Fetal response to maternal exposures of environmental chemicals: Utility of a four-cell human feto-maternal interface organ-on-chip

Abstract

Epidemiological studies suggest that maternal exposures to environmental pollutants may be linked to spontaneous preterm birth. Mechanistic studies are needed to provide support to these hypotheses; however, existing in vitro models do not replicate human feto-maternal barriers beyond placenta. The recently developed four-cell Feto-Maternal interface Organ-On-Chip (FMi-OOC) model enables studies of chemical effects that are critically important for maintaining full-term pregnancy. We tested four environmental pollutants that have been associated with preterm birth – dichlorodiphenyltrichloroethane (DDT-o,p’), bisphenol A (BPA), 2,2′4,4′-tetrabromodiphenyl ether (PBDE-47), and perfluorooctanoic acid (PFOA). Concentration-response effects of these chemicals were first tested on maternal decidua cells in 96-well plates. Then, using the 4-cell FMi-OOC that mimics the in utero tissue topology, compounds were added to the maternal (i.e., decidua) chamber, and chemical propagation, cell viability, and cytokine production (IL-6, IL-8, GM-CSF, TNF-α) were measured in decidua, chorion trophoblast, amnion mesenchymal, and amnion epithelial cell chambers for up to 72 h. Minimal chemical propagation to the fetal chambers was observed. Treatment-associated increase in cytokines was observed for all compounds tested, with PFOA and BPA showing the strongest effects and amnion epithelial cells being most responsive. We demonstrate how the multi-cellular FMi-OOC can be used to study paracrine signaling in feto-maternal interface tissues. We show that upon maternal exposure, albeit at concentrations exceeding human blood levels by 1–2 orders of magnitude, fetal membranes attain pro-inflammatory state, a trigger for preterm birth. These studies support the biological plausibility of the epidemiological associations between exposures to tested compounds and preterm birth.

1. Introduction

Preterm birth (delivery before 37 weeks of gestation) is a condition that affects 10.4 % of live birth pregnancies in the United States [1]. Globally, this rate is similar to that in the United States, albeit it varies widely across countries from 4 % to 16 %, with the greatest incidence in sub-Saharan Africa and southern Asia [2]. Births that occur preterm have been linked to a variety of adverse health outcomes [3]. For example, preterm birth is the leading cause of death in neonates and children younger than 5 [4]. Long term adverse health outcomes that have been associated with preterm birth are increased risks of hypertension, asthma and other respiratory diseases, type 2 diabetes, cardiovascular diseases, chronic kidney disease, and neurocognitive disorders [5,6]. Therefore, preterm birth is a significant public health burden [7].

Many risk factors have been identified as potentially resulting in preterm labor and premature birth [3]. Some of the risk factors are related to medical conditions during pregnancy or other maternal health issues that may be difficult to modify [8]. However, certain lifestyle and environmental factors, including exposure to environmental pollutants, are regarded as “modifiable” risk factors for preterm birth and associated child mortality as well as adverse health outcomes [9]. The potential role of environmental pollutants in preterm birth has been suggested by studies of air pollution, occupational, and other exposures to lead, pesticides, perfluoroalkylated and polyfluoroalkylated substances, phthalates, and other toxicants [3]. Most of the evidence comes from observational epidemiological studies that are limited by sample size and potential confounders. For example, the data on the potential linkages between preterm birth and maternal exposure to environmental contaminants of concern, such as dichlorodiphenyltrichloroethane (DDT), bisphenol A (BPA), 2,2′4,4′-tetrabromodiphenyl ether (PBDE-47), and perfluorooctanoic acid (PFOA) is suggestive but fragmented.

For example, BPA was widely used in the production of plastics and other resins found in consumer products, where it can leach from these materials into water or food [10]. BPA is detectable in virtually all human samples, and after increased public awareness of its potential endocrine disruption effects and phase out from many consumer products, it’s levels in humans have been on the decline [11]. High BPA concentrations in the maternal blood and urine have been suggested as a risk factor for preterm birth [12–14]. Similarly, DDT is an insecticide that was used widely in mid-20th century to control malaria, typhus, and other insect-borne diseases [15], and is a classic persistent organic pollutant because of its resistance to degradation [16]. While already banned for most uses, DDT still persists in the environment and is approved for disease vector control in some countries [16]. While some studies suggested that DDT use increases the risk of preterm birth [17], this finding was not replicated in other cohorts [18]. PBDEs are another class of persistent organic pollutants – they are brominated flame retardants, which are produced as mixtures of congeners with varying degrees of bromination. PBDE-47 is a major component of commercial pentaBDE mixtures, which were widely used as flame retardants in consumer products like furniture foam, electronics, and textiles before being phased out in many countries due to environmental and health concerns [19]. Toxicokinetics of PBDEs is characterized by slow metabolism, high lipophilicity, and high bioaccumulation potential, which make them persistent in the environment and in humans [20]. High plasma concentrations of PBDE-47 in pregnant women during the first trimester was shown to increase the risk of preterm birth [21]; however, heterogeneity among studies was also reported [22]. Finally, PFOA is a completely fluorinated organic acid that has been manufactured since the mid-20th century with production volumes reaching thousands of metric tons annually in the late 20th century [23]. Because of persistence and bioaccumulation, including in humans [24], as well as concerns about its toxicity and carcinogenicity [25], PFOA production largely ceased in 2002. Despite these efforts, legacy contamination and continued use in some regions have sustained environmental and human exposure to PFOA [26]. Contradictory studies have been published with regards to the relationship between maternal blood levels of PFAS and adverse birth outcomes, with most of the authors acknowledging difficulties with controlling for confounders in these observational studies [27–29].

Because most published human studies on association between environmental pollutants and preterm birth suffer from inconsistencies in the findings across cohorts, further research is needed to delineate mechanisms through which chemical exposures may cause preterm birth. Because fetal membranes formed during pregnancy are complex multi-cellular tissues, they present a challenge for mechanistic studies using either rodent, human explant placentas, or traditional in vitro models; however, novel tissue chip models (also called organ-on-a-chip models or microphysiological systems) provide opportunities to study responses across multiple cell layers [30]. Also, even though the mechanisms of preterm birth are complex, the role of inflammation of the feto-maternal interface tissues – decidua, chorion, and amnion – has been firmly established [31,32]. This study used four environmental compounds that have been associated with preterm birth – DDT, BPA, PBDE-47, and PFOA – and tested their effects on maternal and fetal cells using the Feto-Maternal interface Organ-On-Chip (FMi-OOC). By using this device, we were able to isolate chemical vs paracrine effects on the decidua, chorion, and amnion – maternal and fetal tissues responsible for maintaining full-term pregnancy.

2. Materials and methods

2.1. Cells and general culture methods

All cells used in these studies were provided by the Menon Laboratory at the University of Texas Medical Branch. Previous publications describe isolation and immortalization of human maternal decidua (DEC) cells [33], chorion trophoblast (CTC) cells [33,34], amnion mesenchymal (AMC) cells [35], and amnion epithelial (AEC) cells [36]. DECs and AMCs were cultured in DMEM/F12 medium (FisherScientific, Hampton, NH; Cat no. MT10092CV) supplemented with 10 % FBS (FisherScientific, Cat no. MT35010CV), 1 % penicillin/streptomycin (FisherScientific, Cat no. MT30001Cl), and 1 % amphotericin B (Sigma-Aldrich, St. Louis, MO; Cat no. A2942). AECs were cultured in KSFM medium (Gibco, Grand Island, NY; Cat no. 17005042) supplemented with 1 mL primocin (Invivogen, San Diego, CA; Cat no. ant-pm-1). CTCs were cultured in DMEM/F12 medium supplemented with 0.20 % FBS, 0.01 mM β-mercaptoethanol (Sigma-Aldrich, Cat no. M3148), 0.5 % penicillin/streptomycin, 0.3 % BSA (GeminiBio, West Sacramento, CA; Cat no. 700-100P), 1 × ITS-X (Gibco, Cat no. 51-500-056), 2 μM CHIR99021 (Sigma-Aldrich, Cat no. SML1046-5 MG), 0.05 μM A83-01 (Sigma-Aldrich, Cat no. SML0788-5 MG), 1 μM SB431542 (Sigma-Aldrich, Cat no. 616464-5 MG), 1.5 μg/mL l-ascorbic acid (Sigma-Aldrich, Cat no. A4544), 50 ng/mL epithelial growth factor (Sigma-Aldrich, Cat no. E4127–.1 MG), 0.8 mM VPA (Sigma-Aldrich, Cat no. P6273–100 ML), and 5 μM Revitacell (Rock inhibitor/Y27632, MilliporeSigma, Cat no. 68-800-05 MG). All cells were grown at 37 °C and 5 % CO₂ in standard “tissue culture” T-75 flasks (Corning, Corning, NY: Cat no. 43064U) until they reached 80 %–90 % confluency. Cells under passage 30 were used for all experiments.

2.2. Cell culture and chemical treatments in 96-well plates

Test compounds (Table 1) were prepared from powdered stocks: BPA (Sigma-Aldrich, Cat no. 42088–100 mg), PFOA (Sigma-Aldrich, Cat no. 171468–5G), PBDE-47 (Sigma-Aldrich, Cat no. 91834–10 MG), and DDT-o,p’ (ChemService, West Chester, PA; Cat no. N-12708). Chemical stocks were prepared at 400 × in cell culture grade dimethyl sulfoxide (DMSO, CAS#67-68-5; Santa Cruz Biotechnology, Dallas, TX; Cat no. sc-358801). Before experiments, dilutions in media were prepared for each tested concentration (Table 1). Final test concentrations for PBDE-47, BPA, and PFOA were 10, 30, 100 and 300 μM. For DDT, these were 1, 3, 10, and 30 μM. Thus, in all subsequent experiments, 0.25 % (v/v) of DMSO was used as vehicle control. Fresh stocks of each compound were prepared for each set of experiments.

Table 1.

Chemicals tested in these studies.

Chemical	PBDE-47	PFOA	BPA	DDT-o,p’
CAS#	5436-43-1	335-67-1	80-05-7	789-02-6
Supplier	SigmaAldrich			Chem Service
Catalogue #	91834-10 MG	171468-5G	42088-100 MG	N-12708
MW (g/mol)	485.79	414.07	228.29	354.5
LogP	6.2a	6.3	3.3	6.7a
Nominal conc. (μM)	3b, 10, 30, 100, 300b			1, 3, 10, 30b
Human blood levels (μM)c	1.26E-04 – 1.40E-02	7.00E-03 – 4.58E-01	3.24E-02 – 2.91E-01	2.11E-04 – 2.59E-04
Bioavailable concentration in 2D media (μM)d	7.70E+00 – 1.95E+02	2.73E+00–8.20E+01	6.86E+00 – 2.06E+02	7.66E-01 – 2.30E+01
Bioavailable concentration in FMi-OOC media (μM)e	2.49E-01 – 2.49E+00	7.91E+00 – 7.91E+01	n/a	2.07E-01 – 2.07E+01

^aIndicates predicted LogP values.

^bIndicates the concentrations that were tested only in 96-well plates [2D]. See Methods for exact concentrations tested in each model.

^cRange in total reported human blood levels. See Supplemental Tables 2–3 for data sources and calculations.

^dBioavailable concentration in media in 96-well plates was predicted using the method of Armitage et al. [57].

^eBioavailable concentration in media in FMi-OOC was derived from analytical assays as detailed in Methods.

DECs were plated (10,000 cells per well in 90 μL of corresponding media) in a 96-well plate (FisherScientific, Cat no. 07-200-588) and incubated overnight to allow cells to adhere. Then, chemicals (BPA, PFOA, PBDE-47, or DDT-o,p’) were added in dose-response (Table 1) and plates returned to the incubator. After 24 or 72 h, medium was collected from each well for cytokine analysis and stored at −80 °C. Intracellular ATP concentration was measured in DECs as a marker for cell viability. The CellTiter-Glo reagent (25 μL; CellTiter-Glo assay kit; Promega, Madison, WI) was added to each well. Plates were first mixed for 2 min on a plate shaker to induce cell lysis and then incubated at room temperature for 10 min in dark. Luminescence signal was measured using FLIPR Tetra and Screenworks 4.0 (Molecular Devices, San Jose, CA).

2.3. FMi-OOC device fabrication

The four-chamber Feto-Maternal interface Organ-On-Chip (FMi-OOC) device was used in these studies [37]. Devices were fabricated using a two-step soft-lithography technique as previously described [38]. Briefly, devices were made from polydimethylsiloxane (PDMS, SYLGARD^™ 184 Silicone Elastomer Kit; Dow Chemical Company, Horgen, Switzerland) and treated with oxygen plasma (Harrick Plasma, Ithaca, NY; Cat no. PDC-001) for 90 s, followed by bonding the layer onto a glass substrate (VWR, Radnor, PA; Cat no. 48366–067). The process was repeated to bond additional PDMS layers together. When complete, each device consists of four concentric chambers connected by arrays of microchannels, which allows for culturing of four unique cell types separately while allowing biochemical communication between the cell culture chambers through the interconnected micro-channels. Specifically, the FMi-OOC contains 4 cell culture chambers each 250 μm in height, and the width of each chamber was designed to mimic the thickness of each maternal and fetal layer as seen in utero (maternal: decidua – 3000 μm; fetal: chorion – 2000 μm, AMC – 2000 μm, and AEC – 600 μm) [39,40]. The center chamber in this model contains DECs followed by CTCs, AMCs, and AECs. This design allowed the four different cell types to be cultured in four separate microenvironments (e.g., different culture media). FMi cells ratios were based on those seen in utero (AEC: 10 cells, AMC: 1 cell, CTC: 20 cells, and decidua: 1.7 cells; [39]) with slight adjustments made to accommodate device dimensions. The chambers are interconnected through an array of 24 microchannels (5 μm in height, 30 μm in width, and 300–600 μm in length). The microchannel arrays perform multiple independent functions, including: 1) preventing the flow of cells between compartments during the initial cell loading process, 2) allowing localized treatment of each cell layer with infectious or other stimulants while limiting their diffusion to the neighboring chambers, 3) enabling independent elution of supernatant from each cell chamber, and 4) allowing biochemicals to diffuse between chambers in a time-dependent way and also permit active cell migration that may involve cellular transitions.

