Advancing oral squamous cell carcinoma research: the evolving role of patient-derived cell lines, xenografts, and organoid models

Abstract

Oral squamous cell carcinoma (OSCC) is a prevalent and aggressive malignancy associated with high morbidity and mortality. Despite advancements in therapeutic strategies over the past decades, patient survival has not improved significantly due to underlying genomic heterogeneity, drug resistance, and late-stage diagnosis. Traditional preclinical models have contributed significantly to OSCC research, but often fail to fully recapitulate the complexity of the tumor microenvironment (TME) and patient-specific responses. This review comprehensively outlines the diverse preclinical models used in OSCC research, including two-dimensional (2D) monolayer cultures, three-dimensional (3D) spheroids, organoids, animal models, and patient-derived xenografts (PDXs). Each model is discussed in terms of its biological relevance, advantages, limitations, and contributions to translational research. Particular attention is given to the emerging potential of organoid systems and PDXs in personalized medicine and high-throughput drug screening. Furthermore, we underscore the importance of integrative approaches combining genomics, proteomics, and pharmacological data to better recapitulate tumor heterogeneity and predict therapeutic response. This review aims to guide researchers in selecting appropriate preclinical platforms to accelerate therapeutic development and precision oncology in OSCC.

Background

Oral cancer is a subset of head and neck cancers that originates in the oral cavity, involving subsites such as the buccal mucosa, floor of the mouth, anterior tongue, alveolar ridges, retromolar trigone, hard palate, and the inner surface of the lips (Scully et al. 2008; Rivera 2015). Most oral cancers arise from squamous epithelial cells and are therefore commonly referred to as oral squamous cell carcinoma (OSCC). Globally, oral cancer accounts for over 389,846 new cases and 188,438 deaths annually, making it the 16th most common cancer worldwide, according to GLOBOCAN 2022 (Bray et al. 2024). The most well-established risk factors for oral cancer include the consumption of tobacco and alcohol (Huang et al. 2013; Grigoryeva et al. 2015). Additional contributing factors include exposure to viral, bacterial, and fungal pathogens, which may facilitate carcinogenesis (Yen et al. 2009).

Current treatment regimens for oral squamous cell carcinoma (OSCC) typically involve a combination of surgery, radiotherapy, immunotherapy, photodynamic therapy, and targeted therapies, such as EGFR inhibitors and COX-2 inhibitors (Gharat et al. 2016). Treatment strategies are decided based on several clinical factors, including tumor stage, lymph node metastasis, HPV status, smoking history, and functional impairments in the oral cavity. Despite these structured and multimodal approaches, treatment outcomes remain suboptimal for many patients due to high tumor heterogeneity, intrinsic or acquired chemo resistance, and off-target effects that result in non-specific cell death (Feller et al. 2023).

Moreover, most existing diagnostic tools detect OSCC only at advanced stages, limiting the effectiveness of therapeutic interventions and complicating disease management. Nevertheless, a significant proportion of OSCC patients continue to experience tumor progression or recurrence, often within two years post-treatment, with reported recurrence rates ranging from 15 to 60% (Sasaki et al. 2011). Major contributors to recurrence include a lack of early-stage diagnosis, the presence of drug-resistant cell populations, and the emergence of subclones with drug efflux capabilities(Law et al. 2021). These challenges emphasize the importance of advancing molecular insights into OSCC to report the development of more effective and individualized therapeutic strategies.

In vitro cell culture models are widely used to investigate and elucidate the molecular mechanisms underlying cancer progression and facilitate drug target identification and discovery (Arango et al. 2013). Research typically progresses from in vitro studies to animal models, and ultimately to clinical applications. One of the key advantages of using established in vitro cell culture models is their homogeneity and reproducibility, which allows for the generation of consistent and comparable experimental results (Segeritz and Vallier 2017). In this review, we highlight the various in vitro and in vivo models employed in OSCC research, with a particular focus on 3D models that more accurately mimic the tumor microenvironment to enhance drug discovery and translational success. We also discuss the advantages and limitations of each model system in the context of OSCC biology and therapeutic development.

Cell line models

Cell lines have been the cornerstone of in vitro cancer research due to their accessibility, ease of handling, and scalability. Traditional 2D culture systems, where cells are grown as monolayers in flasks or plates, remain the most widely adopted approach. These models offer a cost-effective platform for studying drug responses, cellular signaling, and gene expression (K et al. 2017).

