Based on its pore size and interface effect, the porous membrane material can provide a support surface for cell culture, realize the free permeation of solution, and especially can culture different cells separately without affecting the exchange of chemical substances between cells. Therefore, cell culture based on porous membrane has been widely used in the fields of tissue engineering and biomedicine^[29,30]. Hegde et al. Developed a microfluidic device based on a porous membrane for "sandwich" culture of hepatocytes^[31]. The device is designed to separate two layers of PDMS chip channels by a porous membrane, thus forming a two-chamber device separated by a porous membrane. Among them, the top chamber is used to continuously perfuse the culture medium, and the bottom channel is used to culture hepatocytes. Compared with the static model, the hepatocytes cultured in dynamic fluid have stronger function and can maintain the function of the liver for a long time. Physiologically, the liver is mainly composed of hepatocytes, Kupffer cells, hepatic stellate cells and hepatic sinusoidal endothelial cells. Because of the interaction between various cells in vivo and the specific physiological and biochemical environment, it is difficult to simulate the specific structure and function of liver tissue in vitro^[32⇓~34]. Therefore, Long et al. Developed a microfluidic liver organ chip based on porous membrane to simulate liver sinusoidal structure in vivo (Fig. 1A)^[35]. Liver related cells (including vascular endothelial cells, hepatic stellate cells and Kupffer cells) were cultured in the upper channel of the chip, and liver parenchymal cells were cultured in the lower channel, and the dynamic culture of cells was carried out by external fluid of the chip. The chip integrates two key factors of hydrodynamic and visual tracking and four types of primary hepatocytes, highly integrates the key structures of the liver, realizes the simulation of the microphysiological environment of the liver, and can be used to study the immune response in a short time. In a word, the construction of microfluidic three-dimensional liver chip based on porous membrane method is one of the common methods, which can well simulate the liver sinusoidal structure in vivo, and the dynamic culture environment can provide similar physical and chemical conditions as in vivo, which has a wide range of applications in exploring the relationship between liver physiological structure and function.

The method of multicellular clusters is also an important three-dimensional cell culture technique. The basic principle is that under the condition of inhibiting cell attachment, cells tend to form multicellular clusters, which have a multi-layer cell structure and are closely connected with each other. Compared with two-dimensional monolayer cells, it presents a three-dimensional structure close to the body, and can maintain the phenotype and function of cells, and has similar characteristics with solid tissues. Cell clusters can be formed quickly and efficiently based on the micro-pit structure on the microfluidic chip^[36]. Lee et al. Designed honeycomb-arrayed dimple arrays for the formation and culture of three-dimensional hepatocyte clusters^[37,38]. Compared with two-dimensional culture, three-dimensional clustered hepatocytes showed higher cell activity and function. Luo and Liang et al. Developed a method for dynamic culture of hepatocyte clusters on a chip. The results showed that the metabolic function of hepatocyte clusters in dynamic culture was stronger than that in static culture^[39]. Shen et al. Designed a microfluidic liver chip based on hydrogel materials^[40]. In the hydrogel chip, liver cells formed hepatocyte clusters in the micropores of the channel outside the chip, and the endothelial cells were cultured with collagen in the inner channel. The results showed that the liver cells showed higher activity within 8 days, and compared with the liver cell cluster culture alone, the co-culture of liver cells and endothelial cells could better reproduce the physiological function of the liver in vivo. Khademhosseini et al. Proposed to construct a pathological model of nonalcoholic fatty liver disease by co-culturing hepatocytes and endothelial cells into multicellular clusters in microchip pits. The results showed that when the hepatocyte cluster contained 20% endothelial cells, the pathogenesis of nonalcoholic fatty liver disease could be better simulated^[41]. In addition, the construction of vascular system in vitro is an important method to simulate human tissues and realize the functions of tissues and organs. Therefore, Bonanini et al. Developed a microfluidic organ chip platform integrated with a three-dimensional vascular network and primary liver cells (Fig. 1B)^[42]. The platform mainly consists of 384-well plates and 64 microfluidic chips, each of which can induce angiogenesis. On top of the chip is an open system for manual placement of primary hepatocyte clusters. Hepatocyte clusters continue to culture in collagen and eventually interact with vascular endothelial cells to form a stable, perfusable vascular network. In conclusion, three-dimensional culture based on cell clusters can mimic the characteristics of tissues and organs in vivo, and can maintain cell activity and tissue-related functions. The micropit array integrated in the microfluidic chip can quickly and efficiently form hepatocyte clusters, which will play a positive role in the exploration and construction of liver in vitro models and high-content drug screening.