2.4. Collagen and cell loading into FMi-OOC

Before use, devices were wiped with 70 % EtOH (FisherScientific, Cat no. BP82011), transferred to a biosafety cabinet and sterilized by injection of 70 % EtOH into each chamber for 15 min at room temperature. Then, PBS (FisherScientific, Cat no. AAJ61196AP) pre-warmed to 37 °C was used to rinse all chambers, and 25 % Matrigel (Corning; Matrigel Basement Membrane Matrix, DEV-free, Cat no. 356234, dissolved 1:25 in complete DEC media) was added to fill microchannels between the AEC-AMC chambers and between the CTC-AMC chambers. Active suction was used to facilitate Matrigel loading into microchannels. Then, the devices were incubated for 30 min at room temperature. Matrigel was removed and chambers were rinsed 2 × with warm media. Then, devices were filled with warm media and kept in a tissue culture incubator until cell loading.

Cells were detached from T75 flasks using 0.25 % Trypsin (ThermoFisher, Cat no. 12605036) and counted using the Cellometer^™ Auto T4 Plus (Nexcelom Bioscience, Lawrence, MA) to confirm live cell count and viability prior to loading. Cell loading was performed as detailed elsewhere [38] from the outer ring (AEC) to the inner ring (DEC) with the following cell numbers: outer ring (AEC), 120,000 cells in 20 μL of corresponding media; second ring (AMC), 62,500 cells with 20 % primary collagen (see below) and 25 % Matrigel in 35 μL of corresponding media; third ring (CTC), 200,000 cells with 5 % primary collagen and 25 % Matrigel in 35 μL of corresponding media; and the inner ring (DEC), 50,000 cells in 50 μL of corresponding media. Primary collagen from human amnion membranes was provided by the Menon Laboratory at UTMB, which was prepared as previously described [41]. Following cell seeding, devices were incubated at 37 °C with 5 % CO₂ overnight to allow cell attachment before any chemical treatment or imaging.

2.5. Chemical treatments in FMi-OOC

One day after cell loading, cell attachment in each chamber was confirmed by bright-field microscopy (Fig. 1) and media was removed from all chamber reservoirs. DEC media was used to prepare 1 × solution of each (BPA, PFOA, PBDE-47, or DDT-o,p’) test compound (100 μL) that was added to the center (DEC-containing) chamber of the FMi-OOC. Final test concentrations in FMi-OOC for PBDE-47 and DDT were 3, 10, and 30 μM; and for BPA and PFOA they were 10, 30, and 100 μM. Vehicle for all test chemicals was 0.25 % (v/v) DMSO. Effectively, chemical exposures were initiated through the maternal chamber of the FMi-OOC device to mimic maternal environmental exposures reaches the fetal-maternal interface through blood vessels in the decidua. All other reservoirs were filled (approximately 60 μL of media per reservoir) with fresh media corresponding to each cell type. Exposures and drug propagation were allowed to continue for up to 72 h. At the end of each experiment, medium was removed from each chamber and stored separately at −80 °C for further analyses. Cells in the devices were stained for immunocytochemistry as detailed below.

Fig. 1.

Study design and representative images of cells cultured in the FMi-OOC model and multi-well plates. (A) Schematic of study design for the experiments performed and the range of experimental conditions that were utilized in this study to evaluate the effects of different chemicals. Day −1 corresponded to the day when cells were seeded into multi-well plates (2D) or FMi-OOC devices (3D). Media changes with or without other analyses and the timing of chemical additions are indicated (see symbols legend in the inset). (B) Representative phase-contrast microphotographs of cells in 2D and 3D cultures on Day 0.

2.6. Immunocytochemical detection of necrosis and apoptosis

FMi-OOC chambers were gently flushed with warm PBS and then staining mix (prepared according to manufacturer’s protocol) was gently added into chambers. Cells were stained without prior fixation with a live/dead staining kit (Abcam, Waltham, MA; Cat no. ab176749). Cells were stained to differentiate cells being live (CytoCalcein Violet), or undergoing apoptosis (Apopxin Green), or necrosis (7-Amino-Actinomycin D). Devices were placed on a plate rocker for 1 h at room temperature. After incubation, PBS was gently flushed through cell chambers twice before fluorescent imaging at 4 × magnification (BZX800E, Keyence, Osaka, Japan). To visualize all device chambers in one image, the automated XYZ stitching software (Keyence) was used. Briefly, edge points were set to DEC and AEC chambers in one quadrant of the device and overlapping images at each wavelength were taken and stitched together to create one comprehensive image of the device.

Cell viability was quantified using open-source ImageJ (Fiji version) software. Cell chambers were isolated using the selection tool and cell coverage (using CytoCalcein stain) was measured as area of the isolated section. Using the same method, area of the cells that were stained positively for 7-Amino-Actinomycin D was measured as the area of cells stained for necrosis. The area of necrosis was then calculated by using Equation (1).

$$\%\mathit{\text{Cytotoxicity}}=\frac{\mathit{\text{Necrosis area}}}{\mathit{\text{Total cell coverage area}}}\times 100\%$$ [Eq. 1]

2.7. Analytical chemistry methods

Media samples collected from all experiments were stored at −80 °C until extraction and analysis. PFOA was analyzed using liquid chromatography-mass spectrometry method on a 1290 Infinity II LC (Agilent Technologies, Santa Clara, CA) coupled with a triple quadrupole mass spectrometer (Agilent, 6470A) that was operated in negative ionization mode. PBDE-47 and DDT were analyzed using gas chromatography-mass spectrometry methods on an Agilent GC (7890B) coupled with a triple quad mass spectrometer (Agilent, 7010B).

For analysis of PFOA, medium samples were prepared using extraction with a liquid-liquid protein precipitation method. Briefly, 100 μL of ice-cold acetonitrile (ThermoFisher, Cat. No. A998–4) containing internal standard (1 μM C¹³-PFOA; Wellington Labs, Guelph, ON, Canada) was added to a 50 μL aliquot of each sample. The samples were then vortexed and centrifuged, the supernatant was transferred to a prelabeled microcentrifuge tube and dried under vacuum. Samples were reconstituted in 50 μL of mobile phase A (5 mM ammonium acetate in water) and stored at −20 °C until analysis. For instrumental analysis, samples (5 μL) were injected onto a ZORBAX SSHD Eclipse Plus C18 column (3.0 × 50 mm, 1.8 μm, Cat. 959757–302, Agilent) equipped with a guard column (2.1 × 5 mm, 1.8 μm, Cat. 821725–901, Agilent) using a 1290 Infinity II LC (Agilent) and then analyzed using a triple quadrupole mass spectrometer (Agilent 6470A). Samples were analyzed in negative electrospray ionization mode with gas temperature, sheath gas temperature, and nebulizer pressure set to 130 °C, 375 °C, and 35 psi, respectively. Mobile phase composition started with 70 % mobile phase A (5 mM ammonium acetate in water) and 30 % mobile phase B (acetonitrile) until minute 3.5, then changed to 10 % A and 90 % B. At minute 4.5, the amount of mobile phase B increased to 100 %, The gradient was changed to 60 % B at minute 5. The gradient was returned to initial conditions at minute 5.5 and remained until the end of sample analysis at 7 min.

PBDE-47 and DDT samples were prepared for analysis via GC-MS/MS as follows. Each sample (50 μL) was mixed with 10 μL of internal standard (Terbutryn, Sigma-Aldrich, Cat no. 886-50-0) and 200 μL of 1:1 (v/v) pentane:diethyl ether. Samples were then homogenized 3 times in 30 s increments using an Omni Bead Rupter 24 (Omni International, Kennesaw, GA). Supernatant was then transferred to glass vials with inserts (Ibis Scientific, 4400-FIV2W) and dried down with nitrogen at room temperature. Samples were stored at −20 °C until analysis and immediately following analysis, punctured vial caps were replaced. Samples were analyzed using splitless injection and ionized with electron ionization, with electron multiplier voltage set at 1884 V. A total of 1 μL (for DDT) or 2 μL (for PBDE-47) of each extracted sample was injected onto a VF-5ms GC column (60 m × 250 μm × 0.25 μm, Cat No: CP8960; Agilent Technologies). Sample runtime for DDT was 42.1 min with column head pressure of 21.5 psi and flow rate of 1.2 mL/min with helium gas. The temperature ramp up was as follows: initial temperature of 70 °C for 5 min, increased to 150 °C at 50/min, then ramped to 280 °C at 4C/min and held for 15 min. The injector temperature, ion source, and auxiliary transfer line was set to 250 °C, 300 °C and 300 °C, respectively. For analysis of PBDE-47, sample run time was 15 min with column head pressure of 18.5 psi and a flow rate of 1 mL/min with helium gas. Initial oven temperature was 80 °C and then ramped to 230 °C at a rate of 37 °C/min and then at 30 °C/min until reaching 325 °C where temperature was held until the end of analyses. Ultra-high purity nitrogen was used as the collision gas for all MS/MS experiments, with collision gas pressure was set at 16.8 psi.

2.8. Cytokine analyses

A multiplexed cytokine assay (Millipore, Cat no. HCYTA-60K, HCYTA-60K- PX38, HCYTA-60K-PXBK38, HCYTA-60K-PX48, HCYTA-60K-PXBK48) for detection of Interleukin 8 (IL-8), Interleukin 6 (IL-6), Interleukin 10 (IL-10), Granulocyte-macrophage colony-stimulating factor (GM-CSF), and Tumor-necrosis factor-α (TNFα) was used to analyze the amount of each cytokine in the cell culture supernatant. Assays were performed on the supernatant samples (25 μL) collected from individual culture chambers following 48 and 72 h of exposure according to the manufacturer’s instructions. Standard curves of known quantity of each protein provided by the vendor were prepared alongside samples. Instructions for preparation were provided by vendor and plates were washed using a BioTek 405 TS (405TSUVS, Agilent Technologies, DE) and analyzed using LX200 (Luminex, Austin, TX). Sample concentrations were determined by relating the absorbance values that were obtained to the standard curve by linear regression analysis.

2.9. Gene expression library preparation and sequencing

Cell lysates were collected from frozen stocks of 50,000 (DEC), 100,000 (CTC), 62,500 (AMC), and 120,000 (AEC) cells per 50 μL media and mixed with 50 μL of TempO-Seq^™ Enhanced Lysis Buffer (BioSpyder Technologies, Carlsbad, CA; Cat no. SU-01–100) in microcentrifuge tubes. Lysates were incubated for 10 min at room temperature to allow for complete cell lysis. The lysates were stored at −20 °C until later analysis. The Templated Oligonucleotide Sequencing Assay human S1500+ targeted transcriptome assay, which includes 2982 unique probes for human transcripts (BioSpyder Technologies; Cat no. KT-01–96), was used for mRNA quantitation [42,43]. The detailed protocol for TempO-seq is provided by the manufacturer and was previously described [44]. Briefly, the mRNA content of the cell lysates was hybridized by incubating 2 μL of the cell lysate with 2 μL of hybridization mix. Excess oligonucleotides were then digested in a nuclease-catalyzed reaction and the hybridization products were incubated with DNA ligase, followed by heat denaturation of nuclease and ligase. A total of 10 μL of each ligation product was then mixed with an equal volume of PCR amplification mix and amplified in a LightCycler 96 (Roche, Basel, Switzerland) using the manufacturer-recommended settings. Subsequently, 5 μL of each amplified sample was pooled (up to 96 samples in one pool) and purified using a commercial PCR clean-up kit (Clontech, Mountain View, CA, USA). The pooled libraries were sequenced with custom primers on a NovaSeq X Plus 1.5B sequencer (Illumina, San Diego, CA).

2.10. Gene expression data analysis

Raw sequencing reads for each pooled sample were combined to generate a single FASTQ file per sample. Next, Fastp (version 0.21.0) [45] was used to trim sequencing reads to 50 base pairs using default parameters. These FASTQ files were then processed using the TempO-Seq data analysis pipeline [42] with the corresponding human S1500+ probe manifest file (BioSpyder Technologies). The FASTQ files are available from the Gene Expression Omnibus (accession #GSE295995).

Raw count matrix data was first aggregated to the gene level when a gene (HGNC symbol) was associated with more than one probe in the TempO-Seq library. Prior to transcriptomic analysis, quality control was applied to the raw counts using the following criteria: (i) samples were excluded if the percentage of uniquely mapped reads was below 50 %, (ii) samples were excluded if total mapped reads were fewer than 200,000, (iii) potential outliers were identified using principal component analysis (PCA) with the prcomp function in R, and (iv) genes were excluded if they had counts in fewer than 10 % of all retained samples. The final dataset included 9 samples (n = 3 for each of DEC, AMC, and AEC) and 2732 genes. This dataset was processed using the DESeq2 package in R [46] for normalization, data visualization, and differential gene expression analysis across cell types.

To compare gene expression profiles with those from the previously published studies, transcriptomic data for the same cell types was downloaded from Gene Expression Omnibus (GSE206401) [47]. Ensembl IDs were converted to HGNC symbols using the org. Hs.eg.db package in R, and counts were aggregated for HGNC symbols associated with more than one Ensembl ID. To enable cross-study comparison, only genes that matched those used in the current study and had counts in more than 10 % of tested samples were retained. This dataset was also processed using DESeq2 for normalization and differential gene expression analysis across cell types.

For both datasets, significantly differentially expressed genes (DEGs) were identified for each pairwise comparison (AEC vs. AMC, AMC vs. DEC, and AEC vs. DEC) using the results function in DESeq2, with contrasts specified by cell type. Significance thresholds were set at a false discovery rate (FDR) q-value <0.01 and an absolute log₂-fold change >3. The union of DEGs from all pairwise comparisons within each dataset was used for heatmap visualization and gene clustering, based on z-scaled converted normalized counts, using the pheatmap package in R. Enrichment analysis of gene clusters was performed using the clusterProfiler package [48] to identify associated Reactome pathways. The gene set background included all interrogated genes retained in each dataset, as defined above. Comparisons between the two studies were conducted by: (i) computing Spearman correlation coefficients for each cell type across the datasets, and (ii) identifying overlapping DEGs that were most highly expressed in AEC, AMC, or DEC, along with their associated significant Reactome pathways.

2.11. Statistical analysis

Data analysis for all data other than gene expression results (see above) was performed using GraphPad Prism (version 10.3.1, San Diego, California). Data was analyzed using 2-way ANOVA with post hoc testing as indicated in the figure legends.

3. Results

While maternal-fetal transfer of environmental pollutants and drugs is fairly well understood [49], the effects of environmental exposures on the maternal-fetal membranes that are involved in parturition are far less understood [50]. During pregnancy, the fetal membrane (amniochorion) lines the intrauterine cavity (maternal decidua) and provides mechanical, immune, and endocrine support to the growing fetus [51]. Studies of such multi-cellular interfaces are difficult when using traditional in vitro models; therefore, we used the FMi-OOC chip that has four concentric circular cell culture compartments for maternal and fetal cells (Fig. 1) [37]. Each chamber of the chip was designed to mimic the thickness of each maternal and fetal layer (decidua, chorion, and amnion [mesenchyme and epithelium]) as seen in utero. First, the effects of selected chemicals were tested on maternal decidua cells using 96-well plates (2D) because this is the site of chemical addition for the experiments in FMi-OOC (3D). Second, selected concentrations were used for the experiments in FMi-OOC, devices that were seeded with four different cell types in separate microenvironments (different culture media) (Fig. 1A). Representative phase-contrast microphotographs (Fig. 1B) show morphology of each cell type in 96-well plates and in FMi-OOC, which were consistent with previous reports using these cells and chip [37,52–54].

To characterize the molecular state of the cells used in these experiments, we conducted gene expression profiling using a targeted S1500+ gene set designed to represent multiple human tissues and also include toxicologically relevant genes [43]. Fig. 2 shows the results of these experiments using frozen cell stocks (i.e., untreated cells) that were further expanded and used in both the 2D and 3D experiments. CTC were not included in these analyses because of an unacceptably low mapped read counts for this targeted gene expression assay; this is not surprising given that CTCs are very different in their gene expression profile from other placental cells and other adult tissues [55]. Other cell types – DEC, AMC and AEC – demonstrated considerable differences among their transcriptomes as shown by the Principal Component Analysis (Fig. 2A). Our gene expression data was also compared to the previous bulk RNA sequencing data on these cells and related human tissues [47]. We found that each cell type we used in our study was highly correlated with gene expression data from the same cell type published previously (Fig. 2B). Also, the transcriptomes of these cells were less corelated with BeWo cells or human fetal membrane or placenta tissues. Based on the differential gene expression analysis (genes that were significantly different between each pair of cells used herein) performed on our data and that from Kammala et al. [47], we find genes that were over-expressed in each cell type as shown by clusters in the dendrogram (Fig. 2C). From these analyses, we identified the individual transcripts and pathways (using the Reactome gene set) that were unique or overlapping between the two datasets (Fig. 2D and Supplemental Table 1A–B). These results show that about one-half of the individual transcripts were found to be over-expressed for each cell type in both studies, which is not unexpected given the different gene expression profiling methods. However, at the pathway level, the overlap was pronounced for both AEC and AMC, but not for DEC.

Fig. 2.

Gene expression analyses of untreated AECs, AMCs, and DECs. (A) PCA of gene expression data. Colors and symbols depict cell types as indicated in the legend. (B) Rank-based (Spearman) correlation between transcriptional profiles of the tested cells (depicted in different colors and symbols as indicated in Panel A) and representative human tissues and same cell types examined in a previous publication [data from Kammala et al., 2022]. (C) A supervised heatmap visualizing expression of 291 (left panel) and 251 (right panel) genes that were differentially expressed among these cell types (see Methods and Supplemental Table 1). Colors depict Z-score values (for each transcript). Cell types (columns) are indicated on the plots. (D) Venn diagrams showing overlapping genes (left plots) and pathways (right plots) for each of the three distinct gene clusters indicated in Panel C. Left circles represent data from this study and right circles are those from a previous publication (Kammala et al., 2022).

In subsequent studies, four representative environmental pollutants – PBDE-47, PFOA, BPA and DDT – substances that are generally regarded as persistent, bioaccumulative, and toxic, were used (Table 1). We selected a range of concentrations for these studies to be commensurate with the concentrations that have been tested previously in other large-scale in vitro studies [56]. While the nominal concentrations selected for testing in our study may appear high, it is important to evaluate the bioavailable concentrations of these chemicals in each in vitro assay. To do so, the bioavailable concentrations in both 96-well plates and the FMi-OOC system were first assessed. For the 96-well plates, the in vitro mass balance model developed by Armitage et al. [57] were used to estimate the distribution of chemical concentrations in the media. Based on the assay design, the model inputs included 5 % FBS and 10,000 cells. However, due to the lack of cell content data for the DECs, default cellular parameters from the Armitage model were used. The model simulations for the Armitage model were conducted using Excel tools provided by Armitage and colleagues. Detailed model inputs and the Excel tool used for the simulations are provided in the Supplementary Materials. For the FMi-OOC system, concentrations were experimentally measured as described in the Methods section (Supplemental Fig. 1). Both the predicted media concentrations in the 96-well plates and the measured concentrations in the FMi-OOC system suggest that the total bioavailable concentrations of these chemicals are approximately 3–10 times lower than the nominal concentrations (Supplemental Fig. 2 and Supplemental Tables 2–3). In addition to comparing the in vitro bioavailable concentrations, human serum/plasma concentrations were also collected for reference. Reference doses (mg/kg-day) for the four target chemicals were obtained and converted into steady-state plasma concentrations (μM) using the calc_analytic_css function from the httk R package (ver. 2.3.1). Detailed information and references are summarized in Supplemental Table 3. When compared to human-relevant concentrations, the in vitro bioavailable concentrations were approximately 1–3 times higher than the reported or predicted levels in human blood, depending on the chemical (Supplemental Fig. 2).

All four compounds were first tested in 2D on DECs, because these cells represent the maternal compartment and also because the chemicals are added in the DEC chamber of the FMi-OOC. Fig. 3A shows that all compounds exhibited effects on cell viability, being most pronounced at the highest concentrations tested. Pro-inflammatory cytokines play a key role in decidua’s involvement in parturition [58]. Therefore, we evaluated secretion of various pro-inflammatory cytokines, including IL-6, IL-8, GM-CSF, and TNFα upon exposure to test chemicals (Fig. 3B). Statistically (p < 0.05) and biologically (greater than 2-fold) induction of IL-6 was observed at 72 h timepoint for all compounds at concentrations lower than those causing cell death. Response to PBDE-47 included induction of GM-CSF and IL-8, also most pronounced at 72 h after treatment. Reduced levels of cytokines were observed at concentrations approaching cytotoxicity indicating that cell death, not pro-inflammatory response, was predominant.

Fig. 3.

Dose-response effects of test chemicals on DECs cultured in 96-well plates after 24 and 72 h of exposure. (A) Effects on cell viability. (B) Effects on cytokine (as indicated) release into cell culture media. Data is visualized using colored heat-maps according to the legend beneath each column. Raw data is included as Supplemental Table 4. Asterisks (*, p < 0.05; **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001) indicate statistically significant pairwise differences between groups (asterisks inside each colored cell are for the difference from vehicle [VEH]-treated cells and asterisks to the right of the heatmap are for a comparison between time points at the same concentration) from two-way ANOVA followed by Tukey’s multiple comparisons test. Red edges indicate that an increase in a respective cytokine level was greater than 2-fold as compared to that in vehicle-treated cells.

Based on these 2D studies, we selected concentrations in the lower range of those tested in 2D for the experiments in FMi-OOC (Table 1). Because the FMi-OOC is made of PDMS, material that is known to both adsorb and absorb drugs and chemicals [59], we first evaluated binding of tested chemicals in the cell-free devices (Supplemental Table 2). Because of increased binding to PDMS as compared to 96-well plates, our test concentrations are even more human relevant (Table 1) because they are only about 1–2 orders of magnitude greater than reported human blood levels. We also found that test compounds are largely confined to the DEC chamber in the FMi-OOC (Supplemental Fig. 1), the finding commensurate with previous reports and the intentional design of this chip to limit passive diffusion of drugs and chemicals between chambers [37,52–54].

Fig. 4 shows representative combination images of the live-dead stain in the FMi-OOC for each chemical tested at the highest nominal concentration. Some effects on cell viability were observed in the fetal chambers, with quantification of cell viability shown in Figs. 5–8 for each chemical. Both PFOA and DDT elicited effects on cell viability in CTC and AMC chambers, albeit the effects were significant only for PFOA.

Fig. 4.

Representative fluorescent microscopy images showing the effect of chemical exposures on cell viability in the FMi-OOC. Left panel shows cells 48 and 72 h after treated with vehicle control (0.25 % DMSO). All four chambers of the FMi-OOC are shown and labeled with the cell type seeded into each chamber. To visualize all four chambers in one image, individual images were combined as detailed in Methods. Chemical treatment of DECs within the FMi-OOC induced cell death (red fluorescence). Cell stain: CytoCalcein Violet 450 (blue, nuclei), 7-amino-Actinonmycin D (red, dead cells). Scale bars are shown to demonstrate the length of the micro-channels between each chamber in the device (marked white on the images).

Fig. 5.

Dose-response effects of PBDE-47 on four cell types in the FMi-OOC after 48 and 72 h of exposure. (A) Effects on cell viability. (B) Effects on cytokine (as indicated) release into cell culture media. Data is visualized using colored heat-maps according to the legend beneath each column. Raw data is included as Supplemental Table 4. Asterisks (*, p < 0.05; and **, p < 0.01) indicate statistically significant pairwise differences between groups (asterisks inside each colored cell are for the difference from vehicle [VEH]-treated cells and asterisks to the right of the heatmap are for a comparison between time points at the same concentration) from two-way ANOVA followed by Tukey’s multiple comparisons test. “X” mark test conditions with no data. Red edges indicate that an increase in a respective cytokine level was greater than 2-fold as compared to that in vehicle-treated cells.

Fig. 8.

Dose-response effects of DDT on four cell types cultured in FMi-OOC after 48 and 72 h of exposure. (A) Effects on cell viability. (B) Effects on cytokine (as indicated) release into cell culture media. See Fig. 5 for data visualization heatmap, statistical significance and other symbol legends. Raw data is included Supplemental Table 4.

Next, we quantified pro-inflammatory cytokine levels in the media from each chamber collected at 48 or 72 h after treatment. PBDE-47 (Fig. 5) induced cytokine release in DEC, CTC, and AMC chambers. TNFα increases were observed at both time points, most pronounced at 72 h, but only in the DEC chamber. In the CTC chamber, IL-6 was the early response and GM-CSF a late response to treatment. Greater than 2-fold induction of IL-6 was also observed in the AMC chamber. However, no apparent responses were observed in the AEC chamber. PFOA (Fig. 6) had effects on all four cytokines across all chambers. TNFα and IL-8 increases were observed in all chambers, predominantly in early responses. IL-6 was also induced in the DEC and CTC chambers early and in the AEC chamber later. GM-CSF response was most pronounced in the DEC chamber, with some increases in CTC and AMC, but not AEC. Effects of BPA (Fig. 7) were also robust across all chambers, with most prominent responses seen at both time points in the AEC chamber. It is also noteworthy that these effects in the AEC chamber were observed across all concentrations, time points, and cytokine types. Also, the fold increase in the cytokines in response to BPA was the greatest among all 4 chemicals tested. DDT effects (Fig. 8) were most pronounced for IL-6 and IL-8, observed in all 4 chambers of the FMi-OOC. No induction of TNFα was observed and GM-CSF increases were confined to the DEC and CTC chambers. Table 2 shows a summary of the chemical effects on different cell types in 2D and FMi-OOC testing, with “up” arrows showing effects on cell death or induction of cytokine release.

Fig. 6.

Dose-response effects of PFOA on four cell types cultured in FMi-OOC after 48 and 72 h of exposure. (A) Effects on cell viability. (B) Effects on cytokine (as indicated) release into cell culture media. See Fig. 5 for data visualization heatmap, statistical significance, and other symbol legends. Raw data is included Supplemental Table 4.

Fig. 7.

Dose-response effects of BPA on four cell types cultured in FMi-OOC after 48 and 72 h of exposure. (A) Effects on cell viability. (B) Effects on cytokine (as indicated) release into cell culture media. See Fig. 5 for data visualization heatmap, statistical significance and other symbol legends. Raw data is included Supplemental Table 4.

Table 2.

Summary of the chemical effects observed in these studies. Shown are effects of chemicals on induction of cell death or cytokine release into the media. An up arrow is added if either statistically (p < 0.05) or biologically (>2-fold increase) significant responses were observed as shown in Figs. 3 and 5–8. Shaded columns indicate studies in 2D model as compared to FMi-OOC model.

	Cell type & model	DEC (2D)		DEC (3D)		CTC (3D)		AMC (3D)		AEC (3D)
	Time of exposure, h	24	72	48	72	48	72	48	72	48	72
PBDE-47	Cell death	↑	↑
	GM-CSF	↑	↑	↑			↑
	IL-6		↑			↑			↑
	IL-8		↑
	TNFa			↑	↑
PFOA	Cell death	↑	↑			↑
	GM-CSF			↑			↑	↑
	IL-6		↑	↑		↑					↑
	IL-8				↑	↑		↑		↑	↑
	TNFa			↑		↑		↑		↑	↑
BPA	Cell death	↑	↑
	GM-CSF			↑					↑	↑
	IL-6		↑	↑		↑			↑	↑	↑
	IL-8			↑	↑	↑		↑	↑	↑	↑
	TNFa			↑		↑	↑	↑	↑	↑	↑
DDT	Cell death	↑	↑
	GM-CSF				↑	↑
	IL-6		↑	↑	↑				↑	↑	↑
	IL-8						↑		↑	↑	↑
	TNFa

4. Discussion

The mechanisms of both physiological and pathological shift from the quiescent state of pregnant uterine tissues, one that is maintained through autocrine-paracrine actions of inhibitors of uterine contraction, to myometrial activation and ultimately post-partum involution, are highly complex and involve multiple mediators and molecular pathways [60]. Preterm parturition is often associated with rupture of the fetal membranes (i.e., amnio-chorion), a complication that occurs in 40 % of all spontaneous preterm births [61]. It is well accepted that preterm labor may be initiated and propagated by inflammation [62] and there is bi-directional link between prostaglandins, key mediators of labor through induction of progesterone withdrawal and initiation of parturition [63]. The fetal membranes and maternal decidua produce cytokines that induce COX-2 leading to prostaglandin release and onset of labor [64]. If this inflammatory cascade happens before term, for example due to infection or chemical exposure, it can trigger birth before full development occurs [65]. Environmental chemicals can act via oxidative stress and induce inflammation, common key characteristics for many organ-specific pathological states [66]. They also have been strongly associated with preterm birth [67].

Epidemiological studies of potential environmental causes of preterm birth have been suggestive of the association, but they are also highly heterogeneous [3,9]. There is a strong consensus that additional studies are needed to improve understanding of the cellular and molecular mechanisms underlying reported associations between environmental exposures and preterm birth [67]. Studies in cell-based models can be particularly informative and can provide additional evidence for causality that is lacking in observational studies. They can aid in disaggregating preterm birth into more granular molecular phenotypes and cell drivers, improving our ability to explain associations with exposure to specific compounds. Therefore, in this study we have taken advantage of a unique four-chamber FMi-OOC model to study effects of maternal exposure on fetal membranes that are responsible for maintaining and initiating parturition, namely decidua, chorion, and amnion. This model has been previously developed and used to demonstrate mechanistic sequelae of maternal exposure to various noxious compounds, specifically the role of the paracrine cytokine-mediated signaling [37,52,53].

As detailed in the Introduction, the four toxicants selected herein are both persistent, bioaccumulative, and toxic, and have been associated with preterm birth. PBDE-47, DDT, and PFOA are classified as persistent organic pollutants by the United Nations Environment Programme [68]. BPA is not on the Stockholm Convention list [69]; however, it has been shown to bioaccumulate and is widely detected in humans [70]. It is also considered to be a classical endocrine disruptive compound [71], and has been associated with preterm birth [14]. Still, there are many studies that failed to demonstrate associations with preterm birth for these compounds [18,22,27–29], a strong rationale to conduct cell-based studies to determine if they can indeed elicit pro-inflammatory cascades. Human blood levels for all four test compounds have been reported from different sub-populations, and they have been shown to vary widely. For our studies, we initially selected chemical concentrations spanning 1.5–2 orders of magnitude, reaching fairly high to micromolar levels. We first conducted a range-finding study to ensure that while the site of exposure (i.e., DEC) concentrations are in a wide range they are not overtly cytotoxic – effectively adopting the “maximum tolerated dose (MTD)” concept to these in vitro studies. Also, while the nominal concentrations were high, the bioavailable fractions were lower in both 96-well plates and especially in the FMi-OOC. Because of the expected partitioning of the compounds in cell-based model systems [72], and the fact that human biomonitoring studies typically report steady-state total circulating blood levels [11], the extrapolations from in vitro concentration to human exposures is a challenging task [73]. Even though the comparison of the biomonitoring equivalents in blood and urine for benchmarking against in vitro data is beyond the scope of this study, we show that tested concentrations are within 1–2 orders of magnitude from the reported blood levels in the general population. Thus, our studies can be deemed as relevant to inform interpretation of the findings in epidemiological studies, especially where blood levels of these chemicals are reported.

With respect to the acute cytotoxic effects of the tested compounds, all of them exhibited effects at the highest concentrations in the experiments with DECs in 96-well plates. This allowed us to select concentrations below the MTD for subsequent studies to distinguish between overt toxicity and other biologically meaningful effects. For example, we observed that release of IL-6, a general stress marker and not necessarily indicative of any risk, was robust in DECs exposed to all chemicals in the 96-well plate study, especially at concentrations that were noncytotoxic. This finding is commensurate with the reports of biomarkers of inflammation observed in maternal blood in association with exposure to environmental chemicals [22,29]. For PBDE-47, we observed a more generalized pro-inflammatory response in the form of increases in GM-CSF, IL-6, and IL-8 building up by 72 h after exposure, suggesting a pronounced decidual response. This finding is commensurate with the observations of a positive relationship between serum concentrations of PBDE-47 and serum concentrations of pro-inflammatory cytokines in pregnant women [74].

