Ex vivo model for assessing fetal membrane integrity and therapeutic strategies

Abstract

The fetal membranes (FM) are essential for maintaining the fetal–maternal interface and ensuring mechanical and barrier integrity throughout pregnancy. Preterm prelabor rupture of FM (pPROM) and iatrogenic pPROM (iPPROM) following fetal surgery, remain major causes of preterm birth. To better understand potential FM repair mechanisms and develop preventive strategies, experimental models that remain stable and functional beyond only a few days of culture are particularly needed. Here, we present a modular ex vivo model that enables up to 21 days culture of full-thickness human FM. Quantification of DNA content, ATP levels, glucose consumption, and lactate production suggest sustained but gradually adapting metabolic activity. The largely intact epithelial and mesenchymal organization and FM barrier function reveal maintained overall structural integrity. Additionally, mechanical testing demonstrates highest FM rupture pressure resistance close to the placental region with a progressive decrease with increasing distance from the placenta and preserved mechanical stability after 14 days of culture. Finally, puncture experiments show a time-dependent collagen fiber realignment around puncture-induced FM defects. Together, this accessible and adaptable platform provides a human-based experimental toolbox for investigating FM biology, evaluating biomaterials, and studying defect healing in a clinically relevant context. By enabling extended culture and integrated functional readouts, the model may help to bridge the translational gap between short-term in vitro systems and in vivo studies in maternal–fetal medicine.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-46366-4.

Introduction

Preterm births, occurring before 37 weeks of gestation affect approximately 12% of all pregnancies worldwide. Among these, a dysfunction of the fetal membrane (FM), preterm prelabor rupture of the FM (pPROM), accounts for 40% of preterm births¹. Another contributor to preterm birth is iatrogenic preterm prelabor rupture of the FM (iPPROM) that, dependent on the type of intervention, occurs in 30–100% of the cases after surgical treatments of pregnancy complications or fetal malformations². Since intrauterine surgical treatments become routine in specialized centers and likely a broader range of malformations will be treated in the future, the number of iPPROM cases is expected to increase and treatments to prevent this are urgently needed^2,3.

FM consist of two distinct layers, the amnion and the chorion, both of which are of fetal origin⁴(Fig. 1). The amnion, the innermost layer, serves as the primary load-bearing structure and is in direct contact with the amniotic fluid. Its epithelium rests on a basement membrane and a fibroblast-rich mesenchymal layer, which contains extracellular matrix (ECM) molecules critical for structural integrity^5,6. The spongy layer connects the amnion with the subjacent chorion and absorbs mechanical stress through its proteoglycan-rich composition^5,7. The chorion, the thicker layer, is primarily composed of trophoblast cells that interface directly with the maternal decidua and provides immunological protection⁶. Together, the unique and complex FM tissue composition constitutes a physical and immunological barrier between the fetus and the mother, essential for maintaining fetal health and sustaining pregnancy. Single-cell transcriptomics data from Pique-Regi et al.8 revealed a wide variety of cell types in the full fetal membrane, including immune cells (T cells, B cells, NK cells, monocytes, macrophages), trophoblasts (syncytiotrophoblasts, cytotrophoblasts, extravillous trophoblasts), stromal cells, fibroblasts, hematopoietic stem cells, lymphatic endothelial cells, endometrial cells, and decidual cells. Epithelial cells are also frequently reported in literature^9,10.

Fig. 1.

Structural Organization of the Human Fetal Membrane Graphic showing a pregnancy with the FM lining the uterine cavity (left; graphic created with Gemini AI). Pseudo-coloured SEM image showing the FM structure and corresponding layer denotation (right; figure adapted from Avilla et al., 2022⁵).

Factors described to influence the occurrence of iPPROM are the duration of fetoscopic intervention, the number and size of employed catheters, chorioamnion separation, and bleeding during the operation². Most of these factors seem to relate to the degree of FM damage, which however do not spontaneously heal after surgical fetal interventions^11,12. Moreover, in iPPROM, the failure of FM often does not occur through the surgery-induced defect. These observations suggest that instead of mechanical FM failures, the loss of FM barrier functions is responsible for iPPROM. Therefore, treatments to stabilize operation-induced FM defects and enable their closure by plugging, sealing or even healing-inducing strategies have been investigated for the prevention of iPPROM and consequently the prolongation of pregnancies⁵.

Initial approaches for plugging and sealing of FM have relied on naturally occurring biomaterials such as collagen sponges, or blood components such as fibrin or even platelet and fibrin mixtures (amniopatch)⁵. However, when used in human patients, these materials showed insufficient stability or unreliable closure of the FM defect, such that no major treatment benefits could be observed^13,14,15. Similarly, ex vivo and preclinical animal models confirmed the insufficient stability and sealing properties of fibrin¹⁶. Therefore, various engineered synthetic biomaterials, such as tissue glues, cyanoacrylate-based sealing patches, or semi-rigid patches integrated with hydroxypropyl methylcellulose (HPMC) or dopaminated-hyaluronic acid, have been employed^17,18. While some of these biomaterials have shown promising results in ex vivo and in vivo models, the development of strategies and devices for their precise and reliable application is still ongoing^5,17,19. Additionally, these approaches need to be tested in models that mimic the human situation as closely as possible. While for such tests in vivo studies remain important, each animal model comes with significant limitations. In general, in most mammals, early embryonic support involves a choriovitelline placenta that is later replaced by a chorioallantoic placenta, whereas in humans and other haplorrhine primates, early development of the extraembryonic mesoderm leads to formation of a secondary yolk sac within the exocoelom. In contrast to humans, many rodents possess an inverted visceral yolk sac that persists throughout gestation and performs functions assumed in humans by the syncytiotrophoblast, and most mammals develop a prominent allantoic cavity, which is absent in humans²⁰. Additionally for instance, rodents and rabbits have shorter gestational periods and give birth to large litters, making them less suitable for studying human-like fetal membrane dynamics^21,22. Swine and ovine models due to their longer gestational periods allow the long-term testing of biomaterials. However, swine FM heal spontaneously²³ and ovine pregnancies exhibit significant anatomical differences, such as multi-cotyledonary placentas and detached uterine connections²⁴. Although rhesus monkeys have FM that anatomically and functionally resemble those of humans, their use is ethically debatable and therefore not suitable for the testing of various different biomaterials.

To investigate FM healing functions amnion mesenchymal cells (AMCs) and amnion epithelial cells (AECs) have been used in different two-dimensional (2D) and three-dimensional (3D) cultures^{25,26,27,28,29}. In these cultures, cytokines and growth factors with the capacity to promote migration and extracellular matrix deposition in AMC and AEC have been identified³⁰. However, to further improve treatments for the healing of FM, it will be key to understand the inherent potential of FM cells, the extent the ECM controls and potentially inhibits their function, the role of the maternal and the fetal immune system, and the change in biomechanics at the edge of the defect³¹. The foundation to answer these questions can be established by analyzing the composition of human FM ^{32, 33, 34, 35, 36} and cultures of human amniotic mesenchymal stromal and epithelial cells in both 2D and 3D systems^26,28,37. However, 2D and 3D in vitro models lack the dynamic interactions between the different cell types and ECM components that characterize the intact FM.

Since the 1990s, multiple groups have established short-term FM explant cultures using floating or hydrogel embedded tissue^38,39,40,41 sections or transwell-based systems^42,43,44,45, allowing stimulation from the amniotic or chorionic side and analysis of cytokine secretion, matrix metalloproteinase (MMP) activity, and prostaglandin production. These approaches have been instrumental in defining key inflammatory cascades and degradation pathways that contribute to membrane weakening. In terms of fetal membrane defects, Costa et al.⁴⁶ developed an ex vivo system using amniotic tissue mounted on CellCrown inserts to investigate healing mechanisms in small defects (e.g., 0.8 mm punctures mimicking amniocentesis). This model revealed key processes such as collagen alignment around defects from clinical samples and the ex vivo model^31,46. Although these insert-based systems are well established, widely used, and of substantial value to the field, they present several limitations that create a niche for improved platforms. First, the defect sizes typically investigated (e.g., sub-millimeter punctures) are considerably smaller than the ~ 3 mm incisions generated during fetoscopic interventions, where defect size is a critical determinant of healing dynamics and mechanical failure risk⁴⁷. Second, assessment of mechanical stability in conventional setups often requires specialized and costly equipment, which may limit accessibility and broader adoption of biomechanical testing approaches. Third, while transwell inserts are available in multiple formats (e.g., 6-, 12-, or 24-well plates), their geometry and dimensions remain restricted, potentially constraining the evaluation of certain sealing strategies, such as umbrella-type closure devices developed by Devaud et al.⁴⁸ Finally, as biomaterial-based sealing approaches are intended to provide durable closure, there is a clear need for culture systems that support longer-term structural and functional assessment, which is not the primary focus of most existing acute explant models.

To address these challenges, we designed a novel ex vivo model to maintain the native structure and to replicate healing dynamics of human fetal membrane using 3D-printed devices for the clamping, the culturing and mechanical stimulation of small patches of fetal membrane. This model enabled long-term investigations, while reducing reliance on animal studies, and provide a cost-effective platform for advancing research into sPPROM and iPPROM.

Results

Design and development of a novel ex vivo full-thickness FM model

To establish a physiologically relevant alternative to vivo systems, we developed an ex vivo model that enables the culture and mechanical testing of full-thickness fetal membrane (FM) (Fig. 2a). The system accommodates both amnion and chorion layers and allows maintenance of identifiable epithelial and stromal layers during culture. The modular design integrates several functional requirements within a single platform. The lower component of the device contains micro-spikes that secure the FM by puncturing, allowing optional pre-tensioning and preventing displacement during experimentation. The upper component locks the tissue in place (Fig. 2a, label 1). This configuration enables stable clamping of both amnion and chorion layers. For applications requiring enhanced sealing, three auxiliary openings allow additional fixation by suturing (Fig. 2a, label 2). The upper structure includes small cylindrical compartments that permit differential exposure of each membrane side to distinct media, enabling side-specific stimulation. The culture platform ensures sufficient nutrient diffusion to both surfaces of the FM (Fig. 2a, label 3). In addition, the device is compatible with modular add-ons, including a custom 3D-printed pressure chamber for mechanical testing. Together, this design enables long-term culture, biochemical manipulation, and biomechanical assessment of full-thickness human FM within a single adaptable system (Fig. 2a, label 4–6). Figure 2b presents a direct comparison between conventional insert systems and the newly developed 3D-printed devices.

Fig. 2.

Overview of the FM ex vivo model as a toolbox. (a) On the top left (1), the 3D-printed lower and upper parts, made from polylactic acid (PLA), are shown. On the right (2), the assembled ex vivo model is illustrated: The fetal membrane is placed onto the 3D-printed part, punctured with a 3 mm hole, and secured with a nylon suture as described. On the second row on the right (3), the 3D-printed culture platform is depicted, where the device is positioned and subsequently immersed in culture medium. The possible readouts include temporal monitoring of cell culture medium composition (4), such as glucose consumption over time, immunohistochemistry and confocal imaging (5), and mechanical stability measurements (6). (b) Comparison of the well-established state-of-the-art transwell⁴² model and the newly developed device, highlighting selected distinguishing features.

Long-term preservation of FM cellular viability in the ex vivo model

To evaluate tissue viability during extended culture, live/dead staining was performed at Day 0 and Day 21 (Fig. 3a). Viable cells were detected throughout the chorion layer at both time points, due to bottom-up imaging of the device and incompatibility of live/dead staining with tissue clearing, assessment was limited to the chorion. Quantification of dead nuclei following tissue clearing and DAPI staining (Supplementary Information Fig. 2) further supported sustained cellular viability over 21 days. Histological analysis at Day 21 demonstrated largely preserved tissue architecture, including an intact epithelial layer, although localized disruptions were observed in some samples (Fig. 3b). To further assess structural preservation, a tissue clearing protocol compatible with confocal microscopy was applied. E-cadherin staining confirmed overall persistence of the epithelial layer at Day 21, with occasional focal areas of reduced epithelial continuity observed in some samples (Fig. 3c). Vimentin staining demonstrated the continued presence of mesenchymal cells throughout the membrane. These findings indicate that the ex vivo system supports maintenance of cellular viability and overall tissue structure for up to three weeks.

Fig. 3.

Long-term viability and structural preservation of full-thickness fetal membrane in the 3D device (a) Live/Dead staining of fetal membrane at Day 1 (left) and Day 21 (right). Viable cells are shown in green (Calcein-AM), and dead cells in red (ethidium homodimer-1). Imaging was performed from the chorion side. Scale bar: 200 μm. (b) Histological section of a fetal membrane cultured in the device for 21 days and stained with hematoxylin and eosin (H&E). The amnion is visible in the upper region of the image, and the chorion in the lower region. Scale bar: 200 μm. (c) Representative confocal images of immunofluorescence-stained fetal membrane on Day 0 and Day 21 of culture. Left panel: Epithelial integrity shown by E-cadherin (magenta) and nuclei stained with DAPI (blue). Middle panel: Higher magnification of the region indicated by the white box in the left image. Right panel: Mesenchymal cytoskeleton visualized by vimentin (yellow) with nuclei stained by DAPI (blue).

Scale bar: 100 μm.

Metabolic activity and barrier integrity of FM in long-term culture

To further assess the functional state of the FM during long-term culture, we evaluated DNA content, ATP levels, barrier integrity, and metabolic activity.

DNA content showed considerable inter-donor variability (Fig. 4a). However, no consistent increase or decrease over time was observed. Linear mixed-effects modeling, with Day as a fixed effect and Donor as a random effect, revealed no significant time-dependent change in DNA levels (β = 0.016, p = 0.215; all pairwise comparisons vs. Day 0, p > 0.05). In contrast, the dead control exhibited a marked reduction in DNA content, confirming assay sensitivity (Fig. 4a). These findings indicate that overall DNA levels remained stable throughout the culture period. Representative images of dead cells and nuclei staining and the corresponding quantification are provided in Supplementary Fig. 2.

Fig. 4.

Longitudinal assessment of viability, metabolic activity, barrier integrity, and glucose consumption in cultured full-thickness fetal membrane. a) Relative DNA content (measured using the PicoGreen fluorescence assay) at multiple time points from Day 0 to Day 21 in tissue punches obtained from the device. Individual data points represent the average of technical replicates per donor. Bars indicate the median of the three donors tested, and error bars represent the standard deviation. A dead control served as a negative control. All DNA values were normalized to Day 0. Statistical analysis was performed using a mixed linear regression model with ‘Day’ as a fixed effect and ‘Donor’ as a random effect to account for inter-donor variability. No significant change in DNA amount was observed over time (β = 0.016, p = 0.215, all pairwise comparisons vs. Day 0, p > 0.05). b) Relative ATP amount, quantified using the CellTiter-Glo luminescence assay, was assessed at multiple time points between Day 0 and Day 21 using tissue punches from the device. Individual data points represent the mean of technical replicates per donor. Bars indicate the median of three donors, with standard deviation shown. A dead control prepared at baseline served as a negative control. All values were normalized to Day 0. Statistical analysis was performed using a mixed linear regression model with ‘Day’ as a fixed effect and ‘Donor’ as a random effect to account for inter-donor variability. A significant decrease in ATP amount was observed over time (β =-0.021, p = 0.004, pairwise comparisons vs Day 0 significant results for day 21 (p = 0.002)). c) Barrier integrity assessed by fluorescence permeability at multiple time points between Day 0 and Day 21. A fluorescent tracer was applied to one side of the membrane, and fluorescence intensity on the opposite side was measured after 20 minutes. Individual data points represent the mean of technical replicates per donor. Bars indicate the median of three donors, with standard deviation shown. Dead control served as a reference for maximal permeability. Barrier integrity was calculated as 1 − (sample signal / dead control signal). All values were normalized to Day 0. Statistical analysis was performed using a mixed linear regression model with ‘Day’ as a fixed effect and ‘Donor’ as a random effect to account for inter-donor variability. No significant increase/decrease was observed over time (β =-0.014, p = 0.077, pairwise comparisons vs Day 0 significant results for day 21 (p = 0.022)). d) Longitudinal glucose consumption of fetal membrane samples over time. Samples from three independent donors were analyzed, each with three technical replicates. Donors are depicted in different colors. Glucose levels were measured every second day, and consumption was calculated relative to fresh medium. The black line represents the mean across all donors, and the grey shaded area indicates the standard deviation. While the overall 21-day trend was negative (β = -8.33 × 10^− 5, p = 0.013) due to a Day 0 to 1 acclimation spike, a Linear Mixed-Effects Model (LMM) restricted to the stable phase (Days 1–21) revealed a significant positive maturation trend (β = 6.91 × 10^− 5, p = 0.001). Categorical analysis confirmed that all post-starting timepoints were significantly lower than the Day 1 stress peak (p < 0.001).

ATP measurements likewise demonstrated substantial inter-donor variability (Fig. 4b). Pairwise comparison revealed a significant reduction in ATP levels at Day 21 compared to Day 0 (–55.6%, p = 0.002), while no other time points differed significantly from baseline. Dead controls showed markedly reduced ATP activity (Fig. 4b). Mixed-effects modeling identified a significant time-dependent decline in normalized ATP levels (β = − 0.021, p = 0.004), corresponding to an average decrease of approximately 2.1% per day relative to Day 0. These results suggest a gradual reduction in metabolic activity over time, while measurable ATP levels were maintained throughout the culture period.

Barrier integrity was assessed by applying FITC-dextran to the amniotic side and quantifying fluorescence on the opposite side after 20 min. Donor variability was again evident. The overall linear trend across the culture period did not reach statistical significance (β = 0.014, p = 0.077), indicating no consistent time-dependent increase or decrease in permeability. However, categorical comparison showed a significant increase in barrier integrity at Day 21 relative to Day 0 (mean increase 48.2%, p = 0.022). One donor exhibited barrier integrity values below those of the dead control, indicating that barrier function may be compromised in individual samples (Fig. 4c).

Analysis of glucose consumption revealed a biphasic metabolic pattern. Across the full 21-day period, a significant negative linear trend was observed (β = − 8.33 × 10⁻⁵ mmol/day, p = 0.013), largely driven by an elevated consumption between Day 0 and Day 1, consistent with an acute adaptation response to culture conditions. When the analysis was restricted to the post-acclimation phase (Days 1–21), a slight but statistically significant positive trend was detected (β = 6.91 × 10⁻⁵, p = 0.001), indicating sustained and moderately increasing glucose consumption during prolonged culture (Fig. 4d). A similar trend was observed in the lactate production (Supplementary Fig. 3).

Collectively, these data demonstrate preserved DNA content, measurable metabolic activity, and largely maintained barrier function over 21 days of culture, despite inter-donor variability and a gradual decline in ATP levels.

Mechanical stability and regional variation in FM strength

For the long-term mechanical stability testing of FM and the evaluation of the capacity of biomaterials to seal FM defects, the ex vivo models with spanned FM were subjected to pressure measurements. FM maintained their mechanical stability during 14 days in culture (Fig. 5a). However, significant variability was observed between technical replicates. Thanks to the small sample size required in our setup, the variability of mechanical stability across FM locations could be systematically examined. Results showed a distance dependent decrease in stability in FM, with increasing distance to the placental region (Fig. 5b).

Fig. 5.

Rupture pressure measurements reveal stability of cultured and treated FM. a) Comparison of FM rupture pressure after 0 and 14 days in culture (not significant). Data are shown for three different donors (depicted in different colors) and summarized as a boxplot. b) Rupture pressure at test sites located at increasing distances from the placental rim, using samples from four different donors (shown in different colors), combined and shown as a boxplot. Statistical analysis was performed using a mixed linear regression model with ‘Distance’ as a fixed effect and ‘Donor’ as a random effect to account for inter-donor variability. A significant decrease was observed over time (β =-0.084, p < 0.001, pairwise comparisons vs 2 cm significant results for 8 and 10 cm (p = 0.039 and p < 0.001)). c) FM rupture pressure after 0 or 2 days of treatment with thrombin (not significant), with data from six different donors (depicted in different colors) shown as a boxplot. d) Representative images of 3 mm × 3 mm FM defects sealed with fibrin glue and cultured for 0 and 14 days. Scale bar: 5 mm.

Sensitivity of the model to biochemical perturbations

To further assess the model’s sensitivity to biochemical perturbations, the FM were treated with thrombin, following the method described by Moore et al.⁴². The median rupture pressure of the thrombin treated group (1.079 bar) was 28% lower than the median rupture pressure of the control group by (1.494 bar) (Fig. 5c). However, the observed large standard deviation made the reduction with a sample size of six donors with each five technical replicates non-significant. The model was also tested on FM treated with fibrin glue after simulating an incision site. However, after 14 days the fibrin-based material was completely dissolved, which made stability measurements impossible (Fig. 5d).

Collagen fiber realignment around FM defects during culture

It is known that in fetoscopically treated FM collagen aligns around the incision³¹. Therefore, we investigated whether in our model we could replicate this behavior. To examine collagen fiber orientations, multiphoton microscopy was conducted on day 0, 3, 7, 14 and 21 of culture. A trend toward collagen alignment around the hole was evident (Fig. 6a). Fourier transformation of the images revealed a peak, reflecting enhanced fiber alignment on day 21 that was not observed on day 0 (Fig. 6b). Entropy analysis, which correlates with the distribution of alignment, indicated a significantly increased alignment over time, consistent with findings by Barrett et al. ³¹ These data suggest that the most pronounced alignment occurred between days 3 and 7 (Fig. 6c). Furthermore, immunohistochemistry images for E-cadherin and vimentin revealed morphological differences around the defect site (Fig. 6d and e).

Fig. 6.

Rearrangements of cells and collagen in the vicinity of FM defects. (a) Second harmonic generation (SHG) images of collagen fibers in the vicinity of FM defects at day 0 (top) and day 21 (bottom). Arrows indicate regions where collagen alignment is subtly observable. Scale bar = 100 μm. (b) SHG image-based collagen fiber orientation distribution on day 0 (top) and day 21 (bottom) and (c) Normalized entropy plot of collagen fiber alignment around the defect over time. The individual donors are depicted in different colors as dot. Statistical analysis using OLS regression shows a β =-0.0188, and a p-value of 0.003. (d) Representative immunofluorescence-stained FM at days 0 and 21 of culture. Nuclei are stained with DAPI (blue), and epithelial cell adhesion sites with E-cadherin (magenta). Scale bar = 300 μm. (e) Time-course immunofluorescent staining of FM from day 0 to day 21 of culture. Nuclei are stained with DAPI (blue), and mesenchymal cell cytoskeleton with vimentin (yellow). Scale bar: 300 μm.

Discussion

In this study, we introduce a versatile ex vivo model of human fetal membrane (FM) designed to complement and extend existing experimental platforms. While several ex vivo systems have successfully addressed specific aspects of FM biology, many are optimized for short-term assessment or tailored to narrowly defined applications. Our approach emphasizes extended culture duration, structural accessibility, and experimental flexibility, thereby broadening the scope of investigations that can be performed within a single standardized platform.

A key feature of our model is its modular and adaptable design. The system allows adjustment of membrane size and defect geometry, making it particularly suitable for tissue engineering applications and for studying repair strategies following fetoscopic interventions. The ability to generate controlled defects of variable dimensions, including defect size of the fetoscopic surgery provides a reproducible framework for investigating healing responses and for evaluating biomaterials or sealing devices, including larger constructs such as the umbrella device⁴⁸ or newly developed collagen plugs⁴⁹. This flexibility enables testing conditions that more closely approximate clinically relevant defect sizes. By providing a single adaptable platform for multiple experimental questions, the model may also facilitate methodological standardization and cross-study comparability.

Long-term tissue viability represents a central requirement for ex vivo systems intended to study membrane repair, biomaterial integration, or chronic tissue responses. To our knowledge, this is the first ex vivo FM device systematically evaluated for functional performance over a 21-day culture period. DNA content remained stable throughout the observation period, and no significant decline in its overall amount was detected during the first 14 days. A significant reduction in ATP levels became apparent only after three weeks, indicating a gradual decline in metabolic activity. Interestingly, this decrease occurred during the acclimated phase and coinciding with a slightly increasing glucose consumption, together with stable DNA levels. Therefore, this pattern may reflect metabolic adaptation rather than acute loss of cell function although it could also indicate the onset of cellular senescence, a phenomenon previously described in fetal membrane biology⁵⁰. Toward the third week of culture, cells may experience metabolic stress, leading to increased glucose consumption despite declining ATP levels, potentially reflecting the emergence of senescent cells with altered energy metabolism⁵¹. Further mechanistic studies will be required to clarify this aspect. However, if this is the case, it would provide an interesting opportunity to study aging processes in FM.

Barrier function analysis yielded heterogeneous results. While no consistent linear deterioration in permeability was observed, inter-donor variability was pronounced, and individual samples displayed compromised barrier properties at specific time points—even below dead control levels. These findings suggest that barrier integrity can be maintained in culture but may be sensitive to donor-specific factors, tissue site or local structural alterations. Histological and confocal analyses confirmed the persistence of epithelial and mesenchymal layers after 21 days. The epithelial layer, visualized by E-cadherin staining, remained detectable but exhibited increased discontinuities compared to Day 0, indicating partial structural alteration over time.

Ideally, metabolic and structural parameters would be preserved over three weeks of FM culture. However, given the absence of practical long-term in vitro alternatives and the ethical and logistical constraints of in vivo experimentation, a model that maintains structural integrity and measurable activity for up to 14 days, and partially up to 21 days, already represents a meaningful advance. This extended culture period may offer biologically relevant insights, particularly for applications requiring evaluation of long-term device stability, material–tissue interactions, or delayed healing responses.

For studies specifically targeting epithelial integrity, reducing the culture duration to 14 days or supplementing culture media with growth factors such as EGF may improve long-term epithelial stability. Future optimization of culture conditions may further extend functional longevity.

So far, biomaterials such as fibrin glue have been tested in rabbit models as a sealing agent for incision sites²²,⁵². The extended culture period with our ex vivo model could be particularly beneficial for evaluating the long-term stability of biomaterials ex vivo. This would help to preselect suitable biomaterials and reveal potential material instabilities prior to initial in vivo testing. For instance, the fibrin degradation reported by Papadopulos et al. ⁵² might have been anticipated earlier with the support of such an ex vivo approach, potentially reducing the need for some animal experiments.

Understanding the mechanical properties of FM is critical for studying PPROM and FM biology. Vergote et al.⁵³ and Robmann et al.⁵⁴ developed excellent systems to assess mechanical stability; however, these require large tissue samples and are restricted to testing within hours after delivery. By minimizing the required tissue size and extending the duration of tissue viability, our model provides several advantages. We demonstrated that FM, when cultured for two weeks, maintain their mechanical properties. Additionally, we observed a decreased mechanical stability of FM with increasing distance from the placenta, with their strengths halving between 2 cm and 10 cm away. To further challenge the system, we evaluated thrombin treatment, which has previously been reported by Moore⁴² to reduce membrane stability. In our model, thrombin exposure resulted in decreased mechanical strength. However, this effect did not reach statistical significance, largely due to pronounced inter-donor variability. These findings highlight both a biological reality, namely, substantial donor-dependent differences in FM mechanics, and a technical consideration regarding assay sensitivity. The choice of mechanical testing platform should therefore be guided by the specific experimental question. For exploratory studies, comparative screening, or experiments requiring multiple conditions per donor, the present low-cost and easily adaptable system provides a practical and efficient solution. In contrast, studies requiring maximal sensitivity to detect subtle mechanical differences may benefit from more sophisticated and higher-resolution devices, such as those described by Kumar et al.⁴² While our system exhibited larger variability compared to existing models⁵³, these limitations are balanced by its simplicity and rapid implementation.

To improve treatment options for iPPROM, an accurate representation of the clinically induced FM defects is essential. Animal studies have shown that smaller defects tend to heal spontaneously, whereas larger ones remain unhealed²¹. Our model’s 3 mm puncture size closely mimics the defects created during human fetal interventions⁵⁵, providing a clinically relevant testing environment. Collagen alignment, a critical factor in healing inhibition, was observed in our model, consistent with Barrett et al.’s findings from clinical samples³¹. This alignment highlights the model’s ability to replicate clinical conditions, which are often overlooked in simpler in vitro systems. Our study further demonstrated that collagen alignment is not a gradual process but rather occurs predominantly between days 3 and 7, possibly identifying a critical timeframe for therapeutic interventions targeting alignment. Morphological changes at the wound edge, including altered vimentin and E-cadherin expression, were also observed. This may reflect active cellular rearrangement, as previously described by Barrett et al., or could result from enhanced staining or illumination at the tissue edges. However, further experiments will be required to clarify the mechanisms.

Despite its many advantages, one of the limitations of this ex vivo model relates to the minimally required FM tissue size. This might not be compatible with high-throughput screening approaches or with investigation of developmental aspects, which require large enough FM tissue pieces of early gestational ages. Another major limitation, common to many ex vivo models, is the inherent donor-to-donor variability. This variability can obscure biological effects due to large standard deviations, as observed in our study, where the limited number of donors and replicates per condition constrain statistical power. Additionally, the temporal changes in the composition and behaviour of different cell types, particularly immune cells, within the cultured tissue are not fully understood and represent an important area for future study. This omission may significantly influence cellular growth and behavior. Although overall DNA content remained stable and glucose consumption was sustained during culture, the observed decline in ATP levels over time indicates that functional alterations occur despite preserved cell numbers. Finally, the culture medium in this study relied on fetal bovine serum (FBS), which is associated with ethical concerns and variability. While FBS remains the most effective option for long-term cultures, developing serum-free alternatives is a priority for future research.

Another limitation is the use of FM from preterm or term deliveries, as models using earlier gestational stages are unattainable for ethical and practical reasons. These early-stage membrane would be mandatory and valuable for certain research questions, such as those focused on developmental processes. Nonetheless, the advantages of this model, including its ability to mimic human membrane architecture and dynamics, make it a valuable tool for many research applications.

The ex vivo model was intentionally designed for simplicity and accessibility. Its 3D-printed components and pressure-testing setup are cost-effective and easily reproducible, with 3D printers now widely available for under $200. The straightforward design makes the model accessible to a broad range of laboratories, promoting its adoption and supporting the reduction in animal trials. It is designed as a versatile toolbox that can be easily adapted or supplemented with additional features for any specific research question.

In conclusion, this ex vivo model has significant potential for advancing research on iPPROM and PPROM. It enables investigation into strategies for improving FM stability to prevent PPROM, as well as testing of biomaterials with an increased throughput to assess their long-term stability and healing properties. Moreover, the model allows the exploration of healing-promoting signals and biomaterials in a clinically relevant environment, reducing the translational gap between preclinical and clinical studies. By providing a human-based platform with minimal ethical concerns and great physiological relevance, this model offers a step toward more effective and responsible research practices.

Methods

Sample preparation

The study protocol was approved by the Ethics Committee of the Canton of Zurich (BASEC_Nr: 2023 − 00110), and written informed consent was obtained from all participants. All methods were performed in accordance with the relevant guidelines and regulations. Fetal membrane (FM) was collected from term pregnancies (37–40 weeks) immediately after elective cesarean delivery. Exclusion criteria included infections, premature rupture of membrane (PROM), multiple gestations, and chromosomal abnormalities. Unless otherwise specified, FM samples were taken at least 5 cm from the placenta, stored in phosphate-buffered saline (PBS) and kept on ice until further processing (not more than 15 min). All donors were assigned a unique identifier (1–19). If the same identifier appears in multiple figures, this indicates that the same donor sample was used across those experiments. Tissues were washed three times in cold (4 °C) PBS to remove blood and debris. Afterwards the full thickness membrane was used.

A custom-made, single-use 3D-printed device (Creality Ender-5 S1, polylactide filament) was used to clamp FM samples, as shown in Fig. 1. (STL files in the supplementary) The device, consisting of two interlocking components with an 8 mm central aperture, was sterilized for 15 min in ethanol and then exposed to UV light for 30 min. Afterwards it was incubated in 1% bovine serum albumin in PBS for 30 min before use. Full-thickness FM specimens (amnion and chorion) were mounted between the device components and trimmed to size using a scalpel. Amnion was always facing upwards. Fetoscopic defects were simulated by punching 3 mm holes into the FM using a biopsy puncher. Then the devices were placed on a cell culture platform (Fig. 1) to facilitate nutrient and waste exchange across both sides of the FM and cultured in six-well plates with 5 ml of MEMα supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (10 U/mL penicillin, 10 µg/mL streptomycin) referred to as MEMα++. Plates were incubated at 37 °C with 5% CO₂ in a humidified chamber, and medium was exchanged every 2–3 days.

Tissue clearance and confocal imaging

To facilitate imaging, consecutive clearing steps were conducted by incubating clamped and cultured tissue samples in Tris Buffered Saline (50mM) with Tween 20 (TBST) containing 25%, 50%, and 75% glycerol at room temperature for 5 min each. Samples were then stored in 75% glycerol at 4 °C until imaging. Confocal microscopy (Leica TCS SP5, 10× magnification) was used to acquire Z-stack images (10 μm intervals), which were processed in Fiji (ImageJ 1.54f). Maximum intensity projections were generated from the acquired image stacks.

Cell activity and viability assessment

To evaluate metabolic activity, 1.5 ml culture medium was collected before each media change (every second day), frozen at − 20 °C and later analyzed for glucose levels using an ABL90 FLEX blood gas analyzer (Radiometer). Glucose consumption ([mmol/day]) was calculated as the difference from baseline levels in fresh medium. Membrane from three different donors were used (biological replicate (N) = 3) and tested in three different ex vivo models (technical replicate (n) = 3).

To evaluate cell viability, one cultured ex vivo model from the same donor was analyzed at Day 1 and one at Day 21. Samples were incubated with ethidium homodimer-1 and Calcein-AM (Invitrogen; 1:1000 dilution in culture medium) for 30 min at 37 °C. The ex vivo model was then imaged using confocal microscopy, and images were processed with Fiji (ImageJ). Further to assess cell viability, cultured ex vivo models were incubated with ethidium homodimer-1 (EthD-1, InvitrogenTM, 1:1000 in medium) for 30 min at 37 °C, washed twice with PBS, and fixed in 4% paraformaldehyde (PFA) for 4 h at room temperature on a shaker (300 rpm). Samples were blocked and permeabilized with PBS containing 1% BSA (Sigma-Aldrich) and 0.2% Triton™ X-100 (Sigma-Aldrich) before stained with 4′, 6-diamidino-2-phenylindole (1:1000, DAPI; Sigma). After tissue clearance, (as described above), confocal imaging and generating the maximum projection, cell viability was determined by calculating the ratio of dead cells to total number of nuclei in Fiji (ImageJ 1.54f). FM samples from three different donors were used (biological replicate (N) = 3) and tested in two different ex vivo models (technical replicate (n) = 2).

FITC-Dextran permeability assay

At multiple time points (day 0, 1, 3, 7, 14, and 21), epithelial barrier integrity was assessed using a FITC-dextran permeability assay adapted from Faber et al.⁵⁶ Briefly, devices were washed three times with PBS. Subsequently, 100 µL of 1 mg/mL FITC-dextran (20 kDa; Sigma) dissolved in phenol red-free DMEM/F-12 (Gibco) was added to the upper compartment of the device. The lower compartment was filled with 1 mL phenol red-free DMEM/F-12. Samples were incubated for 20 min at room temperature.

After incubation, the medium from the lower compartment was collected and stored at − 20 °C until analysis. For fluorescence measurement, 200 µL of each sample was transferred to a black 96-well plate (Falcon), and fluorescence intensity was measured using a Cytation 5 plate reader (excitation 485 nm, emission 528 nm).

For dead controls (maximum permeability), devices were incubated in a 60 °C water-bath for 30 min prior to the assay. FM from three independent donors were used (biological replicates, N = 3). For each donor, two independent ex vivo devices were analyzed (technical replicates, n = 2) in all three experiments. All data processing and statistical analyses were performed in Python (version 3.12.7). Barrier integrity was calculated as 1 − (sample signal / dead control signal), and values were normalized to Day 0 where indicated.

Metabolic activity (CellTiter-Glo assay)

To assess metabolic activity, the same ex vivo model used for the permeability assay was washed three times with PBS. Tissue samples were obtained using a 2 mm biopsy punch. Each tissue punch was transferred to a tube containing 100 µL PBS, followed by the addition of 100 µL CellTiter-Glo reagent (Promega). The CellTiter-Glo assay is based on a luciferase reaction that generates luminescence proportional to the amount of intracellular ATP, thereby serving as a measure of metabolically active cells. Samples were vortexed for 1 min to ensure complete tissue disruption and incubated for 30 min at room temperature. Subsequently, 100 µL of the lysate was transferred to a white 96-well plate, and luminescence was measured using the Cytation 5 plate reader according to the manufacturer’s instructions. FM from three independent donors were used (biological replicates, N = 3). For each donor, two independent ex vivo devices were analyzed (technical replicates, n = 2) in all three experiments. All data processing and statistical analyses were performed in Python.

DNA quantification (PicoGreen assay)

To determine DNA content over the 21-day time course, FM samples were collected from the same device, as previously used for the permeability assay, using a biopsy punch (1 mm, KAI) and stored at − 80 °C until analysis.

Frozen samples were lysed in 50 µL lysis buffer (0.56 M 2-amino-2-methyl-1-propanol [Merck], 0.2% Triton X-100, pH 10) for 2 h at room temperature, with vortexing (1 min) performed three times during incubation. DNA quantification was then performed using the Quant-iT™ PicoGreen™ dsDNA Reagent Kit (Invitrogen) according to the manufacturer’s instructions. Briefly, 50 µL of PicoGreen working solution was added to each sample, vortexed, and incubated for 5 min at room temperature protected from light. Subsequently, 50 µL of each sample was transferred to a black 96-well plate (Falcon), and fluorescence was measured using the Cytation 5 plate reader (excitation 485 nm, emission 528 nm). FM from three independent donors were used (biological replicates, N = 3). For each donor, two independent ex vivo devices were analyzed (technical replicates, n = 2) in all three experiments. All data processing and statistical analyses were performed in Python.

Morphological analysis

Ex vivo models with or without defects were cultured for up to 3 weeks and fixed in 4% PFA for 4 h at room temperature on a shaker (300 rpm). The samples were then incubated for 4 h at room temperature in a solution designed for permeabilization and blocking, which contained 1% BSA (fraction V) and 0.3% Triton X-100 dissolved in PBS. After permeabilization, the tissues were incubated in PBS containing 1% BSA, 0.1% Triton X-100, and primary antibodies against vimentin (1:200; Abcam, ab8069) and E-cadherin (1:200; Abcam, ab40772) at room temperature for two days to label mesenchymal and epithelial cells, respectively. Following this, secondary staining was performed over an additional two-day period using Alexa Fluor^® 488 anti-mouse (1:200; Abcam, ab150113), DyLight™ 649 anti-rabbit (1:200; BioLegend, 406406), phalloidin-rhodamine (1:4000; Invitrogen™) and DAPI in PBS. The samples were then cleared and imaged as described above.

Collagen alignment

FM samples in the ex vivo model with or without defects were cultured in the for up to 3 weeks and fixed in 4% PFA for 4 h at room temperature on a shaker (300 rpm). Collagen fiber orientation at defect sites was visualized by second harmonic generation (SHG) imaging using a Leica SP8 MP multiphoton microscope (25×, water immersion). ImageJ was used to calculate orientation distributions with the directionality tool. The entropy was then calculated with the scipy.stats package (Python 3.12.7)⁵⁷. The Shannon entropy was calculated for each dataset (as a measure of alignment randomness) using the alignment values across orientation angles. All values were min–max normalized to ensure comparability between batches. Devices were grouped into batches representing individual donors. Subsequently, entropy values were averaged per device and day. For each batch, mean entropy was normalized and aggregated by time-point to yield donor-specific entropy progression. Average entropy values per donor were plotted using Seaborn (Python 3.12.7), and statistical analysis was conducted with ordinary least squares (OLS) regression (statsmodels, Python 3.12.7). Statistical significance was determined by OLS regression, t-tests, or two-way ANOVA with Tukey’s multiple comparison test (*p < 0.05, **p < 0.01, ***p < 0.001).

FM from three different donors were used (biological replicate (N) = 3) and tested in two different ex vivo models (technical replicate (n) = 2).

Pressure of rupture (POR) testing

A 3D-printed pressure device was developed to measure the mechanical stability of FM samples (Fig. 7). (STL files in the supplementary) During the assembly of the pressure device, a Newton meter was used to tighten the screws to 12 Newtons. Air pressure was gradually applied through manually opening a valve until rupture (Fig. 7), with the pressure of rupture (POR) recorded using a Honeywell ABP sensor connected to an Arduino UNO. Baseline POR measurements were taken on intact FM samples on day 0 or after 14 days of culture.

Fig. 7.

Pressure device for testing mechanical stability. (a) Rendered model of the pressure device. The ex vivo model is positioned in the center, and the pressure chamber can be securely sealed using screws. (b) Schematic side view of the pressure device. The arrow at the bottom indicates the inlet for pressurized air (or optionally water). The pressurized medium increases the load on the fetal membrane. A pressure sensor near the outlet and close to the fetal membrane measures the applied pressure and transmits the data to an Arduino Uno for evaluation.

FM from three different donors were used (biological replicate (N) = 3) and tested in four to five different ex vivo models (technical replicate (n) = 4–5).

Thrombin treatment

To assess FM weakening, ex vivo models were treated with thrombin (10 U/ml, Tisseel Fibrin Sealant, Baxter) for 48 h in MEMα++. POR measurements were compared between thrombin-treated and untreated controls.

FM from three different donors were used (biological replicate (N) = 6) and tested in five different ex vivo models (technical replicate (n) = 5).

Regional Variability

FM samples were collected from varying distances (0–10 cm) from the placental rim, and POR was measured immediately after preparation. Samples were stored in PBS at 4 °C prior to testing and measured as previously mentioned.

FM from three different donors were used (biological replicate (N) = 3) and tested in one to three different ex vivo models (technical replicate (n) = 1–3).

Fibrin glue application

After drying FM tissues with defects in the ex vivo device, 70 µl of thrombin (Tisseel Fibrin Sealant) and fibrinogen (Tisseel Fibrin Sealant) solutions (1:1 ratio, known as fibrin glue) were combined and pipetted onto defects. After fourteen days of culturing the ex vivo models, the fibrin glue had dissolved, preventing any pressure measurement.

Manuscript preparation

ChatGPT (OpenAI) was used to assist in improving the clarity, flow, and language of the manuscript, including spell-checking and stylistic suggestions.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

L.M., N.O. and M.E. conceived the study. L.M., B.T., K.G., M.J. and R.O. performed the experiments. S.R. and L.M. developed the pressure device. L.M and B.T contributed to data analysis. All authors participated in writing the manuscript and reviewed and approved the final version.

Funding

This work was supported by the Innovationspool USZ.

Data availability

The raw data supporting the findings of this study are available from the corresponding author upon request.

Declarations

Competing interests

The authors declare no competing interests.