Microbiota‐gut‐brain axis multi‐organ chip construction and applications in drug evaluation

The intestine, a central organ for digestion and absorption in the human body, exhibits distinct roles across its different segments in terms of physiological function, microbial community structure, and drug metabolism. The small intestine is primarily focused on efficient digestion, absorption, and rapid substance transport, with its villi and microvilli structures significantly expanding the absorptive surface area. In contrast, the large intestine emphasizes water reabsorption, fermentation of undigested materials, and maintaining a stable anaerobic microbial community. Current research often lumps both together under the general term intestine, overlooking the divergent functional needs of each segment in chip design. Indeed, precisely simulating the physiological microenvironments of different intestinal segments is a fundamental starting point for the development of gut‐on‐a‐chip (GoC) technology—small intestine chips need to concentrate on dynamic mechanical stimuli and rapid substance transport, while large intestine chips require adaptation to three‐dimensional structures and stable anaerobic environments.

In the realm of biomimetic mechanical structure design, researchers have explored the synergistic effects of fluid shear stress and mechanical peristalsis. Studies have demonstrated that fluid shear stress generated by liquid flow within a tube can stimulate tissue oxygen exchange, promote extracellular matrix remodeling, facilitate the formation of basement membranes and microvilli, as well as mucus secretion, thereby improving cytoskeleton and intercellular tight junctions and promoting the formation of the intestinal barrier [48, 49]. Kim et al. [50, 51] developed a three‐channel Polydimethylsiloxane (PDMS) GoC that generates low fluid shear stress (0.02 dyne · cm⁻²) through low flow rates (30 μL/h), while applying cyclic suction to the hollow side chamber to expose epithelial cells to mechanical peristalsis with a 10% deformation frequency. This successfully induced the polarization of small intestinal monolayer columnar epithelium and formed villi structures highly consistent with in vivo conditions (Figure 2A,B). Notably, when mechanical peristalsis was ceased and only tubular flow was maintained, insufficient epithelial deformation led to excessive bacterial growth, exhibiting pathological features highly similar to intestinal obstruction and IBD [51]. This inversely validates the critical impact of missing mechanical stimuli on host‐microbe interactions and provides new insights for pathological model construction. With technological advancements, Shin et al. [52] further achieved the regeneration of functional gut microstructures within 5 days using morphogenesis methods, including three‐dimensional epithelial tissue and crypt‐villus spatial structures, highlighting the trend in mechanical design evolving from two‐dimensional polarization to three‐dimensional complex structures (Figure 2C,D). However, this segment‐adapted mechanical simulation still faces challenges: small intestine chips require precise control of the coupling effect of low fluid shear stress and high‐frequency peristalsis, while large intestine chips need to maintain three‐dimensional structural stability in a stable low fluid shear stress environment, posing higher demands on the fluid dynamics control of microfluidic systems.

Maintaining an anaerobic gradient within the lumen is a prerequisite for host–microbe interaction studies. Research has demonstrated that establishing an intraluminal hypoxia gradient in chips can enhance intestinal barrier function and preserve microbial diversity at physiologically relevant levels [53, 58]. To address this challenge, researchers have developed two strategies: spatial gradients and temporal gradients.

Spatial strategies involve physical separation to achieve zonal control of oxygen concentration. Jalili‐Firoozinezhad et al. [53] employed a nitrogen flushing method to reduce oxygen concentration to <0.5% within 30 min, while diffusing oxygen from the lower endothelial vascular channel through a PDMS membrane to maintain the low oxygen environment required for epithelial cell growth (Figure 2E,F). Grant et al. [54], on the other hand, used an impermeable membrane to regulate oxygen permeability in microchannels, combined with natural oxygen reduction from cellular aerobic respiration to sustain a steady‐state oxygen concentration <6 mmHg within the epithelial lumen (Figure 2G,H). Although both methods can construct physiologically relevant oxygen gradients, their technical logics differ significantly—the nitrogen flushing method relies on an external gas source to rapidly establish an anaerobic environment, suitable for acute experiments; whereas the impermeable membrane method requires no external equipment, offering higher system integration and suitability for long‐term dynamic monitoring.