Several OSCC cell lines have been developed and extensively characterized. Amongst them, SCC-4, SCC-9, SCC-25, Cal27, BHY, HY, and OCEM-1 are widely used in OSCC research. SCC-4 and SCC-25 are derived from tongue tissue. SCC-4 originates from a pre-treated patient and exhibits weak colony formation, whereas SCC-25 demonstrates greater invasiveness (Rheinwald and Beckett 1980). SCC-9, obtained from an untreated young patient, shows high resistance to cisplatin and has been widely used in gene function studies(Huang et al. 2017). CAL27, another tongue-derived squamous cell carcinoma line, has been central to studies of oral cancer research, cell proliferation, metastasis/invasion, and apoptosis (Gioanni et al. 1988; Jiang et al. 2009, 2025; Tadijan et al. 2021; Wang et al. 2021a). BHY, derived from a highly aggressive lower jaw tumor, is noted for its strong differentiation capacity (Erdem et al. 2007). HN originates from a soft palate metastasis with subsequent spread to the lungs and brain, offering insights into advanced disease stages.

Another well-characterized cell line is OECM-1, derived from buccal mucosa. It exhibits epithelial morphology and is frequently used to study OSCC progression, drug resistance, and signaling pathways. For instance, treatment of OECM-1 cells with specific compounds has been shown to induce tumor suppressors (p53, p21) and inhibit cell cycle regulators, pointing to their utility in therapeutic screening (Lee et al. 2020). Studies have also demonstrated that IL-1RA promotes malignancy in OSCC through the EGFR/PI3K/AKT pathway, highlighting a potential target for intervention (Yuan et al. 2023). Additionally, IL-1β secreted by macrophages was shown to enhance chemotherapy resistance and correlate with poor overall survival, thereby emphasizing the importance of tumor–immune interactions (Hsieh et al. 2022).

Despite their ease of handling, low maintenance costs, and experimental convenience, 2D cell culture models have significant limitations. One of the primary drawbacks is their inability to accurately recapitulate the native architecture, morphology, and complexity of in vivo tissues. This structural limitation alters critical aspects of cellular behavior, including cell–cell and cell–matrix interactions, which play key roles in regulating cell differentiation, proliferation, and viability, further influencing gene and protein expression profiles (Kapałczyńska et al. 2018). As a result, data generated from 2D systems may not reliably reflect physiological responses, leading to inconsistencies when translated to in vivo models or clinical settings. The limited predictive value of 2D cultures contributes to the high attrition rate in drug development, with pharmaceutical companies incurring substantial costs due to late-stage failures in clinical trials, necessitating the development of more physiologically relevant models.

Animal models

The development and application of animal models are critical for translating in vitro findings into effective clinical interventions. While in vitro cell culture studies provide valuable insights into the molecular and biochemical pathways of cancer, they fail to capture the full complexity of disease, including systemic interactions and tumor microenvironment dynamics. Animal models bridge this gap by closely mimicking the complexity of human disease, making them physiologically relevant models for both basic and translational cancer research.

Species such as mice, rats, and zebrafish are commonly used in cancer research due to their genetic and physiological similarities to humans. These models enable investigation of whole-organism processes, such as immune responses and inter-organ communication, which are often impractical or unethical to study in humans. They also enable the study of complex systemic interactions, including those involving the endocrine, cardiovascular, and nervous systems, as well as cell-to-environment dynamics that cannot be adequately captured by in vitro systems. These animal models are broadly classified into three categories: chemically induced models, genetically engineered models (GEMs), and xenograft models.

Chemically induced models

Chemically induced oral cancer in vivo models are generated by exposing animals to known carcinogens such as tobacco or benzopyrene. These models have significantly advanced our understanding of carcinogenesis by replicating environmental exposure-related tumor development. However, chemically induced models have limitations, including long latency periods for tumor formation (30–50 weeks), off-target mutations caused by the carcinogens, and nonspecific toxicity that interferes with the proper interpretation of the results (De Minicis et al. 2013).

GEM models

GEMs address many of the limitations associated with chemically induced models. These models involve the activation of oncogenes or the deletion of tumor suppressor genes genetically to induce carcinogenesis in a controlled and reproducible manner (Kersten et al. 2017). GEMs offer higher specificity and lower variability, making them more suitable for mechanistic studies and preclinical evaluation of targeted therapies. However, GEMs may still fall short in recapitulating the full heterogeneity and microenvironment of human tumors (Blake et al. 2021). Due to this genetic disparity, and differences in the expression profile of mouse and human genes, the in-depth research for a particular signaling pathway or therapy regimen needs to be carried out on human cell-derived tumors, rather than mouse cell-derived tumors (Richmond and Su 2008).

