Amnion membrane organ-on-chip: an innovative approach to study cellular interactions

Abstract

The amnion membrane that lines the human intrauterine cavity is composed of amnion epithelial cells (AECs) connected to an extracellular matrix containing amnion mesenchymal cells (AMCs) through a basement membrane. Cellular interactions and transitions are mechanisms that facilitate membrane remodeling to maintain its integrity. Dysregulation of cellular remodeling, primarily mediated by oxidative stress (OS), is often associated with preterm birth. However, the mechanisms that maintain membrane homeostasis remain unclear. To understand these mechanisms, we developed an amnion membrane organ-on-chip (AM-OOC) and tested the interactive and transition properties of primary human AECs and AMCs under normal and OS conditions. AM-OOC contained 2 chambers connected by type IV collagen–coated microchannels, allowing independent culture conditions that permitted cellular migration and interactions. Cells grown either independently or coculture were exposed to OS inducing cigarette smoke extract, antioxidant N-acetyl-l-cysteine (NAC), or both. When grown independently, AECs transitioned to AMCs and migrated, whereas AMCs migrated without transition. OS caused AECs’ transition but prevented migration, whereas AMCs’ migration was unhindered. Coculture of cells facilitated transition, migration, and eventual integration in the contiguous population. OS cotreatment in both chambers facilitated AECs’ transition, prevented migration, and increased inflammation, a process that was prevented by NAC. AM-OOC recapitulated cellular mechanisms observed in utero and enabled experimental manipulation of cells to determine their roles during pregnancy and parturition.—Richardson, L., Jeong, S., Kim, S., Han, A., Menon, R. Amnion membrane organ-on-chip: an innovative approach to study cellular interactions.

Fetal membranes (amniochorionic membrane or placental membranes) are the innermost lining of the intrauterine cavity that surrounds the fetus and provides mechanical and immune protection throughout gestation. Membrane homeostasis is vital for the maintenance of pregnancy and fetal growth. Compromise in the fetal membrane’s structural (1), biologic (2), and mechanical (3) functions or chorioamniotic inflammation (4) are often associated with spontaneous preterm birth and preterm premature rupture of the fetal membranes, 2 major complications of pregnancy that affect more than 9.6% of all cases in the United States alone and 11% worldwide (5–7). However, mechanisms that maintain the fetal membrane’s homeostasis during gestation and factors contributing to the loss of its functional ability, which could predispose membranes to labor-associated inflammatory changes at term (physiologic) or preterm (pathologic), are still unclear. A clear understanding of these mechanisms may help fill a major knowledge gap regarding the role of fetal membranes in term and preterm labor, as well as lead to designing better strategies to reduce membrane-associated adverse outcomes.

Fetal membranes are multilayer structures comprising amnion epithelial cells (AECs) and chorion trophoblasts connected by a collagen-rich extracellular matrix (ECM) that also contains amnion mesenchymal cells (AMCs) (8–10). The amnion membrane is the most elastic component of the fetal membranes and is composed of an AEC type IV collagen–rich basement membrane with AMCs embedded in ECM (11). Additionally, the amnion membrane bears a majority of the tensile strength to mechanically keep the tissue intact throughout gestation (3, 12, 13). AECs and AMCs also provide essential immune and endocrine functions that are critical to the maintenance of pregnancy (14–16). Membrane growth and remodeling are essential during gestation and involve both cellular transitions and collagenolytic matrix turnover (8, 17). Cellular-level changes primarily involve AECs and AMCs that are conventionally considered as purely epithelial and mesenchymal in physiognomies. However, recent findings suggest that AECs and AMCs are pluripotent stem cells in a metastate in which they coexpress both epithelial and mesenchymal markers (17, 18). This metastate is thought to allow amnion membrane cells to readily undergo cellular transitions demanded by intrauterine microenvironmental cues to either promote membrane remodeling and maintain integrity during gestation or predispose them to weakening in preparation for labor and delivery (8, 17).

The recent discovery of fetal membrane microfractures (MFs) highlights possible areas of such cellular transitions and remodeling (1, 7, 18). MFs are biologic interruptions in the amnion membrane characterized by AEC puckering or gaps, basement membrane degradation, and tunnels that extend into the collagen matrix with migrating cells (1). Increased number and morphometry (width and depth) of MFs at term and in preterm birth and preterm premature rupture of the fetal membranes suggest that persistence of MFs may indicate lack of remodeling and membrane dysfunction (1, 7). Induction of oxidative stress (OS) in vitro in fetal membrane explants, similar to that seen at term and preterm parturitions, increases MFs, their morphometry, and their collagenolysis, supporting the hypothesis that the persistence of MFs may predispose membranes to dysfunctions and instability. This hypothesis was further supported by in vitro scratch assays (mimicking MFs) (17). Data suggest that AECs can proliferate, migrate, transition, and heal wounds (17, 19), supporting the hypothesis that MFs are likely areas where membrane remodeling occurs. Cell transitions at the scratch site include epithelial-to-mesenchymal transitions (EMTs) or the reverse mesenchymal-to-epithelial transitions (METs) (17). Furthermore, OS prevented cellular transitions and healing (17) and recapitulated similar observations associated with term and preterm parturitions (1).

Despite the recent finding that amnion cells can undergo cellular transitions, it is still unclear if MFs are formed or healed by fetal cells in vivo. Determining a causal relationship between cellular transitions and environmental stimulants will illuminate the role of fetal cells in membrane remodeling. However, studying such phenomena is extremely challenging because of the lack of available experimental approaches and models of the multicellular amnion membrane that can be experimentally manipulated and tested. Current experimental approaches of 2-dimensional (2D) single cell cultures, amnion cell-like organ explant cultures, and transwell coculture systems (of AMCs and AECs) are all insufficient to understand cellular transitions and their roles in tissue remodeling. Conventional mixed culture or coculture methods, where cells are cultured in randomly distributed form or in transwells, often fail to provide means to locally manipulate the physical and biochemical environments of each cell type in culture. Therefore, it is challenging to investigate the interactions between the fetal membrane cells, namely, AMCs and AECs, for detailed mechanistic studies. In addition, migratory cells that are thought to play an important role in collagen homeostasis cannot be easily monitored or studied using these methods. MFs and scratch assay experiments conducted and terminated at discrete time points may show a snapshot of cellular transition but do not convey the cellular mechanisms involved during the course of membrane remodeling; they also create challenges for understanding the dynamic cell-cell relationships.