Polymer gels with good biocompatibility are widely used in three-dimensional cell culture in vitro, and their three-dimensional network structure is very similar to the structure and properties of extracellular matrix molecules in vivo. Cells cultured by gel method are more likely to adhere, promote cell growth and enhance the interaction between cells and matrix, which is a commonly used three-dimensional cell culture method. Compared with traditional gel-cultured cells, cells cultured on microfluidic chips can maintain higher activity. Revzin et al. Developed a microfluidic device for long-term liver culture and maintenance of liver function based on collagen culture^[43]. Collagen and hepatocytes are directly injected into the culture chamber through a syringe, and the cells in the collagen can be continuously cultured for more than one month in a device with fluid flow and maintain good physiological function. However, the liver tissue cultured in collagen in 96-well plates lost its phenotype and function within 3 to 5 days. Therefore, this method can be used for future liver disease modeling or personalized liver-directed therapy. In addition, Qin et al. Combined with droplet microfluidic devices, human induced pluripotent stem cells were encapsulated in droplets to form hydrogel capsules, which were continuously cultured and eventually differentiated into liver organoids^[44]. The method is easy to operate, scalable and stable, and the cultured liver organs show homogeneity and good cell viability, providing a technical platform for stem cell organoids with higher controllability and fidelity. Three-dimensional liver microarray based on gel method is often used to construct liver lobule structure and explore the interaction between liver parenchymal cells and vascular endothelial cells, hepatic astrocytes or fibroblasts^{[45][46][47][48]}. For example, Banaeiyan et al. Designed a high-throughput liver lobule chip^[22]. The chip includes a tissue culture region, a feed region, and a central exit port with a diffusion path between the tissue culture region and the feed region to protect the cells from fluid shear forces. The liver tissue cultured by this method could be continuously cultured for 21 days, and the cells could maintain morphology and function during the culture period. Wang et al. Used an integrated microfluidic chip to fabricate 3D liver lobular microtissue in vitro (Fig. 1C), in which hepatocytes and vascular endothelial cells were cultured in collagen^[49]. The experiment proves that the liver tissue cultured by the device retains great drug metabolism capacity, can accurately analyze the interaction between clinical drugs and the potential adverse drug reaction causing liver injury, and provides a new method for the analysis of drug metabolism and hepatotoxicity. In conclusion, three-dimensional cell culture based on gel method has higher controllability and fidelity, and is widely used in the construction of liver organ chip.

