Microengineered bone models: advances and applications of bone-on-a-chip technology

Abstract

Bone-on-a-Chip (BoC) platforms are emerging microphysiological systems designed to recapitulate the complex cellular, structural, and mechanical microenvironments of human bone. These microfluidic-based models provide enhanced physiological relevance for studying bone remodeling dynamics, disease progression, and drug responses compared to traditional 2D cultures. This review discusses the anatomical and pathophysiological features of bone that must be replicated in vitro, highlighting recent advances in BoC platforms and their applications in regenerative medicine, drug discovery, and personalized therapy. However, despite their potential, current BoC technologies still face significant challenges, including the simplified representation of the extracellular matrix, incomplete recapitulation of the dense osteocyte network, and partial reproduction of native biomechanical stimuli. Furthermore, integrating vascularization, immune components, and ensuring scalability remain critical hurdles. Future directions focus on developing multi-tissue interfaces, advanced biomaterials, real-time sensing technologies, and AI-assisted data analysis. While still evolving, BoC technology serves as a promising tool for bridging the gap between in vitro models and clinical applications in translational bone research.

Graphical Abstract

Introduction

Bone is a dynamic, multifunctional tissue that performs critical physiological roles including structural support, mineral homeostasis, hematopoiesis, and endocrine regulation. Its hierarchical organization spanning from nanoscale hydroxyapatite (HAp) crystals and collagen fibrils to microscale osteons and macroscale cortical and trabecular architectures creates a uniquely complex microenvironment that remains challenging to replicate in vitro [1–3]. Traditional experimental approaches, including two-dimensional cell cultures and animal models, have substantially advanced our understanding of bone biology and pathology. However, these systems exhibit fundamental limitations: monolayer cultures fail to recapitulate the three-dimensional organization, mechanical environment, and multicellular interactions inherent to native bone, while animal models often demonstrate species-specific responses that do not translate reliably to human physiology [4–6]. These constraints have driven the development of more sophisticated in vitro platforms, among which BoC systems have emerged as particularly promising tools for bridging the gap between simplified cell cultures and complex in vivo conditions.

Anatomically, bone tissue is classified into two principal types: cortical (compact) bone and trabecular (cancellous) bone (Fig. 1) [7]. Cortical bone constitutes the dense outer shell of bones, providing mechanical strength to resist bending and torsional forces, while trabecular bone forms a porous, interconnected network that optimizes strength-to-weight ratio and houses the bone marrow microenvironment. This trabecular architecture predominates in vertebral bodies and the metaphyseal regions of long bones, where it facilitates metabolic exchange and accommodates hematopoietic tissue [8]. At the microscopic level, both bone types share a common cellular and matrix composition, comprising specialized bone cells embedded within a mineralized extracellular matrix (ECM) rich in type I collagen and HAp crystals [9, 10].

Fig. 1.

Bone anatomy. This figure illustrates the key structural components of bone, including the compact bone (dense outer layer) and spongy bone (porous inner layer) in the metaphysis. The marrow cavity, which houses hematopoietic tissue, is also depicted. The critical vascular supply is provided by the nutrient artery, metaphyseal arteries, and periosteal arteries, while the periosteum covers the outer bone surface. The osteon structure, featuring concentric lamellae and embedded osteocytes, is highlighted. Created with BioRender.com

The cellular constituents of bone (osteoblasts, osteocytes, and osteoclasts) function not as isolated entities but as an integrated network whose coordinated activities govern bone formation, maintenance, and resorption. Osteoblasts, derived from mesenchymal stem cells, are responsible for synthesizing and mineralizing the bone matrix. Upon becoming embedded within the matrix they produce, osteoblasts differentiate into osteocytes, which constitute over 90% of all bone cells and serve as the principal mechanosensors and orchestrators of bone remodeling [11, 12]. Osteocytes reside within lacunae and extend dendritic processes through canaliculi, forming an extensive lacuno-canalicular network that enables intercellular communication and nutrient exchange. Through this network, osteocytes detect mechanical loading and translate these signals into biochemical responses, secreting factors such as sclerostin, RANKL, and osteoprotegerin (OPG) that regulate osteoblast and osteoclast activity [11, 12].

Osteoclasts, multinucleated cells of hematopoietic origin, are responsible for bone resorption. The dynamic equilibrium between osteoclast-mediated resorption and osteoblast-mediated formation constitutes the bone remodeling cycle a tightly regulated process essential for maintaining skeletal integrity, repairing microdamage, and modulating calcium-phosphate homeostasis. This cellular interplay is modulated by systemic hormones including parathyroid hormone (PTH), vitamin D, calcitonin, and sex steroids, as well as locally produced cytokines and growth factors [13, 14]. Critically, the bone marrow niche within trabecular bone serves dual functions as both a hematopoietic organ and an immunological hub, supporting blood cell development and mediating immune-skeletal crosstalk. Understanding these complex intercellular dynamics is essential for developing physiologically relevant in vitro models, as disruption of any component of this network can precipitate pathological conditions [15, 16].

The intricate balance of bone cell interactions renders the skeleton vulnerable to diverse pathological conditions, each presenting unique challenges for experimental modeling. Osteoporosis exemplifies this complexity: characterized by reduced bone density and microarchitectural deterioration, this condition arises from an imbalance favoring osteoclast activity over osteoblast-mediated formation. Effective in vitro modeling of osteoporosis therefore requires platforms capable of recapitulating the dynamic interplay between bone-forming and bone-resorbing cells under conditions that permit quantification of remodeling imbalances a capability largely absent in conventional culture systems [17, 18]. Similarly, bone metastasis a frequent and devastating complication of breast, prostate, and lung cancers involves intricate tumor-bone cell interactions within the bone microenvironment. Metastatic cancer cells co-opt the normal bone remodeling machinery, establishing a “vicious cycle” in which tumor-derived factors stimulate osteoclastic resorption, releasing matrix-bound growth factors that further promote tumor growth [19, 20]. Modeling this pathological cascade requires systems that integrate tumor cells with functional bone tissue in a manner that preserves physiological cell-cell and cell-matrix interactions.

Fracture healing represents another process inadequately captured by traditional models. While bone possesses remarkable regenerative capacity, complications including delayed union and non-union affect a significant proportion of fractures, particularly in elderly or metabolically compromised patients. The healing process involves a coordinated sequence of inflammation, soft callus formation, hard callus formation, and remodeling each phase involving distinct cellular players and signaling pathways that are difficult to recapitulate in static culture conditions [21]. Additional pathological contexts, including osteoarthritis (where subchondral bone remodeling contributes to joint degeneration) [22] and osteomyelitis (where limited vascularization and biofilm formation complicate infection management) [23], further underscore the need for advanced experimental platforms capable of modeling the spatial heterogeneity, cellular diversity, and dynamic processes characteristic of bone tissue in health and disease.

BoC technologies have emerged as a promising approach to address these modeling challenges. By integrating microfluidics, biomaterials engineering, and advanced cell culture techniques, BoC platforms offer the potential to recapitulate key features of the native bone microenvironment within controlled, miniaturized systems. These platforms enable precise control over mechanical stimuli, fluid flow, oxygen gradients, and cellular organization parameters that are difficult or impossible to modulate in conventional culture systems [24, 25]. However, significant engineering challenges remain. Native bone experiences complex, cyclic, multi-directional mechanical loading (compressive, tensile, and shear stresses), whereas current BoC platforms predominantly employ fluid shear stress as a mechanical stimulus [26]. Faithfully recreating the bone matrix microenvironment presents additional hurdles: native bone ECM comprises a nanoscale calcified matrix with type I collagen and over 200 non-collagenous proteins, whereas most in vitro systems rely on pre-formed ceramic scaffolds (e.g., HAp, β-tricalcium phosphate) that lack the dynamic remodeling properties of natural bone [27, 28].

Despite these limitations, BoC technology has rapidly expanded to encompass applications in osteoporosis research, bone metastasis modeling, fracture healing studies, and pharmaceutical screening. The ability to incorporate patient-derived cells positions these platforms as valuable tools for personalized medicine, enabling individualized assessment of therapeutic responses while reducing reliance on animal models. Concurrent advances in three-dimensional bioprinting, organoid engineering, and artificial intelligence further enhance the fidelity, scalability, and analytical capabilities of BoC systems [29–31].

This narrative review provides a comprehensive examination of the current state of BoC technology, with emphasis on the biological principles underlying platform design, recent technological advances, and emerging clinical applications. We systematically evaluate the strengths and limitations of existing platforms, discuss integration of vasculature, immune components, and mechanical loading systems, and highlight future directions that position BoC technology as a transformative tool for orthopedic research and regenerative medicine.

Organ on a chip technology

Organ-on-a-chip (OoC) technology emerged from the convergence of microfluidics, microfabrication, and cell biology over several decades. The foundational principles of microfluidics precise manipulation of fluids at micro- and nanoscale levels were established through advances in photolithography and microelectromechanical systems (MEMS) during the 1980s [32–35]. The introduction of soft lithography using polydimethylsiloxane (PDMS) by Whitesides and colleagues in the mid-1990s dramatically reduced fabrication barriers, enabling widespread adoption of microfluidic platforms in biological research [36, 37]. The conceptualization of “lab-on-a-chip” systems during this period laid the groundwork for integrating multiple laboratory functions onto miniaturized devices [38, 39]. The landmark development of the “lung-on-a-chip” by Ingber and colleagues at the Wyss Institute in 2010 demonstrated for the first time the capacity to model organ-level physiological functions including mechanical breathing movements and inflammatory responses on a microfluidic platform, thereby establishing the modern OoC paradigm [40, 41]. Subsequent advances in three-dimensional (3D) bioprinting, novel biomaterials, and sensor integration have continued to expand the complexity and functionality of these systems (Fig. 2) [42].

Fig. 2.

Evolution of organ-on-a-chip technology. This timeline illustrates key advancements in organ-on-a-chip systems, from early developments in microfluidics (1950–1960 s) to modern innovations incorporating AI and IoT-based analytics. Created with BioRender.com

OoC platforms represent a paradigm shift in biomedical research by replicating the structural, mechanical, and physiological characteristics of human tissues within precisely controlled microenvironments. Several core features distinguish these systems from conventional in vitro models and underpin their utility for disease modeling and drug development. Unlike traditional two-dimensional (2D) monolayer cultures, OoC devices support 3D cellular organization within architecturally defined compartments that recapitulate the spatial relationships present in native tissues [43, 44]. This compartmentalization enables the co-culture of multiple cell types in anatomically relevant configurations for example, separating epithelial and endothelial layers by a porous membrane or positioning stromal cells within a hydrogel matrix adjacent to parenchymal cells. For bone applications, this feature is essential for modeling the spatial organization of the bone marrow niche, the osteocyte lacuno-canalicular network, and the interfaces between bone, vasculature, and surrounding tissues [44, 45].

Continuous or pulsatile fluid flow through microfluidic channels provides controlled delivery of nutrients, oxygen, and soluble factors while simultaneously removing metabolic waste products [46]. This perfusion mimics the interstitial fluid flow present in vascularized tissues and enables the establishment of physiologically relevant concentration gradients. In bone tissue, interstitial fluid flow through the lacuno-canalicular network is a critical mediator of mechanotransduction; OoC systems can partially recapitulate this feature through controlled flow rates that generate defined shear stress profiles on cultured cells [47, 48].

Many OoC platforms incorporate mechanisms for applying mechanical forces to cultured tissues, including cyclic stretch, compression, and fluid shear stress [49, 50]. These features are particularly relevant for mechanosensitive tissues such as bone, where mechanical loading is fundamental to homeostasis and remodeling. Although current BoC systems predominantly utilize fluid shear stress, emerging designs incorporate flexible membranes, pneumatic actuators, or external loading devices to apply more physiologically relevant compressive and tensile forces [51–53].

OoC devices permit precise manipulation of microenvironmental variables that influence cell behavior, including oxygen tension, pH, temperature, and biochemical gradients. This control is essential for modeling pathological conditions such as hypoxia in the tumor microenvironment or the oxygen gradients present within the bone marrow niche. Additionally, the ability to modulate these parameters dynamically enables investigation of cellular responses to acute or chronic environmental changes [51–53].

Advanced OoC platforms incorporate embedded sensors for continuous, non-invasive monitoring of cellular responses, including transepithelial electrical resistance (TEER), oxygen consumption, pH changes, and secreted biomarkers [54, 55]. Integration with imaging modalities enables real-time visualization of cell morphology, migration, and tissue organization. These capabilities facilitate longitudinal studies and provide rapid feedback during drug screening applications. The modular design of OoC systems allows interconnection of multiple organ compartments to model systemic physiology and inter-organ communication. “Body-on-a-chip” or “human-on-a-chip” configurations linking, for example, gut, liver, and kidney compartments enable investigation of drug pharmacokinetics, first-pass metabolism, and multi-organ toxicity. For bone research, integration with vascular, immune, or tumor compartments is particularly valuable for studying bone metastasis, osteoimmunology, and systemic bone diseases [56–59].

OoC platforms readily accommodate patient-derived cells, including primary cells and induced pluripotent stem cell (iPSC)-derived lineages. This compatibility enables personalized disease modeling and individualized drug response assessment capabilities of particular importance for heterogeneous conditions such as bone cancers or genetic skeletal disorders where patient-specific factors significantly influence therapeutic outcomes [60, 61]. The integrated features of OoC systems confer substantial advantages over traditional experimental approaches (Fig. 3). From a practical perspective, OoC systems offer reduced costs, shortened experimental timelines, and diminished ethical concerns compared with animal studies. Their miniaturized format enables high-throughput screening with minimal reagent consumption, while real-time monitoring capabilities accelerate identification of therapeutic effects or adverse responses during early-stage drug development [51, 62].

Fig. 3.

Comparative analysis of traditional cell culture, animal models, and lab-on-a-chip systems. The comparison is based on key experimental features and capabilities. Lab-on-a-chip platforms demonstrate superior performance in physiological relevance, environmental control, and personalized medicine compatibility, while traditional methods retain certain regulatory advantages. (✓ = advantage; ✗ = limitation). Created with BioRender.com

The core features of OoC technology are particularly well-suited to addressing the challenges inherent in bone tissue modeling. The capacity for 3D compartmentalization enables recreation of the complex spatial relationships between bone-forming cells, bone-resorbing cells, and the bone marrow microenvironment. Dynamic perfusion systems can simulate interstitial fluid flow critical for osteocyte mechanosensing, while emerging mechanical loading modules offer the potential to apply the compressive and tensile forces that drive bone adaptation [62]. Multi-organ integration facilitates investigation of bone interactions with vascular, immune, and tumor compartments essential for modeling conditions such as bone metastasis and osteoimmune disorders [63]. Finally, compatibility with patient-derived cells positions BoC platforms as promising tools for personalized assessment of skeletal diseases and therapeutic interventions [64].

Despite these advantages, significant challenges remain in standardization, manufacturing scalability, and regulatory acceptance of OoC-derived data. Nevertheless, the potential of these systems to transform translational bone research and precision therapeutics is substantial, as detailed in subsequent sections of this review.

Fabrication strategies for microfluidic bone-on-a-chip platforms

The development of BoC systems necessitates precise microfabrication techniques capable of replicating the hierarchical architecture, mechanical environment, and biological complexity of native bone tissue. Unlike many other organ-on-a-chip applications, BoC platforms must accommodate the unique requirements of bone modeling, including integration of mineralized or mineralizable matrices, application of physiologically relevant mechanical loads, recreation of the bone marrow niche, and establishment of functional interfaces between bone, vasculature, and surrounding tissue compartments [63–66]. These requirements impose specific constraints on material selection, geometric design, and fabrication methodology that must be carefully considered during platform development.

Photolithography and soft lithography

Photolithography and soft lithography represent complementary stages of a unified fabrication workflow that has become the predominant approach for academic BoC research. In this integrated process, photolithography is first employed to create high-resolution master molds, which subsequently serve as templates for replica molding in elastomeric materials through soft lithography [67, 68]. The photolithographic process begins with spin-coating a photosensitive resist, most commonly SU-8 epoxy, onto a silicon or glass wafer to a defined thickness corresponding to the desired channel depth. Ultraviolet light exposure through a patterned photomask selectively cross-links the resist, and subsequent development removes unexposed regions to yield a positive-relief master mold with sub-micron resolution and exceptional dimensional reproducibility. For BoC applications, these masters are typically designed with channel dimensions reflecting the physiological scales of bone tissue. Primary culture channels for osteogenic constructs generally range from 200 to 1000 μm in width and 100 to 500 μm in height, dimensions sufficient to accommodate three-dimensional cell-laden hydrogels or scaffold materials [69–71]. Smaller channels of 50 to 200 μm diameter are employed to create vascular networks mimicking bone capillaries and sinusoids, while interconnecting microchannels of 10 to 50 μm can approximate the lacuno-canalicular dimensions relevant to osteocyte communication. When compartments of different vertical dimensions are required within a single device, multi-height masters can be fabricated through sequential spin-coating and exposure steps, albeit with increased process complexity [69–71].

The soft lithography phase utilizes the photolithographically fabricated master as a template for rapid, low-cost replication in poly(dimethylsiloxane) (PDMS), the elastomer that has dominated microfluidic cell culture applications [72, 73]. The PDMS prepolymer is cast over the master, thermally cured at 60–80 °C, and demolded to yield a negative replica containing the microfluidic channel network. Access ports are created using biopsy punches, and the patterned PDMS slab is bonded to a glass substrate or another PDMS layer via oxygen plasma activation, creating enclosed microchannels ready for cell culture [74, 75]. This approach offers numerous advantages for BoC development. Multiple devices can be replicated from a single master within hours, facilitating the iterative design optimization essential during early-stage platform development. The optical transparency of PDMS enables real-time microscopic observation of cell behavior, matrix deposition, and mineralization processes critical to bone studies. Gas permeability allows adequate oxygen and carbon dioxide exchange to support cell viability without requiring complex external gas supply systems, while the established biocompatibility of PDMS supports long-term culture of bone cells including osteoblasts, osteocytes, and osteoclasts [76, 77]. Additionally, thin PDMS membranes of 10 to 50 μm thickness can be incorporated between channel layers to enable mechanical stretching for mechanotransduction studies or to separate tissue compartments while permitting molecular transport. These capabilities have enabled successful development of perfusable channels lined with osteogenic cells, compartmentalized bone marrow niches, and mechanically actuated platforms for investigating osteocyte responses to fluid shear stress [78–80].

However, PDMS also presents notable limitations that must be considered for BoC applications. The material absorbs hydrophobic compounds, potentially confounding drug screening studies involving lipophilic therapeutics such as bisphosphonates or certain chemotherapeutic agents commonly used in bone cancer treatment [81, 82]. The low elastic modulus of PDMS, approximately 1 to 2 MPa, falls far below the stiffness of mineralized bone matrix at 10 to 20 GPa, necessitating incorporation of stiffer scaffold materials within the culture channels to provide appropriate mechanical cues to bone cells [83, 84]. Furthermore, feature sizes below approximately 10 to 20 μm become challenging to replicate reliably due to aspect ratio limitations and demolding difficulties, restricting the ability to recreate the finest details of bone microarchitecture [78–80].

Additive manufacturing

Additive manufacturing has emerged as an increasingly important fabrication approach for BoC devices, offering capabilities that complement and, in some cases, surpass those of conventional lithographic methods. These techniques construct devices directly from computer-aided design files through layer-by-layer material deposition or polymerization, with stereolithography, digital light processing, and two-photon polymerization representing the most relevant modalities for microfluidic applications [85]. Stereolithography and digital light processing systems utilize photopolymerizable resins that are selectively cured by ultraviolet or visible light, achieving feature resolutions of 25 to 100 μm for standard commercial systems and approaching 10 μm for specialized high-resolution platforms. The principal advantage of these techniques for BoC development lies in their capacity for true three-dimensional fabrication. Unlike planar lithography, which is fundamentally limited to extruded two-dimensional patterns, additive manufacturing enables creation of complex architectures including helical channels, branching vascular networks, and trabecular-like porous structures that more closely approximate native bone geometry [86–89]. Bone-mimetic scaffolds with controlled porosity, featuring pore sizes of 100 to 500 μm and overall porosities of 50 to 80%, can be directly printed and integrated with microfluidic channels, eliminating the need for separate scaffold fabrication and incorporation steps [90, 91]. Recent studies have exploited these capabilities to create BoC devices with trabecular-inspired architectures for bone marrow modeling, interconnected pore networks for investigating cell migration during bone regeneration, and even patient-specific geometries derived from micro-computed tomography imaging data [92–94].

Two-photon polymerization represents a specialized additive technique that achieves sub-micron resolution by exploiting nonlinear optical absorption, enabling fabrication of structures at scales approaching the lacuno-canalicular network with canalicular diameters of approximately 0.5 to 1 μm. This technique has been employed to create osteocyte-scale microenvironments for studying mechanosensing at physiologically relevant dimensions that are impossible to achieve with other fabrication methods. However, two-photon polymerization remains limited by slow writing speeds, small build volumes typically less than one cubic millimeter, and high equipment costs, restricting its application primarily to fundamental mechanistic studies rather than high-throughput device production [95–99].

The limitations of current additive manufacturing for BoC must also be acknowledged. Layer-by-layer fabrication produces surface irregularities with roughness values typically ranging from 1 to 10 μm that may affect laminar flow profiles and cell attachment behavior [100, 101]. Many photopolymer resins contain cytotoxic photoinitiators or unreacted monomers that require thorough post-processing including solvent washing and extended post-curing; while biocompatible resin formulations are increasingly available, these may compromise achievable resolution or mechanical properties [102, 103]. Unlike PDMS, most printed materials lack gas permeability, necessitating alternative oxygenation strategies such as media reservoirs or integrated gas exchange membranes for long-term culture applications [104, 105].

CNC micromachining

Computer numerical control micromachining offers a subtractive fabrication approach that complements the additive and replicative methods described above. In this technique, precisely controlled cutting tools remove material from solid substrates following toolpaths derived from computer-aided design models, enabling fabrication of microfluidic devices in materials that cannot be processed by lithographic or printing methods [106].

The primary advantage of CNC machining for BoC applications lies in its material versatility. The technique readily accommodates thermoplastics including polymethyl methacrylate, polycarbonate, cyclic olefin copolymer, and polystyrene, as well as glass and metals, providing access to substrates with superior chemical resistance, optical clarity, or mechanical properties compared to PDMS. Cyclic olefin copolymer is particularly attractive for drug screening applications as it exhibits minimal absorption of small hydrophobic molecules, directly addressing a key limitation of PDMS-based devices. The rigid nature of machined thermoplastic devices also enables compatibility with higher operating pressures and facilitates integration with external mechanical loading apparatus for applying controlled compressive or tensile forces to bone constructs [106–108].

From a manufacturing perspective, CNC machining scales readily from single prototypes to small production batches with consistent dimensional accuracy, and does not require the specialized cleanroom facilities associated with photolithography [109]. However, minimum feature sizes are constrained by cutter diameter to approximately 100 to 200 μm for standard micro-end mills, with dimensions of 25 to 50 μm achievable only using specialized tooling and optimized cutting parameters [110–112]. Complex internal three-dimensional geometries require multi-axis machining capability or assembly of multiple separately machined layers, adding process complexity. Machining operations can also produce tool marks that reduce optical clarity, potentially requiring subsequent polishing steps for imaging applications [109]. Unlike monolithic PDMS devices fabricated by soft lithography, machined components typically require bonding through thermal, adhesive, or solvent welding methods to enclose channel networks, introducing additional fabrication steps and potential failure points [110–112].

Selection considerations and hybrid approaches

The selection of fabrication methodology for BoC devices ultimately depends on the specific biological question being addressed, required channel dimensions and geometries, material performance requirements, and anticipated production scale. For fundamental studies of osteocyte mechanotransduction requiring sub-cellular resolution, two-photon polymerization offers unique capabilities despite limited throughput. Rapid prototyping and iterative design optimization during early development stages favor the photolithography and soft lithography workflow, which remains the most widely employed approach in academic BoC research due to its accessibility, relatively low per-device cost, and compatibility with live-cell imaging [113]. Applications requiring trabecular-like three-dimensional architectures or integrated porous scaffolds benefit from stereolithographic or digital light processing approaches, while studies involving hydrophobic drug compounds or requiring thermoplastic substrates are better served by CNC machining [114, 115].

Increasingly, hybrid fabrication strategies combining multiple techniques are being employed to leverage the complementary strengths of different approaches. Three-dimensionally printed scaffold inserts may be incorporated within PDMS housings fabricated by soft lithography, combining the architectural freedom of additive manufacturing with the gas permeability and optical properties of PDMS. Similarly, CNC-machined thermoplastic chambers may be bonded to micropatterned glass substrates to achieve both chemical resistance and high-resolution surface features [75, 116]. Such hybrid approaches represent a pragmatic response to the reality that no single fabrication technique currently satisfies all requirements for physiologically relevant BoC systems, and their continued development will be essential for advancing the field toward more sophisticated bone tissue models.

A comparison of fabrication approaches is presented in Table 1, summarizing the key parameters relevant to BoC applications.

Table 1.

Comparison of fabrication strategies for BoC microfluidic platforms

Feature	Photolithography	Soft Lithography (PDMS)	3D Printing	Milling & CNC Machining
Resolution	Excellent (sub-micron)	Good (tens of microns)	Variable (tens of microns to hundreds)	Good (tens of microns, tool-dependent)
3D Complexity	2D, challenging for true 3D	Quasi-3D (layer-by-layer bonding)	Excellent (true 3D, internal structures)	Limited (mostly 2D, complex 3D challenging)
Materials	Si, Glass, Quartz	PDMS (most common), other elastomers	Resins (photopolymers), hydrogels, some ceramics	PMMA, PC, COC, Glass, Metals, Polyimide
Cost (Setup)	Very High (cleanroom, equipment, masks)	Moderate (initial master mold)	Variable (moderate to high for high-res printers)	Moderate (CNC machine)
Prototyping Speed	Slow (multi-step, mask fabrication)	Fast (once master mold is made)	Very Fast (direct digital manufacturing)	Moderate (design & machining time)
Biocompatibility	Good (glass/Si), surface functionalization needed	Excellent (PDMS)	Variable (depends on resin)	Good (depends on material)
Key Advantage	High resolution & mass production	Low cost, rapid prototyping, biocompatibility	Design freedom, true 3D, rapid iteration	Material versatility, robust devices
References	[69–71]	[72, 73]	[86–89]	[106–108]

Bone on a chip systems

BoC platforms represent an emerging class of microengineered systems that aim to recapitulate aspects of human bone architecture and physiology within controlled in vitro environments. By incorporating key cellular components osteoblasts, osteoclasts, osteocytes, and endothelial cells into microfluidic devices, these systems provide opportunities to investigate bone biology under conditions that more closely approximate native tissue than conventional culture methods [117, 118]. However, it is important to acknowledge at the outset that the field remains in relatively early stages of development. While numerous BoC configurations have been reported, most current platforms capture only selected aspects of bone physiology, and none yet fully replicate the structural hierarchy, cellular diversity, and dynamic remodeling processes characteristic of native bone tissue [119]. The following sections critically examine the current state of BoC technology across three principal application domains: bone physiology modeling, bone pathology modeling, and drug discovery.

Bone physiology modeling

A central objective of BoC development has been the recreation of bone remodeling processes under controlled in vitro conditions. This capability is of particular importance because bone remodeling involves dynamic, reciprocal interactions among osteoblasts, osteoclasts, and the surrounding extracellular matrix that are difficult to study using conventional static culture systems [24, 120]. By incorporating mechanical stimulation, controlled perfusion, and three-dimensional scaffold architectures, BoC platforms offer potential advantages over traditional two-dimensional cultures, though the extent to which current systems faithfully recapitulate in vivo remodeling dynamics remains an active area of investigation [120–122].

Early efforts to model bone remodeling on microfluidic platforms focused on establishing functional co-cultures of bone-forming and bone-resorbing cells. Middleton and colleagues developed a three-dimensional co-culture system incorporating early-stage osteoblasts and osteoclast precursor cells within a porous β-tricalcium phosphate scaffold on an organ-on-a-chip platform [123]. Following 21 days of dynamic culture, the construct demonstrated cell migration, proliferation, and extracellular matrix formation, with evidence of both bone deposition and resorption activities. While these findings represented an important proof-of-concept for co-culture bone remodeling models, several limitations should be noted. The system lacked osteocytes the principal mechanosensory cells that orchestrate remodeling in vivo and did not incorporate vascular elements or physiologically relevant mechanical loading beyond fluid shear stress. Additionally, the extent to which the observed matrix turnover quantitatively reflects in vivo remodeling rates was not established [123].

Recognizing the importance of the bone marrow microenvironment for skeletal physiology, subsequent work has focused on developing bone-marrow-on-a-chip platforms that integrate stromal and hematopoietic components. Torisawa and colleagues engineered a modular bone-marrow-on-a-chip using a commercially available dual-channel microfluidic platform (Fig. 4.i) [124]. The device features a vascular channel separated from a marrow compartment by a semi-permeable membrane, accommodating osteoblasts, endothelial cells, mesenchymal stem cells, and hematopoietic stem/progenitor cells (HSPCs). This configuration maintained phenotypically defined HSPCs for at least 14 days at frequencies approaching those observed in native marrow. The modular, commercially compatible design offers practical advantages for adoption by laboratories without extensive microfabrication expertise. However, the platform’s reliance on a pre-formed membrane barrier, rather than a self-organized tissue interface, represents a simplification of native bone marrow architecture. Furthermore, the system does not incorporate mineralized bone matrix or the endosteal surface that plays critical roles in HSPC niche regulation [124].

Fig. 4.

Bone marrow-On-a-chip. (i) A dual-channel microfluidic system incorporating a semi-permeable membrane to separate a vascular channel from a marrow chamber containing key stromal (osteoblasts and BMSCs) and parenchymal (endothelial cells and HSPCs) components of the bone marrow niche. This configuration supports long-term maintenance of functional HSPCs in vitro. (ii) A modular microfluidic chip enabling self-assembled, perfusable vascular networks embedded in ECM-mimicking hydrogels. The platform supports hematopoietic differentiation and dynamic drug response analysis, offering high spatial and temporal control to model human granulopoiesis and marrow physiology. Created with BioRender.com, based on data from [124, 125]

More recent platforms have attempted to address these limitations by recreating both perivascular and endosteal niches within integrated devices. Chou and colleagues developed a bone-marrow-on-a-chip supporting self-assembled, perfusable three-dimensional vascular networks embedded within a matrix designed to recapitulate bone marrow extracellular matrix composition (Fig. 4.ii) [125]. CD34 + HSPCs maintained within this system for 14 days subsequently differentiated into mature neutrophils that migrated from the marrow compartment into adjacent fluidic channels, thereby modeling granulopoiesis. The platform also demonstrated niche-specific pharmacodynamic responses to doxorubicin and granulocyte-colony-stimulating factor that qualitatively mirrored in vivo observations. These results suggest improved physiological relevance compared with earlier designs. Nevertheless, several gaps remain: the “endosteal niche” in this system lacks true mineralized bone matrix, the platform has not been validated for long-term culture beyond two weeks, and the extent to which the observed drug responses quantitatively predict clinical pharmacodynamics has not been established [125].

These bone physiology models demonstrate progressive advancement in complexity and physiological relevance. However, critical assessment reveals persistent gaps between current capabilities and the requirements for comprehensive bone modeling. Most notably, no existing platform fully integrates all principal bone cell types (osteoblasts, osteocytes, osteoclasts) with functional vasculature, mineralized matrix, and physiologically relevant mechanical loading. The absence of osteocytes from most current systems is particularly significant given their central role in coordinating bone remodeling responses to mechanical and biochemical signals. Furthermore, validation of these platforms against in vivo benchmarks remains limited, making it difficult to assess their predictive value for translational applications.

Bone pathology modeling

The integration of biosensors into BoC platforms has significantly enhanced their analytical capacity by enabling non-invasive, real-time monitoring of cellular responses and biochemical dynamics. Such sensors allow continuous assessment of parameters including mineral deposition, metabolic activity, and gene expression, offering mechanistic insights into bone pathophysiology that are difficult to obtain from endpoint assays alone. Nevertheless, the interpretability of biosensor outputs remains strongly dependent on how accurately the surrounding microenvironment reflects native bone physiology [126].

Oxygen tension is a key regulatory cue for bone cell function and varies spatially within bone tissue, decreasing with distance from endosteal sinusoids [127]. Conventional in vitro cultures are typically maintained under normoxic conditions (~ 18.6% O₂; 141 mmHg), which approximate ambient air but substantially exceed oxygen levels experienced by most bone-resident cells in vivo [127]. This discrepancy is particularly relevant when modeling pathological processes, as hypoxia modulates osteogenic differentiation, bone remodeling, and tumor–bone interactions. Traditional culture systems offer limited control over microscale oxygen gradients, whereas microfluidic BoC platforms enable spatially and temporally defined oxygenation profiles, partially addressing this critical limitation. However, reproducing the full complexity of in vivo oxygen heterogeneity remains an ongoing challenge [128].

Bone represents a preferential metastatic niche for several solid tumors, notably breast, prostate, and lung cancers. Tumor colonization of the bone microenvironment disrupts the tightly regulated coupling between osteoblast and osteoclast activity, resulting in osteolytic or osteoblastic lesions accompanied by skeletal complications and pain [129, 130]. Beyond morbidity, bone metastasis is associated with markedly reduced survival, particularly in breast cancer, where bone involvement poses substantial therapeutic challenges [131]. Epidemiological data indicate that the majority of patients succumb within a few years following skeletal dissemination, underscoring the need for predictive and mechanistically informative in vitro models [132].

The bone perivascular niche-on-a-chip (BoPV) represents an advanced microfluidic approach aimed at capturing key features of the metastatic bone microenvironment. This platform integrates a human tri-culture system comprising endothelial cells, mesenchymal stem cells (MSCs), and breast cancer cells embedded within a native three-dimensional bone matrix, allowing the formation of stable, perfusable vascular networks in the absence of exogenous angiogenic factors (Fig. 5.i) [133]. Controlled interstitial flow establishes oxygen gradients comparable to those reported in bone marrow and supports the emergence of capillary-like structures stabilized by MSC-derived perivascular cells expressing PDGFR-β, NG2, or CD146. Under these conditions, breast cancer cells exhibit reduced proliferation and acquire phenotypes associated with therapeutic resistance, including diminished sensitivity to agents such as sunitinib. While such features align with in vivo observations of slow-cycling, drug-tolerant tumor cell populations, it should be noted that the model captures early-stage metastatic adaptation rather than the full spectrum of long-term disease progression. Nonetheless, the platform enables longitudinal analysis of vascular dynamics, tumor cell extravasation, and stromal interactions in a controlled setting, offering a valuable tool for investigating metastatic dormancy and early therapeutic response mechanisms [133].

Fig. 5.

Bone-On-a-chip for bone metastasis. (i) A Bone-prevascular Niche-On-a-chip platform incorporating a tri-culture of human endothelial cells, mesenchymal stem cells (MSCs), and breast cancer cells seeded within a native decellularized bone matrix. The microfluidic configuration enables controlled interstitial flow, establishing physiologically relevant oxygen gradients and promoting spontaneous formation of stable, perfusable vascular networks without exogenous angiogenic stimulation. This system recapitulates key hallmarks of the metastatic bone microenvironment, including cancer cell dormancy, vascular niche occupancy, and resistance to targeted therapeutics, thereby offering a robust in vitro model for studying early metastatic seeding and immune interactions. (ii) A miniaturized bone-on-a-chip device supporting the self-directed generation of a three-dimensional mineralized osteoid matrix over a 30-day culture period in the absence of exogenous osteoinductive factors. The architecture facilitates high-content imaging and enhances the surface-to-volume ratio, enabling real-time analysis of direct interactions between disseminated tumor cells and nascent bone tissue. The platform effectively recapitulates in vivo bone colonization patterns, providing a physiomimetic environment for probing tumor–bone crosstalk and testing anti-metastatic therapeutics. Created with BioRender.com, based on data from [133, 134]

Complementary to niche-focused approaches, a miniaturized BoC system has been developed to support the spontaneous formation of three-dimensional mineralized bone tissue. Over a culture period of approximately 720 h (~ 30 days), the device generates mature osteoid layers up to ~ 85 μm in thickness, characterized by densely mineralized fibrillar collagen, without the use of exogenous osteogenic supplements (Fig. 5.ii) [134]. The microfluidic design increases the effective bone surface-to-volume ratio and enhances imaging accessibility, facilitating high-resolution observation of tumor–bone interactions. Co-culture studies revealed hallmarks of early breast cancer colonization that had previously been reported primarily in animal models, suggesting that the platform can reproduce selected aspects of early metastatic seeding in vitro. However, the absence of immune components and long-term remodeling processes indicates that the model is best suited for studying initial colonization events rather than advanced metastatic disease [134].

Beyond cancer-related pathologies, BoC platforms have also been applied to model bone degeneration and fracture repair. A vascularized flat BoC was engineered to evaluate bone-regenerative therapeutics by integrating two orthogonal mechanical cues tensile strain and interstitial shear stress [135]. Tensile loading is induced through a static magnetic field acting on a magnetic nanocomposite scaffold, while microfabrication techniques generate defined topographies that impose controlled fluid shear. This mechanically driven environment promotes concurrent vasculogenesis and osteogenesis without exogenous growth factors, recapitulating key aspects of intramembranous ossification observed during fracture healing. Although the system demonstrates tunable scaffold mechanics and loading parameters, its physiological relevance is constrained by simplified tissue architecture and the absence of inflammatory and immune signaling. Nevertheless, the platform provides a robust, growth-factor-independent framework for preclinical screening of bone repair strategies and for dissecting mechano-biological coupling in bone regeneration [135].

Taken together, current BoC-based models for bone pathology provide valuable insight into early-stage disease mechanisms by enabling controlled manipulation of oxygen gradients, vascular cues, and tumor–stromal interactions. These systems are particularly effective in dissecting initial events such as metastatic seeding, dormancy induction, and microenvironment-driven drug resistance. However, their capacity to model advanced pathological states remains constrained by limited cellular diversity, simplified immune involvement, and the absence of long-term remodeling dynamics. As a result, BoC platforms should be viewed not as comprehensive replicas of bone pathology, but as reductionist yet mechanistically informative tools that complement animal models. Future progress will depend on integrating immune components, dynamic remodeling units, and extended culture stability to more faithfully capture the temporal and spatial complexity of bone disease progression.

Drug discovery

BoC platforms are increasingly explored as intermediate systems for drug discovery and, in selected contexts, personalized medicine. By incorporating disease-relevant microenvironments and, in some cases, patient-derived cells, these platforms aim to improve the prediction of drug efficacy and toxicity compared with conventional two-dimensional cultures. However, their predictive value remains closely tied to the degree of biological fidelity achieved and the specific disease context being modeled [136].

To address the need for scalable preclinical models in age-related skeletal disorders such as osteoporosis, a high-throughput biomimetic BoC platform has been developed and coupled with an artificial intelligence–based image analysis pipeline [137]. This system reconstructs key aspects of osteon-like organization by co-culturing murine osteocytes (IDG-SW3) and osteoblasts (MC3T3-E1) within a three-dimensional osteoblast-derived decellularized extracellular matrix (OB-dECM), integrated into micro-gel units compatible with standard multiwell plates. The combination of native matrix cues and dual-cell co-culture enhances osteocyte differentiation and osteoblast maturation, supporting osteogenic functionality. Importantly, the optical transparency and geometrical uniformity of the constructs enable compatibility with high-content screening workflows. Proof-of-concept testing using an anti-sclerostin (anti-SOST) antibody (an emerging anabolic therapy for osteoporosis) demonstrated quantifiable β-catenin nuclear translocation, which was automatically evaluated using deep-learning algorithms. While this AI-integrated, osteon-mimetic platform offers improved throughput and analytical consistency, its reliance on murine cell lines and simplified remodeling dynamics limits direct extrapolation to human bone physiology [137].

Among various stem cell sources, adipose-derived stem cells (ADSCs) are frequently selected for BoC-based drug screening due to their accessibility, proliferative capacity, and osteogenic differentiation potential [138]. In this context, a miniaturized ADSC-based three-dimensional BoC model was developed to support in vitro evaluation of bone-regenerative therapies (Fig. 6.i) [139]. ADSCs were pre-differentiated for either 7 or 14 days prior to incorporation into a microfluidic collagen matrix, enabling temporal resolution of osteogenic maturation. Osteogenic progression was confirmed through alkaline phosphatase activity, calcium deposition, morphological changes, and expression of bone-related markers, including bone sialoprotein 2, DMP-1, and osteocalcin. Notably, constructs formed from 14-day pre-differentiated cells exhibited features of mature bone-like tissue, whereas the 7-day group retained characteristics of early-stage differentiation. The absence of dynamic fluid flow simplifies device operation and enhances experimental accessibility; however, it also constrains the model’s ability to capture shear-dependent remodeling processes that are relevant in vivo. As such, this platform is well suited for early-stage therapeutic screening rather than comprehensive evaluation of bone regeneration dynamics [139].

Fig. 6.

Bone-on-a-chip for drug discovery. (i) A collagen-based 3D bone-on-a-chip model integrating adipose-derived stem cells (ADSCs) pre-differentiated for 7 or 14 days. The system enables temporal resolution of osteogenic progression under static conditions, validated by alkaline phosphatase activity, mineral deposition, cell morphology, and expression of osteogenic markers (e.g., BSP2, DMP-1, and osteocalcin). The 14-day pre-differentiated group formed mature bone-like tissue, whereas the 7-day group displayed early-stage features, demonstrating the model’s utility for screening regenerative therapies in a simplified, flow-independent setup. (ii) A structurally and chemically biomimetic bone-on-a-chip platform constructed using two-photon polymerization (2PP) guided by high-resolution nano-CT scans of human trabecular bone. The polymeric scaffold is coated with hydroxyapatite (HAp) to mimic bone mineralization and embedded in a microfluidic system optimized for long-term culture. Human mesenchymal stromal cells seeded into the device remained viable for 21 days, producing collagen-rich extracellular matrix. This platform provides a physiologically relevant in vitro model for investigating bone remodeling and evaluating candidate drugs for bone regeneration and repair. Created with BioRender.com, based on data from [139, 140]

