Microfluidic devices remain difficult to translate into commercial products, particularly for three-dimensional (3D) applications. While soft lithography offers high-resolution features and biocompatibility, its workflows and planar geometries are limited. Conversely, industry-standard injection moulding enables scalability, but lacks flexibility and incurs high upfront costs. Here, we present an innovative hybrid microfabrication methodology combining stereolithographic 3D printed moulds with flexible wire templating and polydimethylsiloxane (PDMS) casting to create monolithic 3D microfluidic devices. Leveraging injection moulding design principles, this enables the rapid and repeatable fabrication of microfluidic devices with embedded internal channels, variable-height membranes, and curved geometries. Our methodology is validated using a variety of devices, including flexible membranes, plug-based flow distributors for 96-well plates, and high-aspect-ratio bioreactors with integrated electrodes. The resulting devices were leak-proof, reusable, and compatible with standard cell culture systems. This scalable, low-cost methodology retains the advantages of PDMS while allowing for 3D design freedom for both internal channels and external features, making it ideal for emerging applications such as organ-on-chip, bioprinting, diagnostics, and soft robotics.

Organ- and organoid-on-a-chip technologies provide critical human-relevant models for drug screening. Despite their promise, translating their complex biological outputs into reproducible, quantifiable, and pharmacologically interpretable readouts remains a significant challenge. To address this, artificial intelligence (AI) is increasingly employed to process the high-content imaging, sensor, and molecular data derived from these platforms. Crucially, this integration elevates AI from a conventional post-experimental analytical tool into a comprehensive framework that actively drives quality control, response quantification, model integration, and critical screening decisions. This review examines AI-augmented microphysiological systems across the entire drug screening pipeline by connecting biological readouts with specific computational strategies and pharmacological endpoints. We evaluate representative platforms, analytical methodologies, and specific applications where computational frameworks enable model standardization, robust phenotype interpretation, mechanism-informed evaluation, and compound prioritization. Furthermore, we outline primary barriers to clinical translation, including inherent biological and engineering variability, model generalizability, the need for external validation, and clinical dose relevance. Ultimately, these insights establish a comprehensive framework for evaluating the reproducibility, pharmacological applicability, interpretability, and translational potential of AI-driven microphysiological screening.

On-site nucleic acid detection plays a crucial role in disease diagnosis, biosafety monitoring, and food quality control. This study develops a novel nucleic acid detection platform that integrates catalytic hairpin assembly (CHA) with the CRISPR/Cas12a system and utilizes pregnancy test strips (PTS) for result visualization, addressing the limitations of existing nucleic acid detection methods in balancing sensitivity, specificity, and portability with cost and dependence on a cleanroom. The main mechanism involves the following three steps. The presence of target RNA triggers the CHA reaction, generating double-stranded DNA (dsDNA) as an activation unit. Subsequently, this unit activates the CRISPR/Cas12a system to specifically cleave the single-stranded DNA (ssDNA) that has bridged human chorionic gonadotropin (HCG) to a magnetic bead, ultimately releasing HCG that produces a visual result on the PTS. This dual-signal amplification strategy (CHA cycling and Cas12a trans-cleavage) can detect concentrations as low as 10 pM in approximately 50 min, without the need for pre-amplification of the target nucleic acid. This detection system ensures high sensitivity and specificity while effectively avoiding non-specific activation. In practical applications with transgenic maize samples, the detection results are highly consistent with those of real-time quantitative polymerase chain reaction (qPCR), validating its reliability in real-world scenarios. This innovative method offers advantages such as simple operation and low cost, providing an efficient tool for rapid nucleic acid detection while demonstrating broad potential for application in resource-limited settings.