To overcome the limitations of these traditional approaches, we developed an amnion membrane organ-on-chip (AM-OOC), allowing for direct monitoring of amnion cell migration and transition under a coculture condition in which the 2 different cell types could be cultured in 2 different microenvironments while enabling the application of localized chemical cues to only 1 cell type. Microfluidic organ-on-chip (OOC) technologies allow for control and manipulation of multiple cell types and their microenvironments with high accuracy and have been demonstrated as a promising technology to achieve in vitro models that more physiologically mimic in vivo structures and functions (20, 21). Recently, a fetal membrane OOC model was presented (22); however, this model lacks a degradable basement membrane as well as mesenchymal cells. Thus, this device does not accurately recapitulate how cells migrate through a basement membrane and also lacks critical factors from the mesenchymal cells. A truly physiologically relevant fetal membrane OOC model has the potential to recapitulate inter- and intracellular signaling and the physiologic context of tissue dynamics by compartmentalizing the major cellular components of a fetal membrane while still allowing interactions between these chambers. As a first step in establishing a full fetal membrane OOC system, we initially developed the AM-OOC and tested its usefulness in addressing the experimental limitations as previously described. Using this OOC approach, we tested and compared, using AECs and AMCs harvested from human placenta, AECs’ and AMCs’ migration and transitions independently as well as when cultured together under normal and OS conditions. We report that AECs can migrate, degrade basement collagen, and transition to become AMCs. OS induces AECs to undergo EMT and increase collagenolysis and inflammation. Additionally, the presence of AMCs accelerates this process. Conversely, AMCs migrate, degrade basement collagen, and transition to become epithelial cells in the presence of AECs. OS maintains AMCs’ mesenchymal phenotype, promotes migration, degrades basement collagen, and propagates inflammation.

MATERIALS AND METHODS

Institutional review board approval

This study protocol was approved by the institutional review board at The University of Texas Medical Branch (UTMB) at Galveston, TX, as an exempt protocol for using discarded placenta after standard term Cesarean deliveries (UTMB Project 69693). No subject recruitment or consent was required for this study. The AM-OOCs were developed and microfabricated at Texas A&M University (College Station, TX, USA), and cell-based studies were conducted at UTMB.

Clinical samples and cell culture

AEC culture

Primary AECs and AMCs were isolated from amnion membranes obtained from fetal membranes from term, not-in-labor, and Cesarean deliveries (23, 24). Approximately 10 g of amnion membrane, peeled from the chorion layer, was dispersed by successive treatments with 0.125% collagenase and 1.2% trypsin. All cell culture reagents were purchased from MilliporeSigma (Burlington, MA, USA). Details of AEC isolation protocols can be found in our previous report (2). Briefly, the dispersed cells were plated in a 1:1 mixture of Ham’s F12–DMEM; supplemented with 10% heat-inactivated fetal bovine serum, 10 ng/ml epidermal growth factor, 2 mM l-glutamine, 100 U/ml penicillin G, and 100 mg/ml streptomycin at a density of 3–5 million cells per T75 flask; and incubated at 37°C with 5% CO₂ until 80–90% confluencey was achieved.

AMC culture

AMCs were isolated from fetal membranes as previously described by Kendal-Wright et al. (25, 26) with slight modifications. Primary AMCs were isolated from the placental membranes of women experiencing normal parturition at term (i.e., not in labor) and undergoing a repeat elective Cesarean section. Reflected amnion (∼10 g) was peeled from the chorion layer and rinsed 3 or 4 times in sterile HBSS (21-021-CV; Corning, Corning, NY, USA) to remove blood debris. The sample was then incubated with 0.05% trypsin-EDTA (25-053-CI; Corning) for 1 h at 37°C (water bath) to disperse the cells and remove the epithelial cell layer. The membrane pieces were then washed 3 or 4 times using cold HBSS to inactivate the enzyme. The washed membrane was transferred into a second digestion buffer containing Eagle’s minimum essential medium (10-010-CV; Corning), 1 mg/ml collagenase type IV, and 25 μg/ml DNase I and incubated in a rotator at 37°C for 1 h. The digested membrane solution was neutralized using complete DMEM-F12 medium (10-092-CV; Corning), filtered using a 70-μm cell strainer, and centrifuged at 3000 rpm for 10 min. The cell pellet was resuspended in complete DMEM-F12 medium supplemented with 5% heat-inactivated fetal bovine serum (35-010-CV; Corning), 100 U/ml penicillin G, and 100 mg/ml streptomycin (30-001-CI; Corning). The resuspended cells were subsequently seeded at a density of 3–5 million cells per T75 and incubated at 37°C with 5% CO₂ until 80–90% confluencey was achieved.

Microfluidic AM-OOC design

The AM-OCC platform was fabricated in polydimethyl siloxane (PDMS) using a 2-step photolithography and soft lithography technique (27, 28). To create the master mold (Fig. 1C), 2 layers of photosensitive epoxy (SU-8; MicroChem, Westborough, MA, USA) with different thicknesses were sequentially patterned on a 3-in diameter silicon substrate. The first layer forming the 5-μm-deep microchannels was obtained by spin coating SU-8 3005 at 4000 rpm and soft baking at 95°C for 4 min. It was then exposed to UV light through a photomask, followed by a postexposure bake at 95°C for another 4 min. The second layer forming the cell culture chambers was 500 μm thick and patterned by spin coating SU-8 3050 at 1000 rpm, soft baked first at 65°C for 24 h and then at 95°C for 40 min, exposed to UV through a second photomask, and then postexposure baked in 2 steps, first at 65°C for 5 min and then at 95°C for 15 min. The master mold was then coated with (tridecafluoro-1,1,2,2-tetrahydro octyl) trichlorosilane (United Chemical Technologies, Bristol, PA, USA) to facilitate PDMS release from the master mold after replication.

Figure 1.

AM-OOC fabrication and layout. The AM-OOC is designed to recreate the amnion membrane in vitro by coculturing AECs and AMCs separated by a type IV collagen–filled microfluidic channel array (mimicking basement membrane). A) Schematic illustration of the AM-OOC. 3D and cross-sectional view showing the physical isolation of AECs and AMCs in each culture chamber, connected by 24 microchannels filled with type IV collagen. B) Cross-sectional view showing the principle of diffusion barrier formation by liquid height difference. C) Microfabrication and assembly steps for the AM-OOC device. Two SU-8 layers with different thicknesses were patterned on top of a silicon substrate to form the microchannels, and the 2 cell culture chambers (outer chamber: AECs; inner chamber: AMCs). PDMS devices were replicated from the SU-8 master using soft lithography process, and 7-mm diameter reservoirs were punched out followed by bonding onto poly-d-lysine– or matrigel-coated substrates. D) Each AM-OOC fits into 1 well of a conventional 6-well polystyrene culture plate. E) Bright field microscopy images of AEC morphology and AMC morphology inside an AM-OOC device. Scale bars, 10 µm. F) The outer chamber of the AM-OOC was filled with red dye, and the inner chamber was filled with blue dye. Right image shows that a diffusion barrier has been successfully formed between chambers (indicated by the red dye remaining in the outer chamber; scale bar, 300 µm). Microchannels filled with type IV collagen matrigel (stained with Masson trichrome), connecting the 2 culture chambers (scale bar, 30 µm) can also be seen.

