifically, the integration rate and apoptotic index were treated as dependent variables, while gene expression levels were considered as independent variables. Scatter plots were generated to visually inspect the data distribution and determine whether a continuous trend existed between the variables. For variables showing a continuous trend, the Shapiro–Wilk test was applied to assess normality. If both continuity and normal distribution were confirmed, Pearson correlation analysis was performed. A Pearson correlation coefficient |r| > 0.5 was considered indicative of a strong correlation, with a P < 0.05 denoting statistical significance.

The “caret” package in R (version 4.4.1) [25] was used to evaluate the influence weights of various genes on the integration rate and apoptotic index. In this process, recursive feature elimination (RFE) was employed for feature selection, systematically removing less relevant features to identify the optimal feature combination and assess the importance of variables. To ensure robustness, 1000 bootstrap sampling validations were performed during the RFE process to evaluate model performance. Following this, scores ranging from 0 to 100 were assigned to different types of trophoblast cells that were co-cultured with HUVECs based on the weights of independent variables. Finally, the results were visualized using the “ggradar” package.

Statistical analysis

All data were subjected to analysis utilizing GraphPad Prism (version 8.0) statistical software. The results are expressed as the mean ± SD, derived from a minimum of three independent experiments. Inter-group statistical comparisons were executed using either the independent t-test or the Tukey–Kramer test, contingent upon the number of groups under comparison. Multiple comparison analyses were conducted based on data from a minimum of three independent replicate experiments to ensure the robustness and reliability of the results. The P < 0.05 was considered statistically significant.

Results

Comparative analysis of the expression profiles of SAR-associated molecules in primary EVTs and multiple trophoblast cell lines

In this study, the following human trophoblastic cell models were utilized to compare the expression profiles of the SAR-associated molecules and their capability of vascular remodeling. Primary EVTs were isolated from early gestational human decidual tissues using specific HLA-G antibody sorting, and primary CTBs were obtained from early pregnancy placental villi following previously reported methods [23, 26] (Supplementary Figure S1A and B). Additionally, commonly used trophoblastic cell lines, including JEG-3, BeWo, HTR-8/SVneo, and hTSCs, were included. Among which, hTSCs can be induced to differentiate into either EVT (hTSC-EVT) or syncytiotrophoblast (hTSC-STB) (Supplementary Figure S1C), and BeWo cells can be induced to differentiate into syncytiotrophoblast (BeWo-STB) (Supplementary Figure S1D). As a non-trophoblast control, Pr-dNKs from early pregnancy decidual tissues were also involved (Supplementary Figure S1A).

It is widely acknowledged that EVTs exhibit characteristic expression of HLA-G, with those within uterine vessels specifically expressing NCAM1 [17, 18]. We first conducted flow cytometry analysis to evaluate the levels of HLA-G and NCAM1 across various cell types. As anticipated, Pr-EVTs demonstrated significant and homogenous high expression of HLA-G, along with positive NCAM1 expression in a subset population. Primary CTBs were negative for both molecules, while Pr-dNKs exhibit exclusive positivity for NCAM1 but remained negative for HLA-G (Figure 1; Supplementary Figure S1E). Among the cell lines, JEG-3 cells showed consistent and robust expression of both HLA-G and NCAM1. In contrast, hTSC and hTSC-STB cells were negative for these two markers, while hTSC-EVT cells displayed extensive expression of HLA-G and a notable yet modest increase in NCAM1 level. BeWo and BeWo-STB cells were uniformly negative for HLA-G but expressed NCAM1 extensively. HTR-8/SVneo cells did not express either HLA-G or NCAM1 (Figure 1; Supplementary Figure S1E).

Figure 1.

Expression profiles of HLA-G and NCAM1 across diverse cell types. (A) Flow cytometry analysis of HLA-G expression in various trophoblast cells. (B) Statistical analysis of the relative MFI of HLA-G across diverse cell types. (C) Flow cytometry analysis of NCAM1 expression in various trophoblast cells. (D) Statistical analysis of the relative MFI of NCAM1 across diverse cell types. Data are presented as mean ± SD, n = 3 for all parameters. MFI, mean fluorescence intensity.