Acoustofluidic buffer exchange enables continuous, contact-free replacement of the fluid surrounding suspended particles and cells, offering a gentle alternative to conventional centrifugation-based processing. Its performance is governed by the interplay between acoustic radiation-driven particle migration and streaming-induced fluid transport. Here, we investigate buffer exchange in a silicon-glass bulk acoustic wave microchannel actuated by a piezoelectric transducer, combining experimental characterization, numerical modeling, and biological validation. Particle transport was examined across a range of actuation voltages and flow rates to evaluate the competing effects of particle migration, interface displacement, and hydrodynamic residence time. No-flow calibration experiments confirmed the expected quadratic dependence of migration velocity on acoustic actuation, establishing a radiation-dominated reference independent of flow. Under continuous operation, however, finite residence time and acoustically induced interface motion limited achievable buffer exchange. Metrics relating particle migration to interface velocity and residence time enabled identification of operating regimes associated with efficient particle transfer and varying degrees of buffer preservation. Numerical simulations supported the experimental observations and clarified the distinct contributions of acoustic radiation forces and acoustic streaming. Biological validation using U2OS and INS-1 pancreatic β-cells demonstrated effective buffer exchange, high cell viability (>98%), and improved recovery compared with centrifugation-based processing. These findings provide practical design guidelines for balancing particle transfer, buffer retention, and biological compatibility in continuous acoustofluidic cell-processing systems.

Glioma progression is closely associated with vascular-related microenvironmental cues, but conventional in vitro models often lack spatial control over tumor spheroids, extracellular matrices, endothelial cells, and stromal support cells. Here, we developed a compartmentalized vascular-mimetic glioma spheroid-on-a-chip for visualizing glioma-vascular-associated cell interactions in a three-dimensional hydrogel microenvironment. The device integrates a central hydrogel channel for pre-formed glioma spheroids, two adjacent medium channels for human umbilical vein endothelial cells (HUVECs) and normal human dermal fibroblasts (NHDFs), and an L-shaped barrier for stable gel-medium interface confinement. Simulation and dye perfusion experiments confirmed improved hydrogel confinement and stable compartmentalization. Optimized HUVEC/NHDF co-culture supported vascular-like endothelial-stromal organization and invasion into the hydrogel compartment. In the integrated chip, the HUVEC/NHDF co-culture increased the diameters of U87 and U251 spheroids by 26.4% and 21.4%, respectively, and induced endothelial-associated cell accumulation around U87 spheroids. Transwell assays further indicated that glioma cells together with stromal cells promoted HUVEC vascular-like organization. This platform provides a simplified and reproducible model for studying selected reciprocal interactions between glioma spheroids and vascular-associated stromal components under spatially controlled three-dimensional culture conditions.

Industrialization and clinical translation applications based on organoids are often limited by low automation, high matrix consumption, and poor reproducibility. Here we present the VitroOrganoSphere (VOS) system, an automated droplet-engineered organoid sorting platform, which integrates droplet generation, photoelectric detection, fluorescence-based content evaluation, sorting, and automated dispensing into 96-well plates. Monodisperse droplets (CV <3%) are deposited as single droplets per well with Matrigel consumption as low as ∼0.3 μL. Real-time quality control enables rejection of empty or overloaded droplets and selective collection of droplets with appropriate encapsulation of viable organoids. Under the rigorous screening of two droplet selection criteria (size and sample content), the VOS system ultimately achieved a recovery rate of over 90% and accuracy of over 95%. Encapsulated organoids maintain viability and exhibit enhanced growth compared with manual methods. Drug sensitivity testing using patient-derived tumor organoids reliably distinguishes drug-sensitive and drug-resistant samples and reveals interpatient heterogeneity across multiple therapeutic agents. The VOS system establishes a standardized workflow for automated organoid-based drug screening and supports applications in precision medicine.

The lymphatic system is essential for fluid homeostasis and immune surveillance, yet its role in skeletal regeneration remains poorly defined. Here, we introduce a biomimetic organ-on-a-chip (OoC) model that recapitulates lymphatic endothelial cell (LEC) behavior under precisely controlled flow-mediated environments and elucidates their regulatory contribution in alveolar bone repair. Under flow-mediated shear stress, LECs exhibited enhanced sprouting, forming lumenized lymphatic structures in vitro. Furthermore, conditioned medium from perfused LECs significantly promoted the proliferation and osteogenic differentiation of bone marrow mesenchymal stem cells (BMSCs). Mass spectrometry identified WISP1 as a key mechanoresponsive secreted factor with high levels in LECs that activated canonical Wnt signaling in BMSCs, thus contributing to enhanced osteogenic marker expression and mineralization. In vivo validation using a murine molar extraction model with controlled occlusal loading revealed that occlusal loading promoted lymphangiogenesis and accelerated alveolar bone regeneration, accompanied by elevated WISP1 expression within alveolar sockets. Conversely, pharmacological suppression of lymphatic expansion through MAZ51 administration impaired bone formation while reducing WISP1 levels, thereby confirming the functional contribution of lymphatic vessels in alveolar bone regeneration. Together, these findings demonstrate that LECs function as mechanoresponsive regulators that transduce shear-related cues into osteoinductive signals through WISP1-mediated paracrine communication. Our study uncovers a previously unrecognized lymphatic-osteogenic signaling axis and establishes a physiologically relevant OoC platform for systematic investigation of lymphatic mechanotransduction in bone regeneration.

Microfluidic systems incorporating or contained within hydrogels are important in creating microphysiological systems (MPSs). Often naturally-derived hydrogels are used, as their inherent bioactivity supports dynamic cellular behaviors. Hydrogel biomaterials that are partly or fully synthetic are desirable in engineering systems with specific, designed properties, though they typically lack bioactive features of natural materials without additional molecular design. In particular, engineering synthetic biomaterials to support physiologically relevant, dynamic cellular behaviors is an important design goal. Granular hydrogels inherently permit dynamic cellular activity, owning to porosity between particles and dynamic material properties in the absence of interparticle crosslinking. However, using dynamic granular hydrogels in MPSs requires stable channels to perfuse fluid in these dynamic systems. Here, we establish channels within granular hydrogels to enable perfusion through spatially controlled interparticle crosslinking. Selective crosslinking allowed for the formation of stable channels while allowing hydrogel microparticles between two channels to remain uncrosslinked. This allowed spatiotemporal control of signals within an environment established from microparticles without interparticle crosslinking. Fluorescently tagged molecules allowed for the visualization of controlled soluble gradients between two channels within the device. Additionally, embedded 3D printing processes can be used to specify material composition within the system, demonstrating integrated technology for engineering well-defined hydrogel systems. Integrated microfluidic-based control over soluble signals in a system that is compatible with 3D printing processes will establish a basis for building MPSs for broad applications, and the ability to maintain granular systems in culture without interparticle crosslinking will enable design of synthetic hydrogels that access unique dynamic properties within these systems.

Thermochromic materials with temperature-dependent color changes and color memory are widely used in functional coatings, anti-counterfeiting, and smart architectures. Cholesteric liquid crystal (CLC) molecules undergo periodic helical twisting of their orientation, forming a unique helical superstructure with fluidity and long-range orientational order. When confined in microscale spherical droplets with interfacial anchoring, CLC assemblies exhibit dynamic optical response due to thermoresponsive molecular rearrangement. The helical defects within physical confinement and the interfacial molecular anchoring jointly interfere with the cholesteric reorganization, offering intelligent color modulation and structural stability but with limited color saturation and slow response. Herein, we construct molybdenum disulfide (MoS2)-doped double emulsion droplets with CLC as the core using droplet microfluidics, creating near-infrared-activated thermochromic microcapsules (NIR-TCMs) with enhanced response and color saturation. The microcapsules with controllable diameters (75.4-275.0 μm) and shell thicknesses (9.7-43.0 μm) are formed by polymerizing the middle phase of the double emulsion droplets. NIR-TCMs (136 μm core, 16 μm shell) of 0.5 wt% MoS2-doped CLC show approximately 3-fold reduction in total response time from 19.0 ± 2.35 s to 6.2 ± 0.84 s and an increase in excitation purity from 0.16 ± 0.03 to 0.55 ± 0.05. By tuning molecular arrangement via entropic and enthalpic effects and enhancing thermal transport via heat conduction, MoS2 mitigates the thermal resistance of the polymer shell, leading to a rapid thermochromic response. Additionally, MoS2 boosts color saturation and enables NIR photothermal chromism via broadband absorption. The NIR-TCM films display static patterns at room temperature, while they create new visible patterns via NIR excitation. These patternable NIR-TCMs enable remote modulation of the optical properties of individual microcapsules and the overall device, offering a photonic chip for information encryption and optical coatings.