PDMS devices were replicated from the master mold by pouring PDMS prepolymer (10:1 mixture, Sylgard 184; DowDuPont, Midland, MI, USA) on the mold, followed by curing at 85°C for 45–60 min. The reservoirs to hold culture medium were punched out using a 5-mm diameter punch bit (Syneo, Angleton, TX, USA) mounted on a drill press. To improve the bonding of PDMS devices onto glass substrates and to make the device hydrophilic for easy cell and culture medium loading, the PDMS devices were treated with oxygen plasma (Harrick Plasma, Ithaca, NY, USA) for 90 s, followed by bonding onto glass substrates. The PDMS culture devices were then immersed in deionized water. For sterilization, an autoclave was used to sterilize the PDMS culture devices at 121°C for 30 min.

Microfluidic AM-OOC device preparation for matrigel filling of microchannels

Before using the AM-OOC, devices were washed 3 times with PBS, coated with matrigel (Corning Matrigel Basement Membrane Matrix, DEV-free; 1:50 in PBS), and incubated at 37°C with 5% CO₂ overnight. Diluted type IV basement membrane matrigel was used to coat the microchannels connecting the outer and inner culture chambers, which mimics the amnion basement membrane in vivo. Through this process a thin layer of matrigel is left in the outer and inner chamber after microchannels are filled, where this contact with basement membrane mimics amnion cell growth in utero.

Masson trichrome staining for matrigel imaging

Before and after AM-OOC experiments, devices were stained with Masson trichrome stain to image type IV collagen inside the microchannels. To show that our matrigel loading was working, devices were rinsed with PBS and fixed at room temperature with 4% paraformaldehyde for 20 min. The devices were then stained with Biebrich scarlet–acid fuchsin for 10 min and then rinsed with water 3 times. This process stained all the cells and collagen red. Next, phosphomolybdic-phosphotungstic acid was applied for 15 min, which removed the red stain from the collagen. Aniline blue solution was then added for 10 min to stain the collagen blue. Once the device was stained, it was rinsed 3 times with water and imaged. This procedure was additionally carried out on some devices after 48 h of cell culture to monitor collagen degradation caused by cell migration.

Fluorescent dye perfusion assay

To determine to what degree culture medium in the AM-OOC could diffuse from 1 culture chamber to the other, which corresponds to how much inflammatory mediators can propagate from 1 chamber to the other, we conducted a set of perfusion assay experiments. FITC dye was loaded into the inner or outer chambers, and microscopy images were taken over time (0–70 h). Fluorescence intensity was used to measure the degree of diffusion from the center chamber to the outer chamber, or vice versa. ImageJ software (National Institutes of Health, Bethesda, MD, USA) measured the intensity of FITC dye that perfused through the microchannels (with and without type IV collagen matrigel) and into the opposite chamber over 70 h. Intensity values were normalized between replicates using the following formula: [(intensity at time point − intensity at time zero) ÷ intensity at time zero] × 100.

Cell seeding and culture in the AM-OOC

Before using the AM-OOC, devices were washed 3 times with PBS and coated with matrigel as previously described. The next day, devices were washed 3 times with complete DMEM-F12 medium before cell seeding. Primary cells were then trypsinized and stained with live cell dyes for green fluorescent protein (GFP) (for AEC; CellLight Histone 2B–GFP) or red fluorescent protein (RFP) (for AMC; CellLight Histone 2B–RFP) following the protocol provided by the company (10594 and 10595; Thermo Fisher Scientific, Waltham, MA, USA). Then, 120,000 AECs were loaded into the outer chamber, and 40,000 AMCs were loaded into the inner chamber of the AM-OOC. The AM-OOCs were incubated at 37°C with 5% CO₂ for 24 h before localized treatment (see next section).

Cell culture treatments in the AM-OOC

To test the effect of OS on cellular transition in the amnion membrane, we treated each AM-OOC with one of the following for 48 h: 1) normal cell culture conditions (control DMEM-F12 medium); 2) OS conditions [induced by treating cells with cigarette smoke extract (CSE)] (2, 23) diluted 1:25 in AEC medium or diluted 1:75 in AMC medium; and 3) to verify the effect of OS, cells were cotreated with an OS inducer (CSE) and an antioxidant or stress signaler p38 MAPK inhibitor [N-acetyl-l-cysteine (NAC, 15 mM, A7250; MilliporeSigma] (23, 29) and SB203580 (13 mM, S8307; MilliporeSigma) (23, 29), a p38 MAPK inhibitor and a known inducer of EMT.

To induce OS in fetal membrane cells, CSE was used as previously described (2), with modifications. Cigarette smoke from a single commercial cigarette (unfiltered Camel; R.J. Reynolds Tobacco, Winston Salem, NC, USA) was bubbled into 25 ml of AEC or AMC medium. The stock CSE was sterilized using 0.25 mm Steriflip filter unit (MilliporeSigma) and diluted to 1:50 (AEC) or 1:75 (AMC) in cell specific medium before use. This modification was necessary to minimize any drastic effects of CSE in a microfluidic 3-dimensional (3D) culture system than in a much bigger 2D cell culture system.

Because of AMCs’ exaggerated response to OS inducers (30), a different CSE concentration was used to induce OS in AMCs compared with AECs’ treatment. Once cells reached 70–80% confluence, each AM-OOC was rinsed with sterile 1× PBS, serum-starved for 1 h, treated with the respective conditions, and incubated at 37°C, 5% CO₂, and 95% air humidity for 48 h. After 48 h, bright field microscopy (Nikon Eclipse TS100 microscope, ×10 magnification; Nikon, Tokyo, Japan) or confocal microscopy (Zeiss 880, ×10 magnification; Carl Zeiss, Oberkochen, Germany) was performed to determine cell morphology, percentage of microchannels containing cells, and number of cells that migrated through the microchannels to the other side of the chamber for each treatment.

Immunocytochemical localization of intermediate filaments cytokeratin and vimentin

Cell staining

AEC and AMC immunocytochemical staining for vimentin (3.7 μl/ml; ab92547; Abcam, Cambridge, MA, USA) and cytokeratin-18 (CK-18; 1 μl/ml; ab668; Abcam) were performed after 48 h as previously described by Richardson et al. (17). Manufacturer’s instructions were used to calculate staining dilutions to ensure uniform staining. After 48 h, cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X, and blocked with 3% bovine serum albumin in PBS prior to incubation with primary antibodies overnight at 4°C. After washing with PBS, the AM-OOCs were incubated with Alexa Fluor 488–, 594–, and 647–conjugated secondary antibodies (Thermo Fisher Scientific) and diluted 1:2000 in 3% bovine serum albumin for 2 h in the dark. The AM-OOCs were washed with PBS, treated with NucBlue Live ReadyProbes Reagent (Thermo Fisher Scientific), and imaged as previously described.

Image analysis

Three random regions of interest per AM-OOC were used to determine red (CK-18) and green (vimentin) fluorescence intensity. Uniform laser settings, brightness, contrast, and collection settings were used for all images collected. Images were not modified (brightness, contrast, and smoothing) for intensity analysis. ImageJ software was used to measure vimentin and CK-18 staining intensity from 2 focal plans of 3 different regions per treatment condition at each time point. Image analysis was conducted in triplicate for all cell experiments.

Milliplex luminex assays for inflammatory cytokine markers

Supernatant were manually collected from the reservoirs of both chambers after 48 h of treatment. Milliplex assays were performed with the cytokine IL-8 and granulocyte M-CSF (GM-CSF) antibody-coated beads (HCYTOMAG-60K; Merck, Darmstadt, Germany) as indicators of general inflammation in cell supernatant. Standard curves were developed with duplicate samples of known quantities of recombinant proteins that were provided by the manufacturer. Sample concentrations were determined by relating the absorbance values that were obtained to the standard curve by linear regression analysis.

Statistical analyses

All experiments were conducted in triplicate and images analyzed using Prism 7 software (GraphPad Software, La Jolla, CA, USA). One-way ANOVA and independent samples Student’s t test were used, and P < 0.05 was considered significant.

RESULTS

AM-OOC development

The AM-OOC device is composed of 2 circular chambers, 1 for AEC culture and 1 for AMC culture, connected through arrays of microfluidic channels (Fig. 1A). The center circular chamber is for AMC culture, and the outer ellipse-shaped chamber is for AEC culture. The outer and inner chambers measure 500 μm in height and are connected by 24 microfluidic channels (5 μm in height, 30 μm in width, 600 μm in length; Fig. 1B). Each device contained an outer chamber with 2 reservoirs and an inner chamber with 1 reservoir (Fig. 1C). Because of the device’s small height, suspended cells initially loaded into each culture chamber remain in each chamber while still allowing cellular migration. In addition, in some cases, diluted type IV matrigel was used to fill these microfluidic channels to mimic the amnion basement membrane (see next section for more details). When applying localized stimuli to only the AECs, the culture medium level of the center AMC culture chamber was maintained at a higher fill level compared with the outer AEC culture chamber, preventing any biochemicals and metabolites from AECs from diffusing into the AMC chamber because of the hydrodynamic pressure difference created by the different fluidic level. When applying localized stimuli to only the AMCs, the fluid height difference was reversed. Such hydrodynamic pressure difference–based localized coculture has been previously utilized for neuron-glia cell coculture (27, 28).

After sterilization, each device was placed in a conventional 6-well plate (Fig. 1D) and coated with type IV collagen matrigel. Primary cuboidal AECs were loaded into the outer chamber, and primary AMCs were loaded into the inner chamber. Cells were cultured in good health over 2 d, each showing representative morphologies (Fig. 1E). The inner AMC culture chamber was filled with blue dye, and the outer AEC culture chamber was filled with red dye (Fig. 1F). Matrigel coating successfully produced collagen-filled microchannels (blue stain) to mimic amnion basement membrane (Fig. 1F), enabling isolation of the 2 different culture conditions while still allowing molecular communication as well as cell migration.

Fluidic isolation over time between the AMC and AEC chambers in the AM-OOC

In the multicellular AM-OOC system, fluidic isolation between the 2 chambers is important to independently control and manipulate the 2 cell types and their microenvironments, while still allowing interactions between the 2 cell types. The efficiency to maintain fluidic separation between the 2 chambers was tested using a FITC-based perfusion assay in the AM-OOC by creating a minute fluidic level difference between the outer and inner chambers that resulted in hydrostatic pressure differences. FITC or PBS was loaded into either the outer or inner chamber, and fluorescence microscopy was used to monitor the rate of perfusion between the 2 chambers. When the inner chamber had a higher fluidic level than the outer chamber, it took more than 24 h for the FITC in the outer chamber to diffuse into the inner chamber, successfully demonstrating fluidic isolation between the 2 chambers (Fig. 2A, B). The use of type IV collagen–filled microfluidic channels extended this isolation time to 60 h (Fig. 2C, D). Similarly, fluid pressure from the outer to inner chamber counteracted some of the diffusion and allowed fluidic isolation for at least 20 h (Fig. 2E, F). The use of type IV collagen–filled microfluidic channels extended this fluidic isolation time to 40 h (Fig. 2G, H). Additionally, the difference in fluidic levels was also able to prevent proinflammatory cytokine propagation (GM-CSF) from diffusing from 1 chamber to the other chamber (Supplemental Fig. S1B, C), further confirming successful fluid isolation between the 2 chambers.

Figure 2.

AM-OOC fluid perfusion over time. FITC-based perfusion assay showing fluidic isolation between the 2 culture chambers of the AM-OOC device over 60 h. A–D) Inner chamber fluid level is higher than outer chamber fluid level, countering diffusion from outside to inside. Brightfield and fluorescence microscopy (A) shows FITC diffusing through microchannels without type IV collagen coating. Red boxes highlight regions where fluorescence intensity was measured (n = 3). Graph (B) shows the degree of FITC diffusion from the outside chamber to the inside chamber. C, D) Repeat of A and B, but when the microchannels in the AM-OCC is filled with type IV collagen coating. E–H) Outer chamber fluid level is higher than inner chamber fluid level, countering diffusion from inside to outside. Brightfield and fluorescence microscopy (E) shows FITC diffusing through microchannels without type IV collagen coating. Red boxes highlight regions where florescent intensity was measured (n = 3). Graph (F) shows the degree of FITC diffusion from the inside chamber to the outside chamber. G, H) Repeat of E and F, but when the microchannels in the AM-OCC is filled with type IV collagen coating.

Characteristics of monoculture of AECs and AMCs in the AM-OOC

Amnion cells show migratory and transition capacity

Monoculture of AECs (outer chamber) or AMCs (inner chamber) in the AM-OCC showed that cells can enter into the type IV collagen–filled microchannels, elongate, migrate through type IV collagen, and exit the microchannel (red outline) within 48 h (Fig. 3A, B). Migrated cells either revert to their original epithelial shape or maintain their achieved mesenchymal morphology, clearly showing that direct imaging of cell migration through the collagen-filled microfluidic channel is possible in the developed AM-OOC.

Figure 3.