Temporal strategies involve controlling the sequence of oxygen concentration changes to simulate host‐microbiota symbiosis. Penarete‐Acosta et al. [55] designed a three‐stage model where HCT116 colonic cell spheroids were pre‐cultured at 21% O₂, then transferred to an anaerobic chamber after cell attachment and three‐dimensional proliferation in the chamber. Oxygen levels were reduced to negligible levels within 24 h before introducing microorganisms for coculture, successfully simulating pathogen colonization in colorectal cancer microenvironments (Figure 2I,J). Shah et al. [56] developed the HuMiX chip, integrating oxygen sensors for real‐time dissolved oxygen monitoring. After forming a Caco‐2 epithelial barrier, Lactobacillus rhamnosus GG was added, reproducing in vivo transcriptional, metabolic, and immune responses of human intestinal epithelial cells (Figure 2K,L). These technological advancements highlight the evolution from static oxygen control to dynamic temporal regulation. However, maintaining the ecological complexity of microbial communities remains a core challenge—current models often use single or a few strains for coculture, failing to simulate resource competition among indigenous microbiota. Probiotics may also lose their colonization resistance in complex communities.

Notably, GM play a bidirectional regulatory role in drug absorption, distribution, metabolism, and excretion (ADME) processes: microbial enzyme systems can alter drug bioavailability, while drugs like antibiotics can rapidly disturb microbiota structure, leading to the enrichment of antibiotic resistance genes or disruption of gut–brain axis signaling. Nevertheless, current chips only test the functionality of single probiotic strains and do not evaluate their competitiveness with indigenous microbes in environments simulating host communities, potentially resulting in laboratory‐effective strains failing in real gut environments due to colonization resistance loss. Therefore, developing dynamic competition models involving multiple microbial species is a crucial direction to address this issue.

The application of gut‐on‐a‐chip in drug research highlights its technological advantages in replicating biomimetic microenvironments, following a logical progression from molecular mechanism analysis to systemic interaction revelation, and ultimately to intelligent intervention development. Early research focused on the molecular mechanisms of drug action within the gut. The Duodenum Intestine‐Chip (Table 1), targeting the small intestine, successfully predicted the intestinal transport efficiency of oral drugs and drug–drug interactions mediated by CYP3A4 by simulating the polarized cellular structure, gut barrier function, and specific cell subpopulations of the human duodenum [59]. This chip not only reproduced the high expression induction effects of rifampicin and 1,25‐dihydroxyvitamin D3 on CYP3A4 but also quantified the functional activities of drug efflux transporters (MDR1, BCRP, MRP2) and uptake transporters (PepT1, OATP2B1, OCT1, SLC40A1), providing a humanized model for assessing the bioavailability of oral drugs in the gut [59]. As research extended from the single dimension of drug–gut to the multidimensional interactions of gut‐whole body system, GoCs began to delve deeper into the interaction networks between the gut and distant organs during disease progression. Wang et al. [57] developed the Depression‐Gut‐on‐a‐Chip, integrating patient microbiota to successfully reproduce depression‐related pathological features such as reduced intestinal barrier function, chronic low‐grade inflammation, and decreased serotonin levels, offering an in vitro model based on intestinal pathological simulation for studying the mechanisms of psychiatric disorders (Figure 2M). To further enable high‐throughput screening and precise validation of disease intervention strategies, researchers have deeply integrated artificial intelligence technology with intestinal microenvironment regulation. Wu et al. [67] incorporated machine learning into an intelligent GoC, combining the chip with an environmental control system to provide a standardized and scalable intestinal microenvironment for coculturing various probiotics. They successfully evaluated the functional efficacy of 12 Bifidobacterium strains in alleviating colitis. This innovative chip + algorithm model signifies the functional transition of GoCs from tools for disease mechanism analysis to platforms for therapeutic target screening. This path of technological iteration driving dimensional expansion not only demonstrates the enhanced scenario adaptability of GoCs in disease research but also provides innovative solutions for precise analysis and efficient treatment of complex diseases through the technological integration of biomimetic design → ecological simulation → intelligent analysis. However, the current simulation of multiorgan system‐level ADME processes by chips is still insufficient and needs to be combined with MOCs such as the gut‐liver axis to improve the evaluation system.

