Self‐assembled human placental model from trophoblast stem cells in a dynamic organ‐on‐a‐chip system

Abstract The placental barrier plays a key role in protecting the developing fetus from xenobiotics and exchanging substances between the fetus and mother. However, the trophoblast cell lines and animal models are often inadequate to recapitulate the key architecture and functional characteristics of human placental barrier. Here, we described a biomimetic placental barrier model from human trophoblast stem cells (hTSCs) in a perfused organ chip system. The placental barrier was constructed by co‐culture of hTSCs and endothelial cells on the opposite sides of a collagen‐coated membrane on chip. hTSCs can differentiate into cytotrophoblasts (CT) and syncytiotrophoblast (ST), which self‐assembled into bilayered trophoblastic epithelium with placental microvilli‐like structure under dynamic cultures. The formed placental barrier displayed dense microvilli, higher level secretion of human chorionic gonadotropin (hCG), enhanced glucose transport activity. Moreover, RNA‐seq analysis revealed upregulated ST expression and activation of trophoblast differentiation‐related signalling pathways. These results indicated the key role of fluid flow in promoting trophoblast syncytialization and placental early development. After exposure to mono‐2‐ethylhexyl phthalate, one of the endocrine disrupting chemicals, the model showed inhibited hCG production and disturbed ST formation in trophoblastic epithelium, suggesting impaired placental structure and function elicited by environmental toxicants. Collectively, the hTSCs‐derived placental model can recapitulate placenta physiology and pathological response to external stimuli in a biomimetic manner, which is useful for the study of placental biology and associated diseases.


| INTRODUCTION
The placental barrier is an essential interface to maintain normal pregnancy and support fetal development between the mother and fetus. 1,2 In the first trimester, the placental barrier is composed of multilayered structure including the mononuclear cytotrophoblast (CT), the syncytiotrophoblast (ST), basal lamina, and the fetal capillaries ( Figure 1A). It can not only supply essential nutrient substance, such as glucose, amino acids and vitamins for fetal growth, but also plays a key role in protecting fetus from exposure of virus, bacteria and xenobiotics. [2][3][4] The pathogens or noxious chemicals in maternal blood may pass through the placental barrier and disturb fetal development during early pregnancy. The study of impact of exogenous harmful agents on human placental barrier is instructive for reproductive health.
Currently, various experimental models have been developed to study the response of human placenta to external stimulus.
Generally, animal models have great variations in the architecture of placenta from that in humans due to species differences. [5][6][7] The isolated explant cultures or primary placental cells could mimic the complex composition and structure of human placenta, [8][9][10] but the use of human tissue samples is limited due to ethical concerns of tissue availability, storage and manipulation. In vitro trophoblast cell culture models, such as BeWo 11,12 and Jeg-3 13,14 cell lines and human induced pluripotent stem cells (hiPSCs) 15 have been used to construct placental barrier, but they are often inadequate to recapitulate the complex placental architectures and functions due to abnormal cell phenotype or immature functions of trophoblasts.
Recently, human trophoblast stem cells (hTSCs) were obtained from primary placental tissues or blastocysts. 16 hTSCs have the capacity to give rise to three major trophoblast subpopulations, including CT, ST and extravillous cytotrophoblast (EVT) ( Figure 1B).
It provides a valuable cell source for placenta research. Trophoblast organoids were recently established from hTSCs to recapitulate F I G U R E 1 Schematic of the biomimetic placental barrier model derived from hTSCs. (A) The placenta is located at the interface between maternal and fetal blood. Human placental barrier is composed of fetal endothelial layer, basal membrane and trophoblastic layers in vivo. Trophoblastic epithelium in early pregnancy consists of the ST layer and the underlying CT layer. (B) The hTSCs were derived from trophectoderm of blastocysts, which can differentiate into three major trophoblast lineages (CT, ST and EVT). (C) A bioengineered placental barrier model was constructed in a perfused organ-on-a-chip system. hTSCs were seeded on the upper channel to form the bilayered trophoblastic epithelium. HUVECs were cultured on the other side of collagen-coated membrane to mimic the fetal endothelium. (D) The structure and function of placenta-on-a-chip model were characterized with different methods. placental development and function. 17,18 However, organoid models are often limited to recreate the multilayered structure and functions of placental barrier due to the lack of endothelium or dynamic microenvironment.
Organ-on-a-chip (organ chip) is a microfluidic cell culture device that allows to recapitulate the key architectural and functional hallmarks of living organs in vitro. 19,20 It can mimic the in vivo-like cellular microenvironment by control over tissue-tissue interface, dynamic flow and biochemical signals. Advances in organs-on-chips have enabled the creation of various models of human organs, such as the lung, gut, liver and so on. [20][21][22][23] Although several placental barrier-ona-chip models have been reported previously, 11,12,14,24 these models were commonly established using trophoblast carcinoma cell lines, which do not represent the near physiological features of human placenta.
In this work, we described a new strategy to create a 3D biomimetic placental model from hTSCs by combining stem cell biology and organ chip technologies. The hTSCs were derived from trophectoderm cells of the human blastocyte based on the published protocol. 16 The chip device consisted of upper and lower culture chambers incor-