Bio-3D printing technology is an emerging technology for liver tissue engineering using layered biomanufacturing methods, which is conducive to precise printing of three-dimensional structures and high-throughput manufacturing. Compared with the above three cell culture methods, bio-3D printing technology can directly print personalized tissue cells or organs, which provides convenience for solving the problems of donor shortage, immune rejection and drug treatment^[50]. Lin et al. Used a homemade inkjet printer to print microscale cell patterns and used them to achieve cell culture, stimulation, and analysis on a chip^[51]. Cells were encapsulated in sodium alginate solution, and the cells were precisely distributed on the slide by software control, and then calcium chloride solution was poured into the microfluidic chip channel to form gel immediately and fix it on the bottom of the chip. This method helps to improve cell patterning efficacy in microfluidic chips and reduce laborious experimental work. In order to achieve a variety of cell layered printing, Cho et al. Used 3D bioprinting to wrap liver cells and endothelial cells in different hydrogels, and finally layered printing. The results showed that compared with the traditional liver cell culture, the liver function of the cells cultured after 3D bioprinting was significantly enhanced^[52]. Because 3D printed hepatocytes lack the bile system necessary for bile acid excretion, they can not overcome the drawbacks of in vitro liver testing in drug development, and they have developed liver chips with multiple cell types^[53]. The chip has a vascular and biliary system that shows better liver function and is expected to be used for in vitro liver organ testing. In order to control the spatial distribution of cells more accurately, Jin et al. Developed an advanced 3D bioprinting method to realize the presetting of multiple cell cultures and structures at the same time, and successfully constructed the liver lobule structure (Fig. 1D)^[54]. The chip has structures similar to hepatic lobules, including liver parenchymal cells, endothelial cells and lumens, showing higher liver-related metabolic activity, such as enhanced secretion of urea and albumin, which proves that this method can maintain structural integrity and improve cell function. In order to print the structure of liver lobule more conveniently, Mandal et al. Developed a bioink based on liver extracellular matrix^[55]. Hepatic parenchymal cells, endothelial cells and hepatic stellate cells were encapsulated in bioink in advance, and the structure of hepatic lobules was printed by 3D bioprinter. The results show that the 3D printed liver structure has high metabolic capacity and can predict drug hepatotoxicity more accurately. This method can help accelerate drug development and provide a powerful platform for hepatotoxicity screening. In conclusion, 3D bioprinting technology can produce more anatomically accurate liver structures, including liver-specific spatial structures and vascular networks, which provides a favorable tool for the construction of in vitro liver models for the study of liver diseases and drug screening.

Liver is the place of human metabolism, and its common pathological changes include alcoholic liver disease (ALD), non-alcoholic fatty liver disease (NAFLD), viral hepatitis and metabolic disease type II diabetes mellitus. The establishment of microfluidic liver chip, which can simulate the physiological function and cell state of the liver, provides a good in vitro model and research platform for in-depth understanding of disease mechanisms and finding appropriate treatment options.

ALD is a liver disease caused by long-term heavy drinking, which is a widespread liver disease and seriously endangers human health. The research and development of targeted treatment for ALD has been one of the research hotspots^[56]. As shown in Fig. 2A, Lee et al. Constructed a three-dimensional in vitro model of alcoholic liver based on microfluidic technology^[57]. They placed rat primary hepatocytes and hepatic stellate cells into a microfluidic chip for dynamic culture under flow, using a self-made osmotic pump to provide a dynamic culture environment. Hepatocytes were treated with different concentrations of alcohol for 3 days, and the results showed that the cells were significantly damaged when the concentration of alcohol was 60 μL/mL and 80 μL/mL, which verified the reversibility and irreversibility of liver function recovery under different conditions. Studies have shown that liver injury can promote the proliferation of hepatic stellate cells, and co-culture of hepatocytes and hepatic stellate cells is more conducive to the recovery of hepatocyte function after alcohol stimulation. This method is helpful to reproduce the physiological environment of ALD, and has great potential in the study of physiological and pathological mechanisms and the testing of drug toxicity.

In recent years, with the change of dietary structure, NAFLD has become the most common liver disease worldwide. The disease is not associated with alcohol intake, but with excessive intake of high-sugar and high-fat foods, which often lead to liver fibrosis, liver failure and even liver cancer. Due to the lack of drug screening platforms for NAFLD, the understanding of its causes has been very limited^[58⇓~60]. As shown in Fig. 2B, Yu et al. Developed a 100-well NAFLD drug screening platform based on microfluidics. The chip can generate liver organoid clusters with uniform size and reshape the main structure of NAFLD^[61]. Compared with the organoid cultured on the chip after pre-differentiation, the organoid differentiated in situ in the chip has a better chemical microenvironment, which can promote liver differentiation, improve liver function, and enhance cell polarity and bile duct structure. The results showed that the cells were induced with free fatty acids, and the cells showed a high level of lipid accumulation, affecting blood glucose regulation and reducing AKT phosphorylation in organoids. As a sensitive high-throughput drug testing platform, this method can help accelerate drug development for NAFLD.