Despite incorporating multiple bone cell types, many BoC systems remain limited in their ability to recapitulate the native structural and chemical complexity of bone tissue. To address this gap, a structurally and compositionally biomimetic BoC platform was developed by combining high-resolution anatomical reconstruction with mineral-phase engineering. Using two-photon polymerization (2PP) guided by nano-computed tomography scans of human trabecular bone, a polymeric scaffold was fabricated to replicate native microarchitecture (Fig. 6.ii). The scaffold was subsequently coated with a HAp layer to mimic bone mineral composition and integrated into a microfluidic device optimized for long-term culture. Human mesenchymal stromal cells seeded within the system remained viable for up to 21 days and produced a collagen-rich extracellular matrix. While this approach substantially improves structural and chemical fidelity, it does not yet incorporate active remodeling units or multicellular coupling under dynamic mechanical loading. Nevertheless, the platform provides a valuable framework for investigating drug–matrix interactions and osteogenic responses in a more physiologically representative in vitro setting [140]. In the context of drug discovery, BoC platforms occupy a strategic intermediate position between conventional in vitro assays and in vivo models by offering improved physiological relevance alongside experimental scalability. Their strengths lie in enabling high-content analysis, controlled microenvironmental cues, and early-stage screening of therapeutic candidates under bone-relevant conditions. Nevertheless, current implementations remain limited by simplified mechanics, incomplete multicellular coupling, and short culture durations, which restrict their predictive power for long-term efficacy and safety. Consequently, BoC-based drug screening should be regarded as a complementary filtering step rather than a definitive validation platform. Continued advances in microfabrication, biomaterials, and data-driven analytics will be essential to enhance their translational value and to define standardized benchmarks for their integration into preclinical drug development pipelines.

This recently inaugurated area of BoC technology is a promising arena for bone biology and regenerative medicine. Such platforms attempt to bridge the gap between classical 2D cultures and in vivo models, providing physiologically relevant and scalable systems for fundamental research purposes, disease modeling, and therapeutic screening. With an increased interdisciplinary collaboration between microengineering, cell biology, and materials science, BoC systems will soon become standard tools for translational and personalized bone research.

Challenges and future directions

Although BoC technologies have exponentially advanced, a number of challenges restrict their wide application and translational fulfillment. One of the most significant constraints of BoC is the under-implemented biomimicking of native bone architecture. The present BoC models often lack the characteristic hierarchy of cortical and trabecular bone along with the highly anisotropic mechanical properties and mineral gradients vital for load-bearing function modeling [120, 141]. Design-wise, sanguine recreation of the dynamic remodeling cycle of bone-a process that requires tightly regulated crosstalk between osteoblasts, osteoclasts, and osteocytes-has hardly been imitated in vitro due to the intricacies involved in cellular interactions and the long-term viability of co-cultured cells [142, 143].

The other major technical hurdle is vascularization. Bone is highly vascularized, and a functional microvasculature is critical for modeling nutrient transport, removal of waste products, and modeling oxygen gradients. While some versions have incorporated endothelialized channels, these systems tend to either be inherently non-functional or not scalable [144, 145]. Moreover, there’s no variable that could allow for integration of immune tolerance mechanisms-which are crucial to bone regeneration, inflammation, and disease-into the present designs. This is a major drawback because it limits the efficacy of BoC models in studying immunological diseases of bone such as rheumatoid arthritis or osteomyelitis [146].

From an engineering perspective, there are still issues with standardization, scalability, and reproducibility. Variability in fabrication methods, biomaterials, and cell sources introduces inconsistencies between different platforms [147]. Along this, real-time monitoring and screening can barely be conducted despite advancement made on biosensors and automation, on the other hand. Also, further technological developments are needed to incorporate approximating physiological loading (that is compression and shear stress) in a controlled and reproducible manner to the systems of mechanical stimulation in these platforms [148, 149].

On the future-academic agenda are the integration of multi-organ systems with special emphasis on bone-vascular, bone-marrow, and bone-muscle interfaces for compensating organs to simulate systemic physiological responses. Patient-specific iPSCs and cutting-edge 3D bioprinting and organoid technologies may yield personalized BoC platforms for disease modeling and precision medicine [150, 151]. AI and machine learning could be harnessed to analyze data, process images, and develop predictive models to interpret complex biological outputs, thereby facilitating drug screening while enhancing the predictive capacity of their respective platforms (Fig. 7) [152].

Fig. 7.

Challenges and future directions in bone-on-a-chip technology. Created with BioRender.com

The real effect of BoC technologies will hinge on their ability to address the prevailing bottlenecks of biomimicry, direct application, and standardization. The ongoing collaborative input of engineers, biologists, and clinicians will be indispensable in translating BoC platforms from the bench to relevant clinical dimensions.

Conclusion

BoC platforms are truly a breakthrough in the development of in vitro modeling tools because they offer a controllable, physiologically relevant setting to study bone biology, disease mechanisms, and therapeutic responses. In BoC technologies, key cellular components, microarchitecture, and dynamic stimuli are integrated into a microfluidic system, overcoming many drawbacks of conventional 2D culture systems and animal models. Such systems provide insight into bone remodeling, regeneration process, and complex bone disease progression and offer exciting prospects for drug screening and personalized medicine.

However, despite significant achievements in the last few years, the full potential in the clinical and industrial realm would require several hurdles, such as vascularization, immune integration, long-term co-culture, and standardization, to be dealt with. Advances in biofabrication, sensor integration, and data analysis tools such as AI must be leveraged to overcome these limitations.

As interdisciplinary innovation continues pushing the field, BoC systems are about to become core instruments in orthopedic research, regenerative therapeutics, and drug development. Their ongoing evolution will not only improve bone physiology and pathology knowledge, but will also lay the groundwork for patient-specific therapies in the not-so-distant future.

Author contributions

HR and AV designed the review study and contributed to writing the manuscript draft. HR searched the literature and contributed to writing the manuscript. AV contributed to the literature search and edited the manuscript. All authors have confirmed the final version of the manuscript and are accountable for the contents of all parts of the work. All authors read and approved the final manuscript.

Funding

This research received no external funding.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent to publish

Not applicable.

AI statement

The authors declare that no AI-assisted technologies were used in the generation of this manuscript.

Competing interests

The authors declare no competing interests.