Engineering Osteosarcoma In Vitro: From Traditional Models to Biofabricated Platforms for Precision Medicine

Abstract

Osteosarcoma is a highly malignant primary bone tumor that predominantly affects adolescents. Despite the widespread application of standard treatment modalities, including surgical resection, chemotherapy, and radiotherapy, the long-term survival rate of patients remains unsatisfactory due to the high metastatic potential and drug resistance of the disease. In recent years, researchers have focused on developing more precise in vitro models that aim to better simulate the tumor microenvironment, thereby enhancing the effectiveness of drug screening and personalized therapy. This review summarizes the latest advances in osteosarcoma in vitro modeling, including the development of conventional two-dimensional culture systems, three-dimensional culture platforms, organoid models, and microfluidic chips designed to mimic the tumor microenvironment. Additionally, this review explores the application value of these models in drug screening, immune coculture systems, and personalized treatment strategies. The integration of multiomics data and artificial intelligence is also discussed as a means to optimize model design and facilitate precision oncology. Biomimetic in vitro models have the potential to more accurately replicate tumor heterogeneity, cell–cell interactions, and the complexity of the tumor microenvironment, thereby increasing the translational value of preclinical drug development. Finally, this review highlights the current challenges in the field, including the lack of standardized protocols, issues with model stability and reproducibility, and the practical integration of these models into preclinical research pipelines.

1. Introduction

Osteosarcoma (OS), the most common primary malignant bone tumor, predominantly affects adolescents. This tumor frequently arises in the metaphyses of the long bones. Although the precise etiology of OS remains incompletely understood, both genetic and environmental factors have been implicated in its pathogenesis. Among these, mutations in tumor suppressor genes such as TP53 and RB1 have been identified as key contributors to OS development, closely associated with enhanced cellular proliferation, invasion, and metastasis. Multiple signaling pathways are involved in the tumorigenesis of OS, including Wnt/β-catenin, PI3K/AKT, MAPK, and Notch pathways. Aberrant activation of these pathways plays a pivotal role in promoting tumor progression and metastasis. , Moreover, the immune microenvironment is recognized as a critical component in OS development. Tumor cells can evade host immune surveillance by suppressing the function of immune cells, thereby facilitating tumor growth and dissemination.

Currently, the treatment of OS primarily relies on radiotherapy and chemotherapy, with the latterparticularly combination regimens involving methotrexate, cisplatin, and doxorubicinconstituting the standard therapeutic approach. These chemotherapy protocols have significantly improved local disease control. However, the highly metastatic nature of OS, especially its tendency to spread to the lungs, remains the most formidable challenge in clinical management. Although chemotherapy has improved patient prognosis to some extent, its efficacy against metastatic or recurrent OS remains limited, and the associated toxicity often leads to severe side effects. Emerging immunotherapeutic strategies, such as immune checkpoint inhibitors (e.g., anti-PD-1/PD-L1 antibodies) and adoptive T cell therapies (e.g., CAR-T cells), have shown potential as novel treatment avenues. Nevertheless, their clinical efficacy in OS has thus far fallen short of expectations. , Therefore, the development of effective immunotherapies for OS remains an urgent unmet need and warrants further in-depth investigation.

To better understand the biological mechanisms of OS and identify novel therapeutic targets, the establishment of biomimetic in vitro models is of critical importance. Although traditional two-dimensional (2D) cell culture systems are widely used in drug screening, they offer a limited ability to replicate the complex tumor microenvironment. Specifically, they fail to adequately mimic the interactions between tumor cells and the surrounding extracellular matrix, immune cells, and vasculature, which can result in inaccurate drug screening outcomes and hinder clinical translation. In contrast, three-dimensional (3D) culture systems, organoid models, and immune coculture platforms have emerged as promising alternatives. These advanced models more faithfully recapitulate the cellular interactions within the tumor microenvironment, particularly those between tumor and immune cells. Studies have demonstrated that such models not only capture the heterogeneity of tumors but also more accurately reflect key biological processes such as tumor growth, metastasis, and immune evasion. − From a translational perspective, these models offer improved predictive power for clinical therapeutic responses. They provide valuable tools for high-throughput drug screening, the development of personalized treatment strategies, and the evaluation of immunotherapeutic approaches.

This review summarizes the current status and future prospects of in vitro models for OS research. First, the commonly used types of OS in vitro models are discussed, with an emphasis on their respective advantages and limitations, particularly focusing on the applications of 3D bioprinting and organ-on-a-chip technologies. Next, the review examines the roles of these models in simulating the tumor microenvironment, facilitating drug screening, and studying metastatic mechanisms. Finally, recent advances in OS in vitro modeling are summarized, key challenges in current research are analyzed, and future directions are proposed with a forward-looking perspective.

2. Current Status of In Vitro Models for OS Research

2.1. Traditional 2D Culture Models

2D culture models are widely used in OS research. Commonly employed cell lines included MG-63, U2OS, and SaOS-2. MG-63, derived from human OS, exhibits a high proliferative capacity but tends to be of low malignancy, making it suitable for studying tumor proliferation and drug screening. U2OS is a well-differentiated cell line with a slower proliferation rate and is frequently used in studies of signaling pathways and mechanisms of drug resistance. SaOS-2 is a highly differentiated osteoblastic OS cell line and is often applied in bone metabolism, mineralization studies, and drug resistance assays. − 2D culture models have significantly contributed to drug screening and mechanistic studies by elucidating the roles of various signaling pathwayssuch as PI3K/AKT, MAPK, and Wnt/β-cateninin tumor cell proliferation, invasion, and chemoresistance. For instance, it has been reported that gastrin-releasing peptide promotes tumorigenesis and metastasis in OS by activating the AKT and Wnt/β-catenin pathways, thereby enhancing cisplatin resistance and tumor invasiveness (Figure A). Additionally, 2D models are valuable in high-throughput drug screening and in evaluating the antitumor efficacy of chemotherapeutic agents, such as cisplatin, methotrexate, and doxorubicin. A recent study integrated RNA sequencing and proteomic analyses to identify 2535 OS-specific genes and several potential therapeutic targets. High-throughput drug combination screening in OS cell lines revealed a synergistic antitumor effect between histone deacetylase (HDAC) inhibitors and doxorubicin (Figure B). However, 2D models have significant limitations. They lack a 3D microenvironment and fail to accurately recapitulate cell polarization, matrix stiffness, and the spatial organization of tumor cell growth, leading to discrepancies between the in vitro findings and clinical outcomes. Moreover, conventional 2D models overlook tumor heterogeneity and do not account for the complex interactions between cancer stem cells (CSCs) and nonstem cells, compromising experimental accuracy. Furthermore, they are inadequate in modeling cell–cell and cell–matrix interactions, which are critical for studying drug sensitivity and signaling pathway activation. To address these limitations, researchers are increasingly turning to 3D models, such as organoids, tumor spheroids, and scaffold-based systems. These models better simulate the OS microenvironment, enhance the accuracy of drug screening, and support the development of personalized therapeutic strategies. ,

1.

Applications of 2D models in OS research. (A) Gastrin-releasing peptide enhances cisplatin resistance and promotes OS cell survival and migration by activating the AKT and Wnt/β-catenin signaling pathways. (Reproduced with permission from ref, Copyright the author 2024); (B) Integrated multiomics analysis combined with high-throughput drug screening identifies potential therapeutic gene targets for OS and reveals the synergistic mechanism of doxorubicin (DOX) and histone deacetylase (HDAC) inhibitors in OS treatment. (Reproduced with permission from ref, Copyright the author 2023).

2.2. 3D Culture Models

3D culture models have emerged as a powerful tool to overcome the limitations of traditional 2D systems in OS research. These models mainly include tumor spheroids and scaffold-based cultures. Tumor spheroid models rely on nonadherent culture conditions, allowing OS cells to spontaneously aggregate into spheroid structures. These models are widely used to study tumor–microenvironment interactions, evaluate resistance to chemotherapy and radiotherapy, and simulate immune–tumor cell interactions in coculture systems, thus serving as effective platforms for assessing immunotherapeutic efficacy. For instance, a coculture spheroid model of OS was developed by modulating the ratio of tumor cells to stromal cells to mimic different disease stages (early/late) and evaluate doxorubicin-induced cytotoxicity (Figure A).

2.

3D culture and PDMs. (A) Construction of early and late-stage OS spheroid models to simulate the cytotoxic effects of the chemotherapeutic drug doxorubicin (Reproduced with permission from ref, Copyright 2021 Wiley); (B) OS model fabricated using 3D-printed polyurethane scaffolds combined with in vitro generated bone ECM, revealing a biomimetic microenvironment suitable for OS cells (Reproduced with permission from ref, Copyright the author 2022); (C) 3D bioprinted anisotropic bilayer live hydrogel (ABLH) promoting osteochondral regeneration by reconstructing the cartilage-bone interface (Reproduced from ref, Copyright the author 2023, with permission from Elsevier); (D) Generation and characterization of patient-derived OS xenografts. (Reproduced with permission from ref, Copyright 2021 Elsevier).

Scaffold-based models utilize biomaterial-based 3D scaffolds to better replicate the extracellular matrix (ECM), supporting studies of OS cell growth and invasion. For example, a 3D-printed polyurethane scaffold combined with in vitro-derived bone ECM was used to construct a biomimetic OS model. This platform effectively simulated the bone tumor microenvironment, promoting tumor cell proliferation and invasion, and offered a promising approach for in vitro OS studies (Figure B).

Common scaffold materials include: (1) Natural biomaterials such as collagen and gelatin, which offer excellent ECM mimicry but suffer from limited mechanical strength. (2) Synthetic polymers like poly­(lactic-co-glycolic acid) (PLGA) and polycaprolactone (PCL), known for their tunable properties and stability, although their cell compatibility is relatively low. (3) Composite scaffolds, such as gelatin–chitosan composites, which provide a balance between biocompatibility and mechanical strength, enhancing cell adhesion and proliferation (Table ). ,, For example, an anisotropic bilayered hydrogel (ABLH) was fabricated using dual-channel extrusion bioprinting to spatially embed articular cartilage progenitor cells (ACPCs) and bone marrow-derived mesenchymal stem cells (BMSCs). This construct enabled spatiotemporal cell-driven tissue regeneration (Figure C). Compared with conventional 2D culture models, 3D in vitro models offer significant advantages in studying drug resistance and the invasive behavior of OS. They better replicate oxygen gradients within tumor tissues, offering a more physiologically relevant microenvironment for accurate drug resistance assessment. Scaffold-based 3D systems also more closely simulate tumor–ECM interactions, facilitating studies on tumor cell invasion and matrix degradation. , Moreover, recent findings suggest that combining immune coculture with 3D models improves the evaluation of tumor cell responses to immunotherapy, thereby enhancing the development of personalized therapeutic strategies.