AEC and AMC characteristics inside an AM-OOC device. Characteristics of monocultured amnion cells in an AM-OOC device under control and OS conductions. A) Composite bright field images show that under a normal cell culture condition, AECs can enter into the collagen-filled microchannels, elongate to show a mesenchymal morphology, migrate through type IV collagen, and exit the tunnel (red outline) while potentially retaining this mesenchymal shape (n = 6). Scale bar, 30 µm. B) Composite bright field images show that under a normal cell culture condition AMCs can enter into the collagen-filled microchannels, elongate, migrate through type IV collagen during that process, and exit the tunnel (red outline), then revert to their original shape (n = 6). Scale bar, 30 µm. C, D) Confocal images show native levels of vimentin and CK-18 expression in AECs and AMCs (AEC: 0.56 ± 0.02 vs. AMC: 1.1 ± 0.02; n = 3). AECs are in a metastate, meaning they coexpress both epithelial and mesenchymal markers. AMCs had significantly higher vimentin:CK-18 levels compared with AECs (P < 0.0001; n = 3). Confocal images were captured at original magnification, ×10. Blue, DAPI; green, vimentin; red, CK-18. Values are expressed as mean intensities ± sem. Scale bars, 10 µm. E) Analysis of bright field microscopy images shows AECs migrated into the opposite chamber more frequently than AMCs. Values are expressed in number of migrated cells in each device (P = 0.0064). F, G) Confocal images show native and OS induced (CSE) levels of vimentin and CK-18 expression in AECs (AEC control: 0.56 ± 0.02, CSE: 0.67 ± 0.02) and AMCs (AMC control: 1.1 ± 0.2, CSE: 1.1 ± 0.2; N = 3). CSE AECs had significantly higher vimentin:CK-18 levels compared with AEC controls (P = 0.02) (white arrow showing vimentin on leading edge of cell), whereas AMC intermediate filament expression remained constant regardless of treatment (n = 3). Confocal images were captured at original magnification, ×10. Blue, DAPI; green, vimentin; red, CK-18. Values are expressed as mean intensities ± sem. Scale bars, 30 µm. H) Analysis of bright field microscopy images shows CSE treatment inhibited migration of AECs compared with AEC controls (P = 0.0005), whereas CSE treatment of AMCs stimulated migration (n = 3). Control AECs contain the most migratory potential of all treatments and cell types. Migratory cells were defined as cells that had migrated through the microchannel and were now resident in the opposite chamber.

Amnion intermediate filament expression and migratory potential

Our results confirm that AECs innately express both epithelial and mesenchymal markers (i.e., CK-18, red; vimentin, green; low vimentin:CK-18 ratio; Fig. 3D), suggesting they are in a metastate or an in-between state of cellular transition (17). AMCs predominantly expressed mesenchymal marker vimentin as shown by the significantly higher vimentin:CK-18 ratio compared with AECs (P < 0.0001; Fig. 3D). Furthermore, AECs in a metastate contained a significantly higher migratory potential (P = 0.0064) than AMCs, likely because of the attainment of mesenchymal transition characteristics and metastate status (Fig. 3E).

OS induces changes in amnion intermediate filament expression and migration

Increased OS at term has been shown to induce labor-associated changes, such as cellular senescence, matrix metallopeptidase 9 up-regulation, and increased proinflammatory cytokine production in fetal membrane cells, including AECs and AMCs (2, 23). CSE, a potent and reliable OS inducer (2, 17, 23, 31–35), has been shown to recreate the labor phenotype (OS experienced at term labor in amnion membranes) in vitro and to induce a static state of EMT in AECs (17, 33). EMT contributes to sustained inflammation that promotes the labor-related cascade of events.

Data from this study indeed confirmed our earlier findings (17) that CSE treatment for 48 h induced a fibroblastoid morphology in AECs and vimentin relocalization to the leading edge (white arrow) of migratory cells and significantly increased vimentin:CK-18 ratio (control: 0.56 ± 0.02, CSE: 0.67 ± 0.02; P = 0.02) compared with AECs under standard cell culture conditions [vimentin (perinuclear)] and CK-18 (cytoplasmic; Fig. 3F, G). These changes are indicative of CSE inducing EMT in AECs while significantly decreasing their migratory potential (P = 0.0005). Although mesenchymal characteristics are attained that should contribute to more cellular kinesis, the CSE-induced loss of migratory potential observed in our study is likely due to senescence of cells and independent of transition status. CSE treatment of AMCs resulted in maintenance of their mesenchymal phenotype (Fig. 3F, G) and increased level of migration compared with AECs (Fig. 3H).

Characteristics of cellular transition in AM-OOC under normal and OS conditions

Innate transition properties of AECs

Confocal microscopy documented amnion cell morphology, intermediate filament expression, and cellular transitions to better understand how amnion cells migrate and degrade basement membrane collagen inside microchannels. Microscopy revealed that AECs under innate conditions express basal levels of vimentin:CK-18 (0.5 ± 0.01), perinuclear vimentin, and an epithelial morphology. Migration of AECs was accompanied by a significant increase of vimentin:CK-18 ratio (0.62 ± 0.02; P = 0.0025), vimentin relocalization to the leading edge (white arrow), and a fibroblastoid morphology, suggesting cellular transition (EMT). Once AECs completely migrated to the inner chamber, however, they reverted to basal expression of vimentin:CK-18 ratio (0.48 ± 0.02; P = 0.0009), perinuclear vimentin, and an epithelial morphology indicative of MET (Fig. 4A, B). Thus, the results shows that AECs must undergo 2 cellular transitions, EMT to migrate and MET to exit, to completely migrate through the type IV collagen–filled microchannels. These transitions are similar to what we have reported in scratch assays that resembled membrane MF healing, in which migrating and healing edges had mesenchymal and epithelial phenotypes, respectively (17).

Figure 4.

OS’s effect on cell migration and transition in monoculture. Confocal images measuring vimentin:CK-18 levels in AECs and AMCs during cell migration (n = 3). Confocal images were captured at original magnification, ×10. Blue—DAPI, green—vimentin, and red—CK-18. Values are expressed as mean intensities ± sem. Scale bars, 10 µm. A, B) AECs in the outer chamber under native conditions express basal levels of vimentin:CK-18 ratio (0.5 ± 0.01), perinuclear vimentin, and an epithelial morphology. Migration of AECs is accompanied by a significant increase of vimentin:CK-18 ratio (0.62 ± 0.02; P = 0.0025), vimentin’s relocalization to the leading edge (white arrow), and an elongated mesenchymal morphology indicative of an EMT. Once AECs completely migrated to the inner chamber, they reverted back to basal expression of vimentin:CK-18 ratio (0.48 ± 0.02; P = 0.0009), perinuclear vimentin, and an epithelial morphology indicative of an MET. C, D) AECs in the outer chamber under OS conditions (CSE) express relatively high levels of vimentin:CK-18 ratio (0.65 ± 0.02) compared with control AECs. Additionally, AECs treated with CSE contained vimentin localization at the leading edge (white arrow) and an elongated mesenchymal morphology. Migration of CSE-treated AECs is not accompanied by changes in the vimentin:CK-18 ratio (0.58 ± 0.03), vimentin relocalization, or morphology variations, which shows that CSE maintains AECs in a static state of EMT. However, the few AECs able to cross the microchannels did, under MET transitions, revert to basal expression of vimentin:CK-18 ratio (0.3 ± 0.02; P = 0.0004), very little vimentin, and an epithelial morphology. E–H) AMCs in the inner chamber under native and OS conditions express relatively high levels of vimentin:CK-18 ratio (control: 1.1 ± 0.2 and CSE: 1.1 ± 0.2) compared with AECs regardless of treatment. Migrating AMCs maintain their vimentin:CK-18 (control: 1.0 ± 0.03, CSE: 1.1 ± 0.09) while relocalizing vimentin to the leading edge and inducing an elongated cell morphology. Migration of AMCs into the outer chamber significantly increases the vimentin:CK-18 ratio (control: 2.1 ± 0.14, CSE: 2.4 ± 0.2), while also inducing native vimentin localization and mesenchymal morphology.