Despite the significant advancements in biomimetic design, microenvironment simulation, and drug research achieved by GoCs, their technological evolution still faces three major bottlenecks. The limitation of a single functional dimension hinders the accurate simulation of neuro‐gut interaction diseases such as congenital megacolon [68]. Additionally, the adsorption effect of PDMS material on hydrophobic small drug molecules interferes with detection precision [69, 70]. Furthermore, simplified microbiota ecology in single‐strain models fails to reproduce the functional redundancy of complex microbial communities. Future breakthroughs should focus on integrating gut‐nervous‐immune multicellular types to enhance the simulation of complex diseases involving multiple systems. There is also a need to develop surface modification materials, such as coating with perfluoropolyether‐based lubricating layers [60], to mitigate adsorption effects. Moreover, constructing multi‐strain competition‐cooperation models is essential to better reflect the ecological complexity and functional redundancy of indigenous microbiota. From the perspective of technological evolution, GoCs have transitioned from single‐parameter optimization to systemic biomimetic integration. With cross‐disciplinary advancements in materials science, microfluidic technology, and computational modeling, they are poised to become core tools in precision medicine research. Particularly in areas such as microbiome‐host interactions and bidirectional risk assessment of drugs and microbiota, GoCs are expected to unleash even greater potential.

The BBB is a complex system comprising brain microvascular endothelial cells (BMECs) and their tight junctions (TJs), astrocytes (ACs), pericytes (PCs), basement membranes, and extracellular matrix. As the brain's protective barrier, the BBB isolates blood from brain interstitial fluid, restricting molecular transport between the bloodstream and brain parenchyma, thereby shielding the brain from toxins and maintaining neural microenvironment homeostasis [71]. BMECs express specific transporters and receptors that regulate transcellular transport, while TJs between cells limit paracellular diffusion [72]. These features confer remarkable selective permeability to the BBB, only lipophilic small molecules (<500 Da) can cross [73, 74], posing significant challenges for drug development.

Microfluidic technology has introduced a novel paradigm for constructing in vitro BBB models. Conventional BBB‐on‐a‐Chip (BBBoC) designs typically use a porous membrane integrated into the chip to form a bilayer structure, where BMECs and neuroglia are cultured on opposing sides to simulate vascular and neural lumens [75, 76]. Alternative designs employ structures such as microgaps, trapezoids, circles, or multiple repeating paths with shared access ports [77]. BBBoCs generally feature 3D vascular‐like structures and endothelial basement membranes, mimicking cell‐cell interactions in human BBB tissues and physiological properties like fluid shear stress [78].

Early research on BBBoC primarily focused on biomimetically recreating fundamental barrier functions. Wevers et al. [79] developed a dual‐channel microfluidic platform capable of housing multiple chips, utilizing surface tension techniques for patterned design of extracellular matrix gels to guide BMECs in forming perfusable vessels. They cocultured ACs and PCs on the glial side (Figure 3A,B). This model strictly limited the passage of 20 kDa FITC‐dextran dye and demonstrated significantly higher permeability for antibodies targeting the human transferrin receptor compared to control antibodies (apparent permeability of 2.9 × 10⁻⁵ vs. 1.6 × 10⁻⁵ cm/min, respectively), confirming robust barrier function suitable for studying BBB permeability to macromolecules and the differential sensitivity of antibody penetration. It is worth noting that permeability is commonly employed as an evaluation metric for constructing in vitro BBB models. Generally, the apparent permeability coefficient (P_app) of 20 kDa FITC‐dextran in physiological BBB is typically <5 × 10⁻⁶ cm/s. In conventional simplified cellular models, P_app often exceeds 1 × 10⁻⁵ cm/s. After optimization with cell coculture and fluid shear stress, the P_app range of models can be reduced to 1 × 10⁻⁶–10⁻⁸ cm/s [82, 83, 84, 85, 86]. However, measured values are often influenced by multiple factors: calibration bias in fluorometer excitation/emission wavelengths may cause fluorescence intensity measurement errors; flow instability in microfluidic systems can directly impact tight junction integrity, inducing deviations in P_app measurements; variations in cell seeding density or extracellular matrix coating selection may also trigger substantial permeability changes due to differences in barrier maturation. Therefore, strict standardization of experimental conditions is essential to ensure data comparability.

