Revealing transport, uptake and damage of polystyrene microplastics using a gut-liver-on-a-chip

Abstract

Microplastics (MPs) are pervasive pollutants present in various environments. They have the capability to infiltrate the human gastrointestinal tract through avenues like water and food, and ultimately accumulating within the liver. However, due to the absence of reliable platforms, the transportation, uptake, and damage of microplastics in the gut-liver axis remain unclear. Here, we present the development of a gut-liver-on-a-chip (GLOC) featuring biomimetic intestinal peristalsis and a dynamic hepatic flow environment, exploring the translocation in the intestines and accumulation in the liver of MPs following oral ingestion. In comparison to conventional co-culture platforms, this chip has the capability to mimic essential physical microenvironments found within the intestines and liver (e.g., intestinal peristalsis and liver blood flow). It effectively reproduces the physiological characteristics of the intestine and liver (e.g., intestinal barrier and liver metabolism). Moreover, we infused polyethylene MPs with a diameter of 100 nm into the intestinal and hepatic chambers (concentrations ranging from 0 to 1 mg mL⁻¹). We observed that as intestinal peristalsis increased (0%, 1%, 3%, 5%), the transport rate of MPs decreased, while the levels of oxidative stress and damage in hepatic cells decreased correspondingly. Our GLOC elucidates the process of MP transport in the intestine and uptake in the liver following oral ingestion. It underscores the critical role of intestinal peristalsis in protecting the liver from damage, and provides a novel research platform for assessing the organ-specific effects of MPs.

---

1. Introduction

In recent years, plastic pollution has emerged as a focal issue concerning the environment and human health globally.¹ As the quantity of plastic waste continues to rise, its persistence and limited reversibility have positioned it as a global threat.² Particularly, microplastics (MPs), ranging in size from 0.1 μm to 5 μm,³ are widely distributed across various natural environments, organisms, and human inhabited areas.^4–6 Oral ingestion stands as one of the primary pathways through which humans consume MPs, with an estimated weekly intake ranging from 0.1 to 5 grams per person.⁷ This intake is anticipated to rise over time. Recent studies have demonstrated that MPs can infiltrate the lungs,⁸ intestines,⁹ placenta,¹⁰ liver,¹¹ testes,¹² and bloodstream,¹³ inducing oxidative stress, energy imbalance, metabolic disruption, and neurotoxicity.^3,14,15 Among these, the intestines and liver are regarded as crucial target organs for ingestion and toxicity, respectively.^16,17 MPs ingested orally undergo transport and uptake across the intestinal epithelium, initially encountering the liver.¹⁸ Therefore, there is an urgent need to understand the potential impacts of MPs on these two organs.

At present, the assessment of MP-induced intestinal/liver toxicity predominantly depends on animal models in vivo and two-dimensional (2D) static models in vitro.^19,20 The physiological status after the ingestion of MPs was examined using a chicken model.²¹ Exposure to 5 μm polystyrene MPs for six consecutive weeks led to a reduction in intestinal tight junction proteins, concomitant with hepatic lipid metabolism disruption and cellular apoptosis. However, due to inherent species differences,²² animal models cannot fully elucidate the effects of MPs on human organs. Traditional 2D static models have shown that following passage through Caco-2 cells, 250 nm polylactic acid particles are significantly taken up by HepG2 cells, resulting in changes to exogenous metabolism.²³ This also suggests that the association between the intestine and liver indeed plays a critical role in the toxicity of MPs. However, 2D static models have failed to effectively mimic many critical features of human physiology, such as villous differentiation and mucus production in the intestines,²⁴ as well as the blood flow environment in the liver.²⁵ These features are crucial in regulating the toxicity of MPs.

Traditional in vitro models fail to accurately replicate the unique physiological microenvironmental characteristics of the human body, thus limiting their ability to predict the complex reactions of MPs within the body. In recent years, organ-on-a-chip technology has provided new possibilities for studying the toxic effects of MPs, by closely mimicking the microenvironmental characteristics of human tissues (including the extracellular matrix,²⁶ geometric shapes,²⁷ mechanical stiffness,²⁸ and flow dynamics²⁹). Various organ-on-a-chip models have been developed to investigate the damage caused by microplastics.^30–34 However, most of these models concentrate solely on individual target organs, lacking detailed multi-organ exposure processes, especially regarding the transport, uptake, and damage of MPs in the intestines and liver.

To address this, we designed a gut-liver-on-a-chip (GLOC) with biomimetic intestinal peristalsis and a dynamic hepatic interstitial flow environment. The chip can replicate physiological intestinal peristalsis (1–5%) and blood shear stress (0.002 Pa) in vivo,³⁵ facilitating the successful differentiation of Caco-2 and HepaRG cells into characteristics of the intestinal barrier and liver metabolism. Furthermore, 100 nm polystyrene MPs circulate between the intestine and liver, mimicking the process of oral ingestion in the human body. Through fluorescence visualization and immunohistochemical analysis, it has been discovered that MPs can penetrate the intestinal barrier and induce damage to the liver. It is noteworthy that under conditions of increased intestinal peristalsis, there is a significant decrease in the transport rate of microplastics. Meanwhile oxidative stress and damage to hepatic cells also decrease accordingly. Our study provides important clues for a deeper understanding of the potential impacts of MPs on human health.

2. Results and discussion

2.1. Construction of the biomimetic gut-liver-on-a-chip (GLOC)

MPs are pollutants present in various environments, which can enter the human body and accumulate in organs, causing harm through inhalation of air and ingestion of food. After undergoing transport through the intestines, MPs entering the human body can also target the liver (Fig. 1a). Inspired by the gut-liver axis, we designed a sequentially connected GLOC, exploring the process of MP transportation through the intestines, uptake in the liver, and subsequent damage. The chip comprises two layers of PDMS, forming two culture chambers (Fig. 1b): the intestinal chamber (left side, width 1.5 mm × height 0.5 mm) and the liver chamber (right side, width 2.0 mm × height 0.5 mm). The upper channel is utilized for perfusing the intestinal chamber, mimicking the shear stress experienced by the human intestine. The lower microfluidic channels (width 1.0 mm × height 0.5 mm) serve to sequentially connect the intestinal and liver chambers, with peristaltic pumps utilized for continuous perfusion. The structure mimics the process by which MPs undergo transport through the intestines and are carried to the liver via the bloodstream. A thin and transparent polycarbonate (PC) porous membrane (10 μm thick, pore size 1 μm) is sandwiched between the upper and lower channels of the intestinal chamber. This setup allows for the separation of the intestinal and liver chambers, while permitting the diffusion of MPs. Caco-2 cells are cultured on the membrane. Fig. 1c depicts a photograph of the assembled chip.

Fig. 1 Design and structure of the GLOC. (a) The process of human ingestion of microplastics (MPs), along with a schematic diagram of the GLOC. MPs enter the intestines through the nasal cavity and esophagus. Some of the MPs are taken up and transported through the intestinal epithelial cells into the bloodstream, where they first accumulate in the liver and cause damage. (b) Schematic diagram illustrating the structure of the GLOC for perfusing MPs. The top view of the left channel illustrates two cell culture chambers: the intestinal chamber (i) and the liver chamber (ii), dedicated to the cultivation of intestinal cells (Caco-2) and liver cells (HepaRG), respectively. A microchannel connects the two chambers, enabling closed-loop circulation between the intestinal and liver chambers. The right side shows a cross-sectional diagram of the intestinal and liver chambers. The cross-section of the intestinal chamber (A–A′) demonstrates the peristaltic movement of Caco-2 cells along the porous membrane, while the cross-section of the liver chamber (B–B′) depicts HepaRG cells cultured at the bottom and subjected to fluid flow stimulation. (c) Photograph of the fully prepared chip. Input 1 and output 1 are the perfusion channels in the upper gut lumen. Input 2 and output 2 are the perfusion channels in the lower gut lumen connecting the liver lumen. Cyclic pressure is used to simulate intestinal peristaltic motion. (d) Experimental system diagram for the GLOC.

To investigate the transport process of MPs along the gut-liver axis, our chip utilizes a perfusion system to induce uptake. The complete chip system is shown in Fig. 1d. Syringe pump B drives an MP-containing medium to continuously perfuse the upper layer of the intestinal chamber, simultaneously constructing the uptake environment and facilitating cell culture. Syringe pump A drives a regular medium through the lower layer of the intestinal chamber, transporting the translocated MPs to the liver chamber. Syringe pump C applies cyclic pressure through vacuum chambers on both sides of the fluid channel (2.0 mm wide × 1.0 mm high), cyclically stretching the porous membrane. This process induces peristaltic-like deformation in Caco-2, mimicking the motility of the gut. To resolve bubble formation in the perfusion tubing caused by temperature differences inside and outside the incubator, a reservoir bottle is set up for flow diversion, which prevents bubble formation and maintains a stable flow within the channels.

In addition, it is widely recognized that PDMS readily adsorbs microscale substances. Common strategies to reduce PDMS adsorption of small molecules include bulk material modification and surface coating.³⁶ To minimize the non-specific binding of microplastics to PDMS, we treated the surface of the microfluidic channels with oxygen plasma to introduce hydroxyl groups and enhance hydrophilicity. This was followed by extracellular matrix (ECM) coating to further prevent direct contact between MPs and the PDMS surface (Fig. S4†).

2.2. Determining the perfusion rate and circulating pressure of the GLOC

To understand the flow state of the culture medium within the perfusion channels of the chip, we numerically modeled the fluid flow and shear stress inside the real chip dimensions based on the Navier–Stokes equations (Fig. S1†). The velocity distribution within the channel exhibits a parabolic shape, consistent with the principles of laminar flow (Fig. 2a and b). Meanwhile the shear stress generated by the flow reaches its maximum value at the channel walls (Fig. 2c and d). Given that cell culture occurs at the bottom of the channels, we subsequently determined the shear stress value experienced by the cells at the bottom edge. To further replicate intestinal shear stress akin to in vivo conditions, we investigated the variation in shear stress experienced by cells at different flow rates (ranging from 100 μL h⁻¹ to 1000 μL h⁻¹) through a combination of simulation and experimentation (Fig. 2e). As the flow rate increases, there is a linearly increasing trend in the shear stress. When the flow rate reaches 500 μL h⁻¹, the shear stress reaches 0.002 Pa, closely resembling the shear stress experienced by intestinal cells in vivo.³⁵

Fig. 2 Establishment of physiological shear stress and cyclic peristalsis within the GLOC. (a) Simulation heatmap of the flow channel velocity distribution within the GLOC. (b) Cross-sections of the intestinal chamber (A–A′) and liver chamber (B–B′) displaying a parabolic velocity distribution, consistent with laminar flow conditions. (c) Simulation heatmap of the shear stress distribution in the perfusion channels. (d) Corresponding cross-sectional shear stress distribution, with maximum shear stress occurring at the wall surfaces. (e) To determine the perfusion rate required to achieve physiological shear stress (0.002 Pa),³⁵ finite element analysis and experimental calibration of the relationship between flow rate and shear stress were conducted. As the flow rate increased from 100 μL h⁻¹ to 1000 μL h⁻¹, the shear stress increased from 0.0004 Pa to 0.004 Pa. The shear stress corresponding to a flow rate of 500 μL h⁻¹ is 0.002 Pa, which is similar to the human intestines (n = 3). (f) Experimental calibration of the relationship between applied air pressure and strain of the porous membrane. When the cyclic air pressure increased from 3 kPa to 21 kPa, the tensile strain increased from 1% to 6.2% (n = 3). All experimental results are presented as the mean ± standard deviation.

Moreover, due to technical challenges in its study, the impact of interstitial blood flow in the space of Disse on liver tissue physiology is often overlooked. In fact, the subendothelial cells of the liver can directly sense changes in interstitial blood flow and respond to variations in sinusoidal hemodynamics, demonstrating mechanosensitive properties. Previous studies have shown that, compared to higher wall shear stress (0.5–2.1 Pa), the synthesis rates of albumin and urea in hepatocytes increase by 2.6-fold and 1.9-fold, respectively, under lower wall shear stress (0.001–0.033 Pa).³⁷ Due to the sensitivity of hepatic cells to shear stress, we increased the width of the liver chamber to reduce the flow rate and the corresponding shear stress. Fig. S2† demonstrates that the average linear velocity and shear stress of the intestinal chamber are 165.05 μm s⁻¹ and 0.002 Pa, respectively, while those of the hepatic chamber are 128.52 μm s⁻¹ and 0.0018 Pa. These shear stress values are significantly below the damage threshold for hepatic cells (0.1 Pa).³⁷

To replicate intestinal peristaltic movements resembling those in vivo, experimental calibration was conducted on the applied cyclic air pressure and stretching strain. We applied cyclic air pressure ranging from 3 to 21 kPa to two vacuum chambers around the intestinal chamber using an infusion pump, while recording the displacement changes of the PC porous membrane under a microscope (Fig. 2f). The results demonstrate that as the applied air pressure increased from 3 kPa to 21 kPa, the deformation of the porous membrane increased linearly from 1% to 6.2%, with a fitted curve (red dashed line) given by: tensile strain (%) = 0.290 pressure (kPa) + 0.104, R² = 0.996. According to previous studies,³⁸ a 1% tensile strain can promote the growth and differentiation of intestinal epithelial cells. Therefore, subsequent application of cyclic air pressure at 3 kPa was employed to mimic peristaltic motion in the intestine in vivo.

2.3. Reconstruction of the intestinal–liver microenvironment in vitro

To investigate the effects of shear stress and tensile strain on the morphology and differentiation of intestinal epithelial cells, we conducted perfusion culture of Caco-2 within the chip at a flow rate of 500 μL h⁻¹ (0.002 Pa). Furthermore, we regulated pressure changes in the vacuum chamber to apply cyclic tensile strain (1%, 0.15HZ). On the 7th day, bright-field microscopy images show that cells cultured under both the chip and static conditions adhere well to the substrate (Fig. 3a). Caco-2 cells cultured within the chip exhibit a folded morphology resembling intestinal villi, while those statically cultured appear flatter in morphology. Furthermore, the expression of ZO-1 protein (representing cell tight junctions) and Ezrin protein (representing cell differentiation) was analyzed by immunofluorescence microscopy. The results demonstrate that compared to statically cultured cells, the expression of ZO-1 and Ezrin proteins within the chip increased by 19.85% and 197.82%, respectively (Fig. 3b). This suggests that cells within the chip demonstrate more complete connectivity and stronger differentiation.

Fig. 3 The phenotypes of Caco-2 and HepaRG cells within the GLOC. (a–d) To explore the impacts of tensile strain and shear stress on the cell morphology and function, microscopic bright-field imaging and immunofluorescence staining were conducted on Caco-2 and HepaRG cells within static and chip cultures (500 μL h⁻¹, 0.002 Pa) for 5 days. (a) Compared to the static culture, Caco-2 cultured within the chip exhibit a more three-dimensional morphology, resembling the folds of the intestine. Additionally, there is a stronger expression of ZO-1 protein (red, represents tight junction) and Ezrin protein (green, represents microvilli differentiation). (b) The fluorescence intensity of immunostaining was quantified. Compared to the static culture, the expression levels of ZO-1 protein and Ezrin protein within the chip increased by 19.85% and 197.82%, respectively (n = 3). (c) HepaRG cultured within the chip display more pronounced aggregation, forming distinct 3D tissue-like clusters. (d) Compared to the static culture, the expression levels of CYP3A4 enzyme (representing the metabolic function) and F-actin protein (representing the cytoskeleton) within the chip increased by 99.45% and 117.49%, respectively (n = 3). (e) Continuously measuring the trans-epithelial electrical resistance (TEER) of Caco-2 for 7 days quantified the integrity of tight junctions. The TEER value in the chip culture was 3.62 times higher than that of the static culture on the 7th day (n = 3). (f) Continuously measuring the secretion of albumin from HepaRG for 7 days quantified cellular functionality. Albumin secretion in the chip culture was 1.47 times higher that of the static culture on the 7th day (n = 3). (g) Alkaline phosphatase (AKP) activity of Caco-2 cells and urea metabolism of HepaRG cells were compared between 7th day static and chip cultures (n = 3). All experimental results are presented as the mean ± standard deviation.

Similarly, we observed the morphology and function of HepaRG under perfusion at a flow rate of 390 μL h⁻¹ (0.0018 Pa) (Fig. 3c). Compared to the static culture, HepaRG cells cultured within the chip exhibit stronger aggregation, resembling 3D tissue-like clusters on the 7th day. To further explore the development of cell multipolarity under physiological conditions, we utilized CYP3A4 and F-actin to investigate functional and structural polarization, respectively. CYP3A4 is a crucial intracellular enzyme system responsible for typical cellular metabolic and detoxification functions. F-actin is a cortical cytoskeleton, ideal for studying the dynamic processes of cell remodeling and assembly during organogenesis. The results show that compared to the static culture, the expression of CYP3A4 enzyme and F-actin protein was 99.45% and 117.49% higher, respectively (Fig. 3d).

To further insight into the temporal changes in the functionality of Caco-2 and HepaRG, we analyzed the value of TEER (Fig. 3e) and the synthesis of albumin (Fig. 3f reactive liver synthesis function) continuously for 7 days. The TEER values of both static and chip cultures gradually increased over time. However, the TEER value within the chip was significantly higher than that of the static culture, reaching 3.62 times that of the latter on the 7th day (17.23 kΩ cm²vs. 4.75 kΩ cm²). It is noteworthy that the TEER values of both cultures peaked on the 5th day and then decline. This was possibly due to damage caused by excessive crowding of Caco-2. Similarly, the albumin synthesis capacity of HepaRG within the chip was higher than that of the static culture, the albumin expression being 1.47 times higher than that of the static culture (71.48 ng h⁻¹ per million vs. 48.49 ng h⁻¹ per million) on the 7th day. Additionally, compared to static cultures, the expression of Caco-2 AKP (reaction differentiation degree) and HepaRG urea (reactive metabolic function) metabolism in chip cultures increased by 3.68-fold and 2.59-fold, respectively (Fig. 3g). In conclusion, the GLOC promotes the restoration of corresponding intestinal differentiation and hepatic metabolic functions, providing a better platform for studying the transport and uptake of MPs.

2.4. Transport and damage of MPs in the intestine and liver

The intestinal mucosa serves as the primary cellular barrier for the uptake of particles via the oral route.³⁹ To investigate the potential damage caused by microplastics to intestinal cells, five different concentrations of MPs (control, 0.25 mg mL⁻¹, 0.5 mg mL⁻¹, 0.75 mg mL⁻¹, and 1 mg mL⁻¹) were individually applied to Caco-2. As shown in Fig. 3e, TEER values peaked on day 5, indicating that the barrier function was fully developed. Therefore, MPs were introduced after 5 days of culturing in the GLOC, and cell viability and TEER values were evaluated 48 hours later (Fig. 4a). The results indicate that MPs did not significantly affect the viability and barrier integrity of Caco-2 within the concentration range of 0–1 mg mL⁻¹ (Fig. 4b). The average cell viability for all five groups was 88.39% ± 2.03%, and the average TEER value remained at 20.09 kΩ cm² ± 0.49 kΩ cm² (Fig. 4c), indicating that the concentrations explored were non-toxic to the cells. To investigate the transport of MPs through the intestine, we employed a single Caco-2 model (without adding HepaRG). Once a complete barrier had formed on day 5, the MP solution was circulated and perfused on the upper layer of the intestinal chamber. After 48 hours, the solution from the lower layer was collected, and the fluorescence intensity of MPs was quantified using immunofluorescence microscopy (Fig. 4d). The results indicate that the cellular transport rates at concentrations of 0.25 mg mL⁻¹, 0.50 mg mL⁻¹, and 0.75 mg mL⁻¹ remained 7.9% ± 0.3%. In contrast, the translocation rate of MPs at a concentration of 1 mg mL⁻¹ decreased to 6.4% ± 0.4%, which was statistically significant (p < 0.05). This may be attributed to the saturation of cellular transport.

Fig. 4 MP transport, uptake, and damage in a biomimetic intestine–liver environment. (a) Microplastic induction protocol. (b–d) To understand the impact of MPs on Caco-2, exposure to different concentrations of MP solutions was conducted for 48 hours within the chip, followed by assessments of cell viability, barrier integrity, and translocation rate. (b) The average cell viability for the five concentrations of MP solutions was 88.39% ± 2.03%, showing no significant variation (n = 3). (c) The average TEER for the five concentrations of MP solutions was 88.39% ± 2.03%, showing no significant variation (n = 3). (d) Within the concentration range of 0.25–0.75 mg mL⁻¹, the cell translocation rate remained at 7.9% ± 0.3%. However, at a concentration of 1 mg mL⁻¹, the translocation rate of MPs decreased to 6.4% ± 0.4% (n = 3). (e) Following the translocation of MPs through Caco-2 at different concentrations, the uptake by HepaRG was assessed using inverted fluorescence microscopy. Green represents MPs, while blue indicates cell nuclei. (f) Corresponding analysis of MP fluorescence intensity. Compared with 0.25 mg mL⁻¹, the MP uptake of 0.50 mg mL⁻¹ and 0.75 mg mL⁻¹ concentrations was 144.84% and 308.75% higher respectively (n = 3). (g) Corresponding changes in HepaRG cell viability. When the concentration increases to 0.75 mg mL⁻¹ and 1 mg mL⁻¹, the cell viability decreases to 76.56% and 66.11%, respectively (n = 3). (h) Immunostaining for reactive oxygen species (ROS) was performed to assess oxidative stress in HepaRG. (i) ROS fluorescence intensity analysis. Compared with 0.25 mg mL⁻¹, the ROS expression at concentrations of 0.50 mg mL⁻¹ and 0.75 mg mL⁻¹ was 52.47% and 161.81% higher respectively (n = 3). (j) Release of the pro-inflammatory cytokine TNF-α. The release at 1 mg mL⁻¹ concentration is 1.76 times higher than that of the control group (without microplastics) (n = 3). All experimental results are presented as the mean ± standard deviation.

After analyzing the translocation of MPs across the intestinal barrier, we integrated a single intestinal model with a hepatic model to investigate the uptake of MPs by HepaRG. After circulating (500 μL h⁻¹) the MP solution in the upper layer of the intestinal chamber, immunostaining analysis of HepaRG in the liver chamber was conducted after 48 hours. Fig. 4e demonstrates the penetration of MPs (green) and their uniform distribution within HepaRG. Compared to the concentration of 0.25 mg mL⁻¹, the MP uptake at concentrations of 0.50 mg mL⁻¹ and 0.75 mg mL⁻¹ was 144.84% and 308.75% higher, respectively (Fig. 4f). It is noteworthy that the accumulation of MPs at a concentration of 1 mg mL⁻¹ showed no significant difference compared to the concentration of 0.75 mg mL⁻¹. This result validates the previous speculation regarding the saturation of cellular translocation, as mentioned earlier. To understand whether the accumulation of MPs leads to liver damage, the viability of HepaRG was assessed (Fig. 4g). The results indicate that within the concentration range of 0–0.5 mg mL⁻¹, the cells maintain a positive viability (>80%). However, when the concentration increases to 0.75 mg mL⁻¹ and 1 mg mL⁻¹, the viability decreases to 76.56% and 66.11%, respectively. High doses of accumulated MPs cause damage to HepaRG.

To further explore the type of damage to HepaRG, the expression of reactive oxygen species (ROS) and tumor necrosis factor-alpha (TNF-α) was assessed. ROS are by-products of cellular metabolism, and their overproduction under stress conditions can lead to oxidative damage. The immunofluorescence images demonstrate a positive correlation between ROS expression and the accumulation of MPs (Fig. 4h). Compared with 0.25 mg mL⁻¹, the ROS expression at concentrations of 0.50 mg mL⁻¹ and 0.75 mg mL⁻¹ was 52.47% and 161.81% higher respectively (Fig. 4i). In addition, the expression of TNF-α exhibits a similar trend, with the release at the highest concentration (1 mg mL⁻¹) being 1.76 times that of the control group (Fig. 4j). The results above indicate that MPs (0–1 mg mL⁻¹ concentration range) do not cause damage to the intestine, but can lead to oxidative stress damage in the liver by passing through the intestinal barrier.

2.5. The protective effect of intestinal peristalsis on liver injury

Alanine aminotransferase (ALT) is a specific marker of liver damage. To investigate the impact of intestinal peristalsis on liver damage induced by MPs, we evaluated the release of ALT from HepaRG under different tensile strains (0%, 1%, 3%, 5%) (Fig. 5a). To construct the injury environment, a solution of 1 mg mL⁻¹ MPs was circulated in the intestinal chamber, and the supernatant from the liver chamber was detected after 48 hours. The results indicate that the release of ALT gradually decreases with increasing tensile strain. Compared to a tensile strain of 0%, the release of ALT decreased by 39.19%, 55.67%, and 71.79% at strains of 1%, 3%, and 5%, respectively.

Fig. 5 Effects of intestinal peristalsis on MP transport and liver injury. (a) The release of ALT (a marker of damage) from HepaRG was measured after the translocation of MPs at different levels of stretching strain (0%, 1%, 3%, 5%). Compared to the stretching strain of 0%, the release of ALT decreased by 39.19%, 55.67%, and 71.79% respectively at strains of 1%, 3%, and 5% (n = 3). (b) The TEER values of Caco-2 under different stretching strains on the 7th day. Compared to the stretching strain of 0%, the TEER values increased by 271.14%, 346.41%, and 410.21% respectively at strains of 1%, 3%, and 5% (n = 3).(c) The translocation rates of Caco-2 under different stretching strains on the 7th day. Compared to the stretching strain of 0%, the translocation rates decreased by 46.69%, 57.28%, and 64.54% respectively at strains of 1%, 3%, and 5% (n = 3). (d) The expression of MP uptake (green, MPs; blue, cell nuclei) and corresponding ROS expression (green) in HepaRG under different stretching strains.

To further investigate the underlying reasons for the changes in damage, we first measured the TEER and transport rate of the first barrier (intestinal epithelial cells) for MPs. Compared to a tensile strain of 0%, the TEER values increased by 271.14%, 346.41%, and 410.21% at strains of 1%, 3%, and 5%, respectively (Fig. 5b). This indicates that increasing tensile strain enhances the intestinal barrier function. In contrast, compared to a tensile strain of 0%, the transport rates decreased by 46.69%, 57.28%, and 64.54% at strains of 1%, 3%, and 5%, respectively (Fig. 5c). This implies that fewer MPs penetrate to the lower layer.

Finally, we further analyzed the uptake of MPs by HepaRG and the expression of ROS (Fig. 5d). The immunostaining results indicate that the expression of both parameters decreased with increasing tensile strain. Compared to a tensile strain of 0%, the uptake of MPs decreased by 28.04%, 37.36%, and 52.55% for strains of 1%, 3%, and 5%, respectively. Similarly, the expression of ROS decreased by 20.03%, 50.55%, and 63.06% for strains of 1%, 3%, and 5%, respectively. Therefore, increasing intestinal peristalsis within a certain range can promote the formation of the intestinal barrier and help reduce the transport of MPs.

3. Discussion

The in vitro model faces significant challenges in replicating living organs, especially in achieving structural and functional similarity, as each type of cell is indispensable for its specific function. Various cell types form unique three-dimensional structures in anatomical systems, necessitating their collaboration within specific mechanical and biochemical microenvironments.^40,41 Our developed GLOC maintained the mature phenotype of different human tissues (intestine and liver) over a seven-day culture period. This was made possible by the dual-channel microfluidic architecture and series of interconnected channels. Compared to traditional static co-culture platforms, this design more closely resembles the physiological microenvironments of intestinal and liver cells in vivo. At the same time, it facilitates communication between intercellular factors and secretions, providing a more accurate simulation of the gut-liver axis (GLA). These features enable us to delineate the transport pathways and toxicological effects of microplastics (MPs) within the digestive system. We believe that this structurally defined GLOC not only offers a platform for fundamental research but also serves as a promising tool for drug screening and disease modeling.

Peristalsis was a fundamental component of the intestinal microenvironment, involving rhythmic contraction and relaxation of intestinal muscles that propel the contents along the gastrointestinal tract.⁴² There are two main types of common peristaltic construction strategies within microsystems. One approach was to directly adjust the periodic changes in fluid medium pressure within the chip channel, thereby driving the basement membrane to generate periodic peristalsis.⁴³ However, this method directly exposes cells to fluid pressure, and the unstable fluctuations in fluid shear stress can further impact cellular functions (such as the formation of intestinal villi).⁴⁴ The other approach utilizes pneumatic control to regulate the contraction and relaxation of the vacuum chamber walls within the chip, indirectly driving mechanical deformation of adjacent culture chambers.^45,46 The dual-channel GLOC designed in this study was based on this latter method. Program-controlled cyclic suction induces rhythmic mechanical deformation (1–5%) of the cell layer, mimicking in vivo intestinal peristalsis. Constant medium pressure ensures stability of fluid shear stress (0.002 Pa) within the channels. Cells were flexibly adhered to the membranes between the upper and lower channels, facilitating the trans-channel transfer of cytokines and secretions. Previous studies have demonstrated that peristaltic culture significantly surpasses traditional culture paradigms (e.g., Transwell and plate formats) in achieving intestinal morphology and function.³⁵

The intracellular transport of MPs typically involves endocytosis, intracellular trafficking, exocytosis, and transcytosis. Among these, endocytosis plays a critical role in the internalization of MPs.⁴⁷ When this transport was inhibited using chlorpromazine, a clathrin-mediated endocytosis inhibitor, the accumulation of MPs in intestinal epithelial cells is reduced by 40%.⁴⁸ This suggests that clathrin plays a dominant role in the uptake of MPs by intestinal cells. Notably, we observed that increased intestinal peristalsis significantly reduces MP transport. Given that cyclic stretching directly influences the mechanosensitive ion channel Piezo 1,⁴⁹ we hypothesize that a negative feedback regulation may exist between Piezo 1 and clathrin. This interaction could contribute significantly to the uptake of MPs. Further investigation was needed to elucidate the relationship between these two factors and clarify the pathways of MP internalization.

For the GLOC, there is big room for optimization. First, while PDMS offers advantages in microdevice fabrication and sample visualization, its adsorption of small molecules presents limitations in drug and chemical testing. Future efforts should focus on minimizing PDMS binding or exploring alternative materials, such as glass. Additionally, the simple dual-layer chip structure falls short in capturing the spatial and physical complexity of the in vivo tissue microenvironment. More biomimetic and sophisticated in vitro models are needed to replicate the fluid dynamics, oxygen diffusion, cell proliferation, remodeling, and viability seen in the gut-liver axis in vivo. Future optimization should focus on the composition, stiffness, anisotropy, and permeability of biomaterials. Additionally, it should incorporate advanced manufacturing strategies, such as 3D printing, to enhance precision in control, cell arrangement, and multi-layer interactions during fabrication.

4. Conclusion

In this study, we developed a gut-liver-on-a-chip (GLOC) featuring biomimetic peristalsis in the intestine and dynamic flow in the liver. This platform was utilized to examine the transport, uptake, and potential damage of microplastics (MPs) within both the intestine and liver after oral administration. Microscopic imaging and immunohistochemical analysis revealed that by culturing within the chip (500 μL h⁻¹; 1%), Caco-2 and HepaRG cells successfully differentiated into intact intestinal barrier and metabolic function characteristics, respectively. The investigation unveiled that 100 nm MPs have the capability to traverse intestinal epithelial cells and amass within hepatic cells, thereby triggering oxidative stress followed by cellular demise. Furthermore, as the tensile strain escalated to 3% and 5%, there was a corresponding decrease in the transportation of MPs by 57.28% and 64.54% respectively. The accumulation of MPs and the expression of reactive oxygen species (ROS) in hepatic cells were correspondingly downregulated. This study elucidates the liver damage potential resulting from the translocation of MPs through the gastrointestinal tract, offering valuable insights into preventive strategies against MP-induced injuries.

5. Materials and methods

5.1. GLOC fabrication

The GLOC was fabricated using polydimethylsiloxane (PDMS; 184 Silicone Elastomer, Dow Corning Co., USA). The chip comprised two primary components: the intestinal chamber, which included dual-layer perfusion channels (1.5 mm wide × 0.5 mm high), along with two vacuum chambers; and the liver chamber, which consisted of a single-layer perfusion channel (2 mm wide × 0.5 mm high).The two layers of the intestinal chamber are separated by a porous membrane (10 μm thick; 1 μm pore size; E.motion membrane; Sweden) and connected to the hepatic chamber via a serial channel (1 mm wide × 0.5 mm high).

The chip structure was designed using modeling software, and a high-precision CNC machining method was employed to fabricate the chip mold. Aluminum alloy was selected as the mold material due to its ease of processing, high strength, and lightweight properties. A surrounding barrier structure was incorporated around the mold to prevent PDMS overflow during casting. Acrylic panels were used for the barrier, as they facilitate PDMS demolding and effectively prevent adhesion. PDMS and a curing agent were thoroughly mixed at a weight ratio of 10:1 to ensure uniform blending, then poured into the mold after complete degassing and cured at 70 °C for 4 hours. After curing, the PDMS layer containing channel structures was gently peeled off with tweezers. A punch was used to create connecting holes in the PDMS layer, facilitating subsequent connection to tubing. A suitable porous membrane was cut according to the channel dimensions. The upper layer, porous membrane, and lower layer were treated sequentially with oxygen plasma bonding for 120 seconds (CY-P2L-B, CY Scientific Instrument Co., Ltd., China) to achieve layer adhesion. Finally, the microchannels within the chip were connected to silicone capillaries (inner diameter: 1.0 mm, outer diameter: 1.5 mm) via stainless steel capillary tubes (inner diameter: 0.8 mm, outer diameter: 1.0 mm). The flow rate was regulated by a high-precision peristaltic pump (BT100LC/DG6, Baoding Ditron Electronic Technology Co., Ltd.) to establish circulation.

5.2. Surface modification treatment

The inner surface of the PDMS microchannels was treated with oxygen plasma for 120 seconds to enhance hydrophilicity. After assembling the gut-liver-on-a-chip, it was sterilized with UV and ozone for 1 hour. Subsequently, an ECM solution was introduced into the microchannels and incubated at 37 °C for 2 hours to ensure uniform coating on the PDMS channel walls. This treatment effectively reduced the adsorption of MPs on the PDMS surface, ensuring the accuracy of experimental results.

5.3. Cell culture

Human Caco-2 intestinal epithelial cells were purchased from RuYao Biotechnology (Zhejiang, China). Caco-2 cells were cultured in high-glucose Dulbecco's modified Eagle medium (DMEM, Gibco, Waltham, MA, USA), supplemented with 10% fetal bovine serum (FBS; A3160801, Gibco, USA), and 1% penicillin–streptomycin (MA0110, Meilunbio, China). Undifferentiated HepaRG cells were obtained from the Key Laboratory of Experimental Teratology (School of Basic Medical Sciences, Shandong University).To induce cellular differentiation, cells were placed in William's E medium (Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum, 1% penicillin–streptomycin, 2 mM glutamine (Procell Life Science & Technology Co., Ltd., Wuhan, China), 0.023 IU mL⁻¹ insulin (Solarbio Science & Technology Co., Ltd., Beijing, China), and 5 μg mL⁻¹ hydrocortisone (MedChemExpress, NJ, USA). Subsequently, the cells were cultured for 2 weeks, followed by an additional 2 weeks of culture in the medium supplemented with 2% dimethyl sulfoxide (DMSO, Macklin Biochemical Technology Co., Ltd., Shanghai, China). Throughout the culture period, the cells were maintained in a CO₂ incubator (37 °C, 5% CO₂).

5.4. Cell loading within the GLOC

Following the fabrication and assembly of the chip, the complete chip along with its perfusion channels underwent a disinfection process. A solution comprising 75% (v/v) ethanol was introduced into the channels for 30 minutes, after which they were placed in a drying oven at 60 °C until fully desiccated. The chip was then stored in a sterile table until required for subsequent procedures. Cell seeding was carried out in accordance with the sequential order of Caco-2 and HepaRG cells. Initially, the inlet and outlet of the lower channel were securely sealed. Subsequently, a serum-free DMEM solution comprising rat type I collagen (5 mg mL⁻¹, Solarbio, Beijing, China) was injected into the intestinal culture chamber. The chamber was then incubated at 37 °C for 2 hours to facilitate the generation of the extracellular matrix (ECM) required for cell loading. The digested Caco-2 cells were subsequently adjusted to a concentration of 6 × 10⁴ cells per cm². They were then slowly injected onto the surface of the ECM using a sterile syringe. This was followed by sealing the upper layer inlet and outlet. Similarly, HepaRG cells were gradually introduced into the liver chamber via the lower channel at a concentration of 1 × 10⁵ cells per cm², after which the inlet and outlet were securely sealed. The chips were placed into a CO₂ incubator set at 37 °C for 4 hours to allow the cells to fully adhere to the walls. Subsequently, a peristaltic pump was employed to perfuse the culture medium between the upper and lower layers at a constant flow rate (500 μL h⁻¹, 0.002 Pa). After 5 days of perfusion culture, bright-field images of the cells were captured using an inverted microscope (Nikon Eclipse Ti2U) at 200× magnification.

5.5. Fluid simulation and experimental verification

The actual-size chip 3D files were imported into COMSOL (Multiphysics 5.5, trial version) for the simulation of fluid flow and shear stress within microchannels. We simulated using the ‘laminar flow’ module. The perfusion channel was configured as the fluid domain, with the chip material being PDMS. A Newtonian incompressible laminar flow with no-slip boundary conditions was implemented in a channel. The outlet pressure was set to 0 Pa. A“refined” physical control grid was utilized. The Navier–Stokes eqn (1) was utilized to simulate the fluid behavior under laminar conditions:Here, u is the flow velocity, ρ = 0.9933 g cm⁻³ and η = 0.692 × 10⁻³ Pa S are the density and the dynamic viscosity of the fluid at 37 °C, respectively.

To validate the accuracy of the simulation results, we calculated the wall shear stress using the Newtonian fluid mechanics friction law (eqn (2)):

Here, τ is the shear stress (dyne cm⁻²), μ is the dynamic viscosity (g cm⁻¹ s⁻¹), Q is the volumetric flow rate (cm³ s⁻¹), W is the channel width (cm), and h is the channel height (cm).

5.6. Calibration of tensile strain

To establish the relationship between the tensile strain of porous membranes in the chip and cyclic air pressure, we calibrated the tensile strain by pushing and pulling a 1 mL sterile syringe using a syringe pump. The intestine chamber was positioned under a microscope, and the gas conduit of the injector was connected to the vacuum chamber and pressure gauge. Subsequently, cyclic air pressures of 3 kPa, 6 kPa, 9 kPa, 12 kPa, 15 kPa, 18 kPa, and 21 kPa were sequentially applied. Measurements of displacement were conducted at multiple sites on the porous membrane, subsequently leading to the calculation of average strain.

5.7. Immunostaining

Caco-2 and HepaRG cells were fixed in 4% paraformaldehyde (Meilunbio, Dalian, China) for 30 minutes, subsequently undergoing three rinses with PBS. Following this, samples were subjected to permeabilization treatment with 0.2% Triton X-100 (Solarbio, China) prepared in PBS for 30 minutes. Cells were incubated with immunostaining blocking solution (QuickBlock™, Beyotime Biotech. Inc., Shanghai, China) for 60 minutes to inhibit nonspecific binding. The primary antibodies (F-actin and ZO-1 obtained from Abcam, UK; Ezrin and CYP3A4 sourced from ABclonal Technology Co., Ltd., Wuhan, China) were diluted in PBS and subsequently incubated overnight at 4 °C. Following this, specimens were incubated with the respective secondary antibodies, Alexa Fluor 488 or 647 (Abcam, UK), for 30 minutes at room temperature, under conditions shielded from light. Finally, cellular nuclei were stained with DAPI (Meilunbio, Dalian, China) for a duration of 10 minutes. We utilized a confocal immunofluorescence microscope (Ti-2, Nikon, Japan) to image the samples and performed quantitative analysis of immunofluorescence images using Image J software.

For the ROS, 2′,7′-dichlorofluorescin diacetate (DCFH-DA) was utilized for labeling. Cells underwent triple rinsing with PBS subsequent to the removal of the culture medium. Add DCFH-DA to the DMEM solution to prepare the staining working solution, ensuring the final concentration of DCFH-DA is 10 μM. Cells were covered with staining solution and incubated in a 37 °C incubator for 30 minutes. Pre-warmed DMEM was thoroughly applied for washing to eliminate any remaining DCFH-DA not taken up by the cells and imaging observation was conducted.

5.8. Cell viability assay

In accordance with the instructions provided by the manufacturer, we conducted the detection using the live/dead cell staining kit (MA0361, Dalian Meilun Biotechnology Co., Ltd, China). Live cells (green fluorescence) and dead cells (red fluorescence) were recorded at wavelengths of 495 and 652 nm using immunofluorescence microscopy. The calculation of cell viability was performed using the ImageJ software (National Institutes of Health, USA).

5.9. MP particles

The polystyrene MP particles labeled with green fluorescence (488/518) were obtained from Aladdin Biotechnology Co., Ltd., (100 nm in diameter, Shanghai, China). MPs were diluted with culture medium to concentrations of 0.25 mg mL⁻¹, 0.5 mg mL⁻¹, 0.75 mg mL⁻¹, and 1 mg mL⁻¹, respectively. A culture medium containing MPs was added on the 5th day of cell culture within the chip (mature stage of cells). After circulating at a flow rate of 500 μL h⁻¹ for 48 hours, cell damage and functionality were assessed.

5.10. Detecting the transport rate of Caco-2

The detection of translocation rates was conducted in a single-cultured intestinal cell model. Quantification of average fluorescence intensity was performed for four concentrations (0.25 mg mL⁻¹, 0.5 mg mL⁻¹, 0.75 mg mL⁻¹, and 1 mg mL⁻¹) of MPs prior to infusion. Subsequently, the upper layer of the intestinal chamber was perfused with the four concentrations of MPs for 48 hours, allowing the cells to undergo full translocation of MPs. The solution from the lower layer of the intestinal chamber was collected, and fluorescence intensity analysis was performed using a microplate reader (SynergyNeo2, BioTek, US).

5.11. Detecting the uptake of HepaRG

Four groups of MP solutions with different concentrations (0.25 mg mL⁻¹, 0.5 mg mL⁻¹, 0.75 mg mL⁻¹ and 1 mg mL⁻¹) were perfused in the upper layer of the gut chamber for 48 hours, and then the perfusion was stopped. A MP-free culture medium was perfused simultaneously in the upper and lower layers for 1 hour to fully remove the MPs that were not internalized by the hepatocytes. We utilized a confocal immunofluorescence microscope (Ti-2, Nikon, Japan) to image the MP uptake by HepaRG and performed quantitative analysis of immunofluorescence images using Image J software.

5.12. Transepithelial resistance measurement

Two Ag/AgCl electrodes were integrated into the microchannels on both sides of the porous membrane and connected to the electrochemical workstation (PGSTAT302N, Herisau, Switzerland). Five groups of MP solutions (control group, 0.25 mg mL⁻¹, 0.5 mg mL⁻¹, 0.75 mg mL⁻¹ and 1 mg mL⁻¹) were perfused and circulated in the upper layer of the gut chamber for 48 hours, and the changes in resistance were recorded by electrochemical impedance spectroscopy (EIS). Using electrochemical impedance spectroscopy in a two-electrode mode, we first measured the impedance of the device without cells with a 10 μA AC signal in the frequency range of 1 MHz to 10 Hz to obtain the baseline resistance. Simultaneously, the optimal frequency range (100 Hz to 10 Hz) was selected based on the relationship between frequency and impedance, and the potential difference between the readout electrodes was recorded. The formula for calculating the TEER is as follows (eqn (3)):

TEER = (R₁ − R₀)·S (3)

where R₁ is the actual measured value, R₀ is the baseline resistance value measured in the absence of cells, and S is the surface area of the cell culture.

5.13. Cellular functional analysis

ELISA kits (Saipei Biotechnology Co., Ltd., Wuhan, China) were utilized to quantify biomarkers (albumin, urea, ALT, TNF-α, and AKP) in the culture medium. All cell culture supernatant samples were collected and stored at −20 °C immediately after collection. Upon thawing to room temperature, sample preparation was carried out according to the manufacturer's instructions. Absorbance was measured at a wavelength of 450 nm using the microplate reader.

5.14. Statistical analysis

All experimental results are presented as the mean ± standard deviation (s.e.m.), with n = 3–6. Data analysis was performed with one-way analysis of variance with Tukey's HSD post hoc tests using GraphPad Prism 9 (trial version) and Origin 2021 (student version) software. Statistical analysis between two conditions was performed using an unpaired Student's t test. P values of <0.05 were considered to be statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The research is supported by the National Natural Science Foundation of China (No. 62371267, 62121003); Key R & D Program of Shandong Province (Major innovation project) (2022CXGC020501); Science, Education and Industry Integration Innovation Pilot Project from Qilu University of Technology (Shandong Academy of Sciences) (No. 2022JBZ02-01); Research Leader Studio in Colleges and Universities of Jinan (No. 2021GXRC083); Innovation Team of Organ-on-a-Chip Manufacturing Key Technologies (No. 202333015, funded by Jinan Science and Technology Bureau); Young Innovative Talents Introduction & Cultivation Program for Colleges and Universities of Shandong Province (Granted by Department of Education of Shandong Province, Sub-Title 1:Innovative Research Team of High-Performance Integrated Device, Sub-Title 2: Innovative Research Team of Advanced Energy Equipment); Shandong Provincial Natural Science Foundation (ZR2023QH405); Qilu University of Technology (Shandong Academy of Sciences) Youth Outstanding Talent Program (2024QZJH03); Talent Research Project of the Pilot Project of Science, Education and Industry Integration (2024RCKY002).

References

1. L. Persson, B. M. Carney Almroth, C. D. Collins, S. Cornell, C. A. de Wit, M. L. Diamond, P. Fantke, M. Hassellöv, M. MacLeod, M. W. Ryberg, P. Søgaard Jørgensen, P. Villarrubia-Gómez, Z. Wang and M. Z. Hauschild, Environ. Sci. Technol., 2022, 56, 1510–1521.
2. M. MacLeod, H. P. H. Arp, M. B. Tekman and A. Jahnke, Science, 2021, 373, 61–65.
3. J. Jing, L. Zhang, L. Han, J. Wang, W. Zhang, Z. Liu and A. Gao, Environ. Int., 2022, 161, 107131.
4. V. S. Koutnik, J. Leonard, S. Alkidim, F. J. DePrima, S. Ravi, E. M. V. Hoek and S. K. Mohanty, Environ. Pollut., 2021, 274, 116552.
5. S. M. Rodrigues, C. M. R. Almeida, D. Silva, J. Cunha, C. Antunes, V. Freitas and S. Ramos, Sci. Total Environ., 2019, 659, 1071–1081.
6. R. Qiu, Y. Song, X. Zhang, B. Xie and D. He, Handb. Environ. Chem., 2020, 95, 41–61.
7. K. Senathirajah, S. Attwood, G. Bhagwat, M. Carbery, S. Wilson and T. Palanisami, J. Hazard. Mater., 2021, 404, 124004.
8. L. C. Jenner, J. M. Rotchell, R. T. Bennett, M. Cowen, V. Tentzeris and L. R. Sadofsky, Sci. Total Environ., 2022, 831, 154907.
9. B. Li, Y. Ding, X. Cheng, D. Sheng, Z. Xu, Q. Rong, Y. Wu, H. Zhao, X. Ji and Y. Zhang, Chemosphere, 2020, 244, 125492.
10. A. Ragusa, A. Svelato, C. Santacroce, P. Catalano, V. Notarstefano, O. Carnevali, F. Papa, M. C. A. Rongioletti, F. Baiocco, S. Draghi, E. D'Amore, D. Rinaldo, M. Matta and E. Giorgini, Environ. Int., 2021, 146, 106274.
11. T. Horvatits, M. Tamminga, B. Liu, M. Sebode, A. Carambia, L. Fischer, K. Püschel, S. Huber and E. K. Fischer, EBioMedicine, 2022, 82, 1–10.
12. Q. Zhao, L. Zhu, J. Weng, Z. Jin, Y. Cao, H. Jiang and Z. Zhang, Sci. Total Environ., 2023, 877, 162713.
13. H. Çobanoğlu, M. Belivermiş, E. Sıkdokur, Ö. Kılıç and A. Çayır, Chemosphere, 2021, 272, 129805.
14. M. Prüst, J. Meijer and R. H. S. Westerink, Part. Fibre Toxicol., 2020, 17, 1–16.
15. R. Qiao, C. Sheng, Y. Lu, Y. Zhang, H. Ren and B. Lemos, Sci. Total Environ., 2019, 662, 246–253.
16. R. Qiao, Y. Deng, S. Zhang, M. B. Wolosker, Q. Zhu, H. Ren and Y. Zhang, Chemosphere, 2019, 236, 124334.
17. G. Ye, X. Zhang, X. Liu, X. Liao, H. Zhang, C. Yan, Y. Lin and Q. Huang, Sci. Total Environ., 2021, 782, 146885.
18. J. Wen, H. Sun, B. Yang, E. Song, Y. Song and G. Jiang, Environ. Sci. Technol., 2024, 58, 1832–1841.
19. C. Wang, M. Hou, K. Shang, H. Wang and J. Wang, Molecules, 2022, 27, 1–16.
20. J. Wang, H. Tian, Y. Shi, Y. Yang, F. Yu, H. Cao, L. Gao and M. Liu, Sci. Total Environ., 2023, 903, 166057.
21. K. Yin, D. Wang, Y. Zhang, H. Lu, Y. Wang and M. Xing, J. Adv. Res., 2023, 52, 3–18.
22. N. T. Doncheva, O. Palasca, R. Yarani, T. Litman, C. Anthon, M. A. M. Groenen, P. F. Stadler, F. Pociot, L. J. Jensen and J. Gorodkin, Nucleic Acids Res., 2021, 49, 1859–1871.
23. M. B. Paul, L. Böhmert, I. L. Hsiao, A. Braeuning and H. Sieg, Environ. Int., 2023, 179, 108172.
24. T. E. Ohara, M. Colonna and T. S. Stappenbeck, Dev. Cell, 2022, 57, 166–179.e6.
25. K. G. Wakim, Am. J. Med., 1954, 16, 256–271.
26. S. Han, Y. Shin, H. E. Jeong, J. S. Jeon, R. D. Kamm, D. Huh, L. L. Sohn and S. Chung, Sci. Rep., 2015, 5, 1–10.
27. X. Li, Y. Zhang, H. Zhao, C. Burkhart, L. C. Brinson and W. Chen, Sci. Rep., 2018, 8, 1–13.
28. L. Davenport Huyer, B. Zhang, A. Korolj, M. Montgomery, S. Drecun, G. Conant, Y. Zhao, L. Reis and M. Radisic, ACS Biomater. Sci. Eng., 2016, 2, 780–788.
29. K. Y. Shim, D. Lee, J. Han, N. T. Nguyen, S. Park and J. H. Sung, Biomed. Microdevices, 2017, 19, 37.
30. J. Shi, S. Liu, L. Zhang, B. Yang, L. Shu, Y. Yang, M. Ren, Y. Wang, J. Chen, W. Chen, Y. Chai and X. Tao, Adv. Mater., 2019, 32, 1901958.
31. R. K. Shukla, A. Kumar, D. Gurbani, A. K. Pandey, S. Singh and A. Dhawan, Nanotoxicology, 2013, 7, 48–60.
32. F. Eweje, H. A. M. Ardonã, J. F. Zimmerman, B. B. O'Connor, S. Ahn, T. Grevesse, K. N. Rivera, D. Bitounis, P. Demokritou and K. K. Parker, Nanoscale, 2019, 11, 17878–17893.
33. H. Wang, F. Yin, Z. Li, W. Su and D. Li, Environ. Int., 2023, 172, 107801.
34. C. Fede, I. Fortunati, V. Weber, N. Rossetto, F. Bertasi, L. Petrelli, D. Guidolin, R. Signorini, R. De Caro, G. Albertin and C. Ferrante, Microvasc. Res., 2015, 97, 147–155.
35. H. J. Kim, D. Huh, G. Hamilton and D. E. Ingber, Lab Chip, 2012, 12, 2165–2174.
36. T. Y. Kim, J. W. Choi, K. Park, S. H. Kim, J. F. Kim, T. E. Park and J. Seo, Small, 2024, 2402431, 1–14.
37. F. T. Lee-Montiel, S. M. George, A. H. Gough, A. D. Sharma, J. Wu, R. DeBiasio, L. A. Vernetti and D. L. Taylor, Exp. Biol. Med., 2017, 242, 1617–1632.
38. L. Wang, J. Han, W. Su, A. Li, W. Zhang, H. Li, H. Hu, W. Song, C. Xu and J. Chen, Microsyst. Nanoeng., 2023, 9, 2.
39. Z. Huang, Y. Weng, Q. Shen, Y. Zhao and Y. Jin, Sci. Total Environ., 2021, 785, 147365.
40. Y. Xiang, H. Wen, Y. Yu, M. Li, X. Fu and S. Huang, J. Tissue Eng., 2020, 11, 2041731420965318.
41. W. Shin and H. J. Kim, Nat. Protoc., 2022, 17, 910–939.
42. D. Kumral and A. M. Zfass, Dig. Dis. Sci., 2018, 63, 2500–2506.
43. B. Jing, Z. A. Wang, C. Zhang, Q. Deng, J. Wei, Y. Luo, X. Zhang, J. Li and Y. Du, Front. Bioeng. Biotechnol., 2020, 8, 1–14.
44. L. C. Delon, Z. Guo, A. Oszmiana, C. C. Chien, R. Gibson, C. Prestidge and B. Thierry, Biomaterials, 2019, 225, 119521.
45. G. Fang, H. Lu, R. Al-Nakashli, R. Chapman, Y. Zhang, L. A. Ju, G. Lin, M. H. Stenzel and D. Jin, Biofabrication, 2021, 14, 015006.
46. H. J. Kim, H. Li, J. J. Collins and D. E. Ingber, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, E7–E15.
47. D. Xu, Y. Ma, X. Han and Y. Chen, J. Hazard. Mater., 2021, 417, 126092.
48. Z. Hou, R. Meng, G. Chen, T. Lai, R. Qing, S. Hao, J. Deng and B. Wang, Sci. Total Environ., 2022, 838, 155811.
49. Q. Zhao, H. Zhou, S. Chi, Y. Wang, J. Wang, J. Geng, K. Wu, W. Liu, T. Zhang, M. Q. Dong, J. Wang, X. Li and B. Xiao, Nature, 2018, 554, 487–492.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4lc00578c

---