OS-induced static state of EMT in AECs

As shown above, 48-h OS-treated AECs in the outer chamber expressed higher basal levels of vimentin:CK-18 compared with control AECs (Fig. 4B vs. Fig. 4D), indicative of a mesenchymal phenotype. Migrating AECs maintained their fibroblastoid characteristics while relocalizing vimentin to the leading edge (white arrow); however, because of OS-induced senescence, most cells that underwent EMT were retained inside the microchannels, and the majority of them were unable to transition to an epithelial phenotype to exit, which is in line with our finding that OS inhibits migration in AECs (control: 7.4 ± 1.7 cells, CSE: 0.2 ± 0.2 cells; P = 0.0005; Fig. 3H). However, the few AECs that crossed the microchannels did undergo MET, inducing basal vimentin:CK-18 ratio (0.3 ± 0.02; P = 0.0004) levels and an epithelial morphology (Fig. 4C, D), suggesting the influential role of microenvironment in transitioning amnion membrane cells.

OS does not change innate transition properties of AMCs

AMCs in the inner chamber both under normal conditions and after OS exposure expressed relatively high levels of vimentin:CK-18 ratio (control AMC: 1.1 ± 0.2, CSE AMC: 1.1 ± 0.2; P = ns). Migrating AMCs maintained their vimentin:CK-18 ratio (control AMC: 1.0 ± 0.03, CSE AMC: 1.1 ± 0.09) while relocating vimentin to the leading edge. AMCs that migrated into the outer chamber had significantly higher vimentin:CK-18 ratio (control AMC: 2.1 ± 0.14, CSE AMC: 2.4 ± 0.2; migrating vs. emigrated AMCs: control: P < 0.001, CSE: P < 0.001) while also inducing native vimentin localization and morphology (Fig. 4E–H). These data suggest that AMCs do not require cellular transitions (EMT to migrate and MET to exit) to migrate through microchannels, contrary to the behavior of AECs.

Characteristics of cocultured AECs and AMCs in AM-OOC

To recreate the physiologic context of the amnion membrane components, AEC and AMCs were cocultured inside the AM-OOCs to study tissue dynamics. Crystal violet stain documented that both amnion cell populations were viable after 48 h (Supplemental Fig. S2). To determine the effect of coculture on cellular transitions and migration, we stained live AECs with a histone 2B–GFP and AMCs with a histone 2B–RFP to track them during and after migration. We treated each chamber individually with control, CSE, or CSE with OS inhibitor NAC plus p38 MAPK functional inhibitor SB203580 (CSE+; Table 1). Both of these inhibitor compounds have been shown to reduce the deleterious effects of OS- and stress-associated signaling in amnion cells (23, 29), which is why they were selected for our experiment.

TABLE 1.

Summary of coculture treatments and abbreviations

Treatment	Outer chamber treatment	Inner chamber treatment	Abbreviation
1	Control	Control	Control/control
2	CSE	Control	CSE/control
3	Control	CSE	Control/CSE
4	CSE	CSE	CSE/CSE
5	CSE + NAC + SB	CSE + NAC + SB	CSE+/CSE+

CSE+, CSE with OS inhibitor NAC plus p38 MAPK functional inhibitor SB203580; SB, SB203580.

Coculture effect on cellar transitions

After examining what occurs during monoculture in either the outer or inner chamber of the AM-OOC, we performed similar coculture experiments under fluidic isolation conditions (fluidic isolation shown in Fig. 2). Under normal coculture conditions, GFP-labeled AECs expressed epithelial morphology in the outer chamber, underwent EMT, migrated, and maintained their mesenchymal morphology to join the AMC population in the inside chamber (Fig. 5A; middle column, left cropped area; yellow outline). Similarly, RFP-labeled AMCs migrated and underwent MET to an epithelial morphology while assembling into the AEC population (Fig. 5A; middle column, center, and right cropped area; yellow outline), emphasizing AECs’ and AMCs’ ability to transition under distinct environmental conditions.

Figure 5.

OS’ effect on migration and transition in AEC-AMC coculture. Confocal imaging of coculture experiments revealed that both cell types can migrate, transition, and integrate themselves into the emigrated environment (n = 3). A) Confocal images show native AECs (green) and AMCs (red), which have transitioned, migrated, and integrated into the opposite population (n = 3). Middle right panel highlights (yellow) GFP-AECs that have migrated through the type IV collagen tunnel, relocalized vimentin, and transitioned into a mesenchymal morphology indicative of EMT. Middle left panels highlight (yellow) RFP-AMCs that have migrated through the type IV collagen tunnel, down-regulated vimentin, and transitioned into an epithelial morphology indicative of MET. Bottom panel is a schematic representing AEC (green) and AMC (red) cellular transitions. Gray arrows highlight migration direction. Confocal images were captured at original magnification, ×10. Pink, vimentin; green, histone 2B AEC; red, histone 2B AMCs. Scale bars, 30 µm. B) Top graph: Analysis of bright field microscopy images shows AECs treated with CSE, in coculture, migrated into the opposite chamber more frequently than the AEC control (AEC control: 5.1 ± 0.06, CSE: 7.3 ± 0.3) (AEC: bar 1 vs. 2; 1.5-fold increase) (AMC: bar 1 vs. 3; 2-fold increase). CSE treatment of AMCs did not affect AECs’ migration (5.1 ± 0.06 vs. 4.3 ± 0.8). CSE cotreatment of AECs and AMCs inhibited AECs’ migration (0.67 ± 0.6), whereas cotreatment with CSE + relieved the effects of CSE (1.3 ± 0.8) (AECs: CSE/CSE vs. CSE+/CSE+ = 2-fold higher) (AMC: CSE/CSE vs. CSE+/CSE+ = 2.5-fold higher). Values are expressed as mean intensities ± sem. Bottom graph: Analysis of bright field microscopy images shows AMCs treated with CSE, in coculture, migrated into the opposite chamber more frequently than the AMC control (AMC control: 2.7 ± 0.6, CSE: 5.5 ± 1.5). CSE treatment of AECs did not affect AMCs’ migration (2.7 ± 0.6 vs. 3.3 ± 0.3). CSE cotreatment of AECs and AMCs inhibited AEC migration (1 ± 0.5), whereas cotreatment with CSE + relieved the effects of CSE (2.7 ± 0.6). Values are expressed as mean intensities ± sem. Migratory cells were defined as cells that had migrated through the microchannel and identified by the opposite color of cell nuclei (i.e., green nuclei AEC cells in the red nuclei AMC population).