Recent single-cell and spatial transcriptomic studies have indicated several genes, in addition to HLA-G and NCAM1, that are specifically expressed by EVTs in association with vascular remodeling [18, 19, 27]. We subsequently evaluated the expression levels of these genes using RT-qPCR across various cells, including HLA-G, NCAM1, ITGB3, ANGPT4, GGT1, JAG1, ITGAV, notum, palmitoleoyl-protein carboxylesterase (NOTUM), MMP12, CXCL12, EPHA1, ADAMTS4, CCR3, ITGB1, and FLT4, as well as the stemness-specific gene TEA domain transcription factor 4 (TEAD4), the syncytialization marker gene CGB. As illustrated in Figure 2, Pr-EVTs exhibited relatively high expression of all vascular remodeling-associated genes, while the expression profiles across different cell types were highly heterogeneous. The mRNA expression patterns of HLA-G and NCAM1 in various cell types were largely consistent with the flow assay results (Figure 2A).

Figure 2.

Expression profiles of SAR-related specific genes across diverse cell types. (A) Relative mRNA expressions of HLA-G, NCAM1, JAG1, and NOTUM across diverse cell types. (B) Relative mRNA expressions of ITGAV and ITGB3 across diverse cell types. (C) Relative mRNA expressions of GGT1, CCR3, ANGPT4, ITGB1, EPHA1, and FLT4 across diverse cell types. (D) Relative mRNA expressions of CXCL12, ADAMTS4, and MMP12 across diverse cell types. (E) Relative mRNA expressions of TEAD4 and CGB across diverse cell types. Data are presented as mean ± SD, n = 3 for all parameters.

The stemness-specific gene TEAD4 expressed at the highest level in hTSC cells, while the syncytialization marker gene CGB was exclusively expressed in BeWo-STB and hTSC-STB cells (Figure 2E).

JAG1 has been implicated in EVT plug formation. Compared to EVTs, primary CTB and cell lines such as JEG-3, hTSC, and its two differentiation derivatives expressed JAG1 at less than half the level observed in EVTs, whereas HTR-8/SVneo, BeWo, and dNK cells showed undetectable JAG1 expression (Figure 2A).

ITGB3 and ITGAV have been implicated in association with the transformation of EVTs into endothelia-like characteristics. HTSC-EVT cells demonstrated a high expression of ITGB3 and ITGAV, comparable to that observed in primary EVTs, followed by JEG-3 cells. BeWo and BeWo-STB cells exhibited moderate ITGAV expression (Figure 2B).

Genes such as GGT1, CCR3, ANGPT4, ITGB1, EPHA1, FLT4, and NOTUM have been proposed as potential mediators in the interaction between EVTs and endothelial cells. The WNT inhibitory gene NOTUM was highly expressed in Pr-EVTs, with lower level in JEG-3 and hTSC-EVTs, while negligible expression in other cell types (Figure 2B). HTSC-EVT cells exhibited expression levels of GGT1 and CCR3 comparable to those in primary EVTs, while demonstrating significantly higher expression levels of ITGB1 and EPHA1 compared to Pr-EVTs. JEG-3 and BeWo cells displayed similar expression levels of GGT1, ITGB1, and EPHA1 relative to Pr-EVTs. In contrast, HTR-8/SVneo cells exhibited scarce expression of all these genes. Notably, ANGPT4 was predominantly expressed in Pr-EVTs and JEG-3, whereas FLT4 was detected in Pr-EVTs, BeWo, and BeWo-STB cells, but not in other cell types (Figure 2C).

Genes associated with immune regulation and invasion, including CXCL12, ADAMTS4, and MMP12, exhibited distinct expression in Pr-EVTs and Pr-dNK cells, while hTSC-EVTs demonstrated moderate MMP12 level (Figure 2D).