Type 2 diabetes mellitus is a chronic metabolic disease, which accounts for 90% of diabetic patients. Qin et al. Proposed a microfluidic multi-organoid system^[62]. As shown in Figure 2C, human induced pluripotent stem cells (hiPSCs) self-organize to form clusters that induce differentiation to form liver and islet organoids. Organoid clusters were perfused in this system and co-cultured continuously for one month. The results showed that the cells of the liver and islet organoids were active and the mediators and secreted metabolites between the organs could be exchanged, which was suitable for exploring the insulin and glucose regulation of the human liver-islet axis in disease. This method is helpful to explore the interaction of multiple organs, and provides a platform for the study of disease mechanism and the screening of disease-related drugs.

Hepatitis B (HBV) is an infectious disease of the liver caused by the hepatitis B virus^[63]. Traditionally, cell lines and two-dimensional cultured primary cells are often unable to simulate the complexity of the liver, resulting in limited research on the process of HBV infection. Dorner et al. Used 3D microfluidic chips to study the process of hepatitis B virus infection^[64]. As shown in Figure 2D, collagen was coated on the styrene scaffold for the culture of primary human hepatocytes, and the system could last for 40 days, thus allowing the reconstruction of various steps of the HBV life cycle. This work proves that the co-culture of liver parenchymal cells and non-parenchymal cells can facilitate the identification of the source of cells producing immune effector factors, and provides a valuable research platform for preclinical HBV.

The development of safe and effective clinical drugs is a time-consuming, difficult and expensive process. Studies have shown that it takes more than 10 years and costs about $1.7 billion from the development of new drugs to their approval for marketing^[65]. In this process, most of the compounds are eliminated due to organ toxicity or lack of efficacy. Hepatotoxicity of drugs is a problem that must be faced and solved in preclinical drug research and development, and it is also an important reason why drugs are withdrawn after being introduced into the market^[66]. Animal models similar to humans are usually used for safety verification such as hepatotoxicity before clinical practice, but the use of animal models in the early stage of drug development is not only expensive and time-consuming, but also has species differences, and the effects of drugs on the cellular level can not be observed at any time, only through the observation of animal phenotype, so the mechanism of disease can not be explored^[67]. Therefore, the development of efficient, convenient and reliable microfluidic liver organ chips for rapid evaluation of drug hepatotoxicity has become an urgent need in current research.

Lin et al. Made a microfluidic device that can simulate the metabolic function of the liver and detect the metabolism of prodrugs^[68]. Capecitabine (CAP), as a prodrug, is non-toxic to cells and can only be converted into 5-fluorouracil (5-FU), an effective chemotherapeutic drug that is toxic to cells, after being metabolized into an active intermediate by the liver. They achieved two-dimensional co-culture of liver cells (HepG2) and breast cancer cells (MCF-7) on a microfluidic chip, and integrated a solid phase extraction column on the chip to extract drug components, which can prove the process of drug metabolism by the liver and the toxic effect of drugs on tumors through online detection by mass spectrometer. During metabolism in vivo, the drug may produce side effects in other organs, even though it appears to have no toxic effect on the liver^[65]. Therefore, the development of multi-organ chip system is conducive to early detection of drug off-target side effects and other risks in the process of drug testing. Hickman et al. Of Hesperos, an organ chip company, built a multi-organ chip system to predict preclinical target efficacy, liver metabolic conversion, and drug off-target toxicity (Figure 3A)^[69]. Chamber 1 is used to culture liver cells to simulate the first step of drug metabolism when drugs enter the human body, and leukemia cells can be directly cultured in other chambers to evaluate the inhibitory effect of drugs on bone marrow proliferation and the relationship between liver and bone marrow. In addition, the chip can also culture cardiomyocytes and microelectrode arrays for measuring myocardium in chambers 2 and 4, respectively, and culture different tumor cells in chambers 3 and 5 to assess the effects of drugs metabolized by the liver on the heart and tumors. This method provides a new platform and protocol for preclinical drug efficacy and safety assessment.