1. Classification and Characteristics of Scaffold Materials for 3D Culture.

Material Type	Representative Materials	Main Advantages	Main Disadvantages	Primary Applications	ref.
Natural Materials	Collagen	main components similar to ECM, excellent biocompatibility, promote cell adhesion and proliferation	poor mechanical properties, slow degradation rate	cartilage and bone tissue engineering, skin repair, nerve regeneration
	Silk Fibroin	excellent mechanical properties, good biocompatibility, nontoxic degradation products	variability in purity and batch-to-batch differences may affect performance; complex fabrication processes	bone and cartilage tissue engineering, ligament repair, controlled drug release
	Bacterial Cellulose	3D fibrous network structure, good biocompatibility and biodegradability, exhibiting excellent mechanical properties	small pore size, degradation rate mismatched with bone regeneration, low osteogenic activity	bone tissue engineering, soft tissue repair, drug delivery
	Demineralized Bone Matrix	retains natural bone bioactive components,good biocompatibility, osteoinductive capacity	limited availability, risk of immune response and pathogen transmission	bone defect repair, spinal fusion, dental implants
Synthetic materials	Polylactic Acid (PLA)	good degradability, excellent mechanical properties, ease of processing and molding	degradation products may induce acidic reactions, biocompatibility lower than that of natural materials	bone and cartilage tissue engineering, vascular scaffolds, and sutures.	,
	Polyethylene Glycol (PEG)	good biocompatibility, tunable physicochemical properties, ease of functionalization	nondegradable, limited mechanical strength, typically used in combination with other materials	soft tissue engineering drug controlled release	,
	Polyether Ether Ketone (PEEK)	excellent mechanical properties, good chemical stability, favorable biocompatibility	biologically inert and requires surface modification to enhance bioactivity	orthopedic implants, spinal fusion devices, dental restorations.
Composite materials	Hydroxyapatite (HA)	good biocompatibility and osteoconductivity, excellent osteogenic performance	high brittleness, insufficient elasticity, slow degradation rate, processing difficulties limit	bone tissue engineering,bone defect repair
	Gelatin Methacryloyl (GelMA)	good biocompatibility and tunable mechanical properties,suitable for 3D cell culture and tissue engineering	photocross-linking process may affect cell viability, low mechanical strength necessitates combination with other materials	bone tissue engineering, cartilage repair, as a cell carrier.
	Chitosan/gelatin composite scaffold	combining the antibacterial properties of chitosan with the cell-adhesive characteristics of gelatin, excellent biocompatibility and biodegradability;	low mechanical strength, requires controlled degradation rates, and optimization of the fabrication process	bone tissue engineering, soft tissue repair, and drug delivery
	Stem cell–3D scaffold composite	combining the regenerative potential of stem cells with the structural support of 3D scaffolds, promotes wound healing, tissue regeneration	strict control of stem cell sources and culture conditions is required, ethical considerations must be addressed, preparation and storage conditions are demanding	wound dressings, tissue repair, chronic wound treatment.

2.3. Patient-Derived Models

In recent years, patient-derived models (PDMs) have become essential tools for OS research, primarily including patient-derived cells (PDCs) and patient-derived organoids (PDOs). PDCs are established by isolating tumor cells from patient tumor tissues for in vitro culture, enabling relatively rapid model generation and facilitating the screening of chemotherapeutic and targeted agents. , In contrast, PDOs are reconstructed using 3D culture techniques to form organoid structures that better preserve the heterogeneity and genomic characteristics of the patient’s tumor, such as mutations, epigenetic modifications, and molecular pathway activation states, making them more precise in vitro models. For personalized therapy and drug screening, patient-derived xenograft (PDX) and patient-derived orthotopic xenograft (PDOX) models have been widely applied. These models have demonstrated an improved ability to predict patient-specific therapeutic responses. For example, a study established nine OS PDX models from surgical resections of 21 primary OS tumors. Morphological and immunohistochemical analyses (including SATB2, MDM2, CDK4, and Ki67) alongside whole-exome sequencing revealed significant similarities between the patients’ tumors and their corresponding PDXs, which remained stable through multiple passages in mice. The OS PDX models faithfully recapitulate the morphological and genetic features of the original tumors, providing reliable platforms for evaluating therapeutic strategies (Figure D). The application of these models not only offers more precise treatment strategies for OS patients but also accelerates novel drug development and biomarker discovery, thereby advancing the field of personalized OS therapy.

3. Tumor Microenvironment and Immune Coculture Models

3.1. Complexity of the OS Microenvironment

The microenvironment of OS (OS) is highly complex, comprising multiple interacting cell types including osteoblasts, osteoclasts, and bone marrow stromal cells. Osteoblasts within the tumor microenvironment (TME) not only influence tumor cell proliferation and invasion but also regulate immune cell activity through the secretion of cytokines. Osteoclasts are activated in OS, promoting bone remodeling and releasing growth factors from the bone matrix, thereby accelerating tumor progression. Bone marrow stromal cells play a supportive role in the TME by facilitating tumor angiogenesis and modulating immune cell functions toward an immunosuppressive phenotype. The immune microenvironment of OS is characterized by pronounced immunosuppression, which plays a critical role in tumor initiation and progression. Studies have demonstrated that the OS TME is enriched with immunosuppressive cells such as M2 macrophages and regulatory T cells (Tregs), which inhibit antitumor immune responses by secreting immunosuppressive cytokines including TGF-β and IL-10. Additionally, OS cells often overexpress immune checkpoint proteins such as PD-L1, contributing to T cell exhaustion and immune evasion. These immune escape mechanisms not only diminish the host’s antitumor immunity but also compromise the efficacy of immunotherapy. Therefore, immunomodulatory strategies targeting the TMEsuch as PD-1/PD-L1 inhibitors and macrophage reprogrammingrepresent promising therapeutic approaches for OS.