Collectively, the expression profiles of these marker gene suggest that JEG-3 and hTSC-EVT cells share relatively similar molecular characteristics with Pr-EVTs.

Real-time monitoring of trophoblast integration into vascular structures in a 3D co-culture system

We previously refined a 3D co-culture system comprising human trophoblasts and HUVECs, allowing for dynamic monitoring of trophoblast migration toward and integration into pre-formed HUVEC tubular structures [10]. Here, we further modified the method by labeling HUVECs with CellTracker (red fluorescence) and trophoblasts with F-actin live cell dye (green fluorescence). After introducing trophoblasts into the pre-formed HUVEC-tube structures, cellular behavior was continuously recorded using a live cell workstation over a 24-h period. Following 4 h of co-culture, the cells were collected for TUNEL analysis and confocal microscopy observation (Figure 3A; Supplementary Video S1).

Figure 3.

Development of an in vitro 3D co-culture model for the dynamic replacement of endothelial cells by trophoblast cells. (A) Schematic diagram illustrating the development of an in vitro 3D co-culture model of trophoblast and endothelial cells. (B) Continuous monitoring of Pr-EVTs, HUVECs, and Pr-dNKs co-cultured with HUVECs over a 24-h period. Scale bars indicate 50 μm. (C) Statistical analysis of the integration rates of HUVECs co-cultured with Pr-EVTs, HUVECs, and Pr-dNKs for a 24-h period (n = 3). (D) Fluorescence staining of TUNEL was performed in HUVECs co-cultured with Pr-EVTs and Pr-dNKs, respectively. View 1 illustrates the contacted perspective, whereas View 2 depicts the uncontacted perspective. Scale bars indicate 10 μm. Data are presented as mean ± SD and comparison between groups was performed with unpaired two-tailed t-test. ^*P < 0.05.

Upon seeding onto HUVEC-tube structures, Pr-EVTs cells rapidly migrated towards the tube walls and underwent a morphological transformation from round to spindle-shaped upon contact with endothelial cells. These Pr-EVTs ultimately replaced the contacted endothelial cells (Figure 3B; Supplementary Video S2). Statistical analysis indicated that after 4 h of co-culture, approximately half of the HUVEC-tubes were replaced by Pr-EVTs, and by 24 h, over 80% of the tube structures were replaced by Pr-EVTs (Figure 3C). TUNEL assay results demonstrated significant TUNEL-positive apoptotic signal in endothelial cells that came into contact with Pr-EVTs, whereas no apoptotic signals were detected in endothelial cells that did not interact with Pr-EVTs (Figure 3D). The mean endothelial apoptotic rate was ~20% in the EVT-HUVEC co-culture for 4 h (Figure 4D).

Figure 4.

The comparative analysis of vessel remodeling capacity across diverse trophoblast cells. (A) Continuous monitoring of JEG-3 and hTSC-STB cells co-cultured with HUVECs over a 24-h period. Scale bars indicate 50 μm. (B) Fluorescence staining of TUNEL was performed in HUVECs co-cultured with JEG-3 and hTSC-STB cells, respectively. View 1 illustrates the contacted perspective, whereas View 2 depicts the uncontacted perspective. Scale bars indicate 10 μm. (C) Statistical analysis of the integration rates of HUVECs co-cultured with various trophoblast cells over a 24-h period (n = 3). (D) Statistical analysis of the apoptosis index in HUVECs co-cultured with various trophoblast cells (n = 3). Data are presented as mean ± SD and comparison between groups was performed with unpaired two-tailed t-test. ^*P < 0.05.

As non-trophoblast controls, either Pr-dNKs or HUVECs were co-cultured with the pre-formed HUVEC-tube structures. As illustrated, Pr-dNKs did not migrate toward the tube walls and exhibited no ability to integrate into the tube structures or induce endothelial apoptosis (Figure 3B–D). HUVECs were capable of integration into the tube structures but did not cause endothelial cell apoptosis (Figure 3B).

