Fabrication of a novel porous silicon biomembrane for applications in organ-on-chip technology

Abstract

Conventional in vitro and preclinical animal models often fail to accurately replicate the complexity of human diseases, limiting the success of translational studies and contributing to the low success rate of clinical trials (Ingber 2016). In response, research has increasingly focused on organ-on-chip technology, which better mimics human tissue interfaces and organ functionality. In this study, we describe the fabrication of a novel biomembrane made of porous silicon (PSi) for use in organ-on-chip systems. This biomembrane more accurately simulates the complex tissue interfaces observed in vivo compared to conventional organ-on-chip interfaces. By leveraging established semiconductor techniques, such as anisotropic chemical etching and electrochemical anodization, we developed a reproducible method to create ultra-thin freestanding PSi biomembranes. These membranes were thinned to approximately 10 μm and anodized to contain nanoporous structures (~ 15 nm diameter) that permeate the entire membrane. The incorporation of these membranes into organ-on-chip-like devices demonstrated their functionality in a lung-on-a-chip (LOAC) model system. The results indicate that the PSi biomembranes support cellular viability and adhesion, and are consistent with the expected diffusion of nutrients and signaling molecules between distinct cell types. This novel approach provides a reliable method for generating PSi biomembranes tailored to mimic tissue interfaces. The study underscores the potential of PSi-based membranes to enhance the accuracy and functionality of organ-on-chip devices in translational research.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s10544-025-00760-3.

Introduction

In the scope of biomedical research, many advances have been made in the past few decades which have been pivotal to our understandings of disease pathology, treatment, and pre-clinical drug therapy response. Many of our discoveries thus far have been made with the use of multiple animal models and traditional cell culture experimentation (Neyt et al. 1998; Chader 2002; Buuse et al. 2005; Hackam and Redelmeier 2006; Pearce et al. 2007; Buhidma et al. 2020). While these methods are useful when studying single cellular pathways or overarching biological processes, they often falter when studying the meticulous nature of human disease and limit our abilities to develop effective therapeutics (Ali et al. 2011; Dranoff 2012; Mak et al. 2014). It has become increasingly clear that conventional animal models and cell culture do not provide us with the necessary information to extrapolate human biology and fail to recapitulate human anatomy and physiology (Infanger et al. 2013; Gill and West 2014; Song et al. 2014).

To address this knowledge gap in biomedical research, researchers have developed organ-on-chip devices. Organ-on-chip devices are in vitro models that incorporate multiple cell types and act to recapitulate the anatomy and physiology within human organ systems (Huh et al. 2010, 2013). By utilizing techniques pioneered in the microchip manufacturing industry, biomedical engineers can create micro-fabricated engineered systems that can be used to culture living cells. These organ-on-chip devices can maintain healthy cell populations with lineage-specific function as observed in human organ systems. Through controlling the micro-environment in which the cell populations reside, cell behavior can be made remarkably similar to cell populations in living tissue (Huh et al. 2010, 2012; Beebe et al. 2013; Bhatia and Ingber 2014). Current studies to date have utilized these organ-on-chip devices to study human skin, bone, cartilage, kidney, liver, blood vessels, intestine, lung, and multiple other organ systems (Kasendra et al. 2018; Poussin 2019; Arrigoni et al. 2020; Hassan et al. 2020; Jeon et al. 2020; Khalid et al. 2020; Vriend et al. 2020).

Conventional organ-on-chip devices are often constructed of polydimethylsiloxane (PDMS), a flexible, biocompatible polymer that serves as the scaffold for in vitro microfluidic organ systems. Within the device, there is an engineered porous PDMS membrane located at the interface between two or more different cell types that allow for the transfer of nutrient media, gasses, and cell signaling molecules. While this design is reliable and easily fine-tuned, there are numerous technical challenges involved in making an extremely thin PDMS barrier (Huh et al. 2013; Halldorsson et al. 2015). In addition, the hydrophobicity of PDMS leads to enhanced binding of cell signaling and small molecules, making this material sub-optimal for cell culture and preclinical drug testing (Toepke and Beebe 2006; Regehr et al. 2009; Mair and Williams et al. 2022). To this end, PDMS interfaces may limit the ability of the device to accurately mimic human anatomy, physiology, and dose-dependent drug response in organ-on-chip devices. Therefore, we sought to evaluate new material substrates (i.e., for membranes) that allow for nutrient, cell signaling, and small molecule diffusion.

To address the shortcomings with PDMS membranes, we examined the extent to which crystalline silicon could be used as an improved structural material in lieu of PDMS membranes. Due to its popularity in the semiconductor industry, natural abundance, cost-effectiveness, process maturity, and beneficial properties related to this work, silicon was an ideal material to pursue (Carter et al. 2017; Croissant et al. 2017). Large areas of crystalline silicon can be thinned to less than 2 micrometers (µm), which dramatically increases its mechanical flexibility (Wang et al. 2013). One particular form of silicon, called porous silicon (PSi), has been shown to exhibit many unique beneficial properties including flexibility, biocompatibility, hydrophilicity, optical transparency, and biodegradability (Park et al. 2009; Agrawal et al. 2010; Martín-Palma et al. 2010; Wang et al. 2013). The structure of PSi can be modified to create micro- and nano-sized pores (Formentín et al. 2014). In addition, PSi membranes have also been shown to have excellent optical transparency which facilitates imaging with fluorescence microscopy (Park et al. 2009; Gislason et al. 2024). These properties, in conjunction with the existing methods utilized to favorably alter the material, further validated the use of silicon as a structural membrane material in an organ-on-chip device. Ultrathin porous silicon membranes with nanometer-scale thicknesses and high permeability have also been demonstrated in prior work, typically involving photopatterning and lift-off fabrication techniques for integration into microfluidic and organ-on-chip systems(Striemer et al. 2007; Chung et al. 2014; Mansouri et al. 2022; McCloskey Molly et al. 2024). While these approaches enable excellent molecular transport, they often involve complex microfabrication workflows and can pose challenges related to scalability or long-term mechanical robustness. Ultra-thin macroporous silicon (macro-PSi) and nanoporous (nano-PSi) silicon membranes have been previously created by a lift-off method which ultimately proved to have limited reproducibility and mechanical stability (Miller et al. 2014). In this work, we introduce a novel method for fabricating and implementing thin PSi membranes. To ensure that membranes were fine-tuned to mimic the physical thickness of the interstitium while simultaneously allowing for fluid and gas exchange, they were characterized via scanning electron microscopy (SEM). To further validate the potential for these PSi membranes in organ-on-chip technology, and to advance this in vitro platform, novel lung-on-a-chip-like fabricated devices were cultured with human pulmonary epithelial and endothelial cells as proof-of-concept.

Materials and methods

Overview

Two-inch double-sided polished (DSP), p-type (boron-doped), (1 0 0) orientation silicon wafers with a resistivity of 0.01–0.1 Ω·cm and 250 ± 25 μm initial thickness purchased from University Wafer (Boston, MA) were cleaned by sonicating in acetone, isopropyl alcohol, and methanol. After cleaning, wafers were thinned to approximately 8–25 μm using anisotropic etching in aqueous potassium hydroxide (KOH). The thinned wafers were then anodized in a hydrofluoric acid-based electrolyte solution to generate nanopores. Following established nanopore fabrication protocols, anodization was optimized to create membranes with pores extending through the full membrane thickness. The resulting thinned, porous membranes were bonded between two identical toroidal PDMS layers to form an annular structure with two culture wells separated by the PSi membrane. Multiple cell types were cultured on opposite sides of the membrane, and cell viability, morphology, and interactions were assessed using fluorescence microscopy. After characterizing cellular interactions, the membranes were integrated into a prototype microfluidic organ-on-chip device. This prototype was fabricated using SU-8 photolithography to produce molds for PDMS microchannel replication. The channels were then aligned and bonded to either side of the PSi membrane to complete the device.

Potassium hydroxide wafer thinning

In a 1000 mL beaker, 600 g reagent-grade potassium hydroxide (KOH) pellets purchased from Fisher Chemical (Pittsburgh, PA) was dissolved in 600 mL of deionized (DI) water (17.9 M, 50 wt%). The KOH solution was suspended in an isothermal water bath where it was held at 90° C as shown in Fig. 1. Wafers were secured in a custom-fabricated polytetrafluoroethylene (PTFE) circular holder and was anisotropically etched in the KOH solution to a desired thickness. The etch rates were determined experimentally for each wafer throughout the process by 5 initial wafer thickness measurements using a micrometer followed by 3 measurements after one hour of etching. Following the etch process, membranes are submerged in a 90° C DI water bath for 30 s, then 50° C DI water bath for 5 min, and finally held under running room temperature DI water for one minute to remove all KOH and dried with a nitrogen gun.

Fig. 1.

Wet chemical etching apparatus allows for uniform thinning of DSP silicon wafers. DSP silicon wafer mounted in polytetrafluoroethylene (PTFE) “etch rings” submerged in wet chemical etching apparatus containing super-saturated solution of potassium hydroxide (~ 17.8 M, 50 wt%)

Electrochemical anodization to obtain nanopores

The thinned wafer was mounted in a dual-tank custom-made polyvinylchloride (PVC) anodization cell (Fig. 2). Both tanks were filled with electrolyte solution consisting of 1.5-parts 48 wt% hydrofluoric acid (HF) in H₂O purchased from Fisher Chemical (Pittsburgh, PA) to 1-part reagent-grade 200 proof ethanol purchased from Decon Laboratories (King of Prussia, PA) (1.5:1-HF: EtOH by volume). One square inch platinum electrodes within the cell were connected to a power source (Keithley 2410 1100 V SourceMeter) and a current density of 90 mA/cm² was applied to induce electrochemical anodization of the thinned silicon wafer. Utilizing calculated anodization rates (~ 1 μm/min), wafers were anodized to ≥150% of the calculated “through-anodization” time, to create thin though-etched porous membranes. Through-etching greatly increases the surface area of silicon exposed to atmosphere and is vulnerable to native oxide growth. To limit oxide growth on the membranes and maintain relative uniformity between all samples, the membranes were submerged in a methanol bath until stored under high vacuum at < 10⁻⁶ Torr. Through-anodization was confirmed utilizing SEM analysis. Percent porosity and average pore diameter of nanopores were calculated utilizing SEM and ImageJ analysis (Genc et al. 2023).

Fig. 2.

Dual-sided PVC cell setup for electrochemical anodization to create nanoporous structures in silicon wafers. The cell incorporates an anodic and cathodic compartment, separated by the silicon wafer mounted on an O-ring. The setup includes a 1.5:1 HF electrolyte solution for anodization, ensuring controlled pore formation under applied voltage conditions

PDMS annuli formation

SylgardTM 184 was mixed with SylgardTM Silicone Elastomer Curing Agent (10% wt) and loaded into a 10 mL BD-Syringe (Luer-Lok tip) and degassed under vacuum. From the resulting PDMS, 0.5 mL was extruded into each individual well within a Falcon^® Multi-Well 24-well cell culture tray and cured at 65° C for 1 h. Cylindrical holes (5 mm diameter) were created in the center of the resulting molded PDMS discs with a biopsy punch to achieve a small annulus (Fig. S1) Small sections of the fully porous membrane were then sandwiched between two aligned annuli, and subsequently bonded using an ozone generator (Electro-Technic Products Inc. model BD-20) and heated at 80 °C for 10 min on a hotplate to further bond the PDMS.

Scanning electron microscopy

All SEM analysis for process development/verification and characterization of the surface and cross section morphology after each step of the process was performed using a JEOL JSM-7100 F instrument.

SEM and ImageJ image analysis

Porous silicon (PSi) membrane morphology was characterized using scanning electron microscopy (SEM). Thickness measurements were obtained from cross-sectional SEM images by measuring the membrane at 20 evenly spaced points across the wafer, from the left to the right edge, with 500 μm intervals between each measurement. These measurements were used to calculate the average membrane thickness and standard deviation, which are reported in Table 1. Porosity analysis was conducted using ImageJ following a threshold-based image segmentation method similar to that previously described (Genc et al. 2023). Briefly, top-down SEM images of the membrane surface were imported into ImageJ and converted to 8-bit grayscale. A binary threshold was applied to differentiate pore space (black) from solid silicon (white), and the percentage of pore area was quantified as the ratio of black pixels area to total image area. Five SEM images were analyzed for each sample to determine the average porosity. One image was taken near the center of the membrane, while the others were captured from surrounding regions in the top, bottom, left, and right directions. A representative raw and processed image pair is shown in Fig. S2.

Table 1.

Average PSi membrane porosity

Sample Number	Front Side Pore Diameter (µm)	Back Side Pore Diameter (µm)	Front Side Porosity (%)	Back Side Porosity (%)	Sample Thickness (µm)
1	11.45 ± 2.4	8.94 ± 1.93	32.15 ± 1.36	11.15 ± 2.06	14.75 ± 0.39
2	12.02 ± 2.43	15.5 ± 3.05	22.67 ± 2.94	8.38 ± 2.74	11.16 ± 0.18
3	12.44 ± 3.70	15.38 ± 4.16	22.57 ± 1.82	3.44 ± 0.45	12.54 ± 0.35
4	9.43 ± 2.18	9.15 ± 0.9	23.45 ± 2.98	5.3 ± 2.07	15.3 ± 0.26
5	10.87 ± 1.97	12.22 ± 2.62	24.03 ± 1.92	5.75 ± 3.12	10.57 ± 0.29

Water contact angle measurements

Water contact angle (WCA) measurements were done using a Nikon D750 camera with a Nikkor AF-S 120 mm lens and + 6 close-up filters for small working distances. The camera was aligned to the sample surface, and images were captured from ~ 8 inches away. A 2.5 µL droplet was dispensed onto the sample using a micro-range pipettor. Contact angles were measured using the ImageJ contact angle plugin (Fig. S3) To minimize human error and maximize accuracy, three individuals independently selected nine points around each water droplet. This study aimed to assess and compare the hydrophilicity of PDMS, bulk silicon, thinned silicon, and our thinned porous silicon membranes.

Biological reagents, cell culture, and staining

H441 cells (human lung epithelial cell line) were purchased from the American Type Culture Collection (Manassas, VA) and cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and antibiotics (Life Technologies, Grand Island, NY). Cells were maintained at 37 °C in a 5% CO₂ incubator, fed every 2–3 days, and passaged using 0.25% trypsin (Invitrogen). Human lung microvascular endothelial cells (HMVECs) were obtained from Lonza (Walkersville, MD) and cultured in Microvascular Endothelial Cell Growth Medium-2 (EGM™−2 MV, Lonza) according to the manufacturer’s specifications. Subculturing was performed using Trypsin/EDTA (CC-5012), Trypsin Neutralizing Solution (CC-5002), and HEPES Buffered Saline Solution (CC-5022). Media was replaced every 48 h. Cells were seeded onto thinned porous silicon (PSi) membranes at a density of 300 cells/µL and incubated overnight to allow for adhesion. Calcein-AM, Alexa Fluor™ 594 Phalloidin, Prolong Gold mounting media, Alexa 488–conjugated goat anti-rabbit antibodies, and Alexa 549–conjugated goat anti-mouse antibodies were obtained from ThermoFisher Scientific (Waltham, MA). Mouse monoclonal anti-occludin antibodies were purchased from Sigma Aldrich (St. Louis, MO), and rabbit polyclonal anti-VE-cadherin antibodies were obtained from Cell Signaling Technology (Danvers, MA).

Fluorescence microscopy and immunostaining

To assess cell adherence and morphology, cells were fixed in 4% paraformaldehyde, permeabilized with Triton X-100, and blocked in 5% bovine serum albumin (BSA) in PBS. Phalloidin and DAPI were used for actin and nuclear staining, respectively, and mounted with Prolong Gold mounting media. Fluorescence imaging was performed using an Olympus BX60 epifluorescence microscope equipped with an Olympus camera. For long-term viability assessment, cells were cultured for 14 days on PSi membranes and stained with 4 µg/mL Calcein-AM in PBS for 1 h at 37 °C. Viable cells were visualized using an Olympus IX50 inverted microscope. To assess co-culture on opposite sides of the PSi membrane, H441 cells were seeded on one side and allowed to adhere for 24 h, followed by seeding of HMVECs on the opposite side. After culture, cells were fixed, permeabilized, and blocked as described above. Membranes were incubated with anti-occludin and anti-VE-cadherin antibodies in PBS containing 3% FBS at 4 °C overnight. After washing, membranes were incubated with species-specific fluorescent secondary antibodies for 2 h at room temperature before mounting and imaging.

Scanning electron microscopy of cellular populations to visualize cell-membrane interactions

Cells were fixed in 4% paraformaldehyde for 1 h on a rocker. Following fixation, the cells were washed with PBS (2 × 4 min). The cells were then treated with 1% osmium tetroxide in PBS for 30 min and then rinsed twice with deionized H₂O. The cells were then washed in increasing concentrations of EtOH for 5 min at each concentration (25%, 50%, 75%, 95%). Next, the cells were washed with 100% EtOH for 5 min two times (2 × 5 min). Following EtOH washes, the cells were treated with reagent grade hexamethyldisilazane purchased from Sigma-Aldrich (St. Louis, MO) for 10 min twice (2 × 10 min). Following washes, the PSi membranes were extracted from the annuli cell culture wells, dried with nitrogen, and sputter-coated with Au/Pd at 75 millitorrs with a current of 45 milliamps for 30 s using a Denton Vacuum Desk II. The resultant sample was then imaged in an SEM.

Results and discussion

Silicon membrane thinning

In this work, we have developed methods for obtaining novel ultra-thin biological PSi membranes for use in organ-on-chip devices. The combination of thinning a silicon wafer using KOH followed by the electrochemical anodization of the thinned-wafer provides a simple and reproducible method of creating thin, fully porous membranes. The basic process parameters were characterized and optimized yielding a tailored membrane with respect to thickness, pore size, and porosity. Adjusting these parameters, we designed the PSi membrane to allow the diffusion of small gas molecules, small signaling proteins, and to facilitate cellular adhesion and proliferation. Furthermore, we utilize these novel PSi membranes in surrogate models for the blood-gas exchange interface observed within the human alveolus. To do so, we evaluated the viability of cell types observed at this interface on each side of the novel PSi membrane, both independently, and in co-culture where cells are cultured on both sides of the membrane within their respective regions. To the best of our knowledge, this novel PSi membrane exhibits many unique properties that have yet to be studied and exploited in organ-on-chip systems.

To mimic the anatomical distance between cellular populations, such as endothelial (vasculature) and epithelial cell membranes (parenchyma) at the pulmonary basement membrane (~ 1 μm), we have developed methods for producing uniform ultra-thin (˂ 10 μm thick) bulk silicon membranes using double-sided polished silicon wafers via anisotropic wet chemical etching (Fig. 1). After recording initial thickness measurements (typically on the order of 280 μm), wafers were mounted in PTFE rings which left both sides of the wafer exposed and then submerged in an apparatus containing a super-saturated solution (~ 17.8 M, 50 wt%) of KOH.

While submerged, the bulk silicon was anisotropically etched enabling uniform thinning of the silicon wafer bilaterally. Following one hour of wet chemical etching, wafers were re-measured and etch rates were determined for each individual silicon wafer. Bilateral anisotropic etch rates of the wafers used varied between ~ 1.5 μm/min and ~ 2.5 μm/min. This methodology yields the ability to precisely control the resultant thickness of the silicon membranes (Fig. 3). The resulting samples were cleaved from the central region and measured at 20 equally spaced points along the cross-section, with an average thickness calculated for each sample.

Fig. 3.

Wet chemical etching of double-sided polished silicon wafers. (a) SEM cross-sectional thickness image of DSP silicon wafers post-wet chemical etching measured at 25 μm. (b) SEM cross-sectional thickness image of DSP silicon wafers post-wet chemical etching measured at an average of 8.45 μm

Porous silicon membrane fabrication

Electrochemical anodization of the thinned silicon substrates yielded permeable silicon membranes with the formation of nanopores. In order to electrochemically etch the wafers, thinned silicon wafers were placed into a dual-tank etch cell with a hydrofluoric acid electrolyte solution introduced on both sides (Fig. 2). Once the etch cell was secured, a current was applied to the system, which resulted in the dissolution of silicon in localized regions directly correlating to porous structures. Electrochemical anodization parameters were optimized to allow for the formation of porous structures that penetrated throughout the thinned silicon wafers at a uniform rate. Experiments were conducted with a current density of 90 mA/cm² in a solution of HF and EtOH for periods of time calculated such that nanopores with an average value of 11.58 nm in diameter span the entire thickness of the membrane which was determined via ImageJ image analysis (Table 1, Fig. S2) (Genc et al. 2023). Following anodization, the thinned wafers were analyzed with scanning electron microscopy to determine electrochemical etch rates. Utilizing estimated electrochemical etch rates (~ 1 μm/min) of the silicon wafers and known thicknesses, thinned wafers were etched to ≥150% completion to achieve relatively uniform “through-etching” (Fig. 4, Fig. S4).

Fig. 4.

Silicon bulk membranes anodized to ≥ 150% completion contain nanopores permeating throughout the entirety of the membrane to afford fully PSi membranes. (a and b) Cross section of a fully anodized silicon wafer where pore striations can be observed spanning the entirety of the wafer imaged at 2,500x. (c and d) Angled images of the “backside” of two different PSi membranes where nanopores are seen to be punching through confirming the PSi membranes are fully porous imaged at 43,000x and 6,000x, respectively

Increased uniformity of through-etching was apparent in wafers with better global thickness uniformity on the two polished surfaces of the DSP substrates. A scalloping effect was apparent in non-uniform wafer surfaces, and at increased KOH etch temperatures (i.e., 90 °C), where porous formations propagated through the backside of the wafer in areas where thicknesses were at a minimum, thus ceasing (or minimizing) pore propagation in other areas (Fig. S5). As a result, improved surface uniformity of the starting material resulted in a higher pore surface density on the backside of the membrane, and therefore more uniform full-thickness pore distributions. Due to this observed effect, we chose to utilize DSP silicon wafers for their improved thickness uniformity and a KOH etch bath maintained at 80 °C to minimize the non-uniformity of the final product membrane. With the continued use of DSP silicon wafers and optimized KOH etch temperatures, we can reliably obtain fully through-etched porous membranes with nanopores approximately 10–20 nm in diameter penetrating through the entire thickness of the membrane, thus allowing the complete transit diffusion of small molecules. Occurrences of these permeating nanopore structures were confirmed via scanning electron microscopy analysis where nanopores can be visualized on the backside of the membrane (Fig. 4 c and d).

Porous silicon membrane morphological characterization

Once fully through-etched porous membranes were obtained, we sought to characterize the surface properties of the membrane as surface morphology and hydrophilicity may have implications in cellular adherence, viability, and organ-specific function in the context of organ-on-chip devices. Utilizing the described optimized parameters for membrane fabrication we found an average surface porosity of 24.97% on the topside and 6.08% on the backside (Table 1). The lower backside porosity is attributed to non-uniform wafer thickness following KOH thinning, which leads to earlier pore breakthrough in thinner regions and limits even current distribution across the membrane during anodization. As a result, back side porosity is reduced relative to the front side and could potentially be improved by using silicon wafers with tighter initial thickness tolerances.

To assess hydrophilicity, water contact angle measurements were conducted on both the anodized side of the PSi membrane (front side) and the side through which porosity fully penetrated (back side), with comparisons made to standard PDMS. Additionally, WCA measurements were obtained for bulk silicon and an ultra-thin silicon sample to serve as control references (Fig. S3). The WCA of the PSi front side (99.59°) was comparable to that of PDMS (99.16°), while the PSi back side exhibited a decreased WCA of 59.73°. In contrast, the control samples showed WCA values of 76.5° for bulk Si and 65.81° for ultra-thin Si (Table S1). The observed difference in WCA between the front side and back side of the PSi membrane is likely due to the variation in surface porosity. The lower porosity of the back side may contribute to its reduced WCA, making it more comparable to the ultra-thin Si sample. Overall, the lower WCA of both porous and bulk silicon compared to PDMS indicates an improved surface hydrophilicity, making it more suitable for cell culture applications.

Porous silicon membrane implementation and cell culture

To replicate a 3D organ-on-a-chip system, specifically a LOAC, a PDMS annulus was used as a housing unit to encapsulate the PSi membrane (Fig. S1). The PSi membrane is suspended between two PDMS outer layers which serves two purposes. First, cells plated in suspension have bilateral fluid transfer as seen in conventional organ-on-chip technologies (Huh et al. 2013; Arrigoni et al. 2020; Hassan et al. 2020; Jeon et al. 2020). Second, multiple cell types could be cultured (epithelial and endothelial) with the membrane separating the cell types, acting as the interstitial space as found in in vivo.

In order to study the cellular interactions of epithelium and endothelium at the alveolocapillary barrier, a well-established pulmonary co-culture model was used including human distal lung epithelial cells (H441 cells) and HMVECs (Chung et al. 2018). This model has been used to validate previously developed seminal LOAC devices (Koceva Hristina 2024). To this end, H441 cells and HMVECs were initially plated on PSi membranes to determine whether these materials were suitable to maintain cell viability and cell adherence. No additional surface treatments that are characteristic of previous organ-on-chip devices, such as collagen or a fibronectin coating were required (Bhatia and Ingber 2014; Kimura et al. 2018). Cells were plated on 25 μm thick PSi membranes—selected for their mechanical stability during handling—allowed to adhere overnight, and then visualized with phalloidin and DAPI to assess initial cell morphology and adherence. After overnight culture, both H441 cells and HMVECs exhibited normal morphology and adhered well to the PSi structures. These results are in accordance with previous studies indicating surface roughness enhances cell adherence due to increased surface area (Zhou et al. 2015; Chung et al. 2018). While the membrane porosity reflects a balance between structural integrity and transport, we observed that culture media readily passed through the PSi membranes—particularly after wetting or under minimal media conditions—indicating sufficient permeability to support bilateral nutrient exchange and cell viability in the co-culture system.

Next, longer-term cell culture viability was also assessed to ensure that cellular growth and proliferation on the PSi membrane was possible for future, extended (> 14 days in culture) cell culture experiments. For these studies, both H441 cells and HMVECs were plated individually on the PSi membranes for up to 14 days. After 14 days in culture, adherent epithelial and endothelial cells were viable, metabolically active, and had normal cell morphology as evidenced by the cellular staining (Fig. 5a and b). These data indicate that thinned PSi membranes provide a cell appropriate, non-toxic, supportive and adherent surface to maintain cellular co-cultures with at least two separate cell types.

Fig. 5.

Cell adherence, sustained viability and cell specificity during co-culturing on thinned PSi membranes. (a) Fluorescence microscope images (100X magnification) of viable epithelial (H441) cells (Calcein-AM) seeded for 14 days on topside of a PSi membrane within a fabricated annulus. (b) Representative fluorescence microscope images (400X magnification) of adhered endothelial (HMVEC) cells on topside of 25-micron PSi membrane on a fabricated annulus (c and d) Fluorescence microscope images (1000X magnification) of H441 and HMVEC cells, respectively, stained with an epithelial-specific marker (Occludin, red) and endothelial-specific marker (VE cadherin, green) during a co-culture experiment. (e and f) SEM images (200X magnification) of H441 and HMVEC cells adhered to thinned, nano-PSi membranes following co-culture experiment. H441 cells cultured on bottom side of nano-PSi membrane and HMVECs were cultured on topside of nano-PSi membrane

Finally, we determined whether H441 cells and HMVECs could adhere to opposite sides of a PSi membrane to successfully create a co-culture system. H441 cells were plated and allowed to adhere overnight. After 24 h, HMVECs were seeded on the opposite side of the PSi membrane and allowed to adhere overnight. The membrane was then fixed and stained with immunofluorescent antibodies to identify the individual cell types; human epithelial cells that preferentially express Occludin and human endothelial cells that preferentially express Cadherin. Each cell type adhered to their respective (opposing) sides of the thinned, PSi membrane (Fig. 5 c and d) and could be uniquely identified. Cell adherence and normal cell morphology were also confirmed through SEM analysis (Fig. 5 e and f). During immunofluorescence visualization, the HMVECs were the first cells to be observed as they were closest to the objective lens. More importantly, the H441 cells could be visualized through the thinned PSi membrane on the opposite side of the membrane, indicating that the PSi membrane was thin enough and of a compatible material to image both sides in studies requiring more than one fluorescent molecule such as DAPI, VE cadherin, and Occludin (Fig. 5 c and d). These results support the tenet that using optically transparent materials such as thin PSi membranes with pore diameters in the nanometer range decreases undesired optical effects which impede bilateral microscopy.

Conclusion

In summary, we report on a novel method to fabricate a viable, reproducible PSi membrane that can be successfully integrated into PDMS-based organ-on-chip systems. This fabrication approach yields flexibility in setting the morphological and structural characteristics of the thin PSi membranes, such porosity and membrane thickness. Our study suggests that our PSi membranes may provide an improved tissue interface, as opposed to conventional PDMS membranes used in current organ-on-chip technologies, because PSi membranes can easily be thinned to anatomically relevant thicknesses, are biocompatible, and enable direct cell adhesion to both sides of the PSi membrane. Previous efforts in the field have demonstrated uSIMs with nanometer-scale thicknesses and high permeability, typically fabricated using photopatterning and membrane lift-off techniques (Striemer et al. 2007). While these membranes offer excellent transport properties, they can pose challenges in mechanical robustness and scalability, which our electrochemically etched PSi membranes aim to address. The use of these fabricated PSi membranes in organ-on-chip technology may allow for more accurate assessments of potential therapeutics and further our understanding of human disease pathology. Future studies will seek to evaluate the efficacy of PSi membranes in complex microfluidic organ-on-chip devices.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

M.W, C.W, S.G. and J.J. wrote the main manuscript text, M.W., S.G. and J.J. prepared figures and tables, All authors mades substantial contributions to the conception or design of the work, the acquisition, analysis and interpretation of data, J.J. provided oversight and guidance, and is PI for the grants that supported this work.

Funding

This project was graciously funded by the NSF PREM for Functional Nanomaterials [DMR Award #1827847], the NSF PREM Partnership for Education and Advancement of Quantum and nano-Sciences PEAQS [DMR Award #2424811], and STROBE NSF Science and Technology Center for Real-Time Functional Imaging [DMR Award #1548924].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.