3.2. Immune Coculture Models

In recent years, researchers have utilized coculture systems to investigate the interactions between immune cellssuch as T cells, natural killer (NK) cells, and macrophagesand OS cellsand aiming to better mimic the TME and evaluate immunotherapeutic efficacy. Studies have shown that, in vitro, OS cells can suppress the antitumor activity of NK cells and T cells by secreting immunosuppressive factors, while simultaneously promoting the polarization of M2-type tumor-associated macrophages (TAMs), thereby reinforcing immune evasion mechanisms. Furthermore, TAMs within coculture systems exhibit pro-tumorigenic and immunosuppressive functions, providing novel insights for TAM-targeted immunotherapy strategies. To more accurately assess the efficacy of immunotherapeutic agentssuch as PD-1/PD-L1 inhibitors and CAR-T cell therapiesresearchers have developed immune coculture models integrated with 3D culture systems or organoids. These 3D coculture models more faithfully recapitulate the OS TME, offering a more reliable platform for drug screening and therapeutic evaluation. Additionally, the application of CAR-T cells within 3D models has been further explored, with findings indicating that 3D coculture systems provide more precise assessments of CAR-T cell infiltration, proliferation, and antitumor activity compared to traditional 2D cultures, thus better approximating the in vivo environment. Consequently, the integration of 3D coculture models with immunotherapy research not only offers critical experimental evidence for personalized OS treatment but also lays a solid foundation for future clinical applications.

3.3. Tumor-Bone Microenvironment Models

OS exhibits complex interactions with the bone microenvironment, wherein coculture models of osteoblasts and osteoclasts have been widely employed to investigate OS-induced bone resorption and remodeling. , Studies have demonstrated that OS cells promote osteoclast differentiation and accelerate bone resorption by secreting various cytokines. Notably, microRNA-19a-3p contained within small extracellular vesicles (sEVs) enhances osteoclast function via the PTEN/PI3K/AKT signaling pathway, thereby contributing to bone degradation. Furthermore, coculture systems comprising osteoclasts and chondrocytes revealed that matrix metalloproteinases (MMPs) within the OS microenvironment play a critical role in cartilage degradation and disruption of bone remodeling balance. To more precisely elucidate the mechanisms underlying OS-associated bone remodeling, researchers have developed a calvarial coculture system based on mouse models, which effectively simulates the bone TME and facilitates dynamic assessment of bone resorption and formation. Additionally, single-cell RNA sequencing has uncovered gene expression patterns between osteoblasts and osteoclasts within OS, further clarifying the role of the bone microenvironment in tumor progression and identifying potential targets for therapy. Therefore, the application of coculture systems combined with emerging technologies to explore the interplay between OS and the bone microenvironment not only advances the understanding of its pathogenesis but also provides a crucial foundation for developing therapies targeting bone remodeling (Table ).

2. Applications of Different In Vitro Models in OS Research.

Research Focus	2D Cell Culture	3D Spheroids	Organoids	Organ-on-a-Chip	ref.
Drug Screening	preliminary Screening: Rapid assessment of cytotoxicity and cell viability	advanced screening: mimics intratumoral drug resistance to identify clinically relevant candidate drugs	precision screening: models patient-specific microenvironments for personalized drug testing	high-throughput screening: incorporates fluid dynamics to simulate in vivo pharmacokinetics and enhance drug response prediction accuracy	,,
Chemotherapy Resistance Mechanism Research	investigate drug responses at the single-cell level and screen for resistance-related genes	recreate the TME to study factors such as hypoxia influencing drug resistance	capable of supporting long-term culture to investigate mechanisms underlying the development of acquired drug resistance	integrate dynamic fluidic conditions to simulate in vivo drug metabolism	,,
TME Simulation	unable to accurately mimic the TME	can establish hypoxic conditions but lacks ECM components	supports tumor cell–matrix interactions, enabling studies on ECM effects	incorporates fluid dynamics to simulate blood flow and oxygen gradients	,
Tumor Metastasis Research	limited to observing migration behavior of monolayer cells	spheroids allow investigation of cellular invasion behavior	accurately mimics the interface between tumor and normal tissues	integrates hemodynamics to study circulating tumor cell migration mechanisms	,
Bone–Tumor Interaction	cannot directly study the effects of bone tissue on tumors	allows coculture with bone tissue to observe tumor effects on bone	contains bone tissue components, enabling investigation of OS-induced bone destruction	integrates mechanical stress and blood flow to simulate the bone tumor growth microenvironment	,
Immunotherapy Research	unable to effectively study the role of the immune system	enables coculture with immune cells to investigate immune evasion mechanisms	allows construction of personalized immune models using PDCs	combines microfluidic technology to simulate dynamic immune cell responses in vivo	,

4. Application of Microfluidic Chip Technology in OS Research

4.1. Technical Principles and Advantages of Microfluidic Chips

Microfluidic chip technology allows the precise manipulation of fluids at the microscale, enabling dynamic simulation of physiological fluid environments that closely resemble in vivo conditions. In OS research, this technology is employed to replicate nutrient transport, cell signaling, and mechanical stimuli within the TME. For example, Ahn developed a microfluidic platform mimicking the bone microenvironment to study interactions between the tumor milieu and hydroxyapatite, supporting the culture of tumor cells within a 3D bone-mimetic composite comprising hydroxyapatite and fibrin (Figure A). Furthermore, microfluidic systems support coculture of multiple cell types, allowing OS cells, osteoblasts, and immune cells to be grown simultaneously within a controlled microenvironment. Combined with high-resolution imaging, this approach enables the real-time monitoring of cellular interactions and functional dynamics, providing more accurate data for drug screening and personalized therapy. For instance, Jaiswal utilized dual-extrusion bioprinting to fabricate a bone tumor model with a peripheral matrix containing HUVECs and osteo-primed WJ-MSCs surrounding MG-63 OS cells cultured on a microfluidic platform, facilitating drug sensitivity assessment (Figure B). In the context of OS invasion, metastasis, and angiogenesis, microfluidic chips have been applied to generate 3D tumor spheroids or organoid models. These models better recapitulate tumor cell invasion into adjacent tissues, intravasation into the vasculature, and metastatic colonization. When integrated with quantitative proteomics, this strategy unveils functional changes in distinct cell subpopulations and their contributions to tumor progression.

3.

Applications of Microfluidic Chips. (A) A 3D microfluidic bone TME composed of hydroxyapatite/fibrin composite, utilized for drug screening and mechanistic studies of bone tumor metastasis, growth, and progression (Reproduced with permission from ref, Copyright the author 2019, licensed under CC-BY-4.0); (B) A 3D bioprinted microfluidic OS chip model that closely mimics the native OS-TME, serving as a preclinical platform for anticancer drug screening (Reproduced with permission from ref, Copyright 2021 Elsevier); (C) Construction of a 3D microfluidic bone remodeling model (Reproduced with permission from ref, Copyright the author 2023, licensed under CC-BY-4.0); (D) Development of an OS chip model to investigate tumor stroma-cell interactions and drug responses (Reproduced with permission from ref, Copyright the author 2023, licensed under CC-BY-NC-ND).

Overall, microfluidic chip technology enhances the physiological relevance of OS models and provides a powerful platform for investigating novel anticancer therapies.

4.2. OS-on-a-Chip Models

Microfluidic chip technology has enabled the development of highly biomimetic OS chip models, allowing researchers to reconstruct the dynamic interactions among OS cells, immune cells, osteoblasts, and vascular endothelial cells in vitro, thereby more accurately simulating the TME. Through precise control of fluidic conditions at the microscale, microfluidic systems facilitate the observation of immune cell chemotaxis toward tumor sites and enable quantification of their cytotoxicity, revealing critical mechanisms underlying tumor immune evasion. Moreover, this technology can be applied to investigate the interplay between osteoblasts and osteoclasts, providing deeper insights into the effects of OS on bone remodeling. For instance, a scaffold-free, fully humanized 3D microfluidic coculture bone remodeling model has been developed, in which human mesenchymal stromal cells differentiate into osteoblast lineages and self-assemble into bone-like tissue mimicking human trabecular structure and dimensions. Human monocytes adhere to these tissues and fuse into multinucleated osteoclast-like cells, establishing a functional coculture system (Figure C). In terms of drug screening and efficacy prediction, microfluidic chips offer a precise platform for rapid and efficient evaluation of antitumor agents. Studies have demonstrated that this technology supports the creation of 3D tumor cell cultures combined with real-time imaging to assess responses to various treatment regimens. For example, an OS-on-a-chip (OOC) model was constructed to study tumor-stroma interactions and drug responses by coculturing patient-derived OS cells with stromal components. This model was used to evaluate the combined therapeutic effects of the chemotherapeutic agent doxorubicin (DOX) and the CXCR4 inhibitor Plerixafor (Figure D). Particularly in immunotherapy research, microfluidic platforms are instrumental in assessing the efficacy of immune checkpoint inhibitors, such as PD-1/PD-L1 blockers, and in predicting individualized treatment responses, thereby optimizing precision therapy strategies for OS.

Taken together, the application of microfluidic chip technology in OS research not only enhances the physiological relevance of experimental models but also provides vital theoretical and practical support for the development of personalized therapeutic approaches.

4.3. Technical Challenges and Future Improvement Directions

Microfluidic chip technology has demonstrated significant potential in OS research due to its ability to precisely control the microenvironment, reduce reagent consumption, and enhance data reproducibility. However, its widespread adoption in conventional laboratories remains limited by high manufacturing costs, technical complexity, and dependence on specialized equipment. In recent years, researchers have integrated 3D printing technology with microfluidics to simplify fabrication processes and reduce production expenses, thereby improving the applicability of these chips for OS cell culture and drug screening. ,

Moreover, the incorporation of nanotechnology has enabled microfluidic chips to more effectively mimic the TME. For example, nanocomposite scaffolds have been utilized to simulate bone defect repair following OS surgery, combined with chemotherapy or photothermal therapy to enhance treatment efficacy. , Concurrently, the integration of high-throughput screening techniques allows microfluidic platforms to simultaneously evaluate multiple anticancer drug combinations, accelerating the development of precision therapies for OS.

Overall, the convergence of microfluidic chip technology with 3D printing, nanotechnology, and high-throughput screening offers promising new directions and applications for personalized treatment strategies in the OS.

5. Integrative Multiomics and Precision Medicine

5.1. Applications of Multiomics Technologies in OS Research

OS is a highly heterogeneous malignant bone tumor characterized by complex molecular features, pronounced genomic instability, and marked resistance to therapy, posing significant challenges for clinical diagnosis and treatment. In recent years, integrative multiomics analysis has emerged as a pivotal tool in precision medicine. By combining data from genomics, transcriptomics, proteomics, and metabolomics, researchers are able to gain a comprehensive understanding of the molecular mechanisms underlying OS. This approach facilitates patient stratification, elucidates drug resistance mechanisms, and supports the development of individualized therapeutic strategies. For instance, a recent study integrated genomic, epigenomic, and transcriptomic data from 121 OS patients. While a wide range of somatic mutations were observed, only TP53 showed a significant mutation frequency. Through unsupervised integrative clustering of multiomics data sets, OS was classified into four molecular subtypes with distinct features and clinical outcomes: (1) immune-activated (S-IA), (2) immune-suppressed (S-IS), (3) homologous recombination deficiency-dominant (S-HRD), and (4) MYC-driven (S-MD) (Figure A).

4.

Multiomics Integration and Precision Medicine. (A) Multiomics analysis classifies OS into four subtypes with distinct molecular characteristics and clinical prognoses (Reproduced with permission from ref, Copyright 2019, licensed under CC-BY-4.0); (B) Identification of six gene markers related to OS survival through multiomics analysis, used for assessing patient survival outcomes (Reproduced with permission from ref, Copyright 2021 Sage); (C) OOCs platform for personalized precision medicine applications (Reproduced with permission from ref, Copyright 2024 Wiley); (D) Drug sensitivity prediction based on multiomics data using a deep learning and SNF approach (Reproduced with permission from ref, Copyright the author 2023, licensed under CC-BY-4.0).

At the genomic level, the OS displays extensive genomic instability, including chromosomal aberrations, copy number variations (CNVs), and somatic mutations. These complex alterations are difficult to fully interpret through single-omics approaches alone. Therefore, integrated analyses combining transcriptomic, proteomic, and metabolomic data have gained traction as mainstream strategies for dissecting OS pathogenesis. Given the high degree of heterogeneity in OS, treatment responses vary considerably among patients. Accurate molecular subtyping is critical for optimizing therapeutic regimens. Multiomics analyses have identified six genes (MYC, CHIC2, CCDC152, LYL1, GPR142, and MMP27) associated with the survival prognosis. A prognostic model based on these markers demonstrated robust predictive performance, with area under the curve (AUC) values exceeding 0.85 across multiple data sets, indicating high clinical applicability (Figure B).

Current research has confirmed the broad applicability of multiomics data in tumor subtype classification, prognostic prediction, and elucidation of drug resistance mechanisms. , In the future, with advancements in single-cell sequencing, spatial transcriptomics, and artificial intelligence-based analytical techniques, the depth and precision of multiomics integration are expected to further improve. These developments will accelerate the progression of precision medicine in OS and offer novel strategies and directions for personalized therapy.

5.2. Integration of PDMs with Multiomics Approaches

The core of precision medicine lies in tailoring optimized therapeutic strategies based on an individual’s specific molecular characteristics. In recent years, PDOs, as highly physiologically relevant in vitro models, have been widely employed in cancer research. By integrating multiomics dataincluding genomics, transcriptomics, proteomics, and metabolomicsPDOs can be comprehensively characterized to enable personalized drug screening and enhance the synergy between clinical biobanks and in vitro models, thereby improving the efficiency and reliability of precision medicine research. PDO technology effectively preserves the genomic and phenotypic features of the original patient tumors. When combined with genomic and transcriptomic profiling, it facilitates drug sensitivity testing, providing a powerful platform for individualized therapy. Moreover, the emerging “organoids-on-a-chip” (OOCs) technologyan integration of PDO culture systems with microfluidic platformshas enabled high-throughput drug screening and multiomics data acquisition (Figure C). This approach not only enhances the stability of PDO cultures but also leverages multiomics analysis to predict personalized therapeutic responses, thereby improving the clinical translational potential of precision medicine.

PDOs, when combined with multiomics analyses, have played a pivotal role in personalized drug screening and precision medicine research. The integration of genomic, transcriptomic, proteomic, and metabolomic data with PDOs enables the development of highly tailored therapeutic strategies for cancer patients. Looking forward, as the integration of PDO technology and multiomics advances, it is expected to accelerate the clinical translation of precision medicine, ultimately offering more effective and individualized treatment options for patients.

5.3. The Role of Artificial Intelligence and Big Data in OS Research

With the rapid advancement of high-throughput omics technologies, precision medicine is increasingly evolving toward a data-driven paradigm. The accumulation of multiomics dataincluding genomics, transcriptomics, proteomics, and metabolomicshas provided a wealth of information for elucidating disease mechanisms, optimizing targeted therapies, and predicting drug sensitivity. However, due to the high dimensionality, heterogeneity, and complexity of such data, traditional bioinformatics approaches often fall short in fully extracting their potential value. In recent years, artificial intelligence (AI) and machine learning (ML) have made significant progress in the mining and modeling of multiomics data sets, offering novel methodologies for advancing personalized treatment strategies in OS.

In the field of precision medicine, AI technologies have significantly facilitated the identification of disease-related biomarkers through approaches such as knowledge graphs, deep learning, and network biology. One study employed four feature selection algorithmsMonte Carlo Feature Selection (MCFS), Boruta, Minimum Redundancy Maximum Relevance (mRMR), and LightGBMto identify key genes from gene expression matrices. Subsequently, four ML algorithmsSupport Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF), and k-Nearest Neighbors (kNN)were used to determine the optimal number of genes for model construction. By integrating interpretable machine learning (IML) techniques, a consensus prediction network was generated to identify decision rules and reveal potential relationships among the selected genes. Survival analysis demonstrated that INHBA, FNBP1, PDE9A, HIST1H2BG, and CADM3 were significantly associated with the prognosis of colorectal cancer (CRC) patients. This study underscores the importance of multiomics data integration in the discovery of disease-related biomarkers and highlights the potential of AI in advancing precision medicine.

Drug sensitivity prediction is one of the core tasks in precision medicine. ML and deep learning techniques have been widely employed to integrate multiomics data and improve predictive accuracy. Recently, researchers developed a novel approach that combines deep learning with Similarity Network Fusion (SNF) to integrate gene expression, CNV, and DNA methylation data. To mitigate the risk of overfitting associated with high-dimensional data, sparse principal component analysis (SPCA) was applied for feature extraction (Figure D). The results demonstrated that this method outperformed existing deep learning models in predictive accuracy across multiple cancer data sets, such as GDSC and TCGA. Moreover, it significantly enhanced the predictive capability for both targeted therapies and nonspecific chemotherapeutic agents.

The application of AI and ML in drug development is rapidly expanding, particularly in areas such as novel drug discovery, indication expansion, and clinical trial optimization. Recent studies have reviewed the roles of ML in drug discovery, target identification, and the refinement of personalized treatment strategies, emphasizing its widespread use in high-throughput drug screening, molecular feature analysis, and clinical decision-making in precision medicine. Moreover, ML is increasingly being employed in the clinical trial phase, particularly in optimizing patient recruitment and predicting therapeutic responses, offering novel solutions for advancing precision medicine.