As technology advanced towards integrating multicellular interactions with 3D microenvironments, Ahn et al. [77] designed a chip featuring a 2D endothelial layer combined with a 3D glial network (Figure 3C). This design employed a layered architecture of vascular channel‐porous membrane‐pericyte/astrocyte channel to simulate the vascular‐perivascular niche, providing physiological contact conditions for intercellular signaling. Experience has validated that the model exhibits a P_app of (1.5 ± 0.2) × 10⁻⁶ cm/s for 4 kDa FITC‐dextran and (0.8 ± 0.1) × 10⁻⁶ cm/s for 40 kDa FITC‐dextran. These values are comparable to BBB substance permeability levels validated in rats in vivo (where P_app was (0.92 ± 0.46) × 10⁻⁶ cm/s for 4 kDa dextran and (0.19 ± 0.11) × 10⁻⁶ cm/s for 40 kDa dextran, respectively) [87]. It also showed high expression of tight junction proteins ZO‐1 and Occludin, as well as barrier function proteins Glut1 and P‐gp in hBMECs. Additionally, the end‐feet of ACs exhibited specific polarized expression of aquaporin‐4, indicating a high degree of physiological relevance to the barrier. Using fluorescence tracking of HDL‐mimetic nanoparticles with apolipoprotein A1, the chip enabled three‐dimensional mapping of nanoparticle distribution. Combined with scavenger receptor class B type 1 receptor blocking experiments, it confirmed that nanoparticle trans‐BBB transport relies on receptor‐mediated transcytosis, further validating the barrier's functional simulation capability in drug delivery scenarios.

The innovation in cell sources has further expanded the physiological relevance of BBBoC. Park et al. [80] utilized human induced pluripotent stem cells (hiPSCs) to differentiate into iPSC‐BMECs. By simulating embryonic BBB development through a hypoxic microenvironment, they induced high expression levels of tight junction proteins Claudin‐5 and Occludin, as well as the functional transporter P‐gp (Figure 3D). Chip‐integrated electrodes enable real‐time monitoring of transendothelial electrical resistance (TEER), a core electrophysiological metric that quantifies transcellular resistance across endothelial monolayers to assess barrier integrity and functionality [88]. Reported TEER values for BBB chips typically range from 3000 to 20,000 Ω · cm², partially overlapping with the physiological range of healthy human BBB (1500–8000 Ω · cm²) [83, 89, 90, 91]. However, measurement system optimization often introduces deviations: excessive spatial distance between electrodes and the endothelial barrier amplifies resistance contributions from other system components; suboptimal electrode‐to‐cell area ratios disrupt current density uniformity, thereby distorting resistance readings. Additionally, electrode coverage of cells impedes visual monitoring, further compromising measurement reliability. Park's team recorded impedance values of ~25,000 Ω, exceeding prior reports of BBB chip impedance (~400 Ω) with primary hBMECs and perivascular cells by an order of magnitude. This stark increase intuitively demonstrates the barrier‐strengthening effects of hypoxia induction combined with microfluidic coculture. Critically, measurements were not normalized by surface area, yielding results as impedance in Ohms (Ω) rather than resistance (Ω · cm²). Leveraging this high‐fidelity barrier model, the team not only replicated cetuximab's selective BBB transcytosis but also pioneered a novel paradigm of patient‐derived hiPSCs for personalized drug screening, providing a critical technical pathway for precision medicine in neurological disorders.

Integrating disease microenvironments has propelled BBBoCs to the forefront of clinical translation, provides a simulation platform for drug delivery research in various brain diseases (Table 2). Glioma, the most common malignant tumor in the brain, exhibits significant drug resistance due to the interaction between its tumor microenvironment and the BBB [96]. BBB chips address the lack of complex tumor microenvironment models in new drug development for tumors. Shi et al. [81] constructed a BBB‐U251 chip that integrates BMECs, ACs, PCs, and glioma cells (Figure 3E). By monitoring drug permeability across the BBB (apparent permeability coefficients of matrine (1.78 ± 0.20) × 10⁻⁷ cm/s and resveratrol (7.89 ± 3.68) × 10⁻⁸ cm/s) and drug responses in the tumor compartment (matrine inhibited U251 proliferation with a cell survival rate of 84.9 ± 3.5%; resveratrol resulted in a cell survival rate of 74.0 ± 13.3%), they confirmed that the BBB significantly attenuates drug efficacy, underscoring the necessity of BBB‐targeted strategies in anti‐glioma drug development. In conclusion, BBBoC technology has evolved along the path of biomimetic structure construction → physiological function simulation → clinical scenario adaptation, providing a near‐physiological platform for elucidating brain disease mechanisms, drug delivery across the BBB, and precision medicine. However, challenges such as the complexity of multicellular ecosystems, material compatibility, and clinical translation barriers remain critical directions for future technological breakthroughs.

As therapeutic strategies and drug delivery modalities evolve, BBBoCs have made significant advancements in evaluating novel drug delivery methods such as nanoparticles, adeno‐associated viruses, and stem cell therapies. By integrating three‐dimensional fluid control and multicellular coculture systems, BBBoCs can systematically assess the efficiency of various nanocarriers in crossing the BBB. For instance, doxorubicin‐loaded targeted liposomes can efficiently penetrate the BBB via transferrin receptor‐mediated transcytosis under physiological fluid shear stress [77]. Membrane‐coated nanoparticles utilize the immune cell coculture system within the chip to validate their ability to evade macrophage clearance and prolong their retention time in the brain [97]. In the realm of gene therapy, BBBoCs incorporate reversible barrier regulation techniques and use iPSC‐derived brain endothelial cells to simulate adeno‐associated virus serotype specificity, enabling high‐throughput screening of gene vector delivery pathways [98]. Stem cell therapy research reveals mechanisms by which stem cells regulate BBB function through paracrine pathways. In Parkinson's disease models, it has been demonstrated that barrier integrity directly impacts the ability of mesenchymal stem cell exosomes to repair nigral dopaminergic neurons [99]. These advancements rely on the chip's precise simulation of physiological fluid shear stress, multicellular interactions, and real‐time drug distribution. They not only accelerate the preclinical validation of novel delivery systems, but also drive the evolution of brain disease treatments towards precision, efficiency, and safety.

Despite the significant advancements, BBBoC technology still faces several challenges. Precise control is required over cell ratios, anatomical origins, seeding sequences, medium formulations, and long‐term phenotypic stability in multicellular co‐cultures to achieve optimal conditions for each cell type and maintain model fidelity. Current models often focus on core BBB cells and lack the incorporation of immune cells, neurons, and other distant interactions, which hinders the ability to fully recapitulate the complex pathological networks seen in neurological diseases. The preparation of personalized chips using patient‐derived hiPSCs is time‐consuming, costly, and exhibits poor compatibility with commercial monitoring devices, restricting clinical translation. BBBoCs also struggle to meet the demands for constant flow rates and high‐throughput requirements. Although Xiao et al. [100] developed the constant speed perfusion array chip to address some of these issues, further optimization of chip structure and integrated design is necessary to promote the convenience and commercialization of BBBoCs. Addressing these challenges will be instrumental in enhancing the functionality, reliability, and clinical applicability of BBBoCs, ultimately leading to more effective treatments for brain disorders.

The brain, as the most complex neurological organ in the human body, is composed of numerous glial cells and billions of neurons. Neurons are responsible for receiving, integrating, and transmitting information, while glial cells provide support, nutrition, and protection for neurons. Through their rich electrophysiological activities, the brain plays a dominant role in coordinating interactions among various internal organ systems and the external environment [101, 102]. Investigating the unique firing rhythms of neurons and their responses to chemical drugs and external stimuli is crucial for understanding the mechanisms of brain health and disease, and for advancing the development of therapeutic drugs for neurological disorders.

Microfluidic brain‐on‐a‐chip (BoC) platforms construct in vitro brain function research models through core modules such as neuron and glial cell culture, synapse formation, cell–cell communication, and immune simulation. In the realm of 3D structural innovation, Cho et al. [103] developed a three‐layer, five‐chamber brain‐on‐a‐chip that utilizes a human brain‐mimicking 3D hydrogel matrix combined with boundary element microfluidic technology to achieve bidirectional cerebrospinal fluid perfusion (Figure 4A). This system successfully fosters mature neurons and extensive neural networks during human brain development, spontaneously exhibiting brain morphological features. With the innovation of chamber structures, micropillar arrays have been integrated into devices for 3D cell culture [109]. Zhu et al. [104] introduced an octagonal micropillar array chip that revolutionizes traditional cumbersome processes involving embryoid body formation, neural induction, cell transfer, and 3D encapsulation (Figure 4B). This design allows hiPSCs to undergo embryoid body formation, neural induction, and differentiation in situ, resulting in functional human brain organoids with neuronal differentiation, brain regionalization, and cortical tissue characteristics. It provides an efficient model for studying early neurodevelopmental toxicity, such as cadmium exposure [110]. Both models prioritize precise 3D microenvironment simulation, but Zhu's technology streamlines the differentiation process through in situ integration via micropillar arrays, significantly simplifying operational procedures. This exemplifies how innovative chip structural design can radically enhance the efficiency of organoid culture.

At the level of cell interactions and functional expansion, the critical regulatory roles of non‐neuronal cells, such as glial cells in synapse formation and cell communication, drive the evolution of brain‐on‐a‐chip platforms towards multicellular coculture systems [111]. When iPSC‐derived neurons are mixed with ACs and transferred to the OrganoPlate® platform (Figure 4C), which features a microdroplet plate containing 96 tissue chips, the platform supports 3D culture, coculture, and noninvasive medium regulation. This environment enables rapid formation of neural networks and detection of spontaneous electrical activity, providing a high‐throughput platform for synaptic function research [105]. This design highlights the accelerated value of microfluidic chamber coculture in constructing cellular networks, while the platform characteristics of OrganoPlate® make it uniquely advantageous for high‐throughput screening of neurotoxicity. Beyond fundamental cell interaction studies, Cui et al. [106] combined a micropillar array brain‐on‐a‐chip platform with cortical organoids (Figure 4D) to explore the impact of prenatal environmental exposures on neurodevelopment. This model simulates exposure during pregnancy to breast cancer‐derived exosomes, revealing that treated organoids not only exhibited neuronal damage but also displayed carcinogenic phenomena associated with breast cancer progression. This design not only continues the efficient support of micropillar arrays for 3D organoid structures but also, through in vitro simulation induced by specific pathological factors, offers a high‐throughput research platform for addressing the critical preclinical question of how the tumor microenvironment during pregnancy interferes with neurodevelopment. This exemplifies the innovative application of BoC technology in analyzing environment‐related neurotoxicity mechanisms.

In the context of disease mechanism analysis and drug development, BoC platforms serve as indispensable core modules in MGBA‐MOC research for CNS disorders, demonstrating irreplaceable application value (Table 3). Take Alzheimer's disease (AD) as an example, its pathological features, such as Aβ plaque deposition, Tau protein hyperphosphorylation, and synaptic dysfunction, are closely interrelated [119]. Kilinc et al. [107] developed a three‐chamber microfluidic device that physically separates synapses from pre‐ and post‐synaptic neurons (Figure 4E). By chronically exposing primary hippocampal neurons to toxic Aβ peptides secreted by cells expressing mutant amyloid precursor protein, they observed significant damage to synaptic connections. Further treatment with antibodies targeting different Aβ forms confirmed that low‐molecular‐weight oligomers are key inducers of AD synaptic toxicity. They also discovered that selective expression of protein tyrosine kinase 2β in post‐synaptic neurons can protect neurons from Aβ1–42‐induced synaptic toxicity. Park et al. [108] designed a 3D microfluidic system integrating neurons, ACs, and microglia, which reproduced Aβ aggregation, Tau protein accumulation, microglial recruitment, axonal severing, and NO‐mediated neurotoxicity (Figure 4F). This system provides a multicellular pathological microenvironment simulation platform for studying neuroglial interaction mechanisms and drug discovery. Both types of AD chips focus on in vitro reproduction of pathological features, but Kilinc achieves precise mechanism analysis through physical synapse separation, while Park restores the complexity of pathological processes through tricellular coculture. This reflects the hierarchical application logic of BoC platforms in disease research, ranging from molecular mechanisms to systemic pathology.

Nevertheless, the development of BoC platforms still faces numerous technical challenges. Existing models struggle to accurately simulate the long‐term dynamic pathological processes of brain diseases in vitro: glial cells in coculture are prone to phenotypic drift due to nutrient depletion or microenvironmental imbalances, making it difficult to maintain stable cell ratios over extended periods. Simulating axonal degeneration relies on precise coordination of microfluidic shear forces and chemical gradients, yet current devices lack sufficient capability to control the mechanical‐chemical coupling required for dynamic remodeling of neural synapses. Disease‐related features such as immune cell infiltration gradients and spatiotemporal release of inflammatory factors are even more challenging to fully reproduce in chips. Future advancements need to focus on developing biodegradable material chips that can simulate brain tissue remodeling through controlled material degradation kinetics, thereby extending the functional stability of cells. Additionally, integrating techniques like photothermal neural stimulation and microelectrode arrays can enable real‐time control and monitoring of neuronal activity, aiding in the analysis of neural regulation mechanisms. Furthermore, coculturing patient‐derived hiPSCs with immune cells to construct personalized brain disease models, combined with AI algorithms to predict changes in pathological parameters, can gradually approximate the complex pathological networks of brain diseases. Overcoming these challenges will be crucial for advancing BoC technology, ultimately leading to more effective in vitro models that can better inform our understanding of brain diseases and facilitate the development of targeted treatments.

Multiorgan chips (MOCs) involve linking organ‐specific chips via microchannels to study the interactions and connections between various physiological systems and different tissues. Compared to single‐organ chips, MOCs offer greater physiological relevance, better simulating whole‐body systemic responses [120].

MOCs are categorized into two types: (1) modular coupling of single‐organ chips and (2) multiorgan‐on‐a‐chip plates (Figure 5A). The modular approach connects individual organ modules using capillary channels or microfluidic motherboards, enabling independent development of single‐organ models and flexible reconfiguration of MOC platforms. This method also supports vascularized organ cultures via organ‐specific endothelial cells [126, 127]. In contrast, multiorgan‐on‐a‐chip plates integrate multiple organs on a single platform, facilitating interorgan communication through microchannels mimicking vasculature. This method provides a paradigm for Human‐on‐a‐Chip, which replicates a miniaturized human circulatory system in a compact design while minimizing leakage risks by eliminating manual interconnections [128].

Chip cascade technology is pivotal for constructing MOCs, with the key challenge being establishing functional vascular networks to deliver nutrients, remove waste, sustain tissue physiology, and enable inter‐organoid communication [129]. Current strategies for in vitro vascularization include 3D printing, vasculogenesis, and angiogenesis [130]. 3D printing constructs large‐scale (>100 µm) vascularized tissues with tunable geometries. Miller et al. [121] created a rigid carbohydrate glass filament network via 3D printing, serving as a cytocompatible soluble template to engineer endothelialized networks mimicking blood perfusion under pulsatile flow (Figure 5B). This method allows independent control of vascular geometry, endothelialization, and extravascular tissue formation, accommodating diverse cell types, matrices, and crosslinking strategies. Sub‐100 µm capillaries can be engineered using endothelial cell emergent behavior via vasculogenesis [131] or angiogenesis [132] strategies. Chen et al. [122] cocultured human umbilical vein endothelial cells and fibroblasts in microchannels, generating microvascular networks via endothelial cell emergent behavior (Figure 5C,D), which recapitulated tumor cell exudation and metastatic behavior.