OS’s effect on migration in coculture

A bright field microscopy analysis showed that 48-h CSE treatment of AECs or AMCs in coculture induced migration more frequently than in respective controls (AEC control: 5.1 ± 0.06, CSE: 7.3 ± 0.3; AMC control: 3.3 ± 0.3, CSE: 5.5 ± 1.5). These results were different compared with our monoculture experiments, highlighting the effect of coculture on cell migration. Localized CSE treatment of each chamber did not affect AECs’ or AMCs’ migration potential in their adjacent chambers, showing that localized CSE treatment in the AM-OOC is indeed possible. Interestingly, when CSE cotreatment was added to both chambers, cellular migration slowed, though not to a significant level. These effects were mildly prevented by cotreatment of CSE with NAC and SB203580 (CSE+; Fig. 5B) (AECs: CSE/CSE vs. CSE+/CSE+ = 2-fold higher) (AMC: CSE/CSE vs. CSE+/CSE+ = 2.5-fold higher), suggesting that OS and p38 MAPK downstream signaling could regulate AECs’ and AMCs’ migration.

Propagation of inflammatory mediators in AM-OOC devices

ELISA for proinflammatory marker GM-CSF was evaluated 48 h after treatment to determine whether migratory cells induced inflammatory changes in the opposite chambers.

Standard and OS-induced inflammatory mediator expression

Consistent with the current literature, under coculture condition, AMCs naturally produced more proinflammatory cytokines compared with AECs (18), although this was not to a significant level (Fig. 6A). Furthermore, though not significant, CSE treatment of AECs and AMCs increased proinflammatory cytokines in both cell types [AEC: control: 5.7 ± 1.5, CSE: 8.7 ± 1.9 (black vs. white bars)] (Fig. 6B) [AMC: control: 2208 ± 1629, CSE: 3835 ± 1541 (black vs. light gray bars)] (Fig. 6C).

Figure 6.

Production and propagation of proinflammatory mediators in the AM-OOC coculture system. OS induced proinflammatory mediator production and propagation in amnion cells (n = 3). A) ELISA-measured media concentrations of GM-CSF from the AEC (outer chamber) and AMC (inner chamber). Though not significant, AMCs naturally have a higher level of GM-CSF expression compared with AECs (AEC: 5.7 ± 1.5 ng/ml, AMC: 2208 ± 1629 ng/ml). B, C) CSE treatment, regardless to which chamber, induced higher expression of GM-CSF in AMCs compared with AECs (CSE AMCs’ effect on control AECs [black vs. light gray bars], control/control: 5.7 ± 1.5, control/CSE: 8.2 ± 2.4) (B) (CSE AECs’ effect on AMCs [black vs. white bars], control/control: 2208 ± 1629, CSE/control: 3835 ± 1541) (C). CSE + treatment lowered GM-CSF in both AECs (2.8 ± 0 ng/ml) and AMCs (81.9 ± 43 ng/ml). Values are expressed as mean intensities ± sem.

OS-induced inflammatory mediator propagation

The CSE treatment of 1 chamber was also shown to induce inflammatory mediator (GM-CSF) response in the opposite chamber [CSE AMCs’ effect on control AECs (black vs. light gray bars), control/control: 5.7 ± 1.5, control/CSE: 8.2 ± 2.4] (Fig. 6B) [CSE AECs’ effect on AMCs (black vs. white bars), control/control: 2208 ± 1629, CSE/control: 3835 ± 1541] (Fig. 6C), whereas CSE+ treatments lowered proinflammatory cytokine production compared with CSE alone in both cell types (Fig. 6B; dark gray vs. striped bars) (Fig. 6C; dark gray vs. striped bars). Because fluid isolation was established (Fig. 2 and Supplemental Fig. S1) and CSE treatment did induce migratory changes in the cell population (Fig. 5B), we postulate that inflammatory changes may be initiated from the migrated cells themselves or by supernatant leaking through cell induced tunnels in the collagen-filled microchannels.

DISCUSSION

The AM-OOC that we developed and utilized in this study revealed amnion membrane cells’ transition and migratory properties under interactive environmental conditions. We determined the following: 1) amnion membrane cells can transition and migrate through type IV collagen–filled microchannels; 2) OS induces a static (nonreversible) state of EMT, decelerates cell migration, and increases proinflammatory mediator production; 3) coculture experiments revealed that both cell types can migrate, transition, and integrate themselves into the emigrated environment; and 4) OS cotreatment propelled transition but inhibited migration of cells in cocultures and induced proinflammatory mediator production in the adjacent cell chamber. Inhibition of OS by antioxidants and functional inhibitors of stress signaler p38 MAPK reversing the changes further confirmed the influence and interaction between the AMCs and AECs.

Recreating the whole organ dynamics using OOCs is an idea that has been around for many years (36); however, only very recently has their usefulness been explored in the field of reproductive science. The female reproductive tract (37), placenta (38), and endometrium-on-chip (39) have been developed and used to study multiple aspects of reproductive health, which have highlighted the importance of cell-cell and cell-blood interactions in vitro. Although development of a fetal membrane-on-chip has been postulated (22), no work has been reported yet. We first focused on creating an in vitro culture model of the amnion membrane because it is a key component of the fetal membrane. Another reason we initially focused on the amnion membrane, rather than the entire fetal membrane, is that it is only composed of 2 cell types, compared with the entire fetal membrane, which is composed of 5 cell types, from the fetal side as well as the maternal side. We have previously developed a model that tested interactive features between feto-maternal interface cells, AECs, and maternal decidual cells (40). The current model was used to address the limitations of the previous model, including the following: its lack of an ECM, because cells are directly cultured on a synthetic nondegradable polymer membrane similar to those used in transwell inserts; its inability to locally stimulate only 1 cell type to properly study cell-cell interactions and causal relations in their effect; and the vertical organization of the device that prevents direct imaging of the culture chambers and migratory cells that move from 1 chamber to the other, to name a few. The AM-OOC we developed overcomes these limitations by more accurately mimicking the amnion component of the fetal membrane, especially the existence of ECMs, while allowing different cellular components of the amnion to be independently controlled and stimulated and also allowing the direct monitoring of cellular migration through the ECMs using microscopy.