AI and ML technologies are revolutionizing the mining of multiomics data and advancing the field of precision medicine. Approaches such as knowledge graph-based target discovery, deep learning-enhanced drug sensitivity prediction, and data-driven personalized treatment strategies collectively demonstrate the transformative impact of AI in biomedical research. With the continuous accumulation of multiomics data sets and ongoing improvements in ML algorithms, AI is poised to play an increasingly pivotal role in the personalized management of OS. These advancements are expected to propel medical research toward more intelligent, precise, and individualized therapeutic paradigms, ultimately improving patient outcomes.

6. Major Challenges in In Vitro Models of OS

OS is a highly aggressive primary bone malignancy characterized by pronounced molecular heterogeneity and invasive behavior that continue to pose significant challenges for therapeutic decision-making. Although in vitro models have become indispensable tools for drug screening, mechanistic investigation, and precision medicine research, several critical limitations remain. These include inconsistencies in model heterogeneity and lack of standardization, insufficient clinical relevance and translational applicability, high technical complexity and associated costs, and the need to balance high-throughput screening with personalized treatment strategies.

6.1. Model Heterogeneity and Lack of Standardization

Research on in vitro models of OS encompasses a wide range of approaches, including conventional 2D cell lines, 3D culture systems, PDOs, microfluidic chips, and animal models. However, significant variability exists across laboratories in terms of cell sources, culture conditions, and experimental methodologies, resulting in poor reproducibility of experimental outcomes. Moreover, the lack of standardized evaluation criteria and operating protocols hampers the data comparability between studies. For instance, variations in scaffold materials, cell density, and medium composition in 3D culture systems can all influence experimental results. Therefore, to enhance the reproducibility and clinical translational value of osteosarcoma research, it is imperative to establish unified technical standards and standardized operational protocols. By defining key parameters, such as cell sources, culture systems, and functional evaluation criteria, systematic bias across laboratories can be minimized at the source, thereby improving data reliability and comparability. In the future, standardized methodologies and operational guidelines will serve as essential foundations for integrating research data across different studies, fostering interdisciplinary collaboration and innovation and ultimately enabling the construction of highly biomimetic osteosarcoma models capable of accurately predicting clinical responses.

6.2. Limited Clinical Relevance and Translational Potential

Although in vitro models offer enhanced controllability and reproducibility, their limited resemblance to the in vivo TME often results in a significant gap between experimental findings and clinical applications. For example, conventional 2D cultured OS cell lines lack the TME, making it difficult to recapitulate the authentic features of tumor growth and drug response. In recent years, researchers have developed 3D culture systems and PDOs to improve the clinical relevance of preclinical models­(Table ). For instance, a patient-derived OS organoid biobank was recently established, demonstrating a superior predictive power for individualized drug responses. In addition, recent studies have employed osteosarcoma organoids and 3D tissue-engineered osteosarcoma models to investigate the mechanisms underlying resistance to clinical anticancer oxidative stress therapies and cisplatin treatment. − Although 3D osteosarcoma models have achieved remarkable breakthroughs in basic research, their widespread clinical application remains limited, and the translational pathway from laboratory discovery to clinical practice is still long and challenging. In the future, clinical correlation studies could be promoted as benchmarks for evaluating the predictive performance of these models. For instance, by simultaneously assessing drug responses in models and comparing them with the actual clinical outcomes of patients receiving the same therapies, the predictive accuracy, sensitivity, and specificity of the models can be quantitatively determined. Furthermore, the integration of model-derived data with clinical data sets at multiple levelslinking molecular subtypes, gene expression profiles, and model phenotypes with patient pathology, imaging, and survival datamay facilitate the identification of novel biomarkers predictive of therapeutic response or prognosis. In parallel, establishing standardized biobanks of osteosarcoma models could minimize interindividual variability, while interventional studies guided by model-based drug sensitivity profiling may inform treatment strategies for patients with relapsed or refractory disease. Ultimately, these efforts will drive the evolution of 3D osteosarcoma models from experimental research tools toward mature clinical predictive systems and decision-support platforms.

3. Clinical Translational Applications of In Vitro OS Models.

Application Area	Specific Content	Applicable Model	Clinical Advantages	Challenges	ref.
Personalized Therapy	construction of models using PDCs to identify the most effective drugs	PDMs, organoids	improves treatment success rates, reduces ineffective therapies	requires optimization of culture conditions and establishment of standardized protocols	,
Precision Drug Screening	evaluation of drug efficacy and toxicity on different in vitro models	3D spheroids, organ-on-a-chip	predicts tumor response to therapeutics improves screening efficiency	in vitro results still require validation against in vivo outcomes	,
Drug Resistance Studies	investigating mechanisms of resistance and identifying combination therapies	spheroids, organoids	identification of novel resistance targets and optimization of therapeutic strategies	in vitro resistance models may not fully replicate clinical scenarios	,,
Immunotherapy Screening	investigating the tumor immune microenvironment and testing immunotherapeutic agents	immune coculture systems, organ-on-a-chip models	optimization of immunotherapy regimens and enhanced efficacy	need to develop more complex immune system models	,
3D-Printed Personalized Models	construction of OS structural models based on patient imaging data	3D bioprinting	assists surgical planning and predicts treatment outcomes	further improvement of biocompatibility of printing materials is required	,

6.3. Technical Complexity and Cost Constraints

Despite the significant progress made by emerging technologies such as 3D bioprinting, microfluidic chips, and organoid cultures in enhancing the physiological relevance of OS models, their widespread application remains limited due to high technical complexity, elevated costs, and lengthy experimental timelines. For instance, the cultivation of PDOs often requires several weeks to months and is highly dependent on tumor tissue origin, culture conditions, and patient-specific variability. To lower the technical barriers and enhance experimental efficiency, research efforts can shift from constructing highly complex model systems toward developing simplified, standardized, and cost-effective alternatives without compromising the predictive performance. For instance, the development of pump-free microfluidic systemssuch as those driven by gravity, siphon flow, or micropressurecould simplify device operation and reduce system complexity. Meanwhile, establishing standardized protocols for bioink preparation and advancing automated bioprinting platforms will enhance the batch-to-batch stability and reproducibility of 3D bioprinting. In addition, artificial intelligence (AI) algorithms can be employed to analyze experimental data and provide data-driven predictions and guidance for optimizing culture conditions, thereby improving model accuracy while effectively controlling experimental costs.

6.4. Balancing High-Throughput Screening with Personalized Therapy

High-throughput drug screening relies on rapid, stable, and reproducible in vitro models; however, these models often lack patient-specific characteristics. In contrast, personalized treatment modelssuch as PDOs and patient-derived xenografts (PDXs)more closely recapitulate the biological features of individual patients, but their establishment and maintenance remain technically challenging and resource-intensive. Recent studies have suggested that PDOX models better mimic the native TME compared to conventional PDX models and offer greater utility for personalized drug screening. Additionally, researchers have developed an optimized OS spheroid culture platform designed to strike a balance between high-throughput screening and personalized therapy. Evidence suggests that this platform not only supports large-scale drug screening but also enables the accurate prediction of individualized drug responses.

7. Future Directions and Perspectives

The future development of in vitro OS models is expected to advance toward more biomimetic and integrated systems that more accurately recapitulate the TME in vivo. The future development of osteosarcoma in vitro models is expected to advance toward more biomimetic and integrated systems that can more precisely recapitulate the in vivo tumor microenvironment. Future research will focus on incorporating multiple cell typesincluding immune cells, osteoblasts, and vascular endothelial cellsinto physiologically relevant composite models. For example, microfluidic chip technology can be employed to enable coculture of multiple cell types. By designing microfluidic devices with multiple culture chambers separated by integrated porous membranes, physical isolation between compartments can be achieved while still allowing the free exchange of small signaling molecules such as cytokines and metabolites, thereby facilitating intercellular communication. Alternatively, 3D bioprinting can be utilized to construct biomimetic multilayered architectures, spatially positioning different cell types within distinct regions of the model for coculture.

The construction of next-generation osteosarcoma models will rely on the integration of multiple cutting-edge technologies with the core objective of achieving breakthroughs in model stability, reproducibility, and clinical relevance. For instance, an automated and intelligent osteosarcoma culture system could be established, combining 3D bioprinting and microfluidic technologies to create and maintain a biomimetic dynamically regulated culture environment. The integration of artificial intelligence (AI) algorithms with embedded microsensors will enable real-time monitoring of organoid growth, viability, and other physiological parameters, providing continuous feedback. Moreover, data collected from such systems could be compared with global databases to systematically evaluate the reliability and predictive value of the platform in modeling disease progression and serving as a high-fidelity tool for drug screening.

In clinical applications, in vitro OS models are poised to become essential tools for new drug development and precision medicine. Strengthening collaborations with pharmaceutical companies will facilitate the integration of 3D culture systems, organoids, and microfluidic chips into preclinical drug screening, thereby improving drug development success rates. Concurrently, combining PDOs with multiomics data analysis can optimize personalized treatment strategies and enhance drug response rates across diverse patient populations. Furthermore, exploring rapid approval pathways based on organoid and in vitro models may help shorten the timeline from laboratory research to clinical application, accelerating the advancement of precision medicine for OS and ultimately providing patients with more effective therapeutic options.

8. Conclusion

Significant progress has been made in the development of OS in vitro models, evolving from traditional 2D cell cultures to more physiologically relevant 3D cultures, organoid models, and immune coculture systems. These advanced models more accurately recapitulate the TME, demonstrating great potential in studying cell–cell interactions, drug screening, and resistance mechanisms. Furthermore, the integration of technologies such as microfluidic chips, 3D bioprinting, and PDOs has accelerated the development of personalized therapeutic strategies, enhancing the feasibility of precision medicine for OS. Despite ongoing challenges related to model stability, standardization, and clinical translation, the application of these innovative technologies has laid a solid foundation for the optimization of future treatment approaches.

Biomimetic in vitro models play an indispensable role in the precision medicine of OS. Compared to traditional models, 3D coculture systems and PDMs better recapitulate tumor heterogeneity and effectively evaluate the efficacy of individualized therapeutic regimens, thereby facilitating the development of novel targeted therapies and immunotherapies. The integration of multiomics analyses with AI further elucidates the biological characteristics of OS, enhancing the accuracy of drug screening and clinical outcome prediction. Therefore, future research should focus on optimizing biomimetic models by integrating advanced bioengineering, nanotechnology, and big data analytics to improve model stability and translational value.

To accelerate the standardization and clinical translation of advanced in vitro models, enhanced collaboration between academia and industry is essential to facilitate the transfer of research findings into clinical applications. Establishing interdisciplinary and cross-institutional international collaboration platforms, standardizing cultivation protocols for in vitro models, and promoting the development of shared databases will improve the reproducibility and comparability of the studies. Furthermore, the integration of resources from governments, research institutions, and enterprises is crucial to promote the widespread application of in vitro models in drug discovery, preclinical validation, and personalized therapy. These efforts will shorten the time from laboratory research to clinical practice, ultimately providing OS patients with more precise and effective treatment options.

∥.

(J.L., B.R.) These authors contributed equally to this work. Conceptualization: Jing Liu, Haochen Liu. Investigation: Bihan Ren, Jing Liu. Methodology: Qian Wang, Dingming Li, Tianma He. Formal analysis: Bihan Ren, Tao Ding. Writing – original draft: Jing Liu, Bihan Ren. Writing – review and editing: Haochen Liu.

The authors declare no competing financial interest.