Complex coacervates (CCs) effectively concentrate biomolecules via liquid-liquid phase separation; however, extracting analytes for detection remains challenging because of the high ionic strength required to dissolve the CC phase. Here, we demonstrate a workflow integrating CC-based pre-concentration with nanopore sensing, which inherently operates under high-salt conditions. We investigated CCs formed by polycations (PLL, PDDA) and nucleotides (ATP, ADP) for the enrichment of microRNAs (miRNAs). While these CC components inhibited reverse-transcription polymerase chain reaction (RT-PCR), the dissolved phase was compatible with lipid bilayer-based α-hemolysin nanopore detection. Optimization of the polymer composition (S-PDDA/ATP) minimized pore clogging and noise. Using specific DNA probes, we successfully detected miRNA enriched in the CCs immediately after salt-induced dissolution. This study establishes a proof of concept for using phase separation as a seamless pre-concentration strategy for nanopore-based biosensing.

Neutrophils, the innate immune system's first line of defense, function in pathogen removal through diverse cellular responses. One critical response is neutrophil extracellular trap (NET) formation, which, despite its importance in inflammation, remains poorly understood and lacks high-resolution tools to effectively study its multiple forms: vital and suicidal. To begin addressing this gap, we introduce a tunable micropatterned platform that enables single-cell analysis of NETosis with spatial precision. Using photopatterned bacterial extracellular vesicles (bEVs) on a functionalized surface, we achieved controlled alignment of human peripheral blood neutrophils and demonstrated that engineered patterns control key NETosis outcomes. Meanwhile, phorbol 12-myristate 13-acetate (PMA) stimulation induced uniform suicidal NETosis. This approach supports additional applications, enabling tunable neutrophil patterning and live-cell imaging for intracellular investigation. Additionally, there is flexibility for different extracellular vesicle and nanoparticle patterning and pre-stimulation studies. Here, we present a novel system to quantify and manipulate NETosis pathways at the single-cell level, opening new avenues to explore immune heterogeneity and develop targeted therapies.

High-throughput phenotypic cell sorting is essential for elucidating disease mechanisms and identifying therapeutic targets. Conventional magnetic sorting methods largely fail to isolate multiple cell subpopulations based on membrane protein expression levels, while suffering from fixed capture thresholds, limited resolution, and harsh cell release. To address these challenges, we developed a modular, programmable electromagnetic microfluidic platform. By synergistically integrating an electromagnetic array with a microfluidic chip, the platform utilizes multi-channel, independently driven currents to generate multilevel, high-gradient magnetic fields along the flow direction on a soft magnetic core array. This architecture successfully transforms conventional physical structural limitations into dynamically adjustable electrical parameters, enabling flexible reconfiguration of cell capture thresholds and release strategies solely through preset current-combination programs. Simulation studies elucidate the underlying physical mechanisms governing the stepped magnetic potential landscape and tunable capture thresholds. Experimentally, the platform can precisely classify target cells into four phenotypic subpopulations (high, medium, low, and negative), according to their varied protein expression levels. Furthermore, by adjusting the combinations of driving currents, the phenotypic expression profile distributions of these four cell subpopulations can be dynamically regulated. Concurrently, the platform achieves spatially localized enrichment of multiple target cell types in distinct capture regions based on differences in their overall expression levels. Benefiting from semiconductor temperature control and a gentle recovery mechanism, cellular physiological functions are maximally preserved. With its excellent modular scalability, this platform holds broad prospects for tumor heterogeneity analysis, rare cell isolation, downstream single-cell omics, and precise drug evaluation.