Within this co-culture system, the vascular structure remained intact regardless of whether EVTs incorporated or not. However, the vascular area gradually decreased as the co-culture duration extended. Notably, this phenomenon was also observed when HUVEC-tubes were co-cultured with HUVECs (Supplementary Figure S2C). Therefore, it is likely that this reduction in vessel area can be attributed to the collapse of some tubular structures due to matrix degradation over time [28].

Collectively, this 3D co-culture system effectively recapitulates the dynamic process of vascular remodeling mediated by trophoblast cells.

Comparative analysis of vessel remodeling capacity among multiple trophoblast cell lines

Utilizing the aforementioned refined 3D co-culture system, we conducted a comparative analysis of vessel remodeling capabilities among various trophoblast cell types. The dynamic interactions between different trophoblast cells and HUVEC tubular structures were documented in Supplementary Videos S2–S11. We quantified and statistically analyzed the integration of trophoblasts into vascular structures, induction of endothelial cell apoptosis, and changes in vascular area. Our results demonstrated that JEG-3, hTSC-EVT, HTR-8/SVneo, BeWo, and BeWo-STB cells exhibited the ability to integrate into HUVEC-tubes, induce endothelial cell apoptosis, and replace the endothelium. JEG-3 cells showed a comparable integration rate with Pr-EVTs within the first 12 h of co-culture, maintaining approximately a 50% integration rate by 24 h (Figure 4A and C). The vascular integration process for hTSC-EVT, HTR-8/SVneo, and BeWo cells was slower, achieving an integration rate of 20% to 40% by 24 h of co-culture. Interestingly, BeWo-STB also demonstrated some vascular incorporation capability, albeit at a lower rate of endothelial replacement compared to BeWo cells. Cell apoptosis analysis revealed that JEG-3 and hTSC-EVT cells induced endothelial apoptosis rates of 15% and 10%, respectively, whereas BeWo or BeWo-STB cells induced <5% endothelial apoptosis (Figure 4B and D). Other trophoblast cell types, including primary CTB, hTSC, and hTSC-STB, exhibited minimal activity in integrating into vascular structures or inducing endothelial apoptosis (Figure 4B and D; Supplementary Figure S2B).

The pattern of vascular area decrease over the duration of co-culture was consistence across all trophoblast cell types co-cultured with HUVEC-tube (Supplementary Figure S2C).

Correlation analysis of SAR-associated gene expression and vascular remodeling behaviors across diverse human trophoblast cell types

To establish evaluation criteria for identifying optimal trophoblast models that can simulate vessel remodeling process in vitro, we performed a correlation analysis between the expression levels of SAR-associated genes, the rate of trophoblast integration into vascular structures, and the induction of endothelial cell apoptosis. Our results demonstrated that the rate of trophoblast integration into vascular structures and the induction of endothelial cell apoptosis were positively correlated with the expression levels of NCAM1, ITGB3, JAG1, GGT1, and ITGAV while negatively correlated with TEAD4, ITGB1, and EPHA1 (Figure 5A and B; Supplementary Figure S3). Conversely, the expression levels of FLT4, CCR3, ADAMTS4, HLA-G, MMP12, ANGPT4, CXCL12, CGB, and NOTUM showed no significant association with vessel remodeling behaviors (Supplementary Figure S3).

Figure 5.

The integrative analysis of SAR-associated genes for evaluating the vessel remodeling capability of human trophoblasts. (A) The correlation analysis between the expression levels of NCAM1, JAG1, ITGB3, ITGAV, GGT1, and TEAD4 and the integration rates of trophoblasts. (B) The correlation analysis between the expression levels of NCAM1, JAG1, ITGB3, ITGAV, GGT1, and TEAD4 and the apoptosis index of HUVECs. (C) Weight analysis of SAR-associated genes in trophoblast cells as potential candidates for HUVECs replacement. (D) Comprehensive radar chart analysis of various trophoblast cells with respect to SAR-associated gene expression, integration rate, and apoptosis index. Data are presented as mean ± SD; n = 3 for all parameters; P < 0.05 indicates a statistical significance; |r| > 0.5 indicates a significant correlation.

