Strategies to control cellular spatial organization in microphysiological systems

Abstract

Spatial organization is fundamental to tissue physiology, as it governs how cells migrate, grow, differentiate, and interact within their native environments. In living tissues, cells are positioned within finely tuned microarchitectures defined by chemical gradients, boundaries, and mechanical cues – features that are essential for proper tissue function and homeostasis. Microphysiological systems (MPSs) aim to replicate key aspects of human tissue in vitro, yet without appropriate spatial control, they often fail to reproduce certain aspects of tissue-level organization and function. In this review, we categorize spatial patterning strategies into two main approaches: direct methods, which involve the physical placement of cells or compartments using techniques such as 3D bioprinting, microfluidic compartmentalization, and physical trapping; and indirect methods, which rely on cellular responses to engineered environmental cues, including extracellular matrix (ECM) composition, mechanical gradients, and soluble factor distributions. While direct methods offer precision and reproducibility, indirect strategies more closely reflect natural developmental and self-organizing processes. We discuss how these approaches are applied across diverse biological structures, from cellular interfaces and barrier tissues to dynamic host–microbe systems. Enhancing spatial fidelity in MPSs is essential for recapitulating tissue complexity, and will be key to advancing disease modeling, developmental biology, and drug screening applications.

Introduction

In the human body, cellular spatial organization is not random but highly orchestrated. Cells reside in distinct tissue niches, where their position relative to chemical gradients, mechanical boundaries, and neighboring cells critically influences their behavior and function. These spatial cues, ranging from extracellular matrix (ECM) properties to dynamic physical forces, provide positional information that guides developmental patterning, maintains tissue homeostasis, and enables region-specific responses to physiological stimuli^1–3. Without these positional inputs, cells lose orientation, fail to integrate properly into tissues, or adopt non-functional phenotypes^4,5. Consequently, reproducing this spatial arrangement is essential for simulating organ-level functions, guiding cell-cell communication, and maintaining tissue-specific phenotypes over time.

In recent years, microphysiological systems (MPSs) have emerged as powerful platforms that aim to replicate key structural and functional features of human tissues in vitro. These systems integrate cells with microenvironments that can emulate organ-level physiology with greater fidelity than conventional culture models. By leveraging different microfabrication techniques, such as soft lithography, and perfusable microfluidics, MPSs enable precise control over environmental parameters such as flow, mechanical strain, and chemical composition of the embedding or coating matrices. Their ability to recapitulate functional tissue units in a controlled setting makes them valuable tools for investigating biological mechanisms, evaluating drug toxicity, and modeling diseases⁶. However, many current systems focus primarily on biochemical and mechanical mimicry while underutilizing the role of spatial organization in replicating physiological complexity.

Incorporating spatial patterning into MPSs is not just an enhancement – it is foundational to capturing in vivo-like behavior. The spatial layout of cells within MPSs determines how they interact, how gradients are established and interpreted, and how function emerges at the tissue level. Such spatial context can affect differentiation, polarization, and intercellular communication, directly impacting the functional output of the system. Without spatial fidelity, even otherwise well-designed MPSs may poorly reflect in vivo physiology, limiting their predictive value for basic research and translational applications.

In this review, we first classify applications of spatial patterning across a spectrum of biological complexity, including single-cell-type systems, complex tissues, multi-organ models, and host–microbe interactions, each presenting distinct requirements for spatial control. We then broadly divide current strategies for achieving spatial organization in MPSs into two main categories: direct and indirect methods (Fig. 1). Direct methods involve the deliberate placement of cells or structural components into specific locations during fabrication or seeding, which includes compartmentalization, bioprinting, and physical trapping. Indirect methods, on the other hand, manipulate the chemical or physical characteristics of the surrounding environment, such as ECM composition, substrate stiffness, or soluble gradients, to influence how cells self-organize over time. The patterning methods are often complementary, and selecting the appropriate strategy depends on the biological question, cell type, and desired level of architectural control.

Fig. 1. Summary of strategies to establish cellular spatial patterns in microphysiological systems.

Cell spatial organization can be established indirectly via controlling the culture environment, such as the chemical/physical properties of extracellular matrices and the gradient of soluble molecules, or directly by compartmentalizing microfluidic chambers, bioprinting or applying physical trapping methods. [The schematic was created using BioRender.com]

This review focuses on studies published from 2020 to 2025, ensuring the recent advancements of cellular spatial organization of microphysiological systems. However, for certain seminal studies, we chose to include them despite being published prior to 2020. The search employed a combination of keywords reflecting the core themes of this review, including “cellular spatial organization”, “microphysiological system”, “microfluidic chip”, “3D bioprinting” and “physical trapping”. Studies not directly related to the microphysiological system and spatial organization were excluded to maintain focus and relevance.

Types of biological systems in microphysiological platforms

Single cell-type systems

In native tissues, cells do not operate within uniform environments - they reside in highly structured, spatially compartmentalized niches where local variations in extracellular matrix (ECM) stiffness, geometry, and topography govern their fate, orientation, and function. In simple single-cell-type systems, where intercellular diversity is minimized, spatial patterning becomes even more essential - not only to induce realistic cell behavior, but also to simulate zonation, symmetry breaking, or boundary formation that would otherwise require multiple cell types or soluble gradients. Without these spatial inputs, cells often adopt randomized, non-functional phenotypes that fail to reflect in vivo organization or function.

Advances in micro- and nano-fabrication technologies, such as photolithography, 3D printing, and soft lithography, have enabled the creation of substrates and scaffolds with precisely defined spatial features, including grooves, ridges, stiffness gradients, and confinement patterns. These engineered environments have been shown to direct cell morphology, polarity, migration, proliferation, and even gene expression. For instance, photolithographically patterned substrates have been used to align cardiomyocytes, neural and muscle cells in vitro, recapitulating the anisotropic contractile behavior of native tissue^7–9. Spatial alignment of cardiomyocytes is crucial for synchronized electrical and mechanical activity. A notable example involves microfabricated PDMS grooves up to 350 μm in width, mimicking native trabeculae carneae. Both H9c2/Nor10 co-cultures and neonatal cardiomyocytes showed improved orientation and more coordinated beating behavior in these aligned 3D constructs, demonstrating the platform’s value for cardiac tissue modeling and drug screening¹⁰. Topographical control through micro- and nano-patterning also plays a pivotal role in single-cell-type systems. For instance, ridged substrates have been used to guide axon alignment in dorsal root ganglion neurons, mimicking neural tract organization, while NIH3T3 fibroblasts demonstrate directional migration on topographically graded surfaces^11–13. These features not only help recapitulate native morphology but also enable controlled investigation of mechanotransduction pathways, such as YAP/TAZ activation, focal adhesion formation, and cytoskeletal reorganization. Further supporting this approach, biomimetic hydrogels with spatially defined mechanical properties (e.g., stiffness gradients or viscoelastic patterns) have been engineered to study durotaxis, regional differentiation, and nuclear mechanosensing^14–16. In such systems, the spatial organization of cells can be precisely directed to induce functional compartmentalization and region-specific behavior within an otherwise homogeneous cell population.

In recent years, iPSC-derived single cells have become a dominant cell source for such bio-microsystems due to their scalability, differentiation versatility, and patient specificity. When combined with spatially engineered platforms, these cells can form isogenic tissues that replicate both genetic identity and spatial architecture, offering powerful tools for disease modeling, developmental studies, and regenerative applications. For example, iPSC-derived hepatocytes cultured under oxygen and nutrient gradients have been used to model hepatic metabolic zonation, a key feature of liver physiology absent in conventional homogeneous systems¹⁷. In neurodevelopmental studies, iPSC-derived cortical neurons integrated into compartmentalized microfluidic chips have enabled the investigation of axonal pathfinding and synaptic formation in a spatially controlled context^18,19. Similarly, iPSC-derived cardiomyocytes aligned on microgrooved substrates within ridged spaces of 10-20 µm and subjected to cyclic stretch of ~5-10% at 1 Hz which closely matches physiological myocardial conditions to accurately model arrhythmic disorders and evaluate drug-induced cardiotoxicity²⁰. In kidney models, iPSC-derived single cell kidney aggregates cultured under controlled fluidic shear stress of 0.1–0.2 dyn/cm² within 100 µm channels in microfluidic devices have demonstrated enhanced vascularization and epithelial maturation. Notably, these conditions also closely match physiological parameters of proximal tubules (0.05-0.2 dyn/cm² and ~100 µm for bundled tubules), enabling more physiologically relevant platforms for modeling renal development and assessing drug-induced nephrotoxicity²¹. While, in vascular applications, iPSC-derived endothelial cells seeded in PDMS-based angiogenesis chips and exposed to VEGF gradients have enabled controlled modeling of angiogenic sprouting, tip–stalk cell dynamics, and endothelial barrier function²². These advances collectively demonstrate the utility of integrating iPSC-derived cells with spatially engineered platforms to recapitulate organ-level microphysiology and provide a robust foundation for precision medicine, drug testing, and tissue regeneration.

Arguably, iPSC MPS’s should not be considered as purely single-cell systems. While they can be seeded as individual cells, they retain a degree of potency that allows them to differentiate into multiple cell types. Even when cultured within spatially engineered platforms, iPSC cells inherently respond to exogenous cues (chemical gradients, mechanical forces or spatial confinement) by differentiating to other cell types²³. This begs the question, would studying single cell types in isolation make sense? Since, in vivo, even isolated elicitation at the single cell level will naturally trigger cell responses beyond one cell type or state^24,25. Consequentially, observations made from isolated single cell systems are difficult to translate into physiological contexts. Therefore, one should promote utilizing multicellular models from the onset which instantly improves the physiological relevance of experimental studies.

Complex tissue systems

Complex tissue systems are defined as multicellular constructs within a single organ or tissue that integrate multiple functionally interdependent cell types, organized in spatially defined architectures to replicate native physiology. Unlike multi-organ systems, which interconnect distinct tissues, complex tissue models focus on the hierarchical organization and crosstalk of heterotypic cells (e.g., neurons-glia in the brain, endothelial-smooth muscle cells in vasculature) critical for tissue-specific functions. These systems demand precise control over cellular spatial organization to emulate structural polarity, mechanical interactions, and biochemical signaling gradients inherent to living tissues.

The principal advantage of using MPSs to model complex tissue systems lies in their ability to replicate organ-specific functions at the microscale level where critical physiological processes inherently occur. While bypassing the complexity of modeling entire organs²⁶. At microscales, MPSs enable precise control of cell-cell interactions between different cell types, reduce resource demands, and simplify the study of mechanisms that are often obscured in larger, less manipulable systems^27–30. MPSs are excellent tools that incorporate multi-cell complex tissue systems to study functional aspects at an appropriate length scale.

The simplest renditions of spatial organization in complex tissue MPSs are spheroids^31,32. Spheroids are widely used 3D culture MPSs for studying interactions between multiple cell types. These are typically differentiated cells derived from patient sources that self-assemble into spheroid structures through intrinsic cell-cell adhesions³³. Enabling multiple cell types within a single aggregate. To promote cell-cell adhesion to form these aggregates, non-adherent cell culture methods, such as hanging drop, low-adhesion plates, or magnetic levitation are employed^34–36. The emphasis of cell-cell adhesions drives self-organized equilibration in multicellular spheroids by balancing receptor-mediated interactions and physical constricts (inability to adhere to surfaces). Spheroid MPSs tend to be small in size (<500 cells), yet even at these dimensions, they exhibit extensive cell-cell interactions and rapidly establish their own spatial organization^37–39.

Spheroid MPSs are formed by self-organizing through the intrinsic adhesive properties of the cells-mediated by molecules like cadherins, which tighten cell aggregates, and integrins, which connect cells to the extracellular matrix and maintain structural cohesion as described in cell adhesion theory^40,41. By this concept, the unique adhesive potential of each cell type dictates the resultant spatial architecture, governed by homotypic (same cell type) or heterotypic (different cell type) adhesive interactions. Single cell spheroids typically form a core-shell structure where cells with stronger adhesion cluster at the core and weaker-adhesive cells are displaced to the periphery based on the stochasticity of adhesion potential in one population⁴⁰. In contrast, multiple cell type spheroids generate complex spatial organization because cells must balance both homotypic and heterotypic adhesion strengths⁴². For example, in spheroids containing breast cancer cells and stromal cells like fibroblast and endothelial cells, the cancer cells which typically exhibit stronger homotypic adhesion, cluster into the core. Meanwhile, the fibroblasts and endothelial cells, with weaker and variable heterotypic adhesion to cancer cells, are displaced to form layered structures at the periphery^43,44. Endothelial cells can further exploit their relative adhesive properties into capillary-like networks that connect the spheroid core to the outer layers^44–48. These spheroid MPSs typically measure 200 micrometers in radius and develop central regions of hypoxia and necrosis, mirroring phenomena observed in solid tumors^49,50. The resulting gradients in oxygen, nutrients, and metabolic waste create distinct microenvironments, making multicell spheroids highly relevant for modeling tumor biology, especially for investigating drug penetration, resistance, and the influence of microenvironmental constraints such as oxygen tension and nutrient gradients.

However, the intrinsic tendency of non-potent differentiated cells to self-organise into generic spherical aggregates with core-shell features highlights the limitations of utilizing differentiated cells to create MPS’s with more complex spatial architecture^44,51,52. Differentiated cells express a specific set of adhesion molecules (such as certain cadherins and integrins), which means that while their interactions are largely predetermined and stable, cell-cell and cell-ECM interactions are fixed, limiting their capacity to create higher order structures⁵³. In contrast, potent cells like stem cells and iPSCs possess the capacity to differentiate to multiple cell types, each potentially expressing new or different types of adhesion molecules^54–56. This plasticity allows them to integrate new cell-cell and cell-ECM interactions, expanding the range of structural arrangements and responses to a given signal⁵⁷, much like in embryogenesis, where highly potent cells can generate remarkably complex structures⁵⁸.

The use of potent cells in multicellular MPS cultures is termed Organoid MPS’s. Organoid MPS’s exemplify how cellular potency and dynamic cell–cell and cell–ECM interactions drive complex tissue architecture^52,58–60. The first organoids demonstrated that intrinsic cell potency amidst simple cell culture conditions were able to recapitulate the intricate spatial organization and functional diversity⁵⁹. Mammary epithelial cell aggregates incorporating both differentiated cells and stem cell subsets could self-organize into lumen spherical structures that generate lactating mammary acini or undergo branching morphogenesis^61,62. Subsequent studies by Eiraku et al. and Sato et al. demonstrated that stem cells and their progeny, when directed following specific lineages, can follow innate developmental programs and self-organize into structures that mimic histological and functional aspects of real tissues in organs^63,64. For instance, intestinal organoids can form crypt-like projections radiating from a central lumen, closely mimicking the architecture of the adult small intestinal mucosa^64,65. These structures reconstitute the principal geometrical, architectural, and cellular hallmarks of the native epithelium and serve as powerful models for studying adult stem cell biology, tissue homeostasis, and regeneration⁶⁶.

Despite these advances, the full reliance on autonomous self-organization of cells in organoid MPSs has not been shown to consistently form distinct layers or compartments⁶⁷. Cells often become intermixed, and vascular models frequently fail to establish stable, functional blood vessel networks^48,68,69. The absence of pre-defined extrinsic patterning instructions leads to stochasticity in organoid formation, resulting in heterogeneity in size, shape, and cell-type composition. For example, intestinal organoids often display fluctuating ratios of different cell types, even under identical conditions, while liver organoids may struggle to expand or maintain function over time^59,70,71. This lack of proper architecture remains a major barrier for modeling organs that depend on spatial cues and compartmentalization, such as the intestine, retina, or pancreas⁷², ultimately, diminishing the physiological proximity of organoid MPSs to tissue.

Tissues are seemingly cells organized within boundaries, incorporating boundaries in vitro could guide even stochastic stem cell patterns. Advanced fabrication techniques, including microchannel patterning and 3D bioprinting, enable the precise placement of multiple cell types within restricted geometries that mirror native tissue organization^29,30,73. Microchannel-based systems are particularly powerful in this context. By manipulating channel layout and connectivity, researchers can control the spatial positioning and contact between different cell types, thus recreating physiologically relevant interfaces and gradients. An early example of this approach is the recreation of the small intestinal epithelium, where microfabrication and molding techniques were used to generate arrays of crypts and villi, mimicking the natural topography and spatial arrangement of stem cells and differentiated cells⁷⁴. Another pioneering example is the biomimetic lung-on-chip microsystem, in which alveolar cells and endothelial cells are restricted in separate microchannels and divided by a membrane⁷⁵. This membrane mimics the spatial interface of the alveolar-capillary interface. As seen, microchannel confinement is particularly valuable, as it allows for the maintenance of well-defined microenvironments, even when using primary cells such as patient-derived samples that may not self-organize on their own. This enables the creation of spatially organized constructs without relying solely on multipotent or pluripotent stem cells. This trait appears even more significant considering the 2022 FDA Modernization Act, which permits alternatives to animal testing in drug screening⁷⁶. Wherein the use of MPS microchannel systems, which enable the use of patient-derived samples to recapitulate spatial interactions at the patient’s tissue level, could potentially be used for personalized drug screening.

Equally important is the ability to replicate tissue spatial organization and function, regardless of whether patient-derived samples are used. MPS microchannel platforms are especially advantageous in their modular design, which facilitates the integration of physical, mechanical, and biochemical characteristics such as matrix micro/nanostructures, stiffness, fluidic forces, mechanical stretch, cytokines, and growth factors^26,28. These parameters are known to profoundly influence tissue architecture, function, and disease pathology by guiding how cells organize and interact within three-dimensional space. For example, matrix stiffness can direct cell alignment and drive tissue morphogenesis, while localized gradients of growth factors can induce processes such as endothelial sprouting or immune cell infiltration, further refining the spatial complexity and functional relevance of engineered tissues^77–82. Central to this modularity includes the integration of dynamic conditions, such as fluid flow, which not only supports nutrient and oxygen delivery but also imposes mechanical cues that shape cellular arrangement, and hence tissue-level organization^83–85. This is termed microfluidic MPS. These systems include liver chips, where the positioning of hepatocytes and non-parenchymal cells under continuous flow establishes oxygen and nutrient gradients, resulting in zonate structures akin to the liver acinus^86–89. Importantly, microfluidics, as the name suggests, allows researchers to carefully tune flow rates of nutrients to mimic physiological shear stress (typically 0.1–1 dyn/cm²), which regulates hepatocyte function and metabolism, making the model highly relevant for studying liver physiology and drug metabolism. For example, heart microfluidic chips use aligned cardiomyocytes on flexible substrates or within microchannels, with mechanical stretch and fluid flow to promote the maturation and synchronous contraction of cardiac tissue, closely reflecting the spatial arrangement of the myocardium^90–93. Another example is blood vessel microfluidic chips with endothelial cells cultured in perfused microchannels, where shear stress (1–15 dyn/cm², depending on vessel type), induces cell alignment and tight junction formation, modeling functional vascular barriers essential for studying permeability and angiogenesis^94–97. Kidney microfluidic chips recreate the nephron’s filtration barrier by co-culturing podocytes and tubular epithelial cells on opposite sides of a porous membrane, with flow supporting cell polarization and the spatial organization necessary for simulating filtration and reabsorption^98–100.

Alternative methods to incorporate boundary conditions include 3D bioprinting that prioritizes direct cell-cell contact over artificial physical barriers. 3D bioprinting generates spatially defined cell cultures via additive techniques^101–104. Cells are embedded into biocompatible ECM gels or bioinks and printed. Gels stack additively without the presence of physical channels or fences, generating integrated tissue interfaces seen physiologically. Continued innovations in extrusion-based bioprinting, such as reduced shear stress, optimized thermal conditions, and biocompatible crosslinking methods, have significantly improved cell viability and precision^105–108. Biocompatible GelMA hydrogels have been used to print human aortic endothelial cells stacked over smooth muscle cell (SMC) layers, enhancing tissue interface adhesion¹⁰⁹, which promoted αSMA and SM22 expression, markers of a functional SMC phenotype, a trait typically observed between directly contacting endothelial-muscle cell layers. Similarly, liver-on-chip platforms fabricated via bioprinting demonstrate how layered hepatocyte, stellate cell, and endothelial cell arrangements replicated the spatial organization of liver lobules including pathological features such as collagen accumulation associated with liver fibrosis^110,111. Together, these advances highlight the emerging role of bioprinting not only as a fabrication tool but as an enabler of spatially organized, functionally integrated microenvironments essential for recapitulating complex tissue physiology in vitro.

Barrier models and multi-organ systems

To accurately capture the functions of biological systems, it is intuitive to incorporate multiple organs or tissues into MPSs models. On a relatively small scale, this can be a barrier between two different systems. Here, we group epithelial–endothelial systems, or any vascularized barrier systems, with multi-organ, as one belongs to the internal lining of an organ, whereas the other belongs to the circulatory system. On a larger scale, multi-organ systems typically comprise joint compartments of various tissues, connected via a closed-loop media perfusion or a single fluidic network adjacent to the culture chambers, reassembling vascular media.

Microfluidic MPSs are uniquely suited for modeling biological barriers profoundly because of the spatial organization that is fundamental to barrier tissue function and pathology^112–114. The spatial compartmentalization of endothelial, epithelial layers, and stromal compartments is fundamental for generating selective permeability as seen in barrier tissues^115,116. Cell layers effectively compartmentalize and control molecular and cellular traffic, supporting vital functions like gas exchange, nutrient absorption, and immune surveillance. Barrier tissues, such as the intestinal lining, blood-brain barrier (BBB), and alveolar-capillary interface, depend on the precise spatial organization of cells, intercellular junctions, and extracellular matrix to maintain selective permeability^115,116. The first organ-on-chip system, the lung–capillary chip model developed by Huh et al.⁷⁵, represents a foundational framework for reconstructing actuated barrier tissues in vitro. Many subsequent barrier models have mimicked or drawn inspiration from Huh’s design. This design ensures that nutrients, ions, and immune cells traverse the barrier in a controlled manner. When this organization breaks down due to disrupted cell-cell contacts, loss of cellular polarity, or mechanical stress, pathological conditions emerge^117–120. For example, inflammatory bowel disease is characterized by “leaky gut,” in which the same characteristics of misplaced tight junction proteins and disorganized intestinal epithelia, which allow harmful substances to infiltrate underlying tissues, were observed in microfluidic intestinal chips^121–124. Similarly, neurodegenerative diseases like Alzheimer’s involve compromised BBB integrity, enabling neurotoxic compounds to enter the brain. In BBB-on-chip systems, disruption of endothelial cell alignment and junctions, which can occur in neurodegenerative diseases like Alzheimer’s, enables neurotoxic compounds to cross into the brain^125–127. These examples demonstrate that faithfully recapitulating the spatial organization of cells, junctions, and extracellular matrix within microfluidic barrier models is fundamental in capturing normal barrier function as well as the pathological consequences of its disruption.

In microfluidic barrier systems, spatial patterning is straightforward - distinct cell types are cultured on opposite sides of a membrane, effectively mimicking biological barriers. However, in joint-organ systems, spatial organization is less apparent. Simply linking compartments does not constitute true spatial design. Instead, spatial relevance arises from the spatial arrangement of culture compartments, or sequence of perfusion, which replicates in vivo solute transport or metabolism. Controlling such an arrangement is valuable for pharmacokinetic/pharmacodynamic (PK/PD) studies, in which the sequence of drug metabolism is captured by the sequence of perfusion. Theobald et al. developed a two-organ model comprising liver and kidney chambers, respectively, from the inlet to outlet direction¹²⁸. This arrangement helped simulate the metabolic activation of vitamin D, in which the liver metabolized vitamin D3 to 25(OH)D3, which was then further metabolized by the kidney chamber to 1,25(OH)2D3. Similarly, Wang et al. studied the absorption and metabolism of midazolam, an oral drug, in a gut–liver-on-chip system¹²⁹. The two channels, gut and liver, were perfused in the aforementioned order, mimicking the absorption and metabolism order. They found that the metabolism was inhibited if ketoconazole was co-administered at the inlet, providing potential pharmacokinetic applications. These demonstrate the potential of using multi-organ models in studying substance absorption and metabolism by arranging the culture compartments in physiologically relevant orders.

Adding more organs to MPSs not only increases the overall complexity of the platform but also significantly raises the labor intensity involved in system preparation and maintenance. Each organ model typically requires distinct cell types with unique incubation times for attachment, differentiation, and maturation before the chips are ready for experiments. When working with cell lines or primary cells, these can often be expanded and seeded into the chips under relatively straightforward and established conditions. However, starting from pluripotent stem cells such as iPSCs introduces an additional layer of complexity, as the differentiation process into specific organ-relevant lineages can take several weeks and often demands precise timing and environmental control. Furthermore, the varied maturation timelines and in-chip incubation requirements across different cell types make the overall workflow difficult to coordinate. This level of scheduling and handling not only makes the process tedious but also increases the likelihood of mistakes, especially when multiple organ models must be integrated within a single system and maintained in parallel.

Physical factors such as flow rate and shear stress play a pivotal role in multi-organ and barrier MPSs, not only supporting individual tissue function but also enabling physiologically relevant inter-organ communication. For instance, in a gut–liver-axis microphysiological system, intestinal epithelial cells experience fluid shear stress ranging from ~8 × 10⁻³ to 2 × 10⁻² dyn/cm² under flow rates of 2 to 5 μL/min, which promotes barrier integrity and differentiation, while hepatic cells are protected from high shear, experiencing much lower stresses around 1.2 × 10⁻⁷ dyn/cm². Such precise modulation of flow and shear across different organ compartments ensures that unique mechanical microenvironments are preserved, allowing sequential metabolic and signaling events to unfold as they would in vivo. This careful orchestration underpins the ability of multi-organ MPSs to faithfully recapitulate complex organ interactions and systemic physiology.

To date, the most complex microfluidic multi-organ MPSs usually include three to four organs. Certain systems were developed with unique arrangements of compartments to mimic physiological functions. For example, the pancreas–muscle–liver was constructed by Lee et al. to simulate glucose homeostasis¹³⁰. The parallel channels in the chip design mirrored the closed-loop glucose metabolism, with adjusted scaling of each channel to reflect the physiological level of glucose and insulin. Another example is the four-organ-chip (intestine, liver, skin, kidney) developed by Maschmeyer et al., which had the kidney compartment in a separate perfusion loop to demonstrate the excretory circuit¹³¹. By using this, the group was able to assess the absorption of glucose by the renal proximal tubule by comparing glucose levels in different perfusion circuits, while also assessing the function of other organs in the system. Overall, expanding the number of compartments enhances the ability to mimic complex inter-organ interactions, making MPS data more reflective of in vivo conditions. However, increased complexity requires careful compartment design and precise tuning of physical conditions, including flow and shear stress, to maintain tissue-specific environments and accurate organ crosstalk.

Host-microbe systems

Microbes-on-a-chip systems offer a powerful platform for studying microbe-host interactions under physiologically relevant and highly controlled conditions. By integrating living microbial communities with human tissue models on microfluidic chips, researchers can closely examine how microbes colonize host surfaces^132,133, trigger immune responses¹³⁴, and influence tissue function^135–137. These platforms allow precise control over factors like flow dynamics, nutrient availability, and oxygen levels - key elements that shape microbial behavior and host responses in vivo. Importantly, microbe-on-a-chip systems enable real-time monitoring of interactions such as microbial adhesion, invasion, inflammation, and barrier disruption, which are critical for understanding infections and chronic diseases linked to the microbiome.

It is important to recognize that the microbiome is unevenly distributed across body systems. In the gut, different microbial species have been shown to segregate and occupy distinct ecological niches. The niche differentiation happened both in along-intestinal-tract direction^138,139, and along-villus-crypt axis^140,141. Skin bacteriome also shows differential composition at different skin depths from epidermal surface to within the epidermis¹⁴². Recapitulating such “spatial footprint” is vital to study bacterial growths and ecological relations, which eventually affect the microbe-host interactions. For example, Bacillus subtilis and Staphylococcus epidermidis have been shown to exhibit niche competition on skin¹⁴³. B.subtilis would produce antimicrobial substances and migrate toward S.epidermidis to engulf them. In MPSs, such responses would change the initial spatial distribution of two bacterial populations if they were introduced in proximity to each other. In the intestine, the 10 µm scale was proposed as the distance required for local competition to happen¹⁴⁴. Similarly, gut pathogens generally need to get in proximity to the epithelial lining to cause disease, termed contact-dependent disease induction. The goblet cells in the intestine produce a thick mucus layer to prevent such contact. Moreover, there exists a microbiome layer between the newly introduced pathogen and the host. Therefore, the interplay between epithelium-bacteria spatial distance, as well as bacteria niche partition, should be considered in microphysiological systems.

Bacteria consist of motile and non-motile groups. Once co-introduced into MPSs, bacteria will exhibit ecological interactions, regardless of their motility. If mutualism happens, they can coexist in the same space, on the other hand, in the case of competition, whether one species ceases to survive, or niche differentiation occurs. Therefore, spatial patterning can form without any external causes but only inter-species interactions. Moreover, the gradient of nutrients and O₂ can facilitate the spatial distribution of bacteria, even for non-motile species, as the lack of nutrients or the O₂ level outside of the toleration range can lead to the eradication of bacteria in a specific region, and vice versa, a high amount of nutrients and a correct O₂ level assisting bacteria’s growth. Zheng et al. fabricated a PEGDA-based hydrogel with a gradient of nutrients and O₂ across different layers and encapsulated gut microbiome¹⁴⁵. As expected, bacteria formed a distinct heterogeneous spatial distribution in different layers, with the lowest layer supporting the growth of anaerobic bacteria such as Akkermansia and Parabacteroides.

For motile bacteria, the taxis phenomena can be exploited to establish movements, and eventually spatial pattern. For example, Escherichia coli has been shown to exhibit chemotaxis toward NCI-H460 lung carcinoma cells in a tri-channel microfluidic system¹⁴⁶. From the study, clusterin was proposed as the target of chemotaxis, explaining the tumor-targeting response of E.coli in lung cancer patients. However, for the non-motile group, dynamic forces may be necessary to facilitate spatial niche formation. Lee et al. designed a two-channel microfluidic gut-on-a-chip system, in which the second layer functioned a pressure reservoir that oscillated periodically mimicking the peristalsis motion¹³². The introduced bacteria community, E.coli and Lactobacillus rhamnosus GG (LGG), changed their dynamic spatial distribution under the influence of peristalsis motion. Without the flow oscillation, LGG, a non-motile species, since to the crypt-like regions of the gut epithelium, where it had to compete with E.coli. The introduction of peristalsis helped disperse LGG to the tip of villus-like structures, where the lack of competition assisted the growth of LGG.

External physical factors such as fluid flow, shear stress, and mechanical dynamics also strongly influence bacterial spatial patterns within these systems. For example, motile bacteria like E.coli respond to shear gradients by localizing to low shear regions^147,148, while non-motile bacteria depend on dynamic flows such as peristalsis-like motions to disperse and occupy spatial niches. These forces modulate bacterial aggregation, adhesion, and community composition, promoting coexistence and ecological stability through spatial niche partitioning.

Careful consideration of the bacterial community introduced into MPSs is essential. Using a patient-derived microbiome is often preferred, as it best replicates the complexity and ecological interactions of the native human microbiota. However, this approach presents challenges, particularly in controlling microbial composition and spatial distribution, due to high variability between individuals. Alternatively, introducing defined bacterial strains offers greater experimental control and reproducibility, but this comes at the cost of physiological relevance. Simplified microbial communities lack the rich interspecies interactions of natural ecosystems, which not only affects spatial colonization patterns but also increases the risk of uncontrolled overgrowth, even when using theoretically benign or probiotic strains. Such overgrowth can quickly disrupt tissue integrity and function, which is one reason most MPS platforms with bacterial components are limited to short-term experiments, often under 24 h, before microbial proliferation compromises the system.

Overall, the spatial distribution of bacterial communities is highly dynamic, largely due to their rapid generation time, high motility, and sensitivity to environmental cues. These characteristics allow bacteria to quickly respond to and adapt within their microenvironments, leading to heterogeneous and evolving patterns across different tissue compartments. To more accurately recapitulate the complex biological functions of the microbiome or pathogenic organisms, host–microbe MPSs should be designed to capture these dynamic spatial distributions. Incorporating such spatiotemporal features is essential for studying host responses, microbial colonization, and the progression of infection or symbiosis in a physiologically relevant manner.

Indirect control of spatial organization in microphysiological systems

Spatial control of chemical properties of the environment

The use of ECM components is integral to MPSs due in part to their ability to model the natural, tissue-specific microenvironment that cells experience in vivo. The innate sensitivity of cells to environmental cues enables researchers to guide spatial patterning in a non-invasive manner, leveraging natural cellular responses to control positioning, organization, and behavior without the need for complex engineering or direct manipulation at the cell level.

ECM patterning

Among the microenvironmental cues provided by the ECM, biochemical signals are especially critical. Biochemical signals govern the initial cell-ECM (integrins) adhesion through bioactive binding sites such as the RGD (arginine-glycine-aspartic acid) sequence within ECM fibers that directly interact with cell surface integrin receptors¹⁴⁹. Such interactions between cells and binding sites on ECM not only anchor cells physically to the matrix but also activate intracellular signaling pathways that regulate important cell behaviors (adhesion strength, migration, proliferation, and survival).

The prescribed method to spatially pattern cells involves first patterning ECM (Fig. 2a). Coating microchannels with thin, surface-level layers of ECM, like fibronectin, has been shown to restrict cell adhesion to predefined 2D regions. This method is especially effective in mimicking vasculature by guiding endothelial cells to form lumen-like spatially organized structures^150–153. Alternatively, full ECM gels which fill entire microchannels permit cells to infiltrate and remodel their entire volume^154–157. Integrating this within any microchannel shape forms gel-confined layers as seen earlier. By simply altering where and how ECM is presented, researchers can steer cell localization and structure formation in an indirect yet controllable manner. By this concept, the type of ECM itself can further refine cell positioning by exploiting cell-type-specific adhesion preferences. Epithelial cells will typically remain confined within laminin-rich zones due to their strong adhesion to laminin, while avoiding collagen-rich areas where adhesion is suboptimal^158–160. Conversely, fibroblasts prefer collagen and thus localize accordingly. Finally, ECM can also be patterned on 2D surfaces with micropattern techniques, such as microcontact printing. Dickinson et al. used this technique to pattern hyaluronic acid and fibronectin on a 2D surface in a grid-like manner¹⁶¹. Their approach can be applied to study interactions at spatially defined distributions, such as between endothelial colony forming cells and breast cancer cells.

Fig. 2. Indirect control of ECM properties enhances the complexity of spatial organization in MPSs.

a (i) Basic spatial organization of tissue such as intestinal lumen-like structures or multi-layers of the intestine can be achieved by directly patterning ECM into preformed structures. (ii) Increased spatial definition is obtained by tailoring ECM composition and presenting specific binding motifs that favor cell-type-specific adhesion. (iii) Increased spatial complexity can also be achieved through indirect control of cell deposition using enzyme-responsive hydrogels. (iv) Direct patterning of bioactive ligands offers high-resolution spatial control. Further fine-tuning of these immobilized motifs could also guide cell differentiation to form more spatially and functionally complex tissue architecture. b Effect of matrix degradability on cell growth and movement – fully degradable GPQ-A gel maximized spheroid protrusion¹⁸³. c Patterned conjugation of ligands led to 2D pattern of crypt-like regions, where lysozyme expression matched Wnt3a micropattern²⁰⁶. All reused figures are under CC-BY-NC-ND 4.0 license

Despite these successes, a limiting factor to these approaches is in reproducing ECM-based patterns across laboratories. Natural matrices (e.g. Matrigel, collagen, fibrin) exhibit batch-to-batch variation in protein concentration, growth-factor content, stiffness, and fiber architecture^162–164. Small deviations in coating concentration, adsorption time, crosslinking chemistry, and humidity between protocols can also influence cell adhesion and migration within the ECM^77,165,166, which alters the spatial organization of MPS’s between laboratories much less within the experiment itself.

To address this, publications have increasingly emphasized the characterization of ECM chemistry and the mechanical benchmarking of materials. These efforts in turn, have motivated a move toward ECM tuning with a precise definition of bioactive motifs at controllable densities on otherwise inert ECM backbones. Reducing the reliance on widely variable ECM cocktails (Fig. 2a). In this framework, researchers employ natural or synthetic scaffolds and hydrogels with precisely engineered ECM compositions and tunable adhesion sites¹⁴⁹, which offers a more transferable, reproducible, and standardized way to refine cell-ECM interactions at the binding site level, where intrinsic cellular preferences for bioactive motifs direct haptotaxis. A phenomenon in which cells migrate toward binding motifs in which they possess a higher adhesion potential to whilst avoiding motifs it has a weaker affinity for^167,168. A prime example of this approach is peptide-driven self-assembly, where scaffolds are functionalized with short, cell-selective peptides (e.g., RGD peptides for integrin binding or IKVAV peptides for neuronal adhesion)¹⁴⁹. As a result, spatially patterned ECM with bioactive binding motifs induces spontaneous cell segregation and organization, generating more well-defined domains within engineered tissues. Altogether, these methods indirectly pattern cells spatially using the directed control of ECM and adhesive motifs.

Following this idea, the additive nature of 3D bioprinting, further expands the toolkit for ECM-based spatial patterning. As mentioned earlier, 3D bioprinting enables the precise deposition of multiple cell types, ECM components, and bioactive factors in patterns, layer by layer over one another. While omitting the use of physical boundaries. In this manner, researchers can match the heterogeneity and architecture of tissues inherently without being restrained by the length scale or micropattern engineering limits. 3D bioprinting of patterns thus generates MPSs that more closely mimic the in vivo context, with an emphasis on cell–cell interactions^169,170.

Dynamic matrices

It is crucial to recognize that within the first 48 to 72 h of culture, cells would actively remodel their immediate surroundings by depositing a pericellular matrix with autocrine or paracrine factors, including matrix metalloproteinases^171–173. This pericellular coating around the cell can override the initially presented cues by altering the bulk properties of the bulk ECM. Fibroblasts, endothelial cells, and epithelial cells, as widely studied MPS cell components, readily assemble fibronectin and collagen networks, recruit TGF-β, and generate proteolytic gradients that collectively bias haptotaxis and durotaxis^173–176. This directly influences the working principle of the indirect control of cell haptotactic migration to spatially organise. However, rather than treating this as a nuisance, researchers have harnessed such cell behavior to indirectly control spatial organization within MPS’s. Coined, Smart Hydrogels, these platforms are designed to be permissive to cell-directed remodeling, translating endogenous secretion to guide cell migration pattern formation^177–179. Unlike traditional scaffolds, smart hydrogels are designed to respond to specific stimuli-such as enzymes, pH, temperature, or light - enabling them to dynamically alter their biochemical properties in situ^180–182. This responsiveness allows for the creation of spatially organized tertiary tissue structures not by direct cellular self-assembly, as seen in organoid formation, but by providing dynamic environmental cues that cells sense and respond to as they migrate into and interact with the hydrogel matrix. As cells embedded in ECM matrices secrete matrix metalloproteinases (MMPs), ECM degrades, actively changing the matrix properties. Thai et al. incorporated this effect of ECM degradability to facilitate stronger cell protrusions and angiogenesis within spheroids containing endothelial cells and mesenchymal stromal cells¹⁸³ (Fig. 2b). Whereas, other smart hydrogels actively release enzymes rather than degrade. technology. As cells infiltrate the hydrogel, they begin to secrete MMPs as part of their natural remodeling behavior. These MMP enzymes subsequently degrade the matrix whilecleaving MMP-cleavable linkers that release VEGF near the active cells^179,184. Established VEGF gradients then cause endothelial cells to differentiate and sprout, forming vascular networks directed towards MMP releasing cells, where the architecture of the vasculature is indirectly dictated by the location and activity of MMP-producing cells. This adaptive self-organization have also been demonstrated in other organoid models, including vascularized liver, gut, and kidney tissues, where enzyme-sensitive hydrogels promoted spatiotemporal endothelial integration and maturation^68,185–189. In a parallel strategy, human intestinal organoids (HIO) were grown in a stimuli-responsive hydrogel composed of polyethylene glycol (PEG) and functionalized with a protease-degradable peptide¹⁹⁰. In response to local cell secreted MMPs, local degradation of PEG hydrogel occurs, but instead of facilitating angiogenesis, they facilitate cellular reorganization of mature intestinal features in the HIOs, including spatially defined crypt-villus tissue architecture (lamina propria, muscularis mucosae, and submucosa) and organized collagen fibres. Harnessing the natural responses of cells to want to assimilate into their environment yet creating defined spatial regions.

Although smart hydrogels elegantly couple cellular trophic activity to localized softening and growth factor release, precise temporal dosing remains difficult to preset^191–193. Heterogeneity in protease expression across cell types and donors makes it difficult to tune cell-ECM interactions to consistently recreate spatial architecture. This is because the rate of hydrogel degradation not only depends on the designed cleavage motifs, but the cell-specific protease activity. For instance, fibroblasts, macrophages, and cancer cells express very different levels and repertoires of MMPs, which lead to different kinetics of hydrogel remodeling even under identical conditions^194,195. Likewise, donor-to-donor variability such as cells from ages or diseased tissues often secrete lower baseline levels of proteases and respond more slowly to activation signals compared to young and healthy cells¹⁹⁵. For example, smart hydrogels designed for controlled VEGF release will react differently to different proteolytic environments. Under low activity settings, the same system may fail to release enough factors within the time frame to trigger spatial reorganization of cells¹⁹⁶. Consequentially, each new cell source would demand fresh re-optimisation of ECM crosslinker and ligand density amongst other parameters^77,197. In practice this process is labor-intensive and time-sensitive, which is especially challenging when working with limited patient-derived samples that may also lose viability during lengthy preparation steps¹⁹⁸.

Ligand patterning

Interestingly, studies show that cells do not need to physically probe deeply into matrix substrates to respond to environmental cues. Rather, cells are responsive to adjacent tethers on any substrate. For example, the length and strength of tethering peptides or proteins influence how cells behave more than the bulk property of the material^77,199,200. This finding underscores that presenting cells with surface-localized cell adhesive ligands is sufficient to control both adhesion and migration. At the same time, depending on the cell adhesive ligand, ligands also possess higher-order functions such as the ability of polarization and even cellular differentiation^{149,201–203}. This way, by the innate responsiveness of cells to survey their environment via haptotaxis, such as sensing and moving toward regions with favorable biochemical signals, direct patterning of adhesion motifs is key to indirect cell patterning as well as differentiation.

Unsurprisingly, there is increasing interest directed toward strategies that allow for the immobilization of peptides or bioactive motifs on 3D and 2D substrates. In these systems, peptides are covalently attached to the substrate, providing stable, spatially defined cues as peptides do not degrade, require continual adjustment, or are subject to matrix remodeling and abrogating spatial organization. Despite this, the use cases of peptide immobilization on substrates have not ventured far beyond mechanobiology and cell migration studies. Here, immobilized gradients of chemoattractants (such as CXCL12 for guiding immune cell migration) or morphogens are widely used to study direct cell migration and fate in a controlled manner^204–207. In tandem, surface topography modifications as well as low-protein-binding materials generate nanoscale patterns, such as PEGylated regions, which restrict protein adsorption to specific zones, guiding where cells adhere and organize^208,209.

The potential applications of peptide/ligand/motif immobilization are vast. It is well documented that ligands can be attached to surfaces and micropatterning can be easily achieved using physical or photolithographic masks. In fact, immobilized ligand patterns of Wnt3a ligands on 2D substrates have successfully replicated the spatial organization of intestinal crypt-villus structures even on a 2D substrate (Fig. 2c)²⁰⁶. Ligands such as Wnt3a and EphrinB immobilized on structured substrates (e.g., nanowells) were shown to form distinct crypt-villus axes. The process of peptide functionalisation onto substrates is remarkably simple^210–213. Whereby peptides can be attached onto substrates even in solution form, with the only real limitation being that of the micropattern or the mask used. Central to this concept is the possibility to directly pattern peptides on substrates to mimic cross-sections of tissue architecture through which cells could adhere to and self-organise indirectly into Ligand MPS’s with complex spatial architectures²⁰⁹. This approach enables the immediate maintenance of cell phenotype in spatially organized regions, which is particularly advantageous for patient-derived cells that need to retain their in vivo characteristics from the outset.

After all, ligand immobilization and patterning strategies complement existing approaches to enhance the spatial organization of ligand MPS through the indirect control of cells. When combined with smart hydrogels that dynamically respond to environmental cues such as enzymes, pH, or light, these bioactive motifs can be selectively presented or concealed, providing an additional layer of spatial and temporal regulation. The ability to pattern multiple bioactive signals within complex, three-dimensional arrangements open unprecedented opportunities for recreating heterogeneous tissue interfaces and investigating intricate morphogenetic processes with high spatiotemporal fidelity. A prime example of this integrated strategy is the use of photopatterned aptamer-tethered vascular endothelial growth factor (VEGF), as demonstrated by Rana et al., where VEGF molecules are immobilized on substrates in light-defined patterns to precisely guide endothelial cells in forming luminal vascular networks with filopodia-like structures that closely mimic in vivo vasculature^214,215. Beyond static spatial control, temporal regulation is achieved through external chemical stimuli that dynamically modulate network properties such as orientation and density, offering a degree of control unattainable by traditional ECM gels or uniformly functionalized hydrogels. Altogether, these advances in ECM scaffolding, immobilized bioactive cues, smart hydrogels, and peptide patterning are transforming tissue engineering and regenerative medicine by providing powerful, versatile tools to orchestrate cellular organization and function within physiologically relevant Ligand MPS’s.

Solute gradients

Soluble molecules play crucial roles in the body by enabling a wide array of physiological processes. These molecules, which include soluble proteins, hormones, enzymes, and signaling molecules, are dissolved in bodily fluids such as blood and cytoplasm, allowing them to move freely in the body and interact with cellular components. Their solubility is fundamental for their function, as it ensures proper transport, communication between cells, and regulation of metabolic pathways^216–218. The ability of these molecules to interact with water and other polar substances is determined by their chemical structure, such as the presence of hydrophilic amino acid residues in proteins, which enables them to form hydrogen bonds and remain stable in aqueous environments. When applied to MPSs, soluble molecules influence spatial organization primarily through two mechanisms: guiding cell movements and driving cell differentiation. To establish solute gradients, microfluidic gradient generators are commonly employed. Alternative and, simpler approaches – such as distance-based gradients from a source channel or chamber – can also be effective.

Soluble molecules influence spatial patterning by activating intracellular signaling cascades that govern cell fate decisions. Gradients of morphogens – soluble factors like Wnt, BMP, or TGF-β – are well-known to induce differential gene expression depending on local concentration, thereby leading to region-specific cell differentiation. When incorporated into MPSs, these gradients can be used to recreate zonated tissues, where distinct cell types coexist in a spatially ordered arrangement. Kim et al. devised a microfluidic gradient generator to create a gradient of growth factors EGF and FGF in a neural stem cell (NSC) culture channel²¹⁹ (Fig. 3a). NSCs near the high concentration region of growth factors exhibited higher proliferation and differentiation, thus the chip generated a strip of cells varying in differentiation level. Soluble cues can also coordinate the timing of differentiation, ensuring that progenitor cells mature into appropriate lineages in the correct spatial context. Moreover, drug toxicity can be studied in spatial diffusion platforms. Griddient, a microfluidic array developed by Sanchez-de-Diego et al., allowed regional introduction of solutes and the study of their gradient-based effects²²⁰ (Fig. 3b). For example, the region of cell death caused by local introduction of puromycin, combined with simulation of drug concentration, can be used to assess dose-based toxicity of the antibiotics. Overall, by incorporating gradient designs, MPSs can capture the dynamic interplay between biochemical signals and cell behavior, enabling more predictive models of tissue development or regeneration.

Fig. 3. Solute gradient influence on cell patterns.

a Gradient of growth factor EGF generated differential cell morphology on two sides of culture channels. Neural stem cells appeared more elongated and higher degree of differentiation toward the higher end of GF spectrum²¹⁹ (CC-BY-NC 4.0 license). b Grid micropattern on Griddient generated localized regions of solutes/drugs. High-puromycin regions have more dead cells then other regions²²⁰ (CC-BY 4.0 license). c Solute gradient was generated across a hydrogel barrier. The platform was used to test migration of MCF7 and MDA-MD-231 breast cancer cells towards FBS²²¹ (CC-BY 4.0 license)

In addition, solutes also guide cell movement through the process of chemotaxis, where cells detect and move along concentration gradients of specific molecules. This directed migration plays a critical role in establishing spatial organization within tissues, as different cell types localize to regions with optimal concentrations of chemotactic factors. In the context of MPSs, microfluidic systems can precisely control these gradients to mimic physiological cues found in vivo. Samandari et al. utilized a source channel of FBS and gel-based separation to induce a gradient in a cell culture chamber²²¹ (Fig. 3c). Using the system, the group compared the invasion of MDF7 and MDA-MD-231 cancer cell lines during their chemotactic movement. Sardarabadi et al., utilizing a unique gradient generator design, studied the migration and activation of T cells across varying concentrations of IL6²²². Stromal fibroblast proliferation being influenced by IL6 concentration and T cells, generated parallel spatially distinct immune-inactivated and immune-activated regions across the microfluidic chips. This demonstrates thatby tuning gradient steepness and molecular identity, one can recreate complex, dynamic migration patterns observed in development and disease, thereby improving the physiological relevance of MPS models.

The ability of soluble molecules to regulate both cell migration and differentiation underpins their central role in organizing spatial architecture within MPSs. While microfluidic tools have made it increasingly feasible to generate stable or dynamic gradients, the main challenge remains identifying the specific molecules and concentration profiles required to mimic in vivo spatial patterning. Nonetheless, solute-mediated spatial control definitely contributes to the fidelity of MPSs in modeling organ physiology, pathology, and drug responses.

Spatial control of physical properties of the environment

The spatial modulation of physical properties within engineered scaffolds/substrates is not merely a materials design challenge - it is a biological imperative. In vivo, tissues exhibit spatial heterogeneity in stiffness, viscoelasticity, surface geometry, and porosity, reflecting the complex, dynamic, and compartmentalized environments that cells must interpret to function properly. Physical spatial heterogeneity can be separated into two parts: mechanical cues and topographical cues. Reproducing this physical complexity in vitro is essential to elicit physiologically relevant cell behavior and to decode how cells interpret geometric and mechanical instructions embedded in their surroundings. Without this spatial fidelity, cells in uniform scaffolds/substrates often lose polarization, reduce matrix remodeling capacity, or adopt non-physiological phenotypes, ultimately compromising the accuracy and effectiveness of in vitro models.

Mechanical cues

Mechanical cues—such as variations in substrate stiffness or viscoelasticity—play a central role in directing cell fate. The mechanical microenvironment of a stem cell niche is inherently heterogeneous: neural stem cells reside in soft (<1 kPa) brain tissue, while osteoprogenitors are found in much stiffer environments (>30 kPa). Engineering spatial stiffness gradients within a single scaffold allows for the simultaneous support of multiple cell phenotypes, such as osteogenic and chondrogenic zones in osteochondral tissue models. This spatial control is biologically necessary to replicate soft-to-hard tissue interfaces, which often fail in traditional implants due to mechanical mismatches. Cells interpret these stiffness cues via focal adhesions and cytoskeletal tension, leading to downstream activation of mechanotransduction pathways such as YAP/TAZ and RhoA-ROCK, which ultimately influence gene expression, migration, and lineage specification^223,224.

To replicate such mechanical heterogeneity in engineered systems, several fabrication strategies have been developed. Light-induced crosslinking enables spatial control over stiffness and viscoelasticity through photopatterning of crosslinking density. In Fig. 4a, Liu et al. developed a visible-light-responsive hydrogel using disulfide-based metathesis that allowed precise tuning of local stress relaxation²²⁵. This enabled spatiotemporal control of cell migration and YAP translocation, showing how dynamic viscoelasticity directly impacts cellular mechanosensing and gene regulation. Temperature-mediated phase separation offers a chemical-free approach to creating continuous stiffness gradients. As demonstrated by Vigolo et al., applying a temperature gradient across gellan gum precursors in a microfluidic channel induces thermophoretic flow and phase separation, forming hydrogels with smooth stiffness transitions (Fig. 4b). This method preserves cell compatibility while providing fine-tuned mechanical gradients ideal for stem cell differentiation studies²²⁶. Furthermore, the integration of functional nanoparticles with external fields provides a powerful strategy for engineering mechanical gradients in biomimetic substrates. Under electric fields, charged nanomaterials such as cellulose nanocrystals (CNCs), Laponite nanoclay, Chitosan, and Silk nanofibers can be directionally migrated within pre-gel solutions, resulting in spatially patterned stiffness after crosslinking^227–230. This method enables precise, contactless control over hydrogel mechanics without altering chemical composition. Magnetic fields offer another route to induce stiffness gradients by redistributing magnetic nanoparticles (MNPs) within the gel matrix²³¹. By embedding MNPs in photo-crosslinkable hydrogels and applying spatially controlled magnetic fields, researchers have created osteochondral-mimetic scaffolds with dual chondrogenic and osteogenic regions, which constructs promote spatially guided stem cell differentiation, providing a biomimetic interface for tissue regeneration²³². These effects are especially useful for forming compartmentalized tissue models within microfluidic chips.

Fig. 4. Controlling physical properties of the environment for spatially heterogenous cell patterns.

a Light-responsive hydrogel using disulfide-based metathesis allows precise, reversible tuning of local stress relaxation²²⁵ (CC-BY-NC-ND 4.0 license). b Temperature gradient across gellan gum precursors induces thermophoretic flow and phase separation, forming smooth stiffness gradients²²⁶ (CC-BY 4.0 license). c Dual-gradient silk-based hydrogel combines orthogonal gradients in stiffness (via photoinitiated crosslinking) and growth factor concentration (via peptide-nanocomplexes) for spatially guided tissue differentiation²³³ (Copyright 2025, Wiley). d Neuronal cells align and extend axons along chitosan microgrooved substrates²³⁴ (Copyright 2017, Elsevier). e Epithelial cells confined to different micropatterns exhibit differential YAP activation and nuclear morphology, influencing proliferation and differentiation²³⁶ (CC-BY 4.0 license). f Bicontinuous macroporous scaffolds with tunable architecture are fabricated via kinetically controlled phase separation to support spatially regulated cell behavior²⁴⁰ (CC-BY-NC 4.0 license)

These strategies can also be extended into multi-parametric gradient systems, where mechanics and biochemical signaling are co-patterned. As shown in Fig. 4c, Wang et al. constructed a dual-gradient silk-based hydrogel with orthogonal gradients in stiffness (via photoinitiated crosslinking) and growth factor concentration (via peptide-nanocomplexes)²³³. This platform enabled spatially distinct chondrogenic and osteogenic differentiation within a single construct, effectively mimicking the osteochondral interface.

Topographical cues

Topographical cues - including microgrooves, ridges, pores, and fiber orientation - provide physical templates that guide cell alignment, polarity, and motility. In native tissues, surface geometry is far from flat: aligned ECM fibers in tendons, villi in the intestine, or grooved basal membranes in the cornea all represent critical topographical features that direct cellular functions. The necessity for reproducing these features arises from the fact that cells interpret surface curvature and confinement as positional information. As shown in Fig. 4d, neuronal cells align and extend axons along microgrooved substrates, mimicking in vivo fasciculation²³⁴. In Fig. 4e, while epithelial cells confined to square versus circular micropatterns exhibit differential YAP activation and nuclear morphology, directly influencing proliferation and differentiation^235,236. From a biological standpoint, topography allows cells to organize into functional architectures, which is crucial not just for development but also for barrier formation, immune signaling, and regeneration. Fabrication methods such as photolithography, electrospinning, and freeze-casting offer robust platforms to spatially vary surface structure, enabling location-specific control of cell shape and matrix production^237–239. Jeong et al. employed a deformation-induced alignment strategy to simultaneously integrate microfluidic channels into aligned 3D collagen scaffolds, guiding neural network formation and enabling localized stimulation via reconfigurable microchannels²³⁹. Zhang et al. applied electrospinning to fabricate aligned conductive composite fibers that guide axonal growth and neural regeneration through topographical and electrical cues²³⁸. Additionally, Joukhdar et al. utilized freeze-casting to produce silk fibroin scaffolds with unidirectional porosity and multiscale anisotropy, mimicking tissue microarchitecture and directing ECM deposition²³⁷.

A particularly critical aspect of topographical control is pore structure, encompassing pore size, porosity, and interconnectivity. These parameters directly affect cell infiltration, nutrient diffusion, and tissue vascularization—all of which are essential for building viable 3D tissues. In native systems like bone marrow or liver sinusoids, microarchitecture allows cells to migrate and rearrange in a context-dependent manner. In vitro, scaffolds lacking appropriate pore structure often result in surface-limited cell colonization, hypoxia, and necrosis in the core. Studies have shown that pore sizes of 10–100 μm support optimal migration and 3D spread of many anchorage-dependent cell types, while high interconnectivity facilitates collective cell behavior and matrix remodeling. For this reason, bicontinuous porous hydrogels have emerged as powerful platforms that allow cells to exert contractile forces, form long-range connections, and modulate cytoskeletal markers like α-SMA in response to the mechanical architecture (Fig. 4f)²⁴⁰. Without such porosity, cells may behave more like in 2D monolayers, which fail to capture the complexity of in vivo tissue dynamics.

Together, these technologies not only allow researchers to regulate lineage specification and cell polarity but also enable functional compartmentalization and region-specific cellular behavior within otherwise homogeneous cell populations. By layering mechanical gradients with spatial biochemical control, it becomes possible to recapitulate dynamic tissue zonation, guide durotactic migration, and engineer highly biomimetic platforms that reflect the in vivo mechanical complexity of developing and diseased tissues. Spatial control over mechanical and topographical properties is not merely a technical advancement in scaffold design, but a biological imperative for replicating the spatiotemporal logic of native tissues. It empowers cells to behave as they do in their physiological niches—polarized, migratory, mechanically engaged, and environmentally responsive—ultimately establishing a foundation for constructing in vitro models and therapeutic implants that faithfully emulate organ-level structure and function. However, there remain many challenges with these two cues. Regarding mechanical cues, current materials have fixed mechanical properties, making it challenging to replicate the dynamic stiffness changes of native tissues. Precise control of stiffness gradients or local mechanical environments at the microscale remains difficult, and in vivo tissues are subjected to multiaxial stress, shear forces, and fluid dynamics, which single mechanical cues cannot fully recapitulate. Therefore, future improvements should focus on developing dynamically tunable materials for spatiotemporal stiffness regulation and integrating microfluidic techniques with mechanical modulation to achieve local gradients and multiaxial mechanical environments at the microscale.

Regarding topographical cues, most microfabrication techniques are suited for 2D or shallow 3D structures, limiting their ability to fully mimic complex tissue porosity and fibrous networks. Microscale features may not perfectly match cellular or tissue scales, affecting signal transduction. Moreover, surface topographies are generally static and cannot replicate temporal changes in porosity, shape, or degradability. Advanced additive manufacturing technologies could enable tunable and complex 3D architecture. In addition, developing degradable or dynamically reconfigurable materials would allow temporal modulation of porosity and structure. Combining mechanical and topographical cues to create multimodal spatial signals could further approximate the native tissue microenvironment.

Despite the promise of dynamic microenvironments in simulating in vivo-like spatial organization, mimicking time-dependent physiological processes such as tissue development, wound healing, or disease progression remains a major challenge. While enzyme-responsive hydrogels can couple cellular activity to local biochemical release, the timing and concentration of these events are difficult to predict or standardize due to variability in cell behavior, protease expression, and matrix degradation kinetics. This lack of temporal precision limits the reproducibility and fidelity of MPS models designed to recapitulate dynamic in vivo processes. To overcome this, emerging strategies now include integrating real-time biosensors to monitor key parameters such as protease activity, growth factor release, or matrix stiffness within the platform. Dornhof et al. developed a microfluidic platform to assess patient-derived tumor organoids in a controlled microenvironment by integrating electrochemical microsensors for oxygen, lactate, and glucose. This setup enables real-time monitoring of key metabolic parameters, ensuring the maintenance of hypoxic conditions and nutrient gradients that closely mimic the tumor microenvironment in vivo²⁴¹. These sensing systems can also be coupled with external feedback control, using light, temperature, or chemical inducers, to adjust matrix properties or trigger factor release on demand, ensuring spatiotemporal coordination. For instance, optogenetic or light-responsive hydrogels enable user-defined activation of morphogen release at specific time points, while microfluidic solute delivery systems controlled by programmable pumps can mimic pulsatile or gradient-based signaling events observed during development and repair. Major et al. utilized a light-sensitive matrix composed of adipose-derived decellularized ECM and silk fibroin, crosslinked via a light-mediated process, to mimic the temporal dynamics of the tumor microenvironment. This approach enabled the replication of breast epithelium structures, highlighting the platform’s potential for studying tumor development and tissue remodeling over time²⁴². Altogether, the combination of responsive matrices with feedback-controlled actuation holds strong potential to improve the predictability, adaptability, and physiological relevance of dynamic MPS platforms.

Direct control of spatial organization in microphysiological systems

Microfluidic compartmentalization

Microfluidic systems are a foundational component of MPSs, offering fine-scale control over the cellular microenvironment. Their microscale architecture enables precise regulation of fluid flow, chemical gradients, mechanical forces, and nutrient or oxygen delivery—all of which are essential for replicating physiological conditions. In the context of spatial patterning, microfluidic platforms are designed to organize cells and tissues in ways that reflect native biological structures. Generally, spatial patterning in MPSs falls into three major categories: (1) metabolic or transport coupling between different compartments, (2) reconstitution of physiological barriers, and (3) localized signaling interactions between distinct cell types. Each application relies on distinct microfluidic designs that facilitate communication or separation between regions. The first and second applications typically involve multi-compartment systems arranged either along the same perfusion line or around a shared vascular channel to simulate metabolic exchange. The third application focuses on barrier functions or cell-cell signaling across compartments and is often realized through either a gel-filled region or a porous membrane that separates culture chambers while permitting molecular communication (Fig. 5a).

Fig. 5. Microfluidic designs to control cell positions in MPSs.

a Three general design schemes for multi-cell-type microfluidic MPSs: (i) Sequential perfusion of different compartments, (ii) Common adjacent channel mimicking vascular circulation or (iii) parallel channels separated by a barrier, such as a channel or a membrane. b A vascular circulation microfluidic model comprised of eighteen organ compartments²⁴⁵ (CC-BY 4.0 license). c Two-compartment microfluidic chip to test gut–liver interaction in non-alcoholic fatty liver disease²⁴⁶ (CC-BY 4.0 license). d Two-level microfluidic chip mimicking gut–brain transport axis²⁵⁰ (Copyright 2021, Elsevier)

Network of compartments

MPSs can be designed with multiple compartments that are connected through a shared adjacent vascular channel, which mimics the circulatory network linking different organs in vivo. In this setup, each compartment houses a distinct cell or tissue type that interacts indirectly via the perfused vascular channel running alongside them. Endothelial cells typically line this vascular channel to simulate the endothelial barrier, allowing selective transport of nutrients, waste, or signaling molecules between the “blood” compartment and the adjacent tissue²⁴³. This design allows for more precise separation between tissue compartments while still enabling dynamic exchange through diffusion or trans-endothelial transport. It is especially valuable in modeling organ-organ communication where the bloodstream is a mediator. For example, Ronaldson-Bouchard et al. developed a system comprising four organs: liver, heart, bone, and skin, differentiated from human induced pluripotent stem cells (hiPSCs) with complementary stromal cells²⁴⁴. The four compartments shared an underneath layer of perfusion mimicking vasculature circulation. The system was used to assess the systemic toxicity of doxorubicin. Recently, a network of eighteen connected compartments was developed, coupling the vascular network and excretion system²⁴⁵. The compartments were connected via two adjacent circulating channels, one in the “artery” layer and another in the “vein” layer (Fig. 5b). The system was used for PK/PD studies of carboplatin – an alkylating drug for cancer treatment.

Alternatively, compartments can be connected via a shared perfusion line, the emphasis is on recreating directional metabolic flow or transport processes between tissues. This approach is particularly useful for modeling absorption, metabolism, and clearance pathways. Cells or tissues from different organs are cultured in parallel compartments, and media is circulated through a continuous microchannel that passes sequentially through each compartment. As a result, metabolites or signaling molecules produced by one tissue can be transported downstream to affect the next tissue, thereby simulating physiological cascades. For example, nutrients absorbed in an intestinal module can be metabolized in a downstream liver module. Yang et al., with this approach, came up with a system that comprises Caco-2 cells in the gut compartment and HepG2 cells in the liver compartment to study non-alcoholic fatty liver diseases, by administering free fatty acid into the system (Fig. 5c)²⁴⁶. They found that the coculture exhibited a protective mechanism against free fatty acids compared to monocultures. The simplicity of this two-compartment design enables easy fabrication and operation, while preserving essential aspects of metabolic coupling. A shared perfusion network can also be parallel, mimicking the in vivo circulatory network. The microfluidic system developed by Ramme et al. had two circuits: a blood circuit and an excretory circuit. The surrogate blood circuit supplies media to the intestine, liver, and brain compartments in a parallel manner, with an estimated flow rate proportionate to in vivo values²⁴⁷. One challenge for shared perfusion lines lies in tuning the flow rate and channel volume to ensure physiologically relevant timing and concentration profiles between compartments, thus parallel design is advantageous for better control of flow dynamics. However, compared to the single-line perfusion model, the aforementioned shared vascular channel still offers more control over barrier properties, shear stress, and localized exposure, making it more well-suited for studies of vascular-mediated signaling, immune cell trafficking, and systemic toxicity.

Parallel channels

To capture local paracrine signaling between neighboring cell types, microfluidic systems often use two adjacent culture channels separated by a gel-filled channel. This central hydrogel compartment mimics the ECM, allowing soluble factors to diffuse between the flanking channels while providing a 3D matrix for cell migration or invasion. Such a configuration enables researchers to study how cells respond to nearby signals without direct contact. For instance, to study the hypovascularity phenomenon in pancreatic ductal adenocarcinoma (PDAC), Nguyen et al. developed a PDAC-on-a-chip model in which both PDAC and endothelial channels were embedded in a 3D collagen matrix²⁴⁸. Using the system, they found the involvement of activin-ALK7 pathway in the endothelial ablation process of PDAC. Moreover, the separating channel in the parallel systems also supports gradient formation over time, making it useful for modeling processes like chemotaxis, angiogenesis, or tissue boundary formation. An example is the horizontal gut–brain axis-on-a-chip systems, which are usually used to establish gut-neuron connection within the connecting region of two culture channels. Such chip can be used to assess microbe-derived metabolites on neural activities²⁴⁹. This “sandwich” design allows fine spatial control over signaling dynamics while maintaining physical separation of cell populations, preserving their distinct identities and microenvironments. Moreover, the ECM-like properties of the gel can be tuned to model different tissue stiffnesses or barrier conditions, further increasing the physiological relevance of the system.

Another widely used configuration for enabling local signaling involves two culture chambers separated by a thin, porous membrane. This design, inspired by Transwell assays, allows soluble molecules to pass between compartments while keeping the cells physically apart. The membrane can be engineered with specific pore sizes, thicknesses, and materials to mimic various barrier properties, such as those of the intestinal epithelium, blood-brain barrier, or alveolar–capillary interface. In MPS applications, this setup is ideal for co-culturing cell types that interact through direct contacting or paracrine signaling, such as epithelial and immune cells, or epithelial and endothelial cells. For example, a gut–brain axis-on-chip system was developed by Kim et al. by stacking a gut compartment on top of an established BBB compartment, simulating the path of absorbed chemicals from gut epithelium to brain barrier²⁵⁰ (Fig. 5d). Yin et al. assessed the effect of clomipramine, an antidepressant drug, on hepatic activity and cardiac viability using a multi-organoid-on-a-chip system. The chip contains a two-layer well system, in which the top layer includes liver organoids, and the bottom one contains heart organoids; both were differentiated from hiPSCs²⁵¹. Overall, compared to gel-separated channels, membrane designs offer more consistent and defined diffusion profiles, making them suitable for quantitative studies of transport and signaling kinetics. Additionally, membranes can be functionalized with ECM coatings or patterned with microstructures to better emulate basement membranes and other tissue-specific features.

One of the key advantages of microfluidic design lies in the precise control over spatial organization, as the layout of cellular compartments, fluidic channels, and interfaces is defined during the design phase. This modularity enables researchers to easily tailor chip architectures for specific physiological models by adding, removing, or rearranging compartments, such as introducing endothelial barriers, immune cell niches, or epithelial-mesenchymal interfaces. Such flexibility makes microfluidics ideal for studying organ-level interactions, drug transport, and barrier function. However, this modularity also introduces design constraints when attempting to replicate highly intricate or dynamic tissue structures. For example, mimicking the gut epithelium, especially features like villi and crypts, requires fabricating fine microstructures, such as micropillars with thin membranes that support cell attachment, nutrient exchange, and perfusion through underlying vasculature-like channels. These systems must also be compatible with live-cell imaging, 3D staining, and functional assays, which adds further limitations in terms of optical clarity, accessibility, and geometry. Balancing physiological relevance with design simplicity remains a significant challenge, particularly when modeling structurally complex or highly vascularized tissues.

While advanced microarchitectures are technically feasible, their fabrication often demands multi-step processes, including precise multi-layer alignment, sub-micron resolution, and integration of materials that are biocompatible, optically transparent, and mechanically robust under continuous perfusion. These requirements can limit throughput and scalability, especially when incorporating features like micropillars, porous membranes, or soft hydrogels within rigid chip housings. Despite these challenges, recent advancements in microfabrication techniques, such as soft lithography, photolithography, high-resolution 3D printing, injection molding, and replica molding, have greatly improved the reproducibility and efficiency of fabricating simpler, multi-compartment microfluidic chips. These methods allow for rapid prototyping and design iteration, making it easier to standardize chip performance across batches. In addition, the growing availability of commercial microfluidic platforms and modular components is reducing barriers to entry for labs without cleanroom access, enabling broader adoption in research, drug screening, and preclinical testing. Continued improvements in materials, resolution, and fabrication automation will be key to scaling up complex designs without sacrificing reproducibility.

Bioprinting

Bioprinting, a rapidly advancing branch of additive manufacturing, enables the precise fabrication of biological structures by depositing living cells and biomaterials layer-by-layer. It encompasses several techniques, including inkjet-based bioprinting, extrusion-based bioprinting, and laser-induced bioprinting, each offering distinct advantages in terms of resolution, cell viability, material compatibility, and functionality. In the context of micro-physiological systems, bioprinting holds promise for spatially controlling cellular organization and tissue architecture, thereby allowing researchers to mimic complex in vivo microenvironments with high fidelity. This spatial precision is critical for replicating tissue-specific functions, studying disease progression, and testing drug responses in physiologically relevant models. As such, 3D bioprinting is emerging as a powerful tool in regenerative medicine, drug screening, and personalized healthcare, significantly advancing our ability to model and manipulate human biology in vitro.

Inkjet-Based bioprinting

Inkjet bioprinting precisely deposits microdroplets containing living cells, biomaterials, or bioinks onto a substrate to construct tissue structures, organ models, or drug screening platforms. This approach offers an effective combination of high resolution and low cost. Jung et al. reported a fully inkjet-printed alveolar barrier model consisting of functional layers of epithelial and endothelial cells separated by a collagen-based basement membrane. High-resolution patterning of four types of human alveolar cells—NCI-H1703, NCI-H441, HULEC-5a, and MRC-5—along with type I collagen, was achieved using on-demand inkjet printing (Fig. 6a). This enabled the automated, high-resolution deposition of alveolar cells to fabricate a trilayer alveolar barrier model with a thickness of ~10 μm. When compared to non-structured 3D models where alveolar cells and collagen were uniformly mixed, the structured model showed superior fidelity in mimicking lung tissue characteristics²⁵². Concerning this technology, it is limited by low printing throughput and potential nozzle clogging due to the high viscosity of the bioinks. Although this technique ensures high printing resolution (80 μm) at low cost, its poor formability limits broader applications. In addition, the generation of thermal bubbles and piezoelectric deformation may cause damage to cells, necessitating more precise control over the printing process parameters. Reproducibility is challenged by droplet size variability, nozzle clogging, and bioink properties. Sources of variability include printer calibration, bioink formulation, and operator training. Improvements can be achieved by standardizing bioink rheology and monitoring droplet formation in real time.

Fig. 6. Cellular spatial organization achieved through diverse printing strategies.

a Fabrication of an alveolar barrier model using inkjet printing technology²⁵² (CC-BY 4.0 license). b Schematic illustration of 3D bioprinting employing a microgel-based biphasic bioink in a stepwise manner²⁵³ (Copyright 2022, Wiley). c In-bath triple coaxial cell printing for constructing artery equivalent and shape-tunable fabrication²⁵⁵ (Copyright 2021, Wiley). d Microscale continuous optical printing for precise cell patterning²⁵⁷ (Copyright 2020, Elsevier). e Deep-penetration focused ultrasound-assisted printing. Biostructures can be printed by selective curing of sono-inks²⁶⁰ (Copyright 2023, American Association for the Advancement of Science). f Magneto-Archimedes effects-driven self-assembly of cells into organized architectures²⁶² (Copyright 2023, American Chemical Society)

Extrusion-based printing

Extrusion-based bioprinting is currently the most widely used bioprinting technique, enabling the layer-by-layer deposition of bioinks composed of cells, biomaterials, growth factor onto a printing platform. This technique offers strong material adaptability and structural flexibility, making it well-suited for the fabrication of both in vitro tissue models and in vivo implants. Xiong et al developed a cell-laden biphasic bioink composed of microgels and a continuous phase, which offers an expanded bioprinting window and excellent printability due to its shear-thinning behavior, yield stress, and self-healing capability. By encapsulating HepG2 cells and HUVECs separately within the microgels and hydrogel precursor of the microgel-based bioink, the researchers successfully printed liver tissue constructs. The results revealed that the microgel-based bioink supported a relatively mature gene and protein expression profile of hepatic cells, thereby enhancing liver-specific functions as well as the expression of liver-related genes and proteins (Fig. 6b)²⁵³. Based on this approach, Zhang et al. proposed enhancements to the deposition substrate. MSCs were deposited onto a cryogenic substrate, achieving long-term preservation under low-temperature conditions. The viability, proliferation, and differentiation capacity of the printed tissues were evaluated by examining the osteogenic, adipogenic, and chondrogenic differentiation potential of the MSCs. Moreover, the research confirmed that tissues produced via low-temperature bioprinting retained their ability to promote angiogenesis, indicating that neither the cells nor the growth factors were adversely affected by the freezing process²⁵⁴.

Furthermore, advancements in printhead design have enabled coaxial bioprinting based on extrusion, where different materials are aligned concentrically to form core–shell structures. For instance, Cho et al. proposed a coaxial cell printing strategy and used it to fabricate an atherosclerosis model. A triple-coaxial nozzle was employed to simultaneously dispense a sacrificial ink and two distinct bioinks containing human umbilical vein endothelial cells (HUVECs) and human coronary artery smooth muscle cells (HCASMCs), respectively, resulting in a three-layered structure that closely mimics the physiological architecture of native blood vessels. Under conditions of constricted and tortuous flow, the presence of multiple vascular constructs facilitated the recapitulation of hallmark events associated with early-stage atherosclerosis under physiological conditions. This platform was further used to investigate the individual and synergistic roles of co-cultured cells and local turbulence in modulating the initiation of atherosclerosis, as well as the dose-dependent therapeutic effects of atorvastatin (Fig. 6c)²⁵⁵. Support bath-assisted printing is also a branch of extrusion-based printing. It retains the fundamental principle of extrusion through a nozzle, while the introduction of a supportive medium significantly extends its applicability and enables the fabrication of structures with greater complexity. Lutolf et al. utilized extrusion-based bioprinting in combination with a support bath to spatially arrange and form interconnected and evolving cellular structures. By precisely controlling geometry and cell density, they successfully generated centimeter-scale tissues with self-organizing characteristics, such as lumens, branched vasculature, and tubular intestinal epithelium that exhibit in vivo–like crypt–villus architecture. Furthermore, they demonstrated how morphogenesis can be modulated by spatiotemporal deposition of supportive cells, and how sequential printing of different epithelial cell types can be employed to mimic organ boundaries observed in the gastrointestinal tract²⁵⁶. Because this technique employs large nozzle diameters, its resolution ranges from 100−500 μm, enabling the printing of highly viscoelastic bioinks and facilitating the construction of 3D solid biostructures. Moreover, by sacrificing precision to increase the volume of discrete units, the survival rate of encapsulated cells can be effectively improved to about 90%. Variations in filament diameter and shear-induced cell damage reduce reproducibility. Nozzle geometry, extrusion pressure, material viscosity, and operator technique contribute to variability. Closed-loop control of extrusion, standardized cartridges, and optimized bioinks can enhance consistency.

Light-induced bioprinting

Light-induced bioprinting is a technique that utilizes the modulation of light at specific wavelengths to achieve the spatial patterning of photoactive bioinks. This approach enables high-resolution, non-contact fabrication of complex biological structures by selectively crosslinking light-sensitive materials in defined regions. Chen et al. employed microscale continuous optical printing (μCOP) to encapsulate neonatal mouse ventricular cardiomyocytes (NMVCMs) within a hydrogel matrix, enabling the fabrication of cardiac tissues with aligned cardiomyocytes guided by the printed microstructures. This alignment closely mimics the anisotropic myofibrillar organization found in native myocardium (Fig. 6d). Results revealed that the aligned 3D myocardial microtissues produced nearly twice the contractile force compared to conventional 2D cardiac cultures. Functional assessment of the cardiac constructs was performed by analyzing calcium transient waveforms, where treatment with isoproterenol—a β-adrenergic agonist known to enhance cardiac contractility—led to significant increases in both the amplitude and decay rate of calcium transients. These findings highlight the tissue’s capacity to respond physiologically to pharmacological stimulation, underscoring the potential of μCOP-fabricated cardiac models for drug screening and disease modeling²⁵⁷. This technique can also be integrated with microfluidic chips to enable gradient digital light processing (DLP) bioprinting. By redesigning the resin vat system and enabling chaotic mixing of multiple bioinks during flow, the platform can generate cellular, chemical, mechanical, porosity, and dual bioink gradients, including gradients formed from PEGDA and GelMA. In a 4-week osteogenesis study, three different flow rate ratios were used to mix pore-forming GelMA-dextran/BMP-2/MSC bioink with GelMA/MSC bioink, creating mixed inks with varying concentration gradients for printing. The results indicate that the combined porosity and BMP-2 gradients effectively promoted the osteogenic differentiation of bone marrow mesenchymal stem cells (BMSCs), closely recapitulating the hierarchical architecture of native bone tissue²⁵⁸.

In addition, light-induced bioprinting exhibits exceptional printing efficiency. The upgraded volumetric additive manufacturing technique can reduce the printing duration from hours to 15 s. Levato et al. demonstrated a rapid visible volumetric bioprinting by integrating living cells directly into the bioink. Trabecular bone models containing mesenchymal stem cells (MSCs) were cultured in osteogenic medium for 7 days, followed by co-culture with endothelial colony-forming cells for an additional 3 days. This led to the formation of early angiogenic sprouts—precursors to capillary networks. In another model, a meniscus construct containing 10⁷/mL of articular cartilage progenitor cells was cultured in vitro for 4 weeks. This resulted in the development of a new fibrocartilaginous matrix and a significant improvement in mechanical properties, with the compressive modulus increasing from 24.63 ± 0.65 kPa to 266.54 ± 4.49 kPa, approaching that of native human fibrocartilage. Abundant type I collagen—a hallmark of native meniscal tissue—was detected, along with limited expression of type II collagen, reflecting a composition similar to that of the human meniscus. This volumetric printing approach offers high precision, rapid fabrication speed, and strong bioink compatibility, although it requires careful control over light intensity and resolution to ensure printing fidelity²⁵⁹. In summary, light-induced bioprinting offers both high printing resolution (~20 μm) and excellent cell viability (95%), with great flexibility in pattern design. Nevertheless, further efforts are required to expand the range of printable biomaterials. Inconsistent photopolymerization and curing depth affect structural fidelity. Sources of variability include light intensity, bioink optical properties, and operator settings. Standardized light calibration and consistent photoinitiator concentration can improve reproducibility.

Other bioprinting methods

Acoustic bioprinting is a non-contact bioprinting technique that utilizes sound waves—particularly ultrasound—to precisely deposit cells or biomaterials. This method is exceptionally gentle and cell-friendly, making it an emerging high-precision bioprinting strategy. However, its application is currently limited by the range of compatible materials and the scale of printable constructs. Zhang et al. developed a deep-penetrating acoustic volumetric printing technology, in which acoustic bioink is delivered to a target region and selectively solidified into complex structures by focused ultrasound emitted from an ultrasonic printing probe (Fig. 6e). This versatile approach was demonstrated through proof-of-concept applications including left atrial appendage occlusion, tissue reconstruction, and targeted drug delivery²⁶⁰. Furthermore, Gao et al developed an imaging-guided deep tissue in vivo sound printing platform, termed DISP. This technique integrates thermosensitive liposomes loaded with crosslinking agents into bioinks, enabling precise, rapid, and on-demand crosslinking of various functional biomaterials through focused ultrasound. The study demonstrated successful printing within lesion sites located in the deep regions of the mouse bladder and rabbit thigh muscle, highlighting the platform’s potential for cellular spatial organization and tissue replacement²⁶¹. Although DISP is claimed to achieve high-resolution printing (150 μm) and rapid printing speeds (up to 40 mm/s), its resolution for in vivo printing remains to be improved. Precision is limited by acoustic wave inconsistencies and sensitivity to bioink properties. Equipment alignment, bioink density/viscosity, and operator calibration introduce variability. Solutions include standardized acoustic parameters and optimized bioink formulation.

Magnetic bioprinting is an emerging technique that combines magnetic field manipulation with bioprinting to control, position, and assemble cells or biomaterials into complex three-dimensional tissue architectures. This approach employs magnetic nanoparticles and external magnetic fields to direct cellular movement, offering a contactless and programmable strategy for tissue fabrication. Wang et al. introduced a strategy based on the magnetic Archimedes (MagArch) effect, which manipulates cell movement directly. By adjusting the concentration of the paramagnetic agent gadobutrol, they achieved tunable pattern dimensions (Fig. 6f). Multilayer assembly enabled the spatial organization of multiple cell types, allowing the fabrication of increasingly complex cellular patterns. As a proof of concept, researchers constructed a tumor–endothelial cell co-culture model within a sealed microfluidic channel to simulate epithelial–mesenchymal transition (EMT) under shear stress, mimicking a cancer-relevant microenvironment²⁶². The spatial resolution of magnetic bioprinting primarily depends on the magnetic field gradient and the degree of cell magnetization, and the resolution is comparable to extrusion-based bioprinting. By avoiding high shear stress and elevated temperatures, this approach can achieve cell viability exceeding 95%. Variability arises from different magnetic field gradients and heterogeneous cell magnetization. Equipment (magnet strength), material (nanoparticle concentration), and operator are key sources. Standardized magnetic profiles, controlled cell magnetization protocols, and automated magnetic actuation can enhance reproducibility.

In all, bioprinting technology enables precise spatial positioning of cells at the microscale, providing a structural foundation for the construction of physiologically relevant organoids and microtissues. By accurately controlling the spatial distribution of cells and biomaterials, this technology facilitates the simulation of in vivo microenvironments, thereby enhancing the physiological relevance of tissue development, signal transduction, and functional expression in micro-physiological systems. Currently, beyond its established roles in investigating cellular behaviors and functioning as an in vitro drug screening platform, this technology is progressively advancing toward non-invasive in vivo bioprinting, with the goal of repairing tissue defects and facilitating integrated disease monitoring and therapeutic interventions.

Physical cell trapping

Physical phenomena are usually used to directly organize cells in microbiome systems. By applying a physical external force on cells, they are manipulated and patterned as desired. Based on the physical principles governing the forces, physical trapping is classified into four main methods, including optical trapping, acoustic trapping, magnetic trapping, and electric trapping.

Optical trapping

Optical trapping is a non-contact technology that uses optical force to manipulate bioparticles and cells precisely. Optical forces are exerted by light, typically in the micron or sub-micron range of wavelength, which is generated from a highly focused laser beam. The utilization of optical manipulation was first recorded in 1970 by Arthur Ashkin and has been completely recognized for its impactful application in biomedical studies after receiving the 2018 Physics Nobel Prize.

Since the intrinsic dielectric characteristics of cells, they experience optical forces derived from Maxwell’s theory using the far-field expression based on the incident electric $E_i$, scattered electric $E_s$ and magnetic fields $B_s$ in a medium of refractive index $n_m$ and dielectric permittivity ${\epsilon }_m$, given by Eq. (1)^263,264.

$$F_{optical}=-\frac{{\epsilon }_m}{4}{r}^2\int \left[{∣E_s∣}^2+\frac{c^2}{n_m^2}{∣B_s∣}^2+2Re\left(E_i\cdot E_s^{*}+\frac{c^2}{n_m^2}{B}_i\cdot B_s^{*}\right)\right]\widehat{r}d\Omega$$ 1

where r is the radius of the particle, $\hat{r}$ is the radial unit vector, and the integration is carried out over the full solid angle Ω of a sphere.

Cells and bioparticles have an arbitrary shape and soft deformation in some cases, which can be altered under the influence of external factors, including the gravitational force, the drag force, optical force, and the surrounding areas such as culture gel or culture chamber wall. Under light expression, the shape change has impacts on the scattered and refracted light, leading to difficulty in evaluating their exertion. To simplify the process, cells and bioparticles are illustrated as electric dipoles in a homogeneous field, with a set of conditions reducing the coverage of the computation^265,266.

Since its utilization in trapping and manipulation based on infrared laser beams in 1987, optical tweezers have been utilized for various applications, such as biology, microbubble manipulations, spectroscopy, or chiral optomechanics^267–270. For cell manipulations, the utility of optical tweezers was limited until their potentially harmful impact on cells, which is photodamage, was revealed to be mitigated by using lower-intensity tweezers, including optoelectronic, plasmonic, and photonic crystal^271–273. Among them, photonic-crystal optical tweezers show the most promising ability to be integrated in MEMS²⁶⁸. Using a wavelength of 1064 nm guiding on the surface of a 2D plasmonic-crystal comprising a square lattice with a period of 5.8 μm and a diameter of 3.6 μm, mammalian, yeast, and Escherichia coli cells were trapped²⁷⁴. There are no significant changes in the trapped cells’ morphology until their blebbing after 30 min, confirming the trapping efficiency without compromising cell viability. Building upon his success in using photonic-crystal optical tweezers for cell trapping, Jing et al. developed an advanced method to pattern human pluripotent stem cells (hPSCs) for lateral culturing (Fig. 7a)²⁷⁵. The photonic crystal was coated with oxygen-plasma-treated parylene-C film before adding culture media and being placed on a temperature-adjustable thermoelectric heater. hPSCs were manipulated using optical tweezers to obtain the desired pattern, then an extracellular matrix was added to the culture media for adherent culture. Without applying optical tweezers, the temperature of 37˚ generated by the heaters forced free hPSCs to attach to the parylene-C surface. Cells were cultured for 6 days, witnessing the formation of colonies on the surface. Increasing the plasma treatment power on parylene-C enhances the average cell density on the culture film. Nevertheless, this also necessitates higher laser intensities for optical cell manipulation, thereby increasing the risk of photodamage. In general, the use of lasers in optical tweezers enables highly consistent and precise manipulation of cells as the laser system could provide exceptional stability and minimal variability over time, facilitating the reproducibility of an individual experiment for specific cells and applications. Given that lasers are widely available equipment in laboratories, this method is accessible, though it requires careful tuning and calibration for a specific laser type, brand, and studied cell line. Despite these advantages, their scalability remains limited due to their high cost to implement and maintain, and the small effective working area, restricting high-throughput or large-scale cell organization.

Fig. 7. Physical trapping methods for cellular spatial organization in MPSs.

a Optical-based manipulation. (i) Schematic image of a photonic-crystal optical tweezers setup. (ii) 3D assembling of cells on the ECM fragment to form cell clusters²⁷⁷ (CC-BY 4.0 license). (iii) Merged transmission and fluorescent macroconfocal image of resulting cultures after 72 h²⁷⁷ (CC-BY 4.0 license). b Hybrid BAW and SAW-based tweezers aligned human umbilical vein endothelial cells and adipose stem cells to form functional collateral cylindroids for ischemia therapy²⁸⁷ (CC-BY 4.0 license). c Holographic acoustic tweezers constructed a hologram potential distribution for cell patterning²⁹² (CC-BY 4.0 license). d Magnetic force creating bi-layer spheroids³⁰⁸ (Copyright 2020, American Chemical Society). e Electrode structure for electric DEP-based trapping of cells³¹⁶ (Copyright 2019, Wiley)

Unlike the traditional optical tweezers use a point-based trap principle to manipulate cells, a lightsheet optical tweezers was developed to trap particles in a line²⁷⁶. The gradient potential of the lightsheet field attracted the nearby live Hela cells thanks to their intrinsic dielectric characteristics, taking an average of 8.57 s to reach the trap center. By rotating the light sheet in the transverse plane, the technique was able to manipulate cells in specific patterns, and the patterned cells were able to stay healthy after 18 h of culturing in the cell medium of DMEM and FBS.

Holographic optical tweezers, an advanced version of optical tweezers with higher resolution, were used to 3D position cells to form an intricate cellular microenvironment²⁷⁷. This approach implied a computer-based modulator to customize the incoming laser beam, creating a programed optical trapping net instantly. Multiple cell types were introduced for the first reorganization to form a desired shape with a stable structure, followed by the manipulation of microparticles and electrospun fibers made of hydrogel within the previously formed cell morphology. This strategy allows the stabilization of the co-culture hydrogel system for 10 days. In addition, these holographic optical tweezers were able to spatially pattern stem cells on a re-cellularized ECM 3D fragment (Fig. 6a), allowing the colonization of stem cells after 72 h of culture.

While continued development of optical may reduce the risk of photodamage through carefully tuning laser power and wavelength. The bulkiness of lasers alongside their high-power consumption, pricy accessories, and the need for technical expertise continues to limit their widespread use.

Acoustic trapping

Acoustic trapping, also known as acoustic tweezers, is a technology that uses acoustic force to manipulate particles via acoustic waves. Acoustic tweezers allow non-contact manipulations of particles and cells via fluid mechanics. Acoustic waves, typically generated from an acoustic wave transducer, are often produced using piezoelectric materials, which can convert electrical signals to acoustic signals. According to the propagating acoustic waves in the fluidic chamber, acoustic tweezers can be classified into two main categories of bulk acoustic waves (BAWs) and surface acoustic waves (SAWs). Both mechanisms create zones of high and low pressure regarding the vortex-shaped waves of nodes and antinodes, trapping particles.

BAW devices utilize bulk acoustic transducers via thickness-mode vibrations to induce acoustic waves, which are propagated through the bulk of the material. Bulk standing waves generate a resonator cavity with a size proportional to several times of the half wavelength, leading to the trapping and positioning of particles, which were utilized in the cell spatial organization for culture in micro-fabricated devices or hydrogel environments²⁷⁸. By using an ultrasonic standing wave, cells were positioned to form multicellular tumor spheroids within a microfluidic platform consisting of 100 micro-wells²⁷⁹. In the absence of ultrasonic waves, re-suspended adherent tumor cells were settled on the bottom of the multi-well due to gravity, which was coated with protein-repellent amphiphilic polymer previously to ensure the free-floating of cells. Once the ultrasonic transducer started to generate waves, cells aggregated precisely at nodes of the waves and formed spheroidal clusters after 24 h culture with the activation of the ultrasonic transducer. With 4 tested cell lines, all of them could form multicellular tumor spheroids after 24 h of ultrasonication, while the majority of them witnessed the spreading of cells after the turning off of the ultrasound generator. Because BAW can induce significant heating due to acoustic absorption and viscous dissipation, BAW tweezers often incorporate with cooling system to avoid the heating of the culture plate temperature over 37°C that can harm cell viability and stabilize the culture system^280,281. For these ultrasonic standing wave tweezers, a PID controller was integrated to cool down the device. However, the addition of these cooling components increases the complexity, size, and overall bulkiness of the system.

Besides BAWs, SAWs have been used to manipulate bioparticles, cells, spheroids, and organoids in 2D and 3D, providing high stability, good biocompatibility, precise controllability, and especially lower thermal effect compared to BAWs^278,279,282. SAW trapping employs SAW resonators to exert acoustic waves that propagate along the surface of a piezoelectric material. Acoustofluidic devices, which are the combination of fluid dynamics and acoustic waves, have gained great interest in serving various applications in MPSs. Wu et al. applied acoustofluidics exerting SAWs on a multi-channel microfluidic PDMS platform to generate heterotypic cell cluster formation²⁸³. A SAW generator, made of a pair of interdigital transducers (IDT) coated with lithium niobate (LiNbO3) piezoelectric material, induced SAWs with a wavelength of 500 µm spreading along multi-microfluidic channels placed in the middle. The transmitted SAWs formed a pressure node distribution inside each parallel microchannel, moving the free-distributed cells to the nearby pressure nodes and laterally gathering into 3D multicellular clusters after 2–3 min expression of SAWs. Following the acoustic organization is the utilization of the pulse wave to maintain the cluster pattern for the next 9 h incubation continuously. Heterotypic spheroids and organoids were successfully assembled and viably proliferated after 10 days of lateral culture on Petri dishes (Fig. 7b). This platform allows various cell cluster formations within minutes and offers an improved approach to reducing the heat generation and energy consumption due to the continuous operation of the SAW generator. Based on the same trapping mechanism, several designs have been developed to organize cells and engineer tissues spatially^284–286.

A combination of BAW and SAW was developed to construct cylindroids within a PDMS chamber (Fig. 7b)²⁸⁷. To form the hybrid acoustic tweezers, the trapping chamber was placed between the IDT electrodes, which generate SAW, and on top of the piezoelectrode substrate, which exerts BAW. The induced BAW was perpendicular to the SAW and had its largest displacement amplitude at the nodes of SAWs with intervals at half of the SAW wavelength. The combination of BAWs and SAWs created a 3D acoustic pressure mapping, enabling the organization of cells horizontally and vertically. The system aligned stem cells and endothelial cells into parallel lines to mimic the collateral vessel structure within 5 s with an optimized cell concentration for continuous cell line formation. The encapsulation maintained its structure after 7 days of culture in hydrogel, secreting growth factors, and was also able to be transplanted into the host vasculature for further studies.

To manipulate cells effectively, several key physical parameters of the acoustic wave generated electrodes must be optimized such as applied voltage and geometry of IDT. Moreover, electrode materials, cell culture medium, and intrinsic cell characteristics also influence the overall assembly. Systematic errors originating from the electrical source further contribute to the variability. Consequently, reproducibility across batches and platforms is achievable if proper calibration is performed and variation is minimized. However, achieving standardization across laboratories remains a major challenge, as differences arise not only from device design and operating conditions but also from material properties. In addition, even though using multiple approaches of BAW and/or SAW, the resulting tissue constructs from these approaches are relatively simple in structure, typically limited to spheroids, organoids, or, at best, cylindroids. By changing the structure, the position, as well as the tilt angle of the IDT, the acoustic pattern can be altered and made more complex, yet not flexible enough to arrange an arbitrary structure^288–290.

The integration of the holographic technique into acoustics has shed light on the development of more advanced structures of cell encapsulation and engineered tissue²⁹¹. The hologram plane is the most vital element of this technique, converting the planar wavefront from the wave transducer into the target phase information, creating the desired acoustic pressure distribution. The plane is produced using the 3D printing technique, with its topography encoded regarding the desired phase. When the ultrasound transducer is activated, it generates acoustic radiation forces that push the particles up until constrained by the top boundary, laterally migrate horizontally toward the high-pressure area, resulting in a flowing pattern of bioparticles. Ma et al. have applied the holographic method to cell patterning in a hydrogel (Fig. 7c)²⁹². HHT-116 human colon cancer cells were mixed with hydrogel at 4˚C to keep the aqueous state of the hydrogel. The acoustic transducer was activated as the sample contained in the cooled Petri dish warmed to room temperature, enabling cell patterning before gelation and ensuring their immobilization. The cell pattern was transferable for further cultivation, witnessing the viability of cells after 7 days. Multiple other patterns of cells were investigated using phase holograms^293–295, though recently limited to 2D patterns, showing its great potential in forming more complicated tissue-mimicking structures. To enable effective 3D cell organization, multi-angle and multi-transducer arrays integrating with the holographic method are promising approaches to sculpt 3D acoustic potential wells, manipulating cells horizontally and vertically²⁹⁶.

Magnetic trapping

For direct control of bioparticles using external physical force, magnetic force-based manipulation is a commonly investigated method²⁹⁷. Under the influence of a gradient magnetic field exerted by permanent magnets or electromagnets, the magnetic force acting on particles is given by the following Eq. (2)^297,298.

$$F_m=\frac{\left({\mathrm{\chi }}_{\mathrm{p}}-{\mathrm{\chi }}_{\mathrm{m}}\right)}{{\mu }_0}{V}_p\left(B·\nabla \right)B$$ 2

Where ${\chi }_p$ and ${\chi }_m$ are magnetic susceptibility of the particle and the surrounding area, respectively, $V_p$ is the volume of the particle, and B is the magnetic field. The direction of ${\mathit{F}}_{\mathit{m}}$ depends on the relation between ${\chi }_p$ and ${\chi }_m$, deciding the attraction or repulsion, which are positive or negative magnetophoresis, respectively. Usually, depending on the intrinsic magnetic characteristics of cells and their specific culture medium cannot provide sufficient force to manipulate cells. Therefore, either bioparticles or the surrounding medium is characterized to boost the magnitude of the difference between two susceptibilities. The first approach is to magnetically binding cells, named labeling method. The latter is immersing cells in a paramagnetic environment, called non-labeling. Hence, magnetic cell organization can be classified into two mechanisms, which are magnetically labeling and non-labeling^297,298.

In the instance of labeling technique, cells undergo binding to paramagnetic or superparamagnetic particles, which exhibit greater magnetic susceptibility than the cells as well as the surrounding medium (${\chi }_p$ >> ${\chi }_m$). Hence, the magnetic-labeled cells are exposed to positive-magnetophoresis, also known as paramagnetism, which pushes cells towards the strong magnetic field. This approach allows the assembly of cells when applying an external magnetic field forward the specific desired positions. One thing to note with this strategy is the biocompatibility and non-toxicity of magnetic particles, which are extremely vital for the viability of cells during lateral cultivation after magnetic organization. Although this strategy provides a convenient approach to manipulate cells directly and precisely thanks to the magnetic bead labeling and magnetic field, cells’ characteristics can be affected during the modification process, influencing their natural development, and possibly leading to a biased following assessment. Hence, a standardized protocol for magnetic labeling and a precise lateral characterization process are essential to provide a pure result of the cell manipulation effect.

Magnetic force-based tissue engineering (Mag-TE) is a commonly studied technique in the application of magnetic trapping in cell organization^297,299. Magnetic nanoparticle (MNP)-functionalized cells of different lines were able to form sheet-like structures within 24 h of culture on Petri dishes placed upon the permanent magnets, showing the signal of viability via their expression of specific growth factors^300,301. Alternations in the magnetic distribution on the culture surface caused by the magnet shape, position, and intrinsic characteristics can lead to differences in the resulting cell pattern^302,303. Besides utilizing an external magnetic field to assemble cells, magnetic scaffolds are potential candidates for forming tissues of more complicated structures³⁰⁴. Huang et al. developed a magnetic nanocomposite hydrogel to co-culture with bone mesenchymal stem cells³⁰⁵. The poly(vinyl alcohol) (PVA) hydrogel included dispersed Fe₂O₃, magnetic nanoparticles, was prepared and had similar water content as natural cartilage, providing good cell proliferation as the pure PVA hydrogel after 10 days of culture. Multiple cell types patterning in vitro within a 3D magnetic scaffold was reported³⁰⁶. The magnetic osteogenic scaffold was formed by injecting and depositing the mixture of Fe-doped hydroxyapatite and the poly(ε-caprolactone). Mesenchymal stem cells (MCSs) and human umbilical vein endothelial cells (HUVECs) were patterned one by one, and each cell type following these following steps. First, the magnetically labeled cells suspended in culture medium were injected into the scaffold, then the external magnetic field exerted by a permanent magnet was applied for 3 next hours with a note that the position of magnets for the two cell lines are opposite. Though not specifically stated, the approach was confirmed to witness the viability and proliferation of cells after 72 h, which were closely comparable to the control monolayer dual-cell culture. With the support of magnetic scaffold and magnetic particles, cells can be assembled as desired, though, its application is limited to in vitro studies other than implantation since the impact of magnetic materials.

For the non-labeling method, paramagnetic buffers are used to facilitate magnetophoresis instead of conventional culture solutions. Since the surrounding environment is paramagnetic, its susceptibility is much larger than that of cells (${\chi }_m$ ≫ ${\chi }_p$), cells exhibit diamagnetism, which is the repulsion of cells toward the weaker area of the magnetic field. Hence, the non-labeling method is considered a diamagnetic manipulation.

Similar to the cell patterning strategy of the labeling method, by customizing the magnetic field to assemble cells as the desired structures, various layouts of permanent magnets determine how cell clusters are formed³⁰⁷. Gadopentetic acid (Gd-DTPA) was used as the paramagnetic medium, combined with multiple permanent magnet configurations to contactlessly assemble cells to rectangular, three-pointed star, and spheroid shapes. The morphologies were maintained 2 h after magnet removal. This group has conducted further development for the organization of multicellular spheroids in 3D culture³⁰⁸. Michigan Cancer Foundation-7 (MCF-7) human breast cancer cells were introduced to the paramagnetic solution for clustering in 3 h, followed by the injection of 3T3 fibroblast cells with the same protocol, forming the multicellular spheroids (Fig. 7d). The strategy was scalable by using a magnet array for a 96-well plate. Ren et al. integrated label-free magnetic aggregation into a microfluidic platform³⁰⁷. Ovarian cancer cells (A2780s) and HUVECs were introduced to the assembling chamber one by one under the dynamic culture medium. The pattern size significantly grew after 24 h. Compared to the labeling method, the non-labeling approach offers a less troublesome way to manipulate cells by using a paramagnetic medium instead of modifying magnetic particles on cell surfaces. However, culturing cells in paramagnetic medium for a long period requires careful consideration of dosage and cell lines to maintain an instinct growth of cells.

The applied magnetic field and the magnetic labeling efficiency of cells have an impact on the reproducibility in magnetic trapping. Variations in cell manipulation arise from the heterogeneity in labeling protocols and the biological variability of cell lines. Factors relating to device and equipment like permanent magnet design, positioning, and field stability also contribute to the inter-laboratory discrepancies. To enhance reproducibility, standardized procedures for nanoparticle synthesis and characterization, cell labeling, and field calibration are essential. In addition to the conventional utilization of permanent magnets, which are limited to size, shape, and remanent flux density, the electromagnetic field induced by an electrical coil can bring more advantages of its flexibility to provide a tunable magnetic field to assemble cells into advanced morphology and structure.

Electric Trapping

Cells are dielectric particles that are subjected to a translational force provided by inhomogeneous electric fields. Hence, an electric field strategy such as dielectrophoresis (DEP) provides a viable method for cell manipulation and organization. DEP is considered as the simplest approach for cell separation and confinement because it requires only an activated electric field ^48–50. Under the activation of a non-uniform electric field, a dielectrophoretic force acts on a dielectric particle as the equations below^309–311.

$$F_{\mathit{DEP}}=2\pi {\epsilon }_0{\epsilon }_m{r}^3\mathrm{Re}\left(\mathit{CM}\right)\nabla {\left|E_{\mathit{RMS}}\right|}^2$$ 3

$$\mathit{CM}=\frac{{\epsilon }_{\mathit{ceff}}^{*}-{\epsilon }_m^{*}}{{\epsilon }_{\mathit{ceff}}^{*}+2{\epsilon }_m^{*}}$$ 4

where ${\epsilon }_0$ is the vacuum permittivity, ${\epsilon }_m$ is the relative permittivity, E is the electric field, r is the external radius of the particle, and CM is the Clausius-Mossotti factor, which is calculated using the above equation.

DEP can be separated into positive DEP (pDEP) and negative DEP (nDEP) regarding how cells are manipulated into the high or low field strength, which is determined by the electrical characteristics of cells and the surrounding area. Not only well-established in cell separation and cell enrichment, DEP has also been studied to organize cells in 3D culturing and tissue engineering. By understanding the theoretical DEP acting on cells, which can be nDEP or pDEP, electrodes are designed and customized to control the electric field to obtain the target pattern of assembled cells^312,313. Various architectures of electrodes were studied, including 2D simple gaps, interdigital, quadruple, and concentric-radial-tip-arrays electrodes, showing their abilities in cell clustering and its potential in 3D cell patterning.

The effect of DEP in aggregating cell clusters for 3D culture in a gel system was investigated³¹⁴. Without a scaffold, dot electrodes induced an nDEP to repel cells from the electrode edges, directing them into circular aggregates. The encapsulations of yeast, HL-1, K562, and HeLa cells were formed in various gel types, remaining intact and viable after >7 days post-gelation. These 3D cell encapsulations exhibit different morphology and drug response profiles versus a conventional 2D culture platform.

DEP has been integrated into a lab-on-chip platform³¹⁵. Wireless bipolar electrode (BPE) arrays were embedded under a microfluidic channel to capture clusters of invasive breast cancer cells MDA-MB-231. MDA-MB-231 cells were attracted to BPE tips, obtaining an average of 4 cells per tip for the longest BPE at the highest applied voltage for 5 min. However, the stability and viability of the clusters after electric inhibition were not stated in the paper. Another approach for cell aggregates in LOC was conducted by Cottet et al.³¹⁶. The DEP system comprised 8 electrodes arranged in a circle, with the trapping position placed in the middle, allowing the assembly of cells within 1 min (Fig. 7e). By optimizing the electrode layout and the microfluidic channel design to optimize the pressure contribution as well as the trapping position, a compact aggregate of 21 cells was formed after 5 min.

Despite its advances in cell separation and enrichment, the application of DEP has not been well-investigated, partly because of its side effect, which is the Joule effect caused by the potential applied to electrodes. The heating generated by the electrode system can rapidly surpass 37°C, posing risks to the cell culture over a long period. The usage of DEP might be integrated with a temperature controlling system to cool down for more potential development of DEP in cell manipulation for lateral culture. Moreover, electric trapping primarily relies on the intrinsic dielectric contrast between cells and the suspending culture medium, which is minimal in many cases, limiting the effectiveness of cell manipulation. As a result, the system may exhibit reduced sensitivity, especially among cells sharing approximate physical characteristics. To achieve stronger responses typically requires higher electrical input, consequently resulting in increased energy consumption and facilitated Joule heating effects.

Having components similar to acoustic trapping with a strong dependence on the dielectric properties of cells, conductivity, and permittivity of the culture medium, and the geometry of the electrodes, reproducibility in electric trapping remains challenging and adoptable for specific cell types only. Even with the same cell type but having different sizes, the effect of manipulation is different, leading to difficulty in controlling a consistent outcome. To address these issues, standardized protocols for device fabrication, medium preparation, and electrical calibration are required, alongside clear reporting of experimental parameters. Such standardization would not only improve reproducibility across laboratories but also accelerate the translation of electric trapping techniques into scalable, energy-efficient platforms for reliable cell manipulation and organization.

Methods to validate spatial organization in MPSs

As microphysiological systems become increasingly complex, methodology to validate spatial organization must evolve in parallel, ensuring they reliably interpret spatial arrangements and the functional consequences these arrangements produce. Here, we highlight that regardless of complexity, methods to validate spatial arrangements in MPS must satisfy one of two conditions: they should either allow analysis while preserving spatial architecture or enable reconstruction of spatial patterns after destructive assays. In the following, we briefly outline representative techniques under each category, as a detailed methodological discussion is beyond the scope of this work. The first category explains itself, wherein analyses are done via direct visualization and correlation between observed spatial cues and biological responses. While, the second category involves recall-based strategies, where spatial information is recorded through molecular identifiers, such as spatial barcodes or in situ hybridization coordinates, that are later decoded computationally to reconstruct tissue organization. The latter approach enables a finer resolution of cellular identity and molecular activity than imaging alone, making it possible to correlate complex biological responses with their spatial origins. Together, these two approaches provide complementary perspectives: imaging-based validation confirms that the spatial pattern exists as designed, while recall-based molecular techniques allow the integration of spatial data with high-dimensional molecular readouts.

Methods to validate spatial arrangements in MPS while preserving spatial architecture involve imaging-based approaches. Imaging provides direct, non-destructive visualization of cell placement, morphology, and interaction within patterned constructs³¹⁷. Imaging can be performed on live cells or fixed samples. In live imaging, cells are genetically modified to express fluorescent reporters such as GFP, RFP, or mCherry, allowing dynamic monitoring of cell migration, differentiation, and lineage commitment³¹⁸. At the same time, these reporters can be coupled with biomolecules that respond to calcium flux, redox state, or mechanical stress to link cellular function to spatial context in real time^319–321. However, these methods are constrained by phototoxicity, limited imagine duration and relatively low number of molecular targets available. In which prevents extensive molecular characterization within complex spatial architecture³¹⁸. Fixed sample imaging through immunostaining or multiplexing, offers stronger spatial resolution and broader molecular profiling, albeit without the temporal dimension of live imaging³²². Multiplex imaging techniques such as cyclic immunofluorescence, Co-Detection by Indexing (CODEX), or Imagin Mass Cytometry (IMC) allow sequential staining an imaging of dozens of protein markers within the same tissue section^323–325. For example, a multiplexed analysis of an MPS model could simultaneously display endothelial markers (CD31, VE-Cadherin), stromal (CD90, vimentin), and inflammatory markers (CD68, ICAM-1) within the same construct. By preserving spatial relationships while expanding molecular coverage, multiplexing enables researchers to delineate tissue boundaries, detect gradients of signaling proteins and assess microenvironmental spatial hierarchy at once. Nonetheless, imaging methods remain limited in their capacity to provide high-dimensional molecular data, as labeling techniques primarily screenshot morphological and phenotypic features instead of comprehensive transcriptomic or proteomic patterns³²⁶.

A core technology for MPS imaging is microscopy, which enables capturing cell images carrying insightful information at a micro resolution²⁸. Light microscopy provides a clear visual at a single-cell level, allowing quantification of specific cellular phenotypes, measurement of cell-specific responses, and characterization of MPSs. Moreover, the noninvasive techniques allow imaging in situ while maintaining the spatial organization without suppressing cell proliferation. For MPS imaging, optical microscopes face several challenges due to the thickness of the 3D cell structure, which is a heterogeneous refractive index field of living cells and tissues, resulting in light scattering, wavefront aberrations, limited penetration depth, and high computational data burden³²⁷. However, there are several powerful commercial microscopy techniques that are commonly used for MPSs.

In general, microscopies in MPS imaging can be classified into two main types, 2D and 3D. Transmission light microscopy (TLM), working based on the absorption of light traveling through the sample and refracting without exogenous labeling and staining, is the most common 2D-based technique³²⁸. TLM allows clear imaging of thin and translucent specimens thanks to the inherent image contrast. While TLM may be partially or totally opaque when capturing over hundreds micrometer-thick samples, epifluorescence microscopy using excitation light and fluorescent labeling is possible to capture specific tissue with the support of non-fluorescent background computational deconvolution³²⁹. Confocal fluorescence microscopy (CFM) is the most popular 3D imaging technique used for MPS high-resolution fluorescence imaging. To achieve optical sectioning of thick specimens at a sub-micrometer resolution, confocal microscopes are equipped with a pinhole to filter the emitted light, removing all out-of-focus light caused by the thickness of the sample, and allowing the pass-through light to be only in-focus³³⁰. However, this approach is significantly hindered by penetration depth because of light scattering, high toxicity and photobleaching due to wasted out-of-focus light, shortening the long-term imaging dynamics. CFM relies on single-photon excitation, where fluorescence absorbs one high-energy photon, which is in the visible light range, to excite fluorophores; hence is limited to a depth of <100 µm²⁸. To allow a stronger penetration below the surface of the living sample, particularly >500 µm, longer-wavelength excitation, typically infrared, is utilized, requiring the simultaneous absorption of multiple photons³³¹. This approach is called multiphoton microscopy. The excitation of multiphoton microscopy is a nonlinear process requiring high photon flux to focus on the focal spot, suppressing the generation of out-of-focus light. Therefore, the phototoxicity and photobleaching are limited, lengthening the live imaging. Multiphoton microscopy can leverage the autofluorescence of cells and tissues, increasing the sampling depth and supporting the label-free imaging method. However, multiphoton microscopy has a more complex setup and a lower resolution compared to CFM. Finally, other than using a small focal point, light sheets are used for stacking 2D images of 3D MPSs, which is the approach of light sheet microscopy (LSM). Different subtypes of LSMs share the same traits of using light sheets for image acquisition rather than small focal points of CFM and multiphoton microscopy³³². This method provides faster imaging via simultaneous activity of many pixels, limiting light waste, resulting in low photodamage and bleaching of the sample. Moreover, there are several other microscopy techniques showed promising results for using with MPSs, including Raman microspectroscopy, optical tomography, correlative light and electron microscopy^28,327.

A new class of approaches to validate cell spatial organization in MPSs involves spatial omics, which combines molecular profiling and spatial localization to provide a comprehensive map of tissue organization at cellular and subcellular resolution. Spatial omics encompasses spatial transcriptomics, proteomics, and metabolomics, each offering complementary insights into cell interactions within microenvironments. Intuitively, spatial transcriptomics mapping can be achieved by imaging-based spatial transcriptomics approaches, such as MERFISH³³³ (Multiplexed Error-Robust Fluorescence in situ Hybridization) and seqFISH³³⁴ (Sequential Fluorescence In Situ Hybridization), hybridize fluorescently labeled probes to RNA targets, enabling the visualization of thousands of transcripts at single-cellular resolution. These approaches are often applied to tissue constructs, but their integration into MPS workflows is technically feasible and suitable for preserving 3D structures. Alternatively, spatial transcriptomics can be studied via sequencing-based methods, such as 10x Genomics Visium³³⁵, Slide-seq³³⁶, Stereo-seq³³⁷ (Spatial Enhanced Resolution Omics-sequencing), and DBiT-seq³³⁸ (Deterministic Barcoding in Tissue for spatial omics sequencing), where spatially barcoded oligonucleotides or microfluidics capture mRNA and other molecules for transcriptome-wide reconstruction of cell positions following sequencing. DBiT-seq, for example, utilizes orthogonal microfluidic channel flows to deposit spatial barcodes across tissue grids, producing high-resolution omics maps that simultaneously measure not only RNA but proteins and epigenetic features^338,339. Sequencing-based spatial transcriptomics is particularly suited for 2D MPSs due to planar geometry, however, these methods can still be applied to 3D structures by sectioning, for example through cryosectioning, enabling spatial profiling within complex tissues. For instance, Lozachmeur et al. utilized barcode-based spatial transcriptomics to analyze human cerebral organoids, achieving a more cost-effective analysis compared to conventional approaches³⁴⁰.

Similarly, spatial proteomics offers a means to visualize protein expression and localization with high multiplexity. Techniques such as Imaging Mass Cytometry (IMC)³⁴¹, Multiplexed ion beam imaging (MIBI)³⁴², and CODEX³⁴³ use iterative cycles of staining and imaging with metal-tagged antibodies or DNA barcodes, enabling simultaneous quantification of proteins across an entire MPS construct. More recently, spatial metabolomics has extended these principles to the small-molecule domain. Approaches such as matrix-assisted laser desorption/ionization (MALDI) imaging³⁴⁴ or desorption electrospray ionization (DESI)³⁴⁵ provide spatially resolved chemical profiles of metabolites, lipids, and ions within microtissues. MALDI, for example, involves applying a matrix to the tissue to assist laser desorption and ionization of metabolites, which can later be combined with Mass Spectrometry Imaging (MSI) for metabolite detections³⁴⁴. Similar to spatial transcriptomics, most of these methods have primarily been applied to thick tissues or biopsy samples, but their implementation in MPSs is highly feasible. A key limitation of these approaches is their incompatibility with live temporal tracking, as they typically require fixation or disruption of cellular structures and activities, effectively halting the dynamic processes within the MPS. Consequently, there remains a fundamental trade-off between spatial resolution and temporal monitoring: validating spatial organization at high molecular and single-cell resolution necessitates stopping temporal observations of the same sample. Despite this limitation, spatial omics technologies represent a powerful new generation of analytical tools capable of providing high-dimensional molecular validation of spatial architecture, offering critical insights into complex cellular organizations within MPSs.

Beyond validating cellular spatial organization, mapping the microenvironmental conditions surrounding cells is equally crucial, as discussed in the passive methods to generate spatial patterns in MPSs. Oxygen gradients, for instance, can be quantitatively mapped using luminescent probes³⁴⁶, fluorescence probes³⁴⁷, or integrated optical oxygen sensors³⁴⁸. Real-time tracking of such gradients allows confirmation that the MPS successfully mimics physiological oxygen tensions, such as the hypoxic core of tumor spheroids or oxygen gradient in intestinal lumen. Similarly, nutrient and pH gradients can be tracked using optical or electrochemical sensors^349–351, enabling correlation between spatial nutrient distribution and metabolic responses. In parallel, the mechanical landscape of an MPS can be validated using techniques like atomic force microscopy (AFM), traction force microscopy (TFM)³⁵², or particle-tracking microrheology³⁵³. TFM or other microrheology techniques enable continuous, real-time monitoring of the extracellular matrix’s mechanical properties alongside dynamic cell movements, providing interactive insights into the reciprocal mechanical interplay between cells and their surrounding microenvironment. Additionally, computation models, such as finite element simulations or computational fluid dynamics (CFD), can be integrated with experimental measurements to predict and validate spatial variations in shear stress, pressure, or solute diffusion across the device. Together, these approaches extend the validation framework beyond cellular positioning to the entire biophysical and biochemical microenvironment.

Conclusions and Future Perspectives

Spatial organization is central to the physiological fidelity of MPSs, affecting how cells behave, interact, and function in engineered environments. Strategies developed to control cellular positioning can be classified into two principal groups – direct and indirect patterning – each offers distinct advantages and serves different experimental needs. Direct patterning techniques, such as bioprinting, microfluidic compartmentalization, and physical trapping, enable precise placement of cells or structures at predefined locations. These methods are particularly powerful when architectural fidelity is critical, such as in barrier models, vascularized constructs, or when establishing interfaces between different tissues. They are also well-suited for applications requiring reproducibility and standardization, such as drug screening platforms or disease models where structural constraints influence function, as they can bypass the potential biological unpredictability during cellular self-organization that indirect methods rely on.

In contrast to direct patterning, indirect methods utilized environmental cues, including ECM composition, mechanical gradients, topographical features, and soluble factor gradients, to guide cell growth, differentiation, and movement. These approaches better mimic the natural developmental processes by allowing cells to interpret and respond to cues in a context-dependent manner. Indirect patterning is advantageous in scenarios where plasticity and long-term remodeling are desired, such as in stem cell differentiation, morphogenesis, or tissue maturation studies. Since these systems harness intrinsic cell behaviors, they often yield dynamic, emergent architectures that evolve over time, closely resembling in vivo processes. Fabrication methods such as 3D printing of ECM, or microfluidic generation of solute gradients, allow controlling the environmental cues with high precision. However, as mentioned, the outcomes can be less predictable, as the patterning precision also relies on biological behaviors of the seeded cells/organoids. In highly complex systems, the degree of spatial control may be limited compared to direct methods. A summary of some spatially patterned MPSs with their patterning strategies is listed in Table 1.

Table 1.

Microphysiological systems with their spatial patterning strategies

Organ(s)/Tissue(s)/Structure(s)		Cell source/Cell lines	Patterning strategies		References
Single cell type	Heart/Vasculature	Primary Cardiomyocytes	Indirect	Substrate nanotopography and rigidity	7
		HUVECs and HBVPCs/UASMCs	Direct	Microfluidic network	22
		NMVCMs	Direct	Light-based bioprinting	257
		HUVECs	Direct	Conjugated signal	184
		Isolated cardiac fibroblasts	Direct	Standing surface acoustic waves	290
	Brain/Neuron Network	PC12	Indirect	Substrate micro/nano-topography	8
		PC12 or dorsal root ganglia cells	Indirect	Substrate alignment	11
		hiPSCs	Direct and indirect	Scaffold structure and topology	18
		hPSCs	Direct	Microfluidic parallel channels Assembloid formation	19
		NSCs	Indirect	Growth factor gradient	219
		NPCs	Direct	Magnetite tissue engineering	302
		Ps1	Indirect	Substrate microtopography	234
	Muscle	Isolated muscle mononuclear cells	Indirect	Substrate nanotopography and conjugation of ligand	9
		vSMCs	Direct and Indirect	Extrusion-based bioprinting Matrix alignment	109
	Bone/Cartilage	hMSCs	Indirect	Gradient of conjugated domain	16
		hTMSCs	Indirect	Matrix structure and composition	357
		MC3T3	Indirect	Substrate mechanical gradient	226
		MSCs	Direct	Light-based bioprinting	259
		Bone MSCs	Direct	Magnetic nanocomposite hydrogel scaffold	305
		Bone MSCs	Direct	Magnetite tissue engineering	300
	Intestine	HCT-116	Indirect	Drug gradient	220
		HCT-116	Direct	Acoustic hologram	292
	Liver	HepG2	Indirect	Oxygen gradient	86
		hESCs	Indirect	Oxygen gradient	87
	Kidney	HEK	Direct	Dielectrophoresis	316
	Fibroblast	NIH 3T3	Indirect	Substrate nanotopography	12
		NIH 3T3	Indirect	Substrate alignment and chemical composition	13
	Immune cells	THP-1	Indirect	Gradient of conjugated signal	204
		T cells	Direct	Dielectrophoresis	313
	Embryo/Germ layers	hPSCs	Direct	Photonic-crystal optical tweezers	275
		hPSCs	Indirect	Morphogen gradient	354
	Tumors	A431	Indirect	Matrix viscoelasticity and solute gradient	15
		MCF7 or MDA-MD-231	Indirect	Nutrient gradient	221
		MDA-MB-231	Direct	Bipolar dielectrophoresis	315
Complex tissues	Heart/Vasculature	H9c2, NOR-10 cells	Indirect	Substrate microtopography	10
		HUVECs, VICs	Direct	Magnetic bioprinting	262
		HUVECs, ASCs	Direct	Droplet bioprinting	104
		HUVECs, HCASMCs, HDFs	Indirect	Coaxial printing	255
		HUVECs, hASCs	Direct	High-resolution acoustophoretic	287
	Brain	N2A, BV-2	Direct	Acoustofluidics	284
		Isolated neurons	Direct and Indirect	Reconfigurable microchannels Matrix alignment	239
	Bone	MSCs, HUVECs	Direct	Magnetically labeled cells and magnetic scaffold	303
	Intestine	hESCs, hiPSCs	Indirect	Matrix composition	190
		Organoids	Indirect	Ligand patterning	206
		Organoids	Indirect	Morphogen gradient	207
	Liver	HepG2, HUVECs	Direct	Extrusion-based bioprinting	253
		HepaRG, HUVECs	Direct	Extrusion-based bioprinting	111
	Fibroblast	hDFs	Indirect	Substrate porosity	240
	Heterotypic cell spheroids	2H11, NIH 3T3, 293FT, EO771	Direct	Acoustofluidics	283
Barrier models	Blood – Gut – Barrier	Caco2, HUVECs	Direct	Parallel microfluidic channels	124
	Blood – Brain – Barrier	iCell Endothelial Cells, HBMECs, PCs, ACs	Direct	Parallel microfluidic channels	125
	Alveolar barrier	NCI-H1703, NCI-H441, MRC5, HULEC-5a	Indirect	Inkjet bioprinting	252
Tumor - vasculature	Breast Tumor - Vasculature	MCF10A, MCF7, MDA-MB-231, ECFC	Indirect	Substrate patterning	161
	Pancreatic Tumor - Vasculature	PDAC, HUVECs	Direct	Parallel microfluidic channels	248
Multi-organ	Liver - Kidney	HepG2, RPTEC	Direct	Serial microfluidic perfusion	128
	Gut–Liver	Primary hepatocytes, Caco2	Direct	Serial microfluidic perfusion	129
		Caco2, HepG2	Direct	Serial microfluidic perfusion	246
	Gut–Brain	Caco2, iNSCs	Direct	Parallel microfluidic channels	249
		Caco2, bEnd.3, hBMECs	Direct	Parallel microfluidic channels	250
	Heart – Liver	Organoids	Direct	Parallel microfluidic channels	251
	Pancreas – Muscle – Liver	C2C12, INS-1	Direct	Parallel microfluidic channels	130
	Intestine – Liver – Skin – Kidney	HepaRG, RPTEC/TERT-1, Reconstructed human small intestinal barrier, Human juvenile prepuce	Direct	Serial microfluidic perfusion	131
	Heart – Liver – Bone – Skin	hiPSCs	Direct	Shared vascular channel	244
	Brain – Liver – Intestine – Kidney	hiPSCs	Direct	Parallel microfluidic network	247
	Eighteen organs		Direct	Microfluidic network	245
Host-Microbes	Gut - Microbes	Caco2 E. coli and LGG	Indirect	Species competition and flow oscillation	132
		Isolated gut tissue Gut microbiota	Indirect	Nutrient and O2 gradient	145
	Tumor - Microbes	NCI-H460 E. coli	Indirect	Bacterial chemotaxis	146

Rather than viewing these two approaches as mutually exclusive, the future of spatial organization in MPSs lies in strategic integration. Combining the deterministic precision of direct patterning with the biological responsiveness of indirect cues can produce hybrid systems that are both structurally robust and functionally dynamic. One prominent example is the 3D bioprinting techniques, which not only allow direct positioning of printed cells inside the matrices, but also the properties of the matrices^109,258. Printed cell arrays embedded in smart hydrogels can reorganize in response to enzymatic remodeling, changing their position in a more controlled manner. Microfluidic compartmentalization can also be overlaid with soluble gradients to modulate cell fate post-placement. In fact, this integration is often inherent to microfluidic systems, where secreted factors from one compartment naturally influence the behavior of neighboring regions. Manfrin et al. introduced a gradient of morphogenetic protein 4 (BMP4) to induce germ layer patterning from hPSC colonies, which were cultured in sandwiched layers of gel channels and cell chamber³⁵⁴. The same approach can be applied to more complex, multi-organ systems in other microfluidic platforms. The next challenge lies in identifying the key signaling molecules involved and refining microfluidic designs to establish controlled and physiologically relevant gradient profiles. Overall, future work should focus on developing adaptive platforms that allow spatiotemporal control, where spatial patterns can evolve alongside tissue function. Such systems would not only enhance physiological relevance for the studies development, regeneration, and disease progression in vitro.

While integrating direct and indirect spatial control in MPSs offers clear advantages, it also introduces significant biological challenges. Aligning precise structural placement with dynamic, biologically driven behaviors is inherently complex. Cells positioned through direct methods may migrate, differentiate, or remodel their environment in ways that conflict with the intended architecture, especially when exposed to gradients or responsive matrices. Conversely, over-constraining the system may limit necessary plasticity for tissue development or maturation. Material selection becomes critical, as substrates must be simultaneously supportive of defined patterning and permissive to cell-driven changes. Moreover, indirect cues such as morphogen gradients can be disrupted by cell-secreted factors or tissue remodeling, altering spatial fidelity over time. These interdependencies make system design, tuning, and interpretation more difficult, particularly in long-term or multi-tissue platforms.

Moreover, despite the promising outlook of integrating direct and indirect spatial patterning strategies in MPSs, several technical challenges must be critically addressed to transition these systems from experimental setups toward practical and scalable platforms. Scalability remains a key bottleneck; for instance, while 3D bioprinting has demonstrated considerable progress in reproducibly fabricating tissue constructs with defined architecture, limitations in printing speed, resolution, and bioink formulation constrain the throughput necessary for large-scale applications such as drug screening or personalized medicine. Microfluidic compartmentalization, similarly, offers high precision and compartmental isolation but faces challenges in scaling device complexity and robustness for routine use. On the other hand, indirect methods relying on biochemical and mechanical cues, such as morphogen gradients or ECM remodeling, are still predominantly experimental due to their inherent biological variability and difficulty in standardizing environmental conditions across batches. Controlling heterogeneity in cell responses within these dynamic environments remains a significant obstacle to reproducibility and predictive modeling.

To address these limitations, coupling MPS platforms with advanced high-throughput imaging modalities can enable continuous, non-invasive monitoring of spatial organization, cell behavior, and matrix remodeling in real-time. Techniques such as live-cell fluorescence imaging, optical coherence tomography, or label-free imaging could provide rich spatiotemporal datasets that capture both expected and emergent phenomena within complex patterned systems. When combined with machine learning algorithms capable of processing large-scale imaging data, this approach could facilitate automated identification of patterning inconsistencies, prediction of cellular dynamics, and optimization of system parameters. For example, supervised learning models trained on imaging data could predict how subtle changes in gradient profiles or matrix stiffness affect cellular migration or differentiation outcomes, guiding iterative refinement without extensive trial-and-error experimentation.

Along with the development of methods for cell spatial organization, ethical and regulatory considerations are also addressed as challenges for practical MPS applications in general^355,356. As in other biomedical research utilizing human cells, especially stem cells and patient-derived cells, ethics have been respected for donors and patients, and general concerns about scientific integrity. Professional guidelines have been generally highly regulated for cells and stem cells, such as ISSCR Guidelines for Stem Cell Research and Clinical Translation. However, the spatial organization of cells in MPS ethical and regulatory systems extends far beyond the traditional cell culture research. Since these systems are mimicking organ-like functions and mechanisms, cells undergo many manipulations, and a study can contain various cells from different origins; hence, the ethical and regulatory systems differ and should require more careful consideration than existing frameworks. First, at the individual level, informed consent must be reconsidered as donor cells may be used in long-term and unforeseen applications. At the collective level, regulatory systems must ensure transparency, fair access, and responsible communication of MPS potential, focusing on the role of replacing animal models and advancing precision medicine. And lastly, at the entity level, complex MPS systems involving the brain and embryo raise questions about moral status, which is understandable since research relating to reproductive systems such as embryos has been facing controversies since the beginning of human stem cell research. Together, these concern highlights the need for an ethics and adaptive regulation specific for bioengineering to instruct the responsible design and use of cell-based spatially organized systems.

Ultimately, we think spatial organization is not a technical addition to MPSs but a biological necessity. As the field continues to advance, refining and integrating patterning strategies will be key to building bio-microsystems that accurately recapitulate the complexity of human tissues.

Author contributions

Hung Dong Truong: conceptualization; investigation; validation; writing – original draft. Zhixing Ge: conceptualization; investigation; validation; writing – original draft. Elgene Chng Junyuan: investigation; validation; writing – original draft. Y-Van Tran: investigation; validation; writing – original draft. Yusheng Zhang: investigation; validation; writing – original draft. Chwee Teck Lim: conceptualization; project administration; supervision; writing – review and editing.

Competing interests

The authors declare no conflict of interest.