Two-dimensional cell culture is the main culture method for drug screening in vitro, but the cells cultured in two dimensions do not have a specific tissue structure, so three-dimensional cell models are usually used for secondary screening. Therefore, scientists have gradually shifted their focus to the development of three-dimensional liver organ microfluidic chips. As shown in Figure 3B, Hamilton et al. Of Emulate, an organ chip company, developed a microfluidic chip based on a porous membrane to simulate the physiological structure of the liver^[70]. The chip is divided into an upper layer and a lower layer by using a porous membrane, wherein the upper layer is used for culturing liver cells by a sandwich method, and the lower layer is used for culturing liver-related cells (including endothelial cells, Kuffer cells and hepatic stellate cells), so that the complex physiological structure and the physical and chemical conditions in the liver are simulated. This method provides a more effective way to detect hepatotoxicity, hepatocyte injury, steatosis, cholestasis and fibrosis, and species-specific aspects after treatment with drugs. In early drug screening, it is very important and necessary for the overall process of drug research and development to be able to screen out drugs with greater side effects in time. Therefore, it is of great necessity to develop a high-throughput and high-content drug screening platform^[71]. Wheeler et al. Developed a microfluidic organoid platform based on a digital microfluidic system for drug screening, as shown in Fig. 3C^[72]. The device can generate addressable free-floating liver organoids to explore the effects of different experimental conditions on liver organoids. The experimental results proved that the organoids exhibited fibroblast-dependent contractile behavior, showing high cell activity and metabolic function. The platform could be a cost-effective tool in the early drug discovery phase. A prodrug is a substance that is less active than the original drug but becomes active after hepatic metabolism, which enhances drug targeting and reduces drug toxicity and side effects^{[73⇓⇓~76]}. As shown in Fig. 3D, Jiang et al. Developed an integrated biomimetic array chip for prodrug screening^[77].

The chip consists of two functional chips: a liver chip and a tumor chip, and both the liver and the tumor are cultured in collagen in three dimensions. Using this device, they evaluated the anticancer bioactivity induced by prodrug metabolism and studied the interaction between different drugs. The chip can be used to simultaneously evaluate the anticancer bioactivity and possible hepatotoxicity of prodrugs. Overall, this approach provides a liver organ-on-chip platform with integrated functionality and operational simplicity and high throughput, which can be used as an auxiliary screening tool in the drug development process. In addition to hepatotoxicity, cardiovascular toxicity is also a side effect of drugs that can not be ignored^[78,79]. As shown in Fig. 3e, Khademhosseini et al. Developed a multi-organ system integrated with microfluidic biosensors to detect the toxic effects of drugs on liver and heart organs^[80]. The device contains a multi-organ chip and multiple biosensors. Liver and cardiomyocyte culture is integrated on the multi-organ chip, which is mainly used to detect the toxic effects of drugs on liver and heart. The biosensor is mainly used for in situ continuous online detection of small doses of biomarkers (such as albumin, CK-MB, etc.) secreted by cells to determine the hepatotoxicity and cardiotoxicity of drugs. In addition, the device can also detect the physiological parameters of the system, such as oxygen, pH and temperature. This method effectively combines multi-organ chips with sensors for online detection, providing a technical platform for studying multi-organ interactions and simultaneously detecting biomarkers.