To further assess the contribution of these genes to vessel remodeling behaviors of trophoblasts, we performed a statistical analysis using the “caret” package in R software. The results indicated that NCAM1, ITGB3, ITGAV, JAG1, and GGT1 exhibited high-weighted values, while CGB, CXCL12, ANGPT4, MMP12, HLA-G, ADAMTA4, and TEAD4 showed rather low weighted values (Figure 5C). Based on this weight analysis, we generated a radar chart for these index across different trophoblast cell types. The data clearly demonstrated that JEG-3 and hTSC-EVT cells closely resembled Pr-EVTs in terms of both high-weighted SAR-associated gene expression and tube remodeling performance in co-culture model (Figure 5D). Primary CTBs, BeWo, BeWo-STB, hTSC, hTSC-STB, and HTR-8/SVneo displayed relatively lower similarity with Pr-EVTs in vessel remodeling activities.

Collectively, the integrative analysis underscores the critical roles of SAR-associated genes, including NCAM1, ITGB3, ITGAV, and JAG1, in assessing the vessel remodeling capacity of human trophoblasts. Furthermore, it highlights JEG-3 and hTSC-EVT cells as optimal alternatives to Pr-EVTs for in vitro investigation of vessel remodeling.

Validation of the importance of NCAM1 in mediating trophoblast–endothelium interaction

To further validate the results of our integrative analysis, we selected NCAM1, one of the critical SAR-associated genes, and performed a loss-of-function experiment. The expression of NCAM1 was specifically knocked down in JEG-3 cells using shRNA (Figure 6C). Subsequently, these cells were co-cultured with pre-formed 3D HUVEC tubular structure. Live-cell imaging analysis revealed that the deficiency of NCAM1 in JEG-3 cells significantly impaired their ability to incorporate into the HUVEC-tube and replace endothelial cells, despite no observable effect on their migratory capacity toward the tube (Figure 6B and C; Supplementary Videos S12 and S13). These findings underscore the essential role of NCAM1 in mediating trophoblast–endothelial cell interactions during the SAR process.

Figure 6.

The deficiency of NCAM1 compromises the capacity of trophoblast cells to replace endothelial cells. (A) Continuous monitoring of shNC JEG-3 and shNCAM1 JEG-3 co-cultured with HUVECs over a 24-h period. Scale bars indicate 50 μm. (B) Statistical analysis of the integration rates of HUVECs co-cultured with shNC JEG-3 and shNCAM1 JEG-3 for a 24-h period (n = 3). (C) The knockdown efficiency of the NCAM1 gene in JEG-3 cells was detected by real-time PCR. Data are presented as mean ± SD and comparison between groups was performed with unpaired two-tailed t-test. ^*P < 0.05.

Discussion

The study of uterine SAR has faced significant challenges due to ethical constraints and substantial specimen variability, leading to a lack of appropriate in vivo animal models. Consequently, the development of ideal in vitro models is crucial for investigating this cellular event [13]. A critical step in SAR involves the recognition and interaction between endothelial cells and EVTs [1], and a sophisticated 3D co-culture system has been established to simulate this process. However, multiple studies utilizing this model have reported inconsistent findings regarding factors that regulate trophoblast differentiation toward endothelial phenotypes, trophoblast–endothelia recognition, and endothelial cell apoptosis [9, 11, 29–32]. Therefore, there is an urgent need to comprehensively evaluate the properties of various lines of trophoblast cells to standardize the model and more accurately simulate the physiological processes of SAR. In this study, we use primary cultured human EVTs as a benchmark to assess the molecular characteristics and cellular performance of several commonly used trophoblastic cell models. Our data emphasize the critical roles of SAR-associated genes, including NCAM1, ITGB3, ITGAV, and JAG1, in evaluating the vascular remodeling capacity of human trophoblasts (Figure 7). Through systematic analysis, JEG-3 and hTSC-EVT cells demonstrate their advantages as optimal alternatives to primary EVTs for in vitro investigations of vascular remodeling.