The amnion membrane provides the structural framework for the intrauterine cavity and contributes to pregnancy maintenance by bearing the tensile strain inflicted by the growing fetus (3, 41–43). The highly elastic amnion layer of the fetal membranes maintains its integrity and function through constant remodeling mediated by cellular transitions and matrix rearrangements. AECs are more dynamic in their transitions because they line the inner surface of the intraamniotic cavity, whereas AMCs serve as reserve cells to fill gaps vacated by AECs in the ECM. Maintenance of membrane integrity during gestation and its mechanical and functional compromise at term involve both cellular and matrix components. ECM turnover by collagenolytic processes is well reported (44), and recent work from our lab has shed some insight into cellular-level changes (1, 17, 33). In that work, we reported that OS’s buildup at term causes stress signaler p38 MAPK-mediated senescence (23, 33) as well as EMT of AECs (17). Both of these conditions cause endogenous inflammatory responses associated with parturition. Histologic examination of senescent membranes revealed biologic MFs that are sites of remodeling. Although cell scratch assays can determine mechanisms of MFs’ healing, those experiments lacked cellular interactions that may influence transitions and migrations. In addition, even though some level of evidence of cellular migration can be observed at distinct time points through microscopy, such migration could not be directly monitored. These limitations were addressed by the use of the AM-OOC model we developed.

Our study provides a novel approach to document sites of remodeling in vitro by visualization of cells migrating through collagen-filled microchannels (Fig. 3). This very well may be facilitated by collagen degradation or even in its absence. Our study did not specifically test this aspect; however, based on the nature of amnion cells, it is likely that they will produce type IV collagen–specific matrix metalloproteinases to propel themselves through these microchannels. During gestation, AECs and AMCs undergo cyclic cellular remodeling to heal gaps and MFs in the membranes, a mechanism required to maintain membrane homeostasis. Membrane remodeling at a cellular level is achieved by EMT of AECs and MET of AMCs, aided by redox radicals, growth factors (e.g., TGF-β), and endocrine mediators (e.g., progesterone) (17, 19, 33, 45). Cellular gaps are created (8) when AECs are shed from the membrane because of cellular senescence, mechanical disruption caused by fetal and amniotic fluid shear stress, or both. These gaps lead to MFs’ formation by shed AECs, which migrate through the ECM. This migration is aided by the mobility attained by AECs when they transition to AMCs. Endogenous progesterone recycles transitioned AMCs back to AECs with the production of nascent collagen to fill any degraded ECM components. These biologic processes maintain membrane integrity and cellular homeostasis during gestation. However, at term, an increased OS-induced static state of EMT increases inflammatory mesenchymal phenotype, leading to collagen degradation and mechanical failure of membranes.

These intrinsic in utero events were recreated in our AM-OOC and documented that treatment of adjacent cell populations in a controlled environment results in OS, inducing a static state of EMT and inflammation (Figs. 3 and 4). In single cell culture, OS has also been previously shown (17) to inhibit migration of AECs because of development of cellular senescence and independent of transition status. AEC single culture data from AM-OOC also reconfirmed these findings, suggesting that OS-treated microchannel cells are in a state of cellular senescence, which could contribute to migration inhibition. However, importantly, our model shows that this inhibition can be partially overcome when AECs are cocultured with AMCs that are maintained in a normal cell culture environment (Fig. 5). Conversely, CSE cotreatment induced EMT in AECs but prevented migration and MET in AMCs. Thus, OS treatment induces a static state of EMT in the AM-OOC devices, similar to what is observed in term amnion membrane.

By recreating the full amnion component of the fetal membranes, this AM-OCC model provides a physiologic context that allows manipulation of multiple cell types and their microenvironments with high levels of accuracy. 3D cell culture models, such as amnion membrane explants, organoids, and transwell systems, offer an alternative approach to multilayer assessment of cell-cell and cell-collagen interactions in vitro. However, the ability to distinguish between individual cell signals and analyze how these signals propagate is lost because of the close proximity of different cell types in explants and organoids, or difficulties in manipulating each cell culture chamber in transwell systems. Additionally, most of these models are not compatible with direct imaging of cellular migration, cellular transition, and ECM degradation and thus cannot provide a detailed view or direct evidence of cell-cell interactions.

Traditionally, transwell culture systems are used to study fetal membrane signaling (46); however, their usefulness is limited because of the following reasons: 1) lack of physiologic separation of cell types in coculture; 2) controlling their respective microenvironments in coculture is not possible, and local application of stimuli to only 1 compartment is difficult; 3) direct monitoring of any collagen matrix degradation caused by cellular migration is not possible; 4) imaging of cellular migrations and transitions is limited; and 5) a low signal-to-noise ratio caused by a large culture volume hampers studies on biomarker kinetics. Compared with transwell cultures, the AM-OOC model uses a significantly fewer number of cells (2-fold lower than 24-well transwell culture), which is important because the cell source is quite limited from the human amnion membrane. Additionally, it allows for better interactions between cell layers while providing sensitive measurement capabilities of membrane permeability, biomolecule propagation (e.g., cytokines, growth factors, extracellular vesicles), and signaling pathways. Overcoming the limitations of conventional approaches, the developed AM-OOC allows for real-time in-depth imaging of cellular processes while controlling fluid and treatment flow in coculture chambers that are physically and fluidically isolated yet still allow cell-cell communication. These unique features of the developed AM-OOC allow analysis of complex interconnected biochemical and physiologic responses while maintaining cell viability.

The particular model presented here is focused on visualizing cellular migration and transition, and, therefore, contains a few limitations for conducting other types of studies. In utero amnion membranes comprise AECs connected to an ECM containing AMCs by a 13-µm thick basement membrane (8). Although we included type IV collagen in between our cell layers, the ECM fabric where AMCs are often located is not included in the model, and the influence of that component in migration and transition is still unclear. Thus, future improvement to the device will include 1) shortening the microchannels to properly represent MFs, 2) adding dynamic medium flow to the AEC chamber to induce cellular shear stress normally impacted on AECs by amniotic fluid, and 3) fabrication of additional chambers to culture primary chorion trophoblast and decidual cells. Such improvements will result in recreating the full fetal membrane on OOC format.

The AM-OOC method developed here allowed us to overcome several limitations of traditional 2D and 3D culture systems in investigating amnion membrane cellular and collagen characteristics and interactions. Future designs of this model will include the fetal membrane cells as well as a decidual layer to represent the full feto-maternal interface to study their functions during physiologic and pathologic pregnancies.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

2/3D

2/3 dimensional

AEC

amnion epithelial cell

AM-OOC

amnion membrane organ-on-chip

AMC

amnion mesenchymal cell

CK-18

cytokeratin-18

CSE

cigarette smoke extract

ECM

extracellular matrix

EMT

epithelial-to-mesenchymal transition

GFP

green fluorescent protein

GM-CSF

granulocyte M-CSF

MET

mesenchymal-to-epithelial transition

MF

microfracture

NAC

N-acetyl-l-cysteine

OOC

organ-on-chip

OS

oxidative stress

PDMS

polydimethyl siloxane

RFP

red fluorescent protein

UTMB

University of Texas Medical Branch

AUTHOR CONTRIBUTIONS

L. Richardson designed and conducted experiments, performed data analysis, and drafted the manuscript; S. Jeong and S. Kim fabricated devices and conducted fluidic isolation experiments and data interpretation; and A. Han and R. Menon conceived the project, designed experiments, helped with data analysis and interpretation, and prepared manuscript.