While studies in 96-well plates allow for testing direct effects of chemicals on cells, this method does not reflect the anatomy of the extra-placental feto-maternal membranes. Outside of the placenta, maternal blood is circulating through the decidua, while fetal chorion and amnion are largely avascular and thus chemicals and biological signals need to be transmitted by active or passive transport mechanisms [75]. Therefore, the physical separation of different cell types afforded by the FMi-OOC allows for studies that best reflect potential paracrine signaling among cells. We found that minimal direct propagation of test compounds from the decidua chamber to fetal chambers occurs. This observation is similar to our previous reports with cigarette smoke extract [41,54], cadmium [52], and lipopolysaccharide (LPS) [37]. Also, we found that while no cytotoxicity was observed in the DEC chamber with either compound, some effects were observed in the CTC chamber, indicating a biological effect on the cells that was not due to a direct chemical effect. Interestingly, treatment-associated increases in cytokine production were observed for all compounds; PFOA and BPA showed the strongest effects, with AEC being the most responsive cell type despite being the farthest removed from the site of chemical addition in the FMi-OOC device.

For PBDE-47, we found release of pro-inflammatory cytokines in both decidua and chorion in the absence of cell death. The chorion barrier remained intact, resulting in minimal amnion inflammation, with IL-6 increase only being observed in the AMC chamber. Thus, we conclude that PBDE-47 likely had a direct effect on the decidual cells, inducing sufficient decidual inflammation to affect the chorion, and to a lesser extent, the amnion. Despite this chorion barrier retention, fetal membrane inflammation was still relatively more pronounced than that observed for PBDE-47 in the placenta OOC [76], suggesting that fetal membranes may be more susceptible to pollutant-induced dysfunction.

For PFOA, we observed less decidual cell death compared to 2D cultures, suggesting improved resilience when cells are part of a full tissue context in the FMi-OOC. We also found that PFOA induced pronounced decidual and chorion inflammation, with chorion cell death correlating with downstream amnion inflammation (i.e., increases in pro-inflammatory cytokines in downstream AMC and AEC). This response profile is consistent with feto-maternal membrane inflammation, a phenomenon reported in several human studies [77,78]. Still, it remains to be established whether this preterm birth-like phenotype leads to preterm birth.

For BPA, we found that decidual cells were more resistant in the FMi-OOC as compared to 2D cultures. Minimal maternal inflammation, but pronounced chorion-amnion response, were observed. Chorion cell death correlated with increased amnion inflammation (increases in all pro-inflammatory cytokines) were found in both AMC and AEC chambers. These data suggest that chorion barrier integrity is key, where once breached, it permits strong amnion responses, also echoing a preterm birth-like phenotype. Indeed, our previous review identified several studies suggesting possible propagation of inflammation in the fetal membranes in response to BPA [79]. The finding that BPA elicited a robust response similar to PBDE-47 is also indicative that the effects of these chemicals on the extra-placental feto-maternal membranes may be different from that on the placenta. In fact, our previous studies with explant placentas [80] or placental OOC [76] showed a less pronounced response to BPA.

For DDT, we also observed that decidual cells were more resilient in the FMi-OOC as compared to 2D culture. DDT also induced cell death in fetal chorion and amnion chambers. The inflammatory responses that were observed were likely overshadowed by widespread cellular collapse, pointing toward a stillbirth-like phenotype rather than a preterm birth model. Human epidemiological studies reporting positive associations between exposure to DDT congeners and adverse reproductive outcomes are observational in nature and suffer from small sample size [81,82]; therefore, our findings shall be interpreted with caution even if they provide mechanistic plausibility to the reports of stillbirth.

Overall, this study highlights several key differences in how environmental toxicants may impact fetal tissues when comparing traditional 2D cell culture to a more physiologically relevant FMi-OOC model. In 2D cultures, chemical toxicity appeared exaggerated, particularly in maternal decidual cells, whereas the FMi-OOC allowed detection of downstream effects on fetal membranes, showing greater resilience and tissue-specific responses. While these differences could be explained, in part, by the higher bioavailability of the chemicals in 2D cultures due to lower adsorption to plastic, they also could reflect the paracrine signaling among cell types that are present together in the FMi-OOC. Indeed, immune dysregulation at the FMi, whether from infection, oxidative stress, or chemical insults, may provoke pro-inflammatory cytokine production and immune cell infiltration. Maternal exposure s may initiate the process that can propagate through the fetal membrane layers, overcome the initial resistance from the chorion trophoblast cells, and reach the highly vulnerable amnion layer to initiate the vicious cycle of events (fetal inflammatory response) as shown in this study, collectively manifesting in adverse pregnancy outcomes. While an intricate interplay of the maternal and fetal immune systems is maintained in normal pregnancies, the decidua can function as an amplifier of biochemical and inflammatory signals creating immune intolerance leading to preterm birth [83]. In this regard, present study, together with our previous work with FMi-OOC [37,41,52,54], shows that each cell type exhibits a unique immune-related response to chemical and biological stimuli. Specifically, this OOC model restricts full propagation of the noxious stimulants present in the maternal compartment, replicating the in utero barrier properties of cells on FMi-OOC. This model creates a more physiological condition where various damage-associated molecular patterns released by stressed cells can propagate through the compartments and simulate intricate and complex cell-cell and cell-matrix interactions at the maternal-fetal interface [84]. Still, because parturition is a very complex process dependent on shifts in homeostatic balance in multiple systems, including the immune system of both the mother and the fetus, our understanding of its mechanisms and responses to various environmental stimuli is incomplete. Additional studies are needed to characterize the proteome and metabolome of the paracrine signals between the chambers in the FMi-OOC to fully determine the mechanistic linkages between exposures and preterm birth.

Overall, this study found that the chorion’s integrity was critical in terms of the chemical effects on downstream amnion cell types. When intact, the amnion was protected, but once compromised, inflammation appeared to propagate downstream, suggesting that chorion breakdown may be a key trigger in preterm birth-like phenotypes. The fetal membranes, especially the chorion and amnion, exhibited more sensitivity to toxicants like BPA, PBDE-47, and PFOA compared to the placenta, underscoring their relevance as targets for pollutant-induced dysfunction studies. Notably, inflammation and cytokine responses were often more pronounced in fetal membranes than in maternal or placental tissues, identifying them as primary sources of inflammatory signaling in this model. Also, our findings may support a “two-hit hypothesis” for fetal responses to chemical exposures during pregnancy: pollutants may reach the amniotic environment both by crossing the placenta and by diffusing from maternal decidua to fetal membranes, reinforcing the need to consider multiple exposure pathways in research of the mechanisms of preterm birth.

Abbreviations:

DEC

Human maternal decidua cells

CTC

Human chorion trophoblast cells

AMC

Human amnion mesenchymal cells

AEC

Human amnion epithelial cells

FMi-OOC

Feto-Maternal interface Organ-On-Chip

DDT-o,p’

dichlorodiphenyltrichloroethane

BPA

bisphenol A

PBDE-47

2,2′4,4′-tetrabromodiphenyl ether

PFOA

Perfluorooctanoic acid

DEGs

Differentially expressed genes

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.cbi.2025.111782.

Data availability

All numerical data included in this manuscript is included as Supplemental Tables. Gene expression data has been deposited into Gene Expression Omnibus (accession #GSE295995).