Figure 7.

The schematic diagram illustrates the mechanism through which SAR-associated core genes participate in the interaction process between enEVTs and endothelial cells.

Uterine SAR in human pregnancy is a complex and multifaceted process that encompasses the infiltration of EVTs into vessels, recognition, and replacement of certain proportion of endothelial cells by EVTs, endothelial cell apoptosis, recruitment of multiple immune cells, dedifferentiation of vessel smooth muscle cells, and degradation of extracellular matrix with subsequent deposition of fibrinoid materials. Traditional studies as well as recent single-cell sequencing analyses have identified several genes associated with uterine SAR [17–19, 33]; however, the precise mechanisms by which these molecules participate in these cellular events remains to be fully elucidated. In this study, we highlight the critical roles of genes such as NCAM1, ITGB3, ITGAV, and JAG1 in the recognition and replacement of endothelial cells by trophoblasts. Proteins encoded by these genes may function as specific receptors on EVTs to interact with ligands on endothelial cells, thereby directly mediating the specific recognition of uterine endothelial cells by EVT. Further investigation is warranted to explore their underlying mechanisms. Other SAR-associated genes, including MMP12, CXCL12, ADAMTS4, and ITGB1, may play roles in the recruitment of immune cells, dedifferentiation of vessel smooth muscle cells, or degradation of extracellular matrix [34–37], processes that cannot be fully presented in the 3D co-culture model utilized in this study. On the other hand, CGB has been established as an essential molecule for STB formation and functions [38], showing little correlation with trophoblast integration into vessel tubular structures, thus further supporting the notion that VTs and EVTs represent distinct differentiation pathways and regulatory mechanisms.

The recently established hTSCs and their induced differentiation products have emerged as ideal models for investigating human trophoblast cell fate and placental development [16]. The findings of this study indicate that hTSC-derived EVT cells can mimic primary EVT cells in interacting with endothelial cells; however, their capacity to remodel blood vessels appears to be relatively limited. It is likely that the current method of inducing hTSC differentiation towards EVT results in a restricted cell population exhibiting pseudo-endothelial phenotypes [17, 21]. Therefore, further optimization of the induction protocol for hTSC is warranted.

Several studies have systematically compared the molecular characteristics of various trophoblast cell lines [13, 39–41]. JEG-3 cells, which express endogenous HLA-G, have been proposed as a suitable model for simulating the invasive behavior of trophoblasts [39]. This study provides evidence supporting the use of JEG-3 as an alternative to primary EVTs when investigating trophoblast–endothelium interactions. BeWo cells represent a heterogeneous population that includes both STBs and EVT-like cells [42]. Although BeWo cells demonstrate good performance in tubular integration in the 3D co-culture model, their heterogeneity may introduce significant uncertainty in the investigation of SAR mechanisms. HTR-8/SVneo cells are heterogeneous populations containing both trophoblasts and mesenchymal cells [15]. They do not express EVT-specific HLA-G and exhibit minimal level of the epithelial marker cytokeratin [39]. Multiple studies have also advised against utilizing HTR-8/SVneo cells as a trophoblastic model.

An interesting question is the cell fate and interactions during human SAR, which has raised longstanding controversies [5]. Since the introduction of the “endothelial mimicry” theory illustrating that EVTs acquire endothelial-like characteristics within uterine arteries, replace endothelial cells, and ultimately occupy the inner linings of the arterial wall, numerous studies have provided substantial evidence supporting this hypothesis [17, 19, 26, 43, 44]. Conversely, other investigations have challenged this theory, arguing that the replacement of endothelium by endovascular EVTs (enEVTs) is a transient event, with stable enEVT occupation within the vascular lumen rarely observed [45]. Importantly, a recent study by Greenbaum et al. [19] conducted a high-resolution, single-cell omics-driven systematic analysis of uterine arteries derived from 66 pregnant individuals across gestational week 6 to 20. Arteries characterizing by the full occupation of the inner arterial linings by HLA-G⁺ CK7⁺ EVTs were identified, and EVT-mediated endothelial replacement occurred persistently throughout gestational weeks 6 to 20. Our previous studies [22, 44] also suggest that EVT-mediated endothelial cell replacement may persist in a certain proportion of uterine arteries during human SAR. However, it is inevitable that the majority of studies to date have been conducted using cross-sectional specimens from either archived collections or aborted materials. Several factors, such as the specimen positioning, section orientation, immunostaining methodologies, and individual variability, may contribute to differing observational outcomes. Moreover, it remains infeasible to trace cellular behaviors within the human uterus during pregnancy. Consequently, it’s unsurprising that significant controversies persist in our understanding of human SAR. Further in-depth investigations are warranted to fully address this question.