MOCs can be classified into static, semi‐static, and flexible systems based on cascading modes [133]. Static cascading involves placing specific tissues or organs in individual chambers of a single microfluidic chip, sequentially connected via a common fluid flow. The connection order between the organs is restricted by the permanent geometry of the chip. Esch et al. [123] developed a static MOC containing gastrointestinal, liver, and other tissue modules to study the composite effect of nanoparticle‐tissue interactions (Figure 5E). Its fixed fluidic configuration limits dynamic adjustments. The concept of semi‐static cascade was proposed by Wagner et al. [134]. This approach integrates microfluidic channels, pumps, and sensors to enable pre‐culturing of individual tissues and pre‐configured tissue combinations within one device [135, 136]. Maschmeyer et al. [124] designed a semi‐static four‐organ‐chip with pre‐formed gut and skin models, a 3D liver lobule mimic, and a renal barrier simulated by a monolayer of human proximal tubular epithelial cells (Figure 5F,G). A peristaltic micropump drives pulsatile media flow through interconnected tissue chambers. The flexible cascade offers maximum flexibility, allowing post‐culture integration of organ modules via microtubing or microconnectors [137]. Loskill et al. [125] proposed a modular plug‐and‐play “μOrgano” MOC, where each organ unit comprises a functional master chip and a connective plug‐and‐play connector arranged on an equidistant grid (Figure 5H). They interconnected multiple cardiac chips in series, successfully recapitulating the cellular viability and spontaneous beating rhythms of healthy cardiac tissue. The complexity of MOCs extends further, with the ultimate goal being Body‐on‐a‐Chip systems that replicate whole‐body physiology via Multiorgan cascade technology.

The technological advancement of MOCs provides a robust platform to replicate continuous mediator circulation and inter‐tissue interactions within physiological pathways, enabling comprehensive modeling of the human internal environment for disease pathology, pharmacokinetics, and toxicology studies (Table 4). Ronaldson et al. [126] developed an organ interstitial chip using polysulfone, a biocompatible inert material. This system cultures human‐derived heart, liver, bone, and skin tissues in their respective optimized microenvironments, interconnected via a vascular flow containing immune cells, cytokines, and extracellular vesicles. A selectively permeable endothelial barrier separates tissues from peripheral vascular flow, creating a multiorgan system with interdependent organ functions (Figure 5I). After 4 weeks of culture, interconnected tissues retained normal structural and functional phenotypes and recapitulated the pharmacokinetic and pharmacodynamic profiles of doxorubicin in vivo.

Overall, MOC cascade technology offers a novel strategy for drug development, complementing animal experiments and clinical trials. By mimicking the human body's complex physiological environment under ethical constraints, this technology enables multiorgan‐level drug screening, metabolic profiling, and toxicology assessments. However, MOCs remain largely in the laboratory phase, with commercial applications still underdeveloped, and face several challenges. First, developing physiologically relevant organ and barrier models, along with optimizing environmental perfusion strategies, requires further refinement. Next, due to chip size limitations, cell type composition and ratios must be carefully calibrated during organ scaling to physiologically appropriate sizes [144, 145]. Otherwise, the normal morphology and function of the organs may be affected [146]. Future advancements should focus on enhancing interorgan communication and integrating sensors for multimodal real‐time analysis of organ functional states and biochemical levels under stable coculture conditions.

No new data were generated or analyzed in this review. Supplementary materials (graphical abstract, slides, videos, Chinese translated version, and update materials) may be found in the online DOI or iMeta Science http://www.imeta.science/imetaomics/↗.