Additive manufacturing has been invaluable for 3D microfluidic integration. We combine high-resolution two-photon printed microfluidic features with fast one-photon printing to reduce overall fabrication time by about 20-fold without compromising the accuracy of performance-critical features. Alignment strategies and interface design tolerances ensure fluidic sealing and the reproducible assembly between both 3D printing techniques. The resulting devices produce highly monodisperse GelMA-based foams with a polydispersity index of below 5% and tunable bubble radii between 235 and 400 μm. Combining multiple print modalities into a unified fabrication workflow will harness efficiency gains by utilizing high precision only where necessary.

Rare cells, such as circulating tumour cells, minimal residual disease-associated populations, and antigen-specific immune cells, carry disproportionate diagnostic and therapeutic value. Yet rarity makes measurements fragile: sampling statistics, background interference, workflow-induced bias, and loss of provenance can dominate signals, making many molecular or functional assays impractical without prior enrichment. Microfluidics offers a route beyond capture-and-count by operating at cellular length scales, enabling deterministic handling, low-dead-volume processing, and integration of enrichment with analytical readouts in closed, automatable formats. In this Tutorial Review, we discuss microfluidic rare-cell platforms as information-preserving workflows that link sample burden, target/background contrast, metric prioritisation, and downstream assay compatibility. We first define rare-cell contexts and workflow constraints, then introduce scenario-dependent design rules that connect input burden, target/background contrast, performance metrics, and downstream assay compatibility. We next compare label-free methods, label-based strategies, and hybrid pipelines, emphasising trade-offs among recovery, purity, throughput, viability, phenotype preservation, and workflow stability. We then introduce analytical modules for captured rare cells, spanning identity-confirming counts, nucleic-acid assays, multiplexed protein measurements, and functional phenotyping. Finally, we discuss omics-enabled workflows and translation requirements for diagnosis, drug screening, cell-therapy development, clinical implementation, and industrial adoption.

The genetic basis for sperm swimming ability in humans is underinvestigated. In particular, selfish genes known as segregation distorters dramatically influence sperm swimming in organisms like mice. Segregation distorters are difficult to screen for in humans because no tool exists to compete sperm by swimming ability and collect the separated fast and slow sperm for further study. We designed, built, and tested a "sperm racetrack", a microfluidic device that races sperm against each other and allows collection of sperm based on the distance travelled. This biologically-inspired PDMS device contains a long, straight channel with counterflow against which sperm swim naturally, and contains regularly-spaced ports along the channel for collection of sperm after the race. We performed a series of test races and found that sperm in the fourth channel segment are significantly faster swimming according to video analysis than those in the second channel segment. Sperm from these same segments are significantly different in speed after extraction from the channel as well. We characterized several swimming behaviors for these faster and slower sperm, including linearity (ability to swim in a straight line) and wobble (amount of side-to-side head movement per beat), and found a significant association between higher sperm wobble and distance travelled along the channel in extracted sperm. Together, we show that the sperm racetrack separates sperm according to swimming speed, and this sorting of sperm is faithfully maintained after extraction from the device as fast- and slow-swimming groups. The sperm racetrack may prove useful in understanding the genetic and physiological causes of sperm immotility, and in identifying swimming-based segregation distorters in humans. The sperm racetrack may also serve as a tool for enriching fast-swimming sperm for selection in assisted reproductive technologies such as in vitro fertilization.