This study acknowledges certain limitations. Firstly, the endothelial cells used in the 3D co-culture model are HUVECs, rather than uterine vascular endothelial cells, particularly spiral artery endothelial cells. This discrepancy may lead to divergent outcomes compared to using spiral artery endothelial cells. The choice of HUVECs was driven by the technical challenges associated with isolating and obtaining human uterine spiral artery endothelial cells, as well as the absence of stable cell lines for these cells. In contrast, HUVECs offer advantages such as ease of accessibility, minimal ethical concerns, and a robust ability to form tubular structure in Matrigel that closely mimic human endothelial cells [46–48]. Secondly, the primary EVTs used in this study represent a mixture of interstitial and endovascular EVT subsets, as evidenced NCAM1 expression pattern (Figure 1C). While these cells exhibit strong potential for endothelial remodeling, further mechanistic studies on SAR should ideally employ purified populations of enEVTs, as demonstrated in our previous research [27, 44]. Thirdly, the origin of enEVTs has been a subject of long-standing debate, with two primary hypotheses proposed: the intravasation hypothesis and the Plug-origin hypothesis. The intravasation hypothesis posits that enEVTs arise from the interstitial invasion of iEVTs at the perivascular region [49, 50]. Conversely, the Plug-origin hypothesis suggests that enEVTs predominantly originate from Plugs that penetrate into spiral arteries through their front openings [43, 51, 52]. To date, an increasing body of evidence supports the integration of these two views, indicating the co-existence of EVTs invading from the interstitial side and those penetrating intramurally. In our study, it’s relatively challenging to distinguish these two pathways using the co-culture model, which is better suited for mimicking the interaction between trophoblasts and endothelial cells.

In conclusion, we have conducted a comprehensive and meticulous comparative analysis and propose an evaluation framework to systematically assess the key attributes of trophoblast cells that accurately represent EVT behaviors in SAR. Our findings underscore the critical roles of specific molecules during SAR. Specifically, CXCL12, ADAMTS4, ITGB1, and MMP12 may play critical roles in regulating extracellular matrix degradation and vascular smooth muscle cells (VSMCs) dedifferentiation by mediating interactions among enEVTs, dNKs, and VSMCs. Additionally, during the essential process of endothelial cell replacement by enEVTs, NCAM1, ITGB3, ITGAV, and GGT1 demonstrate significant positive regulatory effects (Figure 7). Additionally, JEG-3 and hTSC-derived EVT cells demonstrate significant advantages as optimal models for mimicking primary EVTs in vitro, thereby facilitating investigations into uterine vascular remodeling. This study establishes a robust foundation for future exploration of the molecular mechanisms underlying human SAR.

Author contributions

W.L. conducted the experiments, analyzed the data, and prepared the initial draft of the manuscript. Y.M. and S.Z. participated in the experiments of 3D cell co-culture, S.Y., X.F., and X.W. were responsible for collecting and identifying clinical samples of human maternal–fetal interface. Y.W. and Y.M. designed and oversaw the study, interpreted the data, and revised the manuscript.

Conflict of interest: The authors declare no potential or actual conflicts of interest regarding this work.

Funding

None declared.

Data availability

All data required to evaluate the conclusions presented in this paper are included within the manuscript and/or the associated Supplementary Materials.