| Microchip fabrication
The organ chip was fabricated through standard soft lithography techniques as described in our previous study. 26 Briefly, moulds made of SU-8 (MicroChem Corp.) were prepared using photolithography. Then, polydimethylsiloxane (PDMS) (Dow Corning) mixed with monomer and curing agent at a ratio of 10:1 was poured on the moulds. After incubation at 80 C for 30 min, the upper and lower layers with microchannels of 20 mm (length) Â 1.2 mm (width) Â 0.2 mm (height) were prepared. For assembly of the chip, PET nuclear pore membrane with 2 μm pores was sandwiched between the upper and lower layers following the treatment of plasma.

| Construction of placental barrier on chip
The assembled chip was sterilized by UV irradiation overnight. Then, microchannels were coated with Collagen IV (Corning, 0.01 mg/mL in DPBS) at 37 C for 6 h. hTSCs were first seeded on the upper microchannel at a concentration of 6 Â 10 6 cells/mL. TS medium was changed every 12 h to maintain hTSCs proliferation for 2 days.
Next, HUVECs at a concentration of 2 Â 10 6 cells/mL were introduced into the lower microchannel. The chip was then inverted and incubated at 37 C for 2 h. Following the attachment of HUVECs, ECM medium was perfused into the lower microchannel by syringe pump with a speed of 10 μL/h. Meanwhile, hTSCs differentiation medium containing DMEM/F12 basic medium, 0.5% N2 supplement, 1% B27 supplement minus vitamin A, 100 μg/mL Primocin,

| Scanning electron microscopy
Samples were fixed with 2% glutaraldehyde and 1% osmic acid successively for 2 h at room temperature. Then, the fixed samples were dehydrated with a series of diluted ethanol, substituted with isoamyl acetate, dried using critical point dryer (Quorum) and sputter-coated with gold in ion sputtering apparatus (Hitachi). Next, samples were observed and imaged using a scanning electron microscopy (SU8100, Hitachi).

| Real-time quantitative PCR
Total RNA was isolated from cells on the chip using Trizol reagent (TAKARA). RNA quality and concentration were determined by Nano-Photometer (IMPLEN). cDNA was produced after RNA was diluted to 50 ng/μL. Then, cDNA was amplified via qPCR using Ex Taq DNA polymerase (TAKARA) under the following reaction conditions (40 cycles

| ELISA
Human chorionic gonadotropin (hCG) ELISA kit (Cloud-Clone) was used to examine the hCG secreted from trophoblast cells. After construction of the placental barrier model or treatment of MEHP, cell medium was changed and maintained under static culture for another 12 h. Then, the medium was collected and stored at À80 C until use.
For detection of hCG concentration, samples were prepared following the instructions provided by manufacturer and the fluorescence intensity was assessed by a microplate reader.

| Oil red O staining
Trophoblast cells on the chip under static or dynamic culture were fixed in 4% paraformaldehyde for 20 min at room temperature. Next, microchannels were washed with DPBS and filled with 0.2% Triton X-100 for 10 min. Then, the lipid droplets in trophoblasts were stained by Oil red O (Sigma-Aldrich) for 20 min. After the remaining dye was washed away, the stained samples were imaged using a fluorescence microscope (Olympus).

| 2-NBDG based assays
The hTSCs were cultured on the chip with or without differentiation medium perfusion for 4 days. After the microchannels were washed by DPBS, the maternal side was perfused with 1 mM 2-NBDG (Psaitong). Next, the chip devices were incubated at 37 C for 30 min.
Then, the buffer in fetal channel was collected and trophoblast cells were lysed with 1% Triton X-100 for 5 min. Fluorescence intensity of the buffer and lysis solution were analysed by a microplate reader (Tecan).

| Statistical analysis
Data were expressed as the mean ± standard error of the mean (SEM) or mean ± standard deviation (SD) for at least three independent experiments. The difference between two groups was analysed using Student's t-test. Significance was indicated by asterisks: *p < 0.05; **p < 0.002; ***p < 0.001. The image-based fluorescence intensity statistics were sampled from three different chips and each for three different areas. The areas were picked randomly.

| RESULTS
3.1 | Self-assembled bilayered trophoblastic epithelium from hTSCs in a perfused chip system In vivo, human placenta barrier is consisted of the maternal facing trophoblastic epithelium layer, connective tissue and endothelium lining the fetal capillaries in the first trimester. After implantation, villous CT cells proliferate and fuse to form the multinucleated ST, which assembles into the trophoblastic epithelial layer of the chorionic villi exposed to blood flow. 27 hTSCs derived from the blastocyst can differentiate into major trophoblast subtypes including CT and ST. 16 In order to build the biomimetic human placental barrier in vitro, we initially identified the formation of trophoblastic epithelium from hTSCs in a dynamic microenvironment.
Herein, we designed a perfusable chip device comprised of multi- in lipid droplet accumulation. 33 In contrast to the static group, FSSexposed trophoblast layer showed a larger area of PLIN2 expression with the similar cell density ( Figure 5A). Moreover, the result of Oil red O staining showed a bigger size of lipid droplet formation in trophoblasts under fluid flow condition ( Figure 5B), indicating that FSS enhanced the lipid droplet accumulation in trophoblast cells. Apart from PLIN2, the HADH and CPT1A genes, which are closely related to fatty acid catabolism, also showed higher expression in the FSS group as shown by RT-qPCR ( Figure 5C). As to the glucose transport, the immunofluorescence images showed significantly stronger expression of GLUT1 in FSS-exposed trophoblastic epithelium compared to the static group ( Figure 5D). Besides, trophoblasts under perfusion culture showed an increase in mRNA expression levels of GLUT1 (SLC2A1) and GLUT4 genes (SLC2A4) ( Figure 5E,F), suggesting that FSS may facilitate the glucose transport through placental barrier. The 2-NBDG based assay further confirmed that FSS-exposed trophoblasts not only showed higher glucose uptake, but also allowed more glucose transfer to the fetal side ( Figure 5G The image-based fluorescence intensity statistics were sampled from three different chips and each for three different areas. The areas were picked randomly. Image J software was used for statistical analysis. The data are presented as the mean ± SD. Data significance was assessed by unpaired two-tailed Student's t-test; ***p < 0.001. (C) hCG secretion by trophoblast cells collected from the placental barrier under static or dynamic culture. The concentration of hCG in culture medium was analysed by ELISA kit. The data are presented as the mean ± SD. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05. (D) Relative mRNA expression of TP63, ITGA6, CGB, SDC1 and HLA-G in trophoblast cells collected under static or dynamic culture. mRNA expression normalized to GAPDH RNA level was analysed by realtime PCR. The data are presented as the mean ± SEM from three independent experiments. Data significance was assessed by unpaired twotailed Student's t-test; *p < 0.05, **p < 0.002. cultures ( Figure 6A). Volcano plots showed that 2005 down-regulated genes and 1965 up-regulated genes among DEGs were notably modulated in FSS-exposed trophoblast cells ( Figure 6B mRNA expression normalized to GAPDH RNA level was analysed by real-time PCR. The data are presented as the mean ± SEM from three independent experiments. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05. (D) Representative fluorescence image of the trophoblast layer stained with GLUT1 antibody (red) and DAPI (blue) under static or dynamic culture. Scale bars are 50 μm. (E, F) Relative mRNA expression of SLC2A1 and SLC2A4 in trophoblast cells collected under static or dynamic culture. mRNA expression normalized to GAPDH RNA level was analysed by real-time PCR. The data are presented as the mean ± SEM from three independent experiments. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05, **p < 0.002. (G, H) Glucose uptake and transfer to fetal side of trophoblastic barrier under static or dynamic culture. The glucose transport activity was tested by 2-NBDG uptake assays. The data are presented as the mean ± SD. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05, **p < 0.002. were related to the fusion of trophoblasts and calcium ion fluxes. [37][38][39] Moreover, gene ontology (GO) analysis identified enhanced biological processes including calcium ion transport, calmodulin binding, epidermal cell differentiation and hormone secretion in FSS-exposed trophoblasts ( Figure 6E). The data above suggested that syncytialization of trophoblasts were probably facilitated by FSS through calcium-related downstream pathways. RNA sequencing about genes related to Hippo signalling pathway ( Figure 6F) and Rap1 signalling pathway ( Figure 6G) were finally verified by RT-qPCR.