Xenograft models

The term xenograft refers to the transplantation of foreign tissue into a host organism. Xenograft models involve the implantation of human tumor cells or tissues into immunocompromised mice, such as athymic nude or SCID mice, to avoid immune rejection. The source of the transplanted cells determines whether the xenograft model is cell-derived or patient-derived (Fig. 1). Cell line-derived xenografts (CDXs) use immortalized human cancer cell lines that form tumors when implanted into mice (Fig. 1). CDX models are used to study various aspects of cancer biology and therapeutic interventions that including evaluating tumor growth kinetics, metastasis, tumor microenvironment interactions, and assessing response to different treatment modalities, such as chemotherapy, targeted therapies, immunotherapies, or combination therapies. Several Cell Line-Derived Xenograft (CDX) models have been extensively employed in oral squamous cell carcinoma (OSCC) research to study tumor biology and assess therapeutic efficacy. For instance, Gohara et al. used OSCC-derived CDX models to assess the efficacy of oncolytic viruses in radioresistant tumors (Gohara et al. 2022), Xu et al. employed KB cell line-derived xenografts to evaluate prodrug-activating suicide gene therapies(Xu et al. 2023), Wei et al. assessed genipin’s anticancer potential using SAS cell-derived xenografts (Wei et al. 2020), and Xu et al. utilized HSC-3 xenografts to explore therapeutic strategies in OSCC (Xu et al. 2023).

Fig. 1.

Sub-classification of the Xenograft Models based on the cell types. Xenograft models are broadly classified into Cell-Derived Xenografts (CDX), which utilize immortalized cancer cell lines, and Patient-Derived Xenografts (PDX), which involve implantation of tumor tissues directly obtained from patients. Both CDX and PDX models can be established either orthotopically—at the original tumor site—or ectopically, most commonly via subcutaneous implantation. These models are essential tools for investigating tumor biology and evaluating therapeutic responses in preclinical studies

Patient-derived xenografts (PDXs), on the other hand, are generated by implanting fresh tumor tissue obtained directly from OSCC patients into immunocompromised mice (Fig. 1). These models retain the histological architecture, genetic profile, and heterogeneity of the original tumor, making them invaluable for studying clonal dynamics, tumor progression, and personalized therapy responses(Xu et al. 2023). Recent studies have successfully established OSCC PDX models, confirming their fidelity through histopathological and molecular analyses. These models have been instrumental in evaluating therapeutic responses and understanding tumor behavior (Campbell et al. 2018, Eirew et al. 2015, Hammerman et al. 2015, Hidalgo et al. 2014). For instance, research has demonstrated that PDX models retain key features of the donor tumors, including perineural invasion and activation of pathways like PI3K-Akt and epithelial–mesenchymal transition, which are associated with aggressive tumor behavior and poor prognosis (Yen et al. 2022). The establishment of OSCC PDX models also facilitates the identification of potential biomarkers for tumor engraftment and progression. For example, high expression of matrix metalloproteinase-1 (MMP1) has been associated with successful engraftment and may serve as a predictive marker for tumor aggressiveness (Yen et al. 2022). A study investigated the impact of photobiomodulation therapy (PBMT) on OSCC PDX models to measure alterations in tumor growth or histopathological characteristics and concluded that PBMT does not adversely affect tumor behaviour in physical characteristics (Silveira et al. 2023).

While CDX models are ideal for standardized drug efficacy testing and mechanistic studies, PDX models are more suitable for personalized therapy development and translational research. However, both systems have limitations. CDXs fail to capture full tumor heterogeneity and often consist of clonal populations that may not reflect patient tumors. PDXs may not exhibit the full disease course, including metastasis. Additionally, both CDX and PDX models lack a functional immune system, limiting their utility in immunotherapy studies. Engraftment site also matters—ectopic sites may not replicate the behavior of tumors in their native tissue, leading to the adoption of more physiologically relevant orthotopic xenografts, despite their technical complexity(Pompili et al. 2016; Li et al. 2020). These limitations often contribute to discrepancies between preclinical efficacy and clinical outcomes, underscoring the need for more human-relevant models such as patient-derived organoids and organ-on-chip systems. Nonetheless, these models remain foundational tools for preclinical testing, helping assess tumor growth, drug efficacy, toxicity, and guiding personalized treatment approaches. Table 1 summarizes the well-studied xenograft models in oral cancer research.

Table 1.

Xenograft models studied in oral cancer

Gene target	Pathway involved	Mechanism of action
LSL-K-rasG12D	RAS/MAPK	The proto-oncogene KRAS can be activated by a single alteration in an amino acid. Using either the keratin 5 (K5) or keratin 14 (K14) promoter, the oncogene K-rasG12D is purposefully elevated in the oral epithelium of mice in this investigation. While K14 is mostly located in the basal layer of the oral mucosa and tongue, K5 is found in the basal epithelium of the tongue and fore stomach (Caulin et al. 2004)
Tgfbr1/Pten 2cKO mice	AKT/mTORC1	The oral epithelium showed hyperplastic alterations after four weeks of tamoxifen induction (Bian et al. 2012) Following a total of 10 weeks of tamoxifen therapy, there was evident tumour growth in the oral cavity with particular histological characteristics. The loss of PTEN caused the epithelium to proliferