A perfusable decellularized plant-based air-liquid-interface-on-a-chip for investigating inorganic dust aerosol exposure Authors: Omur Sert, Tolga Erdogan, Yagmur Arslan, Ozlem Yesil-Celiktas Published: 2025 ABSTRACT Dust exposure affects an estimated 330 million people worldwide and can penetrate deep into the respiratory tract, disrupting epithelial and endothelial barriers. However, current in vitro models do not recapitulate the perfusable airway-vascular interface required to investigate dust-induced barrier alterations. Bioinspired scaffolds from decellularized plants hold strong potential as new approach methodologies (NAMs), considering the need for physiologically relevant in vitro models with vascular networks. Herein, we developed a decellularized plant-based air–liquid interface (ALI)-on-a-chip to study inorganic dust aerosol exposure in human airway epithelial barrier. The decellularization method achieved ≥ 99% DNA removal and a hydrophilic interface, while preserving vascular network and cellulose-rich extracellular matrix. Subsequently, human umbilical vein endothelial cells and bronchial epithelial cells were seeded to the decellularized scaffold, positioned on a PDMS-based platform to reconstruct a perfusable ALI-on-a-chip. Nanoscale dust particles were administered to the epithelium. The dust particles resulted in a significant decrease in barrier integrity, where the TEER value of the epithelial barrier decreased from ~ 126.42 to ~ 56.11Ω*cm2, with altered cell viabilities, and significant increases in pro-inflammatory cytokines such as TNF-α and IL-6. Our study reports the effects of inhaled dust aerosols on pulmonary barriers, along with offering a plant-based, biocompatible, and tissue-emulating air–liquid interface on-a-chip model. The developed model offers a versatile platform for studying toxicity of diverse particles and a preclinical evaluation of respiratory therapeutics.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-33930-7. Introduction Airborne environmental particulates (e.g., dust aerosols, industrial aerosols, nanoparticles) are among the most critical environmental challenges facing human respiratory health1. These agents can enter the lungs via inhalation and cause both acute and chronic respiratory diseases, such as asthma, chronic obstructive pulmonary disease (COPD), and pneumoconiosis2–4. As an example of an airborne environmental particulate; the amount of dust released into the Earth’s atmosphere is approximately 2 billion tons5, and its main sources are Africa (Sahara Desert), Arabian Peninsula, Asia (Gobi and Taklamakan), and deserts in South America and Australia6. Due to small aerodynamic diameter and low settling velocity, dust particles can remain suspended in the air for months and thus can easily spread long distances7. Nanoparticles (NPs) in ambient air can penetrate deep into the respiratory tract and accumulate in delicate tissues. Although the air–liquid interface (ALI) formed by the respiratory epithelial tissue creates a formidable defence to control the passage of inhaled NP to the underlying tissues, the small sized particulate matter (PM) such as PM1, PM2.5 reaching the alveolar region can diffuse, accumulate, and enter the cardiovascular system by passing through the epithelial-endothelial cells8. NPs entering the human body create a response by immunogenic cells and cause toxicity9. These particles can induce the release of pro-inflammatory cytokines (e.g., IL-6, TNF-α, IL-1β) and chemokines, leading to the recruitment of neutrophils and lymphocytes to the site of deposition10. As the focus moves toward NAMs aimed at reducing animal testing11,12, there is a growing need for reliable, robust and physiologically relevant in vitro models to investigate the effects of inhaled PM on the lung epithelial barrier, mostly formed by single-layered, tightly packed cells13. Microfluidic ALI-on-a-chip platforms basically emulate the hierarchical structure of tissue by two chambers containing a biological or synthetic membrane, which is functionalized with epithelial-endothelial cells14–16. Among these, conventional Transwell ALI systems and lung-on-chip models have been widely employed to evaluate dust or nanoparticle toxicity, providing valuable insights into barrier disruption and cytokine release under air exposure conditions15–17. However, these approaches generally rely on static or simplified microchannel geometries that do not reproduce perfusable microvascular networks. Although these designs mainly provide controllable physiological conditions relevant to tissue, they fall short of recapitulating native three-dimensional perfusable vasculature. Consequently, commercial ALI-on-a-chip designs are unable to fully recapitulate convective transport, shear-dependent endothelial phenotypes, or the physiologically relevant three-dimensional vascular architecture required for realistic nanoparticle clearance and translocation18,19. Current techniques, such as 3D printing and sacrificial templating, allow the fabrication of microchannel networks. However, these techniques are insufficient at spatial resolution, vessel perfusability and recapitulating complex microvascular networks matching native vasculature20–22. On the other hand, pluripotent stem cell-derived human lung organoids recapitulate the cellular heterogeneity of lung epithelium. But they typically lack organized, perfusable vascular networks and microfluidic flow required to study the translocation of NPs23. Vascularized organotypic lung tissue models permit several cell types of the lung, including resident immune cells24 to be studied. However, these systems mostly rely on non-perfused or simplistic channel geometries, which do not reproduce the hierarchical, shear-sensitive microvascular architecture25. As an alternative, decellularized plant-based constructs have become a promising candidate due to their perfusable vascular structures, which obey Murray’s Law, similar to the human vascular network26,27. Harris et al. showed that the surfaces of decellularized plant scaffolds exhibit native-like microtopography and biochemical cues that support mammalian cell adhesion, proliferation, and differentiation28. Indeed, decellularized spinach leaves mimicking complex capillary structures were used to study vascular amyloidosis29and thrombosis30. Advancing beyond existing strategies, we incorporated the decellularized main vein of spinach leaf populated with endothelial cells to the plant-based vasculature-on-chip platform, wherein breast cancer spheroids and fibroblasts embedded in GelMA recapitulated the tissue parenchyma31. Taken together, these findings suggest that plant-derived scaffolds may provide a perfusable three-dimensional vascular template suitable for physiological ALI models. Herein, we developed a gravity-driven method for decellularization of spinach leaves. Then, we recellularized the scaffolds with human umbilical vein endothelial cells and human bronchial epithelium cells. Following recellularization, we transferred the scaffold to the ALI-on-chip platform that we designed and fabricated from PDMS. The validation of the on-chip platform in terms of functionality and barrier integrity was carried out by immunofluorescence staining, TEER analysis and ELISA. Finally, we investigated the effects of well-characterized dust particles on the human respiratory system in the ALI on-a-chip. Results Perfusion-based decellularization and extracellular matrix assessment of spinach leaves A gravity-driven approach was chosen for decellularization of spinach leaves, where chemical agents were delivered by gravity-driven flow from the primary vein into capillaries and then into the non-vascular regions of the leaf, ensuring controlled perfusion of the reagents (Fig. 1A). Although the majority of cellular components had been removed, a 5% (v/v) sodium hypochlorite solution was applied for a transparent, optically clear decellularized spinach leaf (Fig. 1B, Fig. S1C) and leakage assay was performed with standard red dye. It is important to note that detergent-based decellularization can not fully remove endogenous chlorophylls and wall-bound phenolic compounds32. Therefore bleaching treatment steps are commonly included in plant decellularization workflows to remove residual pigments and further sterilize33. Consequently, the bleaching step reduced light scattering and absorption within the tissue, while improving imaging depth and contrast. To further confirm the preservation of structural integrity, red food dye was infused into the decellularized leaf. The dye flowed throughout the entire plant vascular network, including the capillaries, indicating that the vascular network was intact (Fig. S1D). DNA analysis showed that approximately 99% of DNA was removed compared to the native leaf group (Fig. 1C), resulting in a residual content of 20 ng/mg DNA (Fig. 1D). This shows that an effective decellularization procedure was developed with gravity-driven flow. Considering plant extracellular matrix (ECM), mostly polysaccharides dominate the structure along with a small number of proteins. There was approximately 5 µg protein in the decellularized plant (Fig. 1E). In the plant cell wall that forms the plant ECM, cellulose is the main component, followed by hemicellulose, pectin and structural glycoproteins34. Scanning electron microscopy (SEM) analysis demonstrated that the process does not have a significant effect on fibers such as cellulose, while cellular material was successfully removed31 (Fig. 1F). In this context, no significant difference was observed between the cellulose content of the native and decellularized spinach leaves (approximately 0.6 mg cellulose/mg plant) (Fig. 1G), indicating that the decellularization protocol does not alter the cellulose content while removing cellular components. As surface wettability plays a critical role in cellular adhesion, we tested this feature and a contact angle of 72° ± 1.43, 62.21° ± 4.02 was found for the native and decellularized leaves respectively, the decellularization process has not altered the hydrophilicity of the material since this value was less than 90° for the decellularized leaf (Fig. 1H), proving the hydrophilic nature35. This is attributed to the decellularization protocol, which removes the hydrophobic cuticle layer on the plant surface. Additionally, we conducted Attenuated total reflectance Fourier-transform infrared (ATR-FTIR) analysis to determine the functional groups in decellularized leaves for protein adsorption. Native and decellularized samples exhibited similar bands (Fig. 1I, Table S1), which confirms the presence of functional groups such as hydroxyl (–OH), carbonyl/ester (C = O), and cellulose-associated CH₂ (alkyl) vibrations28,36. Around 3300 cm⁻1 broad O–H stretching and ~ 1734 cm⁻1 absorption bands were observed37. C = O ester peaks (~ 1734 cm⁻1), and cellulose-associated CH₂ vibrations (1422, 1317 cm⁻1) were preserved after decellularization. Taken together, the developed decellularized plant can be considered as a scaffold for further tissue engineering applications.Fig. 1Characterization of decellularized spinach leaf. (A) Schematic representation of the decellularization protocol and chemical structure of sodium dodecyl sulfate (SDS). (B) Macroscopic images of a spinach leaf during the decellularization process. (i) native leaf before treatment, (ii) spinach leaf during the decellularization process showing progressive translucency, and (iii) fully decellularized scaffold. (C) Residual DNA content expressed as percentage of native tissue, comparing native versus decellularized leaves (*p < 0.05). (D) DNA content of native and decellularized samples (*p < 0.05). (E) Total protein and (F) total cellulose contents measured in native and decellularized leaves (*p < 0.05; ns: not significant). (G) Contact angle measurements for both groups. (H) ATR-FTIR spectra of the native and decellularized spinach leaf. Characterization of decellularized spinach leaf. (A) Schematic representation of the decellularization protocol and chemical structure of sodium dodecyl sulfate (SDS). (B) Macroscopic images of a spinach leaf during the decellularization process. (i) native leaf before treatment, (ii) spinach leaf during the decellularization process showing progressive translucency, and (iii) fully decellularized scaffold. (C) Residual DNA content expressed as percentage of native tissue, comparing native versus decellularized leaves (*p < 0.05). (D) DNA content of native and decellularized samples (*p < 0.05). (E) Total protein and (F) total cellulose contents measured in native and decellularized leaves (*p < 0.05; ns: not significant). (G) Contact angle measurements for both groups. (H) ATR-FTIR spectra of the native and decellularized spinach leaf. Dose-response assessment of characterized desert dust aerosols The average particle sizes of desert dust aerosols were between 400 and 900 nm (Fig. 2A.i), with an average diameter of 727.8 nm. Thus, they have been attributed as PM1, which refers to particulate matter with a diameter of fewer than 1 μm. Zeta potential of dust aerosols was obtained at a level of -37 mV (Fig. 2A.ii). We also ascertained the chemical composition of the dust sample, which was mainly comprised of Si, Fe, Mg, K, and S (Fig. 2B), with concentrations comparable to those reported in the literature for Saharan and Northern African desert dust aerosols38–40. The next step was to administer the characterized NPs to the human bronchial epithelial cells (BEAS-2B) cultured in a conventional 2D monolayer at doses between 1.56 and 100 μg/ml to determine the cytotoxic concentration range prior to exposure studies. Bright-field images showed a direct relationship between doses and cellular disruption with irregularities in cell morphologies compared to that of the untreated control group (Fig. 2C.i). In parallel, differences were observed in cell viability with MTT analysis (Fig. 2C.ii). The lowest dose of 1.56 μg/mL dust exposure led to a reduction in cell viability of ~ 10% and ~ 49% for 24 and 72 h exposure, respectively. Conversely, 100 μg/ml dose, as the highest exposure, resulted in ~ 55% and ~ 68% reduction in cell viability in 24 and 72-h exposure compared to the untreated control group.Fig. 2Characterization of dust aerosols and dose determination. (A) (i) Particle size distribution and (ii) Zeta potential analyses of desert dust aerosols by dynamic light scattering technique. (B) The X-ray fluorescence spectrometer analysis of dust aerosols. (C) (i) The bright-field images of BEAS-2B cells exposed to different doses of dust aerosols for 72 h (bar scale: 100 µm). (ii) MTT results of BEAS-2B cells exposed to different doses of dust aerosols for 24 h and 72 h. Statistical comparisons between control and dust-exposed groups were performed using a two-tailed unpaired Student’s t-test (ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 different dose groups vs control group). Characterization of dust aerosols and dose determination. (A) (i) Particle size distribution and (ii) Zeta potential analyses of desert dust aerosols by dynamic light scattering technique. (B) The X-ray fluorescence spectrometer analysis of dust aerosols. (C) (i) The bright-field images of BEAS-2B cells exposed to different doses of dust aerosols for 72 h (bar scale: 100 µm). (ii) MTT results of BEAS-2B cells exposed to different doses of dust aerosols for 24 h and 72 h. Statistical comparisons between control and dust-exposed groups were performed using a two-tailed unpaired Student’s t-test (ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001 different dose groups vs control group). Development and validation of a decellularized spinach leaf-based air-liquid-interface-on-a-chip The spinach leaf scaffolds were first recellularized with human umbilical vein endothelial cells (HUVEC) and human bronchial epithelial cells (BEAS-2B) (Fig. S2), then incubated for 7 days (Fig. 3A). After cell proliferation, the leaf lamina was positioned within a custom-fabricated PMMA stand, which left the upper leaf surface exposed to air. The tissue chamber was constructed from gas-permeable PDMS to allow oxygen exchange and was sealed between transparent PMMA plates with screws and bolts to maintain hydraulic tightness (Fig. 3B, Supplementary video 1).Fig. 3Diagram of air–liquid-interface platform for dust exposure. (A) Stepwise experimental timeline of the decellularized plant-based ALI model and dust exposure. (B) 3D design of the platform, which provides an enclosed chamber for accommodating spinach leaves. (C) The macroscopic image of the recellularized spinach integrated into the air–liquid-interface platform. Live/Dead fluorescence images (Calcein-AM = green, EthD-1 = red) of recellularized leaves (Left and right panels correspond to images taken from different focal planes representing the apical epithelial surface and the vascular endothelial layer, respectively). (D) Quantitative cell viability of BEAS-2B and HUVECs analyzed using ImageJ software (NIH, USA). Diagram of air–liquid-interface platform for dust exposure. (A) Stepwise experimental timeline of the decellularized plant-based ALI model and dust exposure. (B) 3D design of the platform, which provides an enclosed chamber for accommodating spinach leaves. (C) The macroscopic image of the recellularized spinach integrated into the air–liquid-interface platform. Live/Dead fluorescence images (Calcein-AM = green, EthD-1 = red) of recellularized leaves (Left and right panels correspond to images taken from different focal planes representing the apical epithelial surface and the vascular endothelial layer, respectively). (D) Quantitative cell viability of BEAS-2B and HUVECs analyzed using ImageJ software (NIH, USA). The primary vein of the leaf was cannulated and connected to the peristaltic-pump-driven closed-loop media flow via tubing and a microcannula to emulate the circulatory system and blood flow (Fig. 3B). In this configuration, the outlet tubing was positioned at the lower level of the chamber to allow continuous medium collection and recirculation, thereby maintaining a constant flow through the vascular network. This setup enabled nutrient and oxygen delivery to epithelial cells via the endothelialized vasculature through diffusion and convective transport, while ensuring that the apical surface of the leaf remained continuously exposed to air within the chamber. Thus, the platform supports both a stable air–liquid interface and continuous vascular perfusion, closely resembling physiological ALI conditions. At day 7 of post-seeding, CellTracker staining images showed that HUVECs were homogeneously distributed throughout the inner surface of the entire vasculature, while BEAS-2B cells proliferated and spread on the surface of the gelatine-coated leaf and formed the epithelial layer (Fig. S3). To assess cell viability, Calcein-AM/EthD-1 Live/Dead staining was performed. Fluorescence images showed widespread green fluorescence and minimal red signals, confirming high viability of both epithelial and endothelial layers (Fig. 3C). Quantitative analysis indicated a viability of epithelial and endothelial cells, approximately ~ 95% and ~ 86% respectively, indicating that the plant-based ALI-on-a-chip provided a stable microenvironment that maintained cell survival (Fig. 3D). These findings demonstrated that the decellularized spinach leaves can be functionalized by human lung epithelial cells and endothelial cells and integrated into the on-a-chip platform, forming an intact air–liquid interface. Cellular and barrier responses of air-liquid-interface-on-a-chip to dust exposure Once we have validated the ALI-on-chip, the next step was to administer dust particles and investigate the effects on epithelial barrier (Fig. 4A). Expression levels of tight junction proteins such as zonula occludens-1 (ZO-1) and epithelial cadherin (E-cadherin), as well as endothelial junctional proteins including vascular endothelial cadherin (VE-cadherin) and endothelial adhesion marker proteins such as cluster of differentiation-31 (CD-31) are known to decrease subsequent to nanoparticle exposure, resulting in altered intercellular connections and barrier integrity41. However, cellular responses vary based on exposure doses and durations. Therefore, we needed realistic doses to emulate real-life exposure as much as possible. The fine particulate matter concentration in the atmosphere in Turkiye is at a level of ~ 20 μg/m3 according to the annual mean fine particulate matter concentration in urban areas report of the World Health Organization42. Human respiratory volume is approximately 0.5 L and respiratory frequency is ~ 0.2 s−1. According to these data, 8 h of exposure of an adult can be calculated as 58 μg using Eq. (1), yet studies show that most of the inhaled particles are cleared by the respiratory system43,44. Therefore, we elicited the lowest dose of 1.56 μg, which was previously tested in cell viability assay.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered} Respiratory\,volume* PM2.5\,Concentration \hfill \\ \quad *Time\,interval*Respiratory\,frequence = Exposure \hfill \\ \end{gathered}$$\end{document}Fig. 4Examination of hazardous effects of desert dust exposure on ALI-on-chip. (A) Schematic representation of dust aerosol exposure on the decellularized plant-based ALI-on-a-chip. (B, C) IF staining of dust-exposed ALI-on-chip on day 3 compared to the untreated control group (Scale bars: 100 μm and 50 μm). Quantitative junction metrics were obtained from three ROIs per sample using Fiji (skeleton/coverage analysis) (*p < 0.05; **p < 0.01, n = 3). (D) SEM images showing epithelial surface morphology (scale bars: 40 and 4 μm). (E) qRT-PCR heatmap demonstrating relative gene expression changes of E-cad, ZO-1, VE-cadherin, CD31, and CASP3 (control vs dust). (F) TEER measurements of recellularized (with epithelial cells) and dust-exposed spinach leaves (****p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test).(G) Quantification of cytotoxicity based on LDH release measured from the basolateral perfusate after 72 h of exposure (**** p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test). ELISA quantification of pro-inflammatory cytokine release after dust exposure: (H) TNF-α and (I) IL-6 (***p < 0.001; ****p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test). Examination of hazardous effects of desert dust exposure on ALI-on-chip. (A) Schematic representation of dust aerosol exposure on the decellularized plant-based ALI-on-a-chip. (B, C) IF staining of dust-exposed ALI-on-chip on day 3 compared to the untreated control group (Scale bars: 100 μm and 50 μm). Quantitative junction metrics were obtained from three ROIs per sample using Fiji (skeleton/coverage analysis) (*p < 0.05; **p < 0.01, n = 3). (D) SEM images showing epithelial surface morphology (scale bars: 40 and 4 μm). (E) qRT-PCR heatmap demonstrating relative gene expression changes of E-cad, ZO-1, VE-cadherin, CD31, and CASP3 (control vs dust). (F) TEER measurements of recellularized (with epithelial cells) and dust-exposed spinach leaves (****p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test).(G) Quantification of cytotoxicity based on LDH release measured from the basolateral perfusate after 72 h of exposure (**** p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test). ELISA quantification of pro-inflammatory cytokine release after dust exposure: (H) TNF-α and (I) IL-6 (***p < 0.001; ****p < 0.0001; independent biological replicates = 3, n = 6; unpaired Student’s t-test). We administered dust particles at the dose of 1.56 μg/mL to the ALI-on-chip, perfused for 3 days, harvested, and stained against anti-CD-31 and anti-ZO-1antibodies, which are specific for endothelial and epithelial cells, respectively45,46. The images were acquired at distinct focal planes for epithelial and endothelial layers using separate fluorescence channels to prevent spectral overlap. Quantitative image analysis (Fiji skeleton and coverage analysis) of the immunofluorescence images revealed a dramatic reduction in junction continuity and signal intensity in the dust-exposed group, indicating disrupted epithelial and endothelial integrity (Fig. 4B,C).Consistent with the immunofluorescence results, qRT-PCR analysis demonstrated downregulation of the cell–cell, cell-surface adhesion markers, E-cadherin, ZO-1 and VE-cadherin by 1.39-fold, 1.77-fold, and 1.35-fold respectively, in dust exposed groups. Among cell apoptosis markers, Caspase expression was increased by 1.55-fold after dust exposure compared to control groups, indicating an increase in cell apoptosis (Fig. 4E). These transcriptional changes further support the observed loss of barrier structure and activation of cell stress pathways. All findings were compared to those of the untreated control group, which demonstrates the epithelial barrier integrity on BEAS-2B cells and vasculature formation on HUVECs (Figs. 4B, C; S4). To further investigate cell dysfunctions caused by exposure, SEM analysis was carried out, which revealed visible adverse effects in the morphology of BEAS-2B cells, including the cell–cell contacts, showing apoptotic-like morphologies (Fig. 4D), supporting the immunofluorescent images. After exposure of dust aerosols, we employed transepithelial /transendothelial electrical resistance (TEER) technique to quantitatively measure the integrity of tight junctions47 and observed a reduction in the TEER value of BEAS-2B cells cultured on the leaf surface compared to the non-exposed cell control group from ~ 126.42 to ~ 56.11 Ω*cm2, after subtraction of the baseline resistance of the decellularized leaf (~ 200 Ω·cm2) (Fig. 4F). To further quantify cytotoxicity, lactate dehydrogenase (LDH) release was measured from the basolateral perfusate after 72 h of exposure and expressed as % cytotoxicity relative to maximal lysis. Dust exposure significantly increased LDH release compared to the control (p < 0.0001) (Fig. 4G). In parallel, we quantified pro-inflammatory cytokines, where TNF-α and IL-6 levels were found to significantly elevate in the dust-exposed group (42.07 and 118.9 pg/ml) compared to the control (21.8 and 81.6 pg/ml), indicating activation of pro-inflammatory signalling pathways in response to the particulate matter (Fig. 4H,I). Discussion Mammalian cells require direct access to nutrients and oxygen within 100–200 µm of the source to remain viable48. Our infusion test showed that the decellularization process did not damage the plant vascular network. This preserved network enables uniform flow of the culture medium through the scaffold, allowing for successful cell seeding even in regions distant from the main veins or with low vascular density. To confirm the success of the decellularization process for cell culture applications, the residual DNA content must be below 50 ng/mg49. In our study, we achieved DNA levels of 20 ng/mg, which indicates that our decellularization method is effective and the resulting scaffold is suitable for use in cell culture studies. As a result of protein analysis, limited amounts of plant proteins were removed together with DNA and other cellular components at a rate similar to the literature. As shown in a study on decellularized spinach scaffolds, Gershlak et al. reported that DNA content decreased from 1,129 ± 217.3 ng/mg (native) to 9.4 ± 1.3 ng/mg (decellularized), while total protein decreased from 19.1 ± 1.9 μg/mg to 2.4 ± 0.6 μg/mg50. These results suggest low immunogenicity, which is potentially the most important advantage of decellularized plant tissue scaffolds for clinical applications33. No significant difference was found in the amount of cellulose in decellularized and native plant tissues, which ensures that the procedure does not negatively alter the cellulose content of the ECM while liberating plant cellular components. The removal of the hydrophobic cuticle resulted in an increase in hydrophilicity and enhanced protein adsorption, which further improved cell adhesion, spreading, and proliferation. Indeed, hydrophilic interfaces are reported to encourage cellular interactions, leading to expression of cell specific genes and maintaining their phenotype51. As such, scaffolds derived from plants that are more wettable preserve their complex vascular microarchitecture52 and efficiently promote proliferation and differentiation in a variety of tissue types. Furthermore, preclinical research shows that these scaffolds can be successfully applied in vascular and osteogenic tissue engineering models, highlighting the importance of surface characteristics53. Thus, by creating an appropriate environment for cell attachment, proliferation, and lineage specification, the hydrophilicity enhancements shown here not only supports recent research but also demonstrates the scaffold’s potential for plant-based tissue engineering. The dust particles smaller than 1 μm are classified as fine particles, while those exceeding this size are called coarse particles54. Particle size analysis showed that the majority of the sampled aerosols were fine particles, whereas 15% were coarse particles. The Zeta potential of particles quantifies the charge of particles developed on the surface and determines the fate of particles for aggregation or stabilization. The values higher than + 30 mV and less than − 30 mV can be assumed as electrostatically stable55. Dust samples with zeta potential values of − 37 mV were assumed as stable in stocking and working solutions, which is in accordance with the existing studies (at neutral pH levels)56. In parallel, XRF analysis was carried out for examination of chemical composition of collected samples. (Ca + Mg)/Fe ratio level is accepted as an indicator of dust particles originating from deserts within a range from 4.29 to 8.4057. High percentiles of Ca and Al elements in the composition cause a deviation from reported levels of (Ca + Mg)/Fe ratio, showing higher calcinate and clay mineral contents in the collected samples. This might be associated with the increased industrial and construction activities, resulting in an abundance of these minerals in urban areas58, which may in turn increase the probability of the interaction or mixing of desert dust and urban aerosols. From an exposure point of view, this deviation recapitulates the real daily respiratory exposure of a person living in urban areas. The next step was to examine the dose and time dependent toxic effects of aerosols on cell viability to elicit the eligible dust exposure dose. The lowest tested dose of 1.56 μg/mL reduced cell viability to 92.85% at the end of 24 h, whereas to 57.48% after 72 h of exposure. This time-dependent effect may be attributed to the accumulation of cytotoxic substances, such as inflammatory mediators and reactive oxygen species (ROS), both of which are known to be induced by the exposure stimulus59,60. In parallel to cell viability, daily exposure levels of an adult living in urban areas were calculated based on data from WHO reports and the exposure dose was determined with a combination of calculations and experimental studies to emulate the pathophysiology of exposure-mediated epithelial barrier dysfunctions in the ALI-on-a-chip. As mentioned earlier, lung epithelial barrier has a mucosal surface for clearance of airborne particles, allergens, and pathogens in cooperation with the innate defense system as a unified structure enabling the formation of the ALI with endothelium61. In lung epithelium, neighbouring epithelial cells are connected with the intercellular junctions (i.e., the tight junctions, the adherens junctions, and the desmosomes) as an adhesive force to form an epithelial barrier51. Nanoparticles entering the body through inhalation first interact with the lung epithelial barrier, while some are eliminated by the innate defence system of the respiratory tract, whereas those that are not detected continue to progress in the body43. During long-term and high-concentration exposures, the possibility of these particles evading immune surveillance increases62, thereby triggering oxidative stress, inflammation, and apoptotic pathways that lead to disruption of epithelial barrier integrity63. Leakages caused by epithelial barrier dysfunctions cause nanoparticles to get into blood circulation by disrupting actin organization and pulmonary endothelial barriers64. Prior to the exposure studies, we achieved a functional scaffold for the development of the ALI model. The decellularized plant-based scaffold supported the spatial organization and functional behaviour of both endothelial and epithelial cells. The scaffold was integrated into a microfluidic chip and actively perfused to generate continuous luminal flow. Under these physiologically relevant flow conditions, the distribution of HUVECs along the hollow constructs of plant vasculature demonstrates the successful emulation of native vessel function for organ-on-chip systems. Compared with conventional PDMS-based chips that rely on user-defined, planar microchannels21, the plant-derived vasculature provides a hierarchical, perfusable 3D network enabling uniform medium distribution, shear-dependent endothelial responses, and continuous nutrient/oxygen exchange under flow19,28. Epithelial barrier-like structure formation on the leaf surface indicates the suitability of our model and utilization as a barricade between the blood circulation and external environment through intercellular adhesive forces mediated by epithelial (ZO-1) and endothelial (CD-31) junctional proteins. We investigated the presence of these proteins before and after exposure to examine the alterations in the structure of the epithelial barrier, in parallel to demonstrating the harmful effects of dust exposure on cell–cell junctions. Subsequent to exposure, the lack of continuity and decreases in ZO-1 protein expression located along the cell membrane and contact in BEAS-2B epithelial cells, shows the disruption of epithelial barrier and impaired intercellular interactions caused by the exposure, which are also evident in SEM images. In parallel, suppression of the CD-31 expression and morphological anomalies in HUVEC cells, reveals that the exposure does not only affect the epithelial barrier integrity but also damages the endothelial vasculature by decreasing the cell viability and adhesion. The quantitative junction analysis further confirmed a significant reduction in junction continuity and signal intensity in both epithelial and endothelial layers following dust exposure, supporting the qualitative fluorescence images. Consistent with these structural findings, qRT-PCR analysis further demonstrated transcriptional downregulation of key junctional genes, including E-cadherin , ZO-1 and VE-cadherin following dust exposure, indicating weakening of both epithelial and endothelial barrier cohesion and integration at the molecular level. The reduction in ZO-1, which is expressed in both epithelial and endothelial tight junction complexes, supports the observed global disruption of junctional architecture across the ALI interface. In parallel, Caspase 3 expression was significantly elevated, suggesting activation of apoptosis-associated pathways consistent with the LDH release and morphological alterations. The observations of the TEER analysis further supported alterations in the epithelial barrier, with a decrease in the functionalized membrane resistance from ~ 126.42 to ~ 56.11 Ω*cm2, whereas the TEER value of the decellularized leaf was 200 Ω*cm2. With average TEER values of 100–200 Ω*cm2, BEAS-2B cells are considered as a suitable epithelial cell line for establishing barrier integrity in ALI models65,66. In addition to physical barrier breakdown, LDH release from the basolateral perfusate markedly increased following dust exposure, confirming membrane damage and cytotoxicity, consistent with TEER reduction and morphological alterations. Furthermore, we evaluated the inflammatory responses by measuring pro-inflammatory cytokines, necrosis factor (TNF-α), interleukin (IL)-6, as reliable markers of systemic inflammation67. Aerosol exposure triggers an early inflammatory response characterized by elevated levels of IL-8, MCP-1, and GM-CSF, which promote macrophage recruitment and activation68. The release of these cytokines contributes to impaired barrier integrity and further modulates inflammatory signalling pathways, including TNF-α, IL-6, and IL-1β. These mediators orchestrate acute inflammatory cascades through NF-κB activation and leukocyte recruitment that exacerbate tissue injury69. Hence, secretion of these cytokines in our model suggest that desert dust not only disrupts barrier integrity but also triggers an immunogenic irritation. Importantly, chronically high levels of TNF-α/IL-6 have well-reported effects on distant organs. For instance, TNF-α signalling in the vasculature and heart is reported to promote endothelial dysfunctions and fibrotic remodelling of cardiac tissue70. Furthermore, systemic TNF-α and IL-6 can compromise the blood–brain barrier and initiate neuroinflammatory cascades71. All these results demonstrated that the ALI model ensured barrier integrity and facilitated vasculature formation by supporting cell viability and adhesion, as well as mitigating the disruption of epithelial and endothelial barrier integrity caused by desert dust exposure. While air pollution is one of the major scourges of the era, dust particles have a significant place in total aerosol loading72,73. Therefore, our findings contribute to a deeper understanding of the dust aerosol exposure effects on human perturbations with a decellularized plant-based perfusable ALI-on-a-chip and validate the promise of plant-based biomaterials for unique applications in tissue engineering. In order to establish a more accurate relationship between exposure dose and biological response, future studies should focus on quantifying aerosol deposition during inhalation within the developed ALI-on-a-chip platform. The interpretation of inhalation effects is limited in the current setup, as the dust particle exposure is applied by liquid suspension placement, however dose-dependent effects are evaluated. Future designs incorporating a nebulization set-up with time-dependent aerosol monitoring and deposition would improve the physiological relevance of the ALI-on-a-chip platform by recapitulating real-life exposure scenarios. Material and methods Decellularization of spinach leaves Fresh spinach leaves were purchased from a local store and were pre-classified according to their geometry (surface area, lamina thickness, and main vein diameter) to minimize biological variability among samples. Leaves were immersed in 100% (v/v) hexane in a sonicated water bath at room temperature for 30 min to remove the waxy cuticular layer from the leaf surface. Subsequently, the leaves were washed with distilled water to remove hexane residues. The petioles were then cannulated and connected to a custom perfusion system (Fig. S.1). Next, leaves were treated with 1% (v/v) sodium dodecyl sulfate (SDS) for 3 days, followed by 1% (v/v) Triton X-100 for 1 day, and finally with 5% (v/v) sodium hypochlorite for 4 h. As a final step, the decellularized leaves were rinsed in deionized water for 2 days to eliminate residual detergents and chemicals9. Quantification of DNA and protein content Decellularized and native plant tissues were first pulverized under liquid nitrogen, and DNA was isolated using the Microprep DNA Isolation method74. The concentration (ng DNA/mg wet tissue) of isolated DNA was measured in the NanoDrop spectrophotometer. Decellularization efficiency was calculated based on the percentage reduction in DNA content relative to native controls. To quantify residual proteins in the decellularized plant and assess protein retention after decellularization, bicinchoninic acid (BCA) assay (Pierce™ BCA Protein Assay Kit, 23,225; Thermo Scientific) was performed. Approximately 100 mg of each plant sample was digested in a radioimmunoprecipitation assay (RIPA) buffer (Thermo Scientific™, 89,900) on an orbital shaker at 4 °C for 3 h. Following digestion, the BCA assay was performed according to the manufacturer’s instructions. Quantification of cellulose content Decellularized and native plant tissues, together with pure bacterial cellulose samples for standard curve generation, were hydrolyzed in 80% (v/v) sulfuric acid at 60 °C for 2 h to depolymerize cellulose into glucose. Reducing sugars released were quantified by the dinitrosalicylic acid (DNS) method75, and the cellulose content of each sample was calculated from the standard curve (Fig. S.1E). Fourier-transform infrared (FTIR) spectroscopy Attenuated Total Reflectance Fourier Transform Spectroscopy (ATR-FTIR) analysis was carried out to evaluate the chemical composition and the presence of functional groups on the surface of the scaffolds. Briefly, decellularized and native leaf samples were analyzed on a PerkinElmer Spectrum Two FT-IR spectrometer equipped with a Diamond/ZnSe ATR accessory (PerkinElmer Spectrum Two, PerkinElmer Inc., Waltham, MA). Samples were placed directly onto the ATR crystal, and constant pressure was applied to ensure a good contact between the sample and the infrared beam. Spectra were collected in the mid-infrared region from 4000 to 600 cm⁻1with 4 cm⁻1 spectral resolution and 32 co-added scans. Contact angle analysis The wettabilities of the non-vascular region (approximately 1 cm × 1 cm) of decellularized and native plant tissues were assessed using a contact-angle goniometer. A 4 µL droplet of deionized water (dH₂O) was deposited onto the sample surface, and the contact angle between the droplet and the scaffold was measured by a camera-based imaging system. Dust particle collection and characterization Dust particles sampled from the Izmir Region on September 23, 2023, were crushed with a mortar and filtered using 35 μm pore size filters. The filtered particles were UV sterilized for 4 h, and mixed with Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM-F12) cell culture medium containing FBS (10% v/v), L-glutamine (1% v/v), and penicillin/streptomycin (0.5% v/v, (100U/100 mg/ml). The mixture was sonicated (Benchtop Model Ultrasonic Cleaners, Everest Ultrasonic, CleanEx-812) for 20 min, and the particle size distribution and zeta potentials were measured by Zetasizer Nano ZS (Malvern Panalytical). Energy-dispersive x-ray fluorescence (EDXRF) spectroscopy Approximately 100 mg of dust sample was lightly spread into a low-background sample holder lined with thin polypropylene film to secure the powder. EDXRF measurements were performed on a Rigaku NEXCG EDXRF spectrometer (Rigaku NEXCG, Rigaku Corporation, Tokyo, Japan) equipped with a Pd-anode X-ray tube operated at 50 kV and 1 mA, coupled to a high-performance silicon drift detector (SSD). The processed data of elements from Na (Na-11) through U (U-92) were exported using the instrument’s QuantEZ software. Cell culture The human bronchial epithelium (BEAS-2B) cells and human umbilical vein endothelial cells (HUVEC) purchased from the American Cell Culture Collection (ATCC, Manassas, VA, USA) were cultured in DMEM-F12 medium supplemented with 10% (v/v) FBS, 1% (v/v) L-glutamine and 0.5% (v/v) penicillin/streptomycin (100U/100 mg/ml) at 37 °C and 5% CO2. All reagents were purchased from Sigma-Aldrich. Exposure dose determination Prior to dose-dependent dust particle administration, BEAS-2B cells were seeded in 96-well plates (NEST Biotechnology, 701,001) at a concentration of 1 × 104 cells/well at 37 °C and 5% CO2. After overnight incubation, the particles were administered to the epithelial cells at 7 different concentrations (0, 1.56, 3.12, 6.25, 12.5, 25, 50 μg/mL) for 72 h. Then, MTT (2,5-diphenyl-2H-tetrazolium bromide, Sigma-Aldrich, M5655) assay was performed to determine cell viability. For this, the growth medium was removed, 100 µl/well MTT solution (Sigma-Aldrich, M5655, ABD) was added to wells, and cells were incubated for 3 h at 37 °C in the dark. The cells cultured in the growth medium and DMSO were considered positive control and negative control, respectively. After removing the MTT solution, 100 µl/well DMSO was added to wells, and optical density (OD) was measured by a microplate reader (BioTek) at 570 nm. Design and fabrication of the platform Poly(methyl methacrylate) (PMMA) molds were designed with the 3D CAD program (Autodesk Fusion) software for the soft-molding-based fabrication. Molds were obtained using a laser cutter (VLS2.30, Universal Laser System) device. Polydimethylsiloxane (PDMS, SYLGARD 184) elastomer base and curing agent were mixed with a ratio of 10:1 (v/v) to obtain a PDMS solution. The solution was placed in a vacuum pump (EduScience) for 30 min to remove air bubbles. The homogenous and transparent solution was poured into prepared molds and cured for 2 h at 80 °C. After the demolding platform, they were tightened using PMMA supports and clamped with screw and bolt sets. Tubings were placed on both sides of the micro platform for inlet and outlet. All the components of the microplatform were sterilized with ethylene oxide for 16 h before the transfer of the functionalized leaves. Formation of human air–liquid interface-on-a-chip Dcellularized spinach leaves were sterilized with UV and coated with gelatine to improve cell attachment. Firstly, leaves were incubated with sterilized 0.2% gelatine (type A) solution for 1 h at 37 °C. Then, leaves were covered with 0.5% glutaraldehyde solution and incubated for 30 min at room temperature (RT). Following cross-linking, glutaraldehyde was aspirated, and 0.1 M glycine solution was added and incubated for 20 min at RT to quench residual aldehyde groups76. Leaves were washed with PBS between each step, and coated leaves were conditioned with DMEM-F12 medium. Once the leaves were ready for cell culture studies, endothelial cells (HUVECs) with a concentration of 5 × 105 cells/mL were seeded by intravenous injection through the main veins of leaves to establish a vascular network for the ALI model (Fig. S2). After that, epithelial cells (BEAS-2B) were seeded on the coated surface of leaves with a concentration of 5 × 105 cells/mL and incubated at 37 °C and 5% CO2 until confluency. Before cell seeding, BEAS-2B and HUVEC cells were stained with 15 µM green and 10 µM red CellTracker fluorescent dye (Life Technologies, USA), respectively, for 30 min to evaluate time-lapse cell adhesions and proliferation. Live/dead cell viability assay Cell viability within the recellularized spinach scaffolds was evaluated using a Calcein-AM/Ethidium homodimer-1 (EthD-1) Live/Dead Viability/Cytotoxicity Kit (Biotium, USA) following the manufacturer’s protocol. After 7 days of cell seeding on the leaf scaffolds, the samples were washed gently with PBS and incubated with a dye solution containing 10 μM ethidium homodimer (EthD-1) and 15 μM calcein AM was prepared in PBS from the stock solutions in the test kit. After 45 min of incubation at RT, the leaves were washed with PBS at least three times. The stained leaves were examined under a fluorescence microscope (Zeiss Axio Vert.A1) to capture live and dead cell images. The fluorescence images were quantitatively analyzed using ImageJ software (NIH, USA). The green (Calcein-AM) and red (EthD-1) channels were separated, and the mean fluorescence intensity of each channel was measured. Background fluorescence was subtracted to remove any non-specific signal. The ratio of live to dead cells was determined by comparing the corrected fluorescence intensities of the two channels. Cell viability was expressed as the proportion of green fluorescence (live cells) relative to the total fluorescence (live plus dead cells). The % viability graph was created by using GraphPad Prism 8.3.0 (n = 2). Dust particle exposure on human air–liquid interface-on-a-chip After both cell types reached confluency on the leaf scaffolds, dust particles were administered for overnight exposure. Next day, the main veins of leaves were connected to the inlet tubing, and the media collecting outlet tubing was placed at the bottom level of the platform for a continuous flow provided by a peristaltic pump (Longer Pump L100-1S-1; Baoding, China) at 1 rpm (~ 45µL/min). During the run time of the platform, the apical surface of the epithelial layer remained exposed to air, while only the basolateral (vascular) channel was perfused in a closed-loop circuit. After 3 days of perfusion of ALI-on a-chip platform at 37 °C and 5% CO2, the platforms were harvested for further analysis. Immunofluorescence staining Harvested leaves were fixed with 4% PFA for 2 h and washed with PBS thrice. Fixed leaves then permeabilized with 0.5% Triton X-100 (Sigma-Aldrich, USA) for 1 h at RT and blocking was performed with blocking solution (containing 10% FBS, 1.15% glycine, 0.5% bovine serum albumin (BSA, Sigma-Aldrich, USA)), and 0.1% Tween-20 (Sigma-Aldrich, USA) diluted in 1X PBS for 1 h at RT. Afterwards, a primary antibody against CD-31 and ZO-1 was diluted 1:250 in blocking solution and applied at 4 °C for 1 day. Primary antibodies were washed with PBS thrice for 30 min. Secondary antibodies (Alexa Fluor 488-Goat Anti-Mouse and Alexa Fluor 647-Goat Anti-Rabbit) were diluted 1:1000 in blocking solution and applied at RT overnight on a rocker shaker. Secondary antibodies were washed with PBS thrice for 30 min, and DAPI (1.5 μg/μL) staining was performed for 15 min. Stained leaves were washed with PBS thrice for 30 min and visualized under a fluorescence microscope (Zeiss LSM 880, Germany). Junction morphology was quantified from immunofluorescence images using a skeleton-based analysis in Fiji (ImageJ, NIH, USA). The images were pre-processed by thresholding and converted to binary format, followed by application of the “Analyze Skeleton” plugin to extract junctional structures. Junction network continuity was expressed as the total skeleton length per unit area (a.u.), averaged from three randomly selected regions per image. The junction continuity graph was created by using GraphPad Prism 8.3.0 (n = 3). Total RNA isolation and qRT-PCR analysis Total RNA was isolated from each sample using the RNeasy Plus Mini Kit (Qiagen, Germany) according to the manufacturer’s protocol. Complementary DNA (cDNA) was synthesized using the RT2 First Strand Kit (Qiagen, Germany). Gene expression analysis was performed by real-time PCR using RT2 SYBR Green qPCR Mastermix with gene-specific primers (Table S2) on a LightCycler 480 Instrument II (Roche, Germany). In total, 5 target genes related to cell–cell adhesion, barrier function, and cell survival were examined. Gene expression levels were normalized using GAPDH as a housekeeping gene, and fold-changes were calculated using the 2-ΔΔCt method. Heatmap was generated using GraphPad Prism 8.3 (San Diego, CA, USA). ELISA assays IL-6 and TNF-a levels as a result of dust particle exposure were analyzed using ELISA kits by subjecting them to samples were collected from the basolateral (vascular) perfusate which was continuously circulated in a closed-loop system according to the kit instructions (BT-Laboratory, Shanghai, China). Subsequently, the absorbances were detected by a microplate reader at 450 nm and concentration graphs were created by using GraphPad Prism 8.3.0. Each condition was analysed in three independent biological replicates, with 6 technical readings (independent biological replicates = 3, n = 6). LDH assay About 100 µL of centrifuged perfusates were transferred to the 96-well plate, and 100 µL of reaction mixture from the Cytotoxicity Detection Kit (LDH) (Roche Diagnostics GmbH, Mannheim, Germany; Cat. No. 11644793001) was added to each well (prepared by mixing 250 µL Buffer A with 11.25 mL Buffer B according to the manufacturer’s instructions). The control group was used as low control (LC) and the group incubated with Triton X-100 were used as positive and high control (HC). After 30 min of incubation at room temperature, the reaction was stopped with 1N HCl (50 µL), and ODs were measured at 490 nm. The cytotoxicity (%) graph was created by using GraphPad Prism 8.3.0 (independent biological replicates = 3, n = 6). TEER measurements TEER analysis was performed for control and exposure groups by using a two-part cell crown insert (Scaffdex; Cat. No. C00005S). Cells were cultured on the coated leaves (described above), which were assembled between upper and lower parts of the insert. After confluency, dust particles were administered at predefined concentration. After 3 days of culture in inserts, TEER measurements were carried out with EVOM™ Manual voltohmmeter (World Precision Instruments, Sarasota, FL, USA) and STX4 electrodes (WPI, Sarasota, FL, USA). Before measurements, electrodes were disinfected with 70% ethanol and sterilized under UV light and calibrated in a warmed complete medium. Two readings were taken at distinct locations per insert and normalized to the membrane surface area to yield TEER in Ω·cm234 (independent biological replicates = 3, n = 6). SEM analysis For SEM analysis, leaves were treated with buffer A (glutaraldehyde with sodium cacodylate solution), buffer B (sucrose with sodium cacodylate solution), and buffer C (osmium tetroxide with sodium cacodylate solution) solutions at 4 °C for 30 min. Then, leaves were washed with distilled water and dehydrated with EtOH series (35, 50, 70, 85, 95, 100 and 100%) for 5 min. All samples were treated with a critical point dryer to remove moisture and sputter-coated with 10 nm gold (Leica EM ACE600). Then, cell morphologies were observed by SEM (Thermo Scientific Apreo S). Statistical analysis Statistical analyses were carried out by one-way and/or two-way variance analysis (ANOVA, Tukey’s or Dunnet’s Multiple Comparison Test) and Student’s t-test via GraphPad Prism 8.3 (San Diego, CA, USA) with ± 95% confidence interval, and p values < 0.05 were considered statistically significant. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2