| Responses of placental barrier model to mono-2-ethylhexyl phthalate (MEHP) exposure
The placental development is susceptible to environmental toxicants, especially in early gestation. 40 MEHP is one of the endocrine disrupting chemicals found in maternal blood, which is the primary metabolite of the common plasticizing agent di(2-ethylhexyl) phthalate (DEHP). The concentration of MEHP in maternal and umbilical cord blood may reach to 1-40 μM. 41 Several studies have reported that MEHP exposure was associated with disturbance of placental and fetal development. 42,43 To verify the irritability of this placental model in response to phthalates, we added MEHP in the trophoblast side on chip to mimic the environmental toxicants exposure in maternal circulatory system ( Figure 7A).
We first examined the viability of hTSCs and HUVECs with MEHP exposure. The two types of cells were exposed to MEHP at different concentrations for 48 h in 96-well plate, respectively.
The result showed that the viability of hTSCs decreased with the dose increase of MEHP, while HUVECs were barely affected ( Figure S4). Then, we tested the expression of genes related to hTSCs differentiation in trophoblasts exposed to MEHP at different dose levels. The data showed that 0.1, 1, and 10 μM MEHP inhibited the differentiation of hTSCs, but 100 μM MEHP promoted the expression of ST markers ( Figure S5). Such a U-shaped effect of MEHP was consistent with previous study. 41  Among them, only IL-1a increased due to the exposure of MEHP, suggesting that 1 μM MEHP barely trigger inflammation on the fetal side ( Figure 7C). To further verified the inhibitory effects of lowdose MEHP in syncytialization process. We examined the expression of CGB on protein level by immunofluorescence. As shown in the fluorescence images ( Figure 7D) and statistics of mean fluorescence intensity ( Figure 7E), the expression of CGB decreased in MEHP-exposed trophoblast layers. Moreover, the result of ELISA showed reduced hCG secretion in trophoblasts exposed to MEHP ( Figure 7F). In order to verify the physiologically relevant responses of this placental model to MEHP exposure, we further compared the expression of genes related to trophoblast differentiation among 2D trophoblast cell model, trophoblast epithelium-on-chip model (chip-TSC) and the placental barrier-on-chip model (chip-coculture). As shown in Figure 7G Placenta is an organ sensitive to mechanical forces, especially the fluid shear stress caused by maternal blood flow. 44 Trophoblast cells are exposed to a broad range of shear stress value from 0.001 to 30 dyn/cm 2 in different periods in vivo. In the first trimester, blood flow in uterine spiral arteries is blocked by extravillous trophoblastic plugs, which results in a small amount of maternal blood passed through the plugs into intervillous space. 45 Therefore, trophoblasts subject to a relatively low FSS during early gestation. In this study, we built the placental barrier on chip using a shear stress of The image-based fluorescence intensity statistics were sampled from three different chips and each for three different areas. The areas were picked randomly. Image J software was used for statistical analysis. The data are presented as the mean ± SD. Data significance was assessed by unpaired two-tailed Student's t-test; **p < 0.002. (F) hCG secretion by trophoblast cells collected from the placental barrier model with vehicle or 1 μM MEHP treatment. The concentration of hCG in culture medium was analysed by ELISA kit. The data are presented as the mean ± SD. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05. (G) Relative mRNA expression in trophoblast cells collected from the 2D trophoblast cell model, trophoblastic epithelium-on-chip model (chip-TSC) and placental barrier-on-chip model (chip-coculture) treated with vehicle or 1 μM MEHP. The data are presented as the mean ± SEM. Data significance was assessed by unpaired two-tailed Student's t-test; *p < 0.05, **p < 0.002.
have been devoted to engineer placental barrier model using hTSCs.
In addition, we confirmed that low FSS ($0.005 dyn/cm 2 ) is more favourable for the differentiation of trophoblasts compared to high FSS ($0.05 dyn/cm 2 ). The results suggested that our placental The extensive use of DEHP in polyvinyl chloride materials caused a wide distribution of phthalates in the environment. MEHP, a metabolism of DEHP in human bodies, has been reported to be more toxic than its precursor. 53 Previous studies have identified the side effects of MEHP on trophoblast differentiation and syncytialization. 41 In our model, we demonstrated a U-shaped dose-response effect of MEHP on hTSCs differentiation, which is similar to previous study of primary trophoblasts.
The results of comparison with 2D model suggested that intercellular interaction between endothelium and trophoblasts might make trophoblastic epithelium more sensitive to MEHP exposure, although MEHP barely showed direct toxicity to HUVECs. Moreover, our study indicated that low-dose MEHP promoted the differentiation of trophoblast cells into EVT, which may be associated with the abnormal trophoblast migration. These responses of the placental model to endocrine disrupting chemicals are often related to placental dysfunctions and disrupted placental development at early stages. It also demonstrated that this bioengineered model could provide a new platform for the investigation of environmental toxicants exposure in human early placenta.
Despite the potential applications of our model, there is still space for improvement. In this work, PDMS is the main material for fabricating chip device, but it may adsorb organic molecules, such as hormones or chemicals, impeding precise drug testing and quantitative response. Although we modified the surface of PDMS channels with PF127 to reduce adsorption of molecules, the chip materials for more biocompatibility and less nonspecific adsorption is needed to be considered. In addition, given the complexity of human placenta, the incorporation of more physiologically relevant placental cells could further improve the functions of placental model. For example, the interactions between trophoblasts and immune cells including natural killer cell and Hofbauer cell may contribute to construct more predictive models and reflect more accurate responses to external stimuli at maternal-fetal interface. We envision that other bioengineered approaches or microfluidic elements could be incorporated to advance the development of placental models with high fidelity, thereby contributing to their applications in studies of human reproductive health and disease.

AUTHOR CONTRIBUTIONS
Jianhua Qin conceived the study and revised the manuscript. Rongkai Cao performed most of the experiments and writed the manuscript.
Yaqing Wang designed the experiments and writed the manuscript.
Jiayue Liu analysed the data of RNA sequencing. Lujuan Rong provided the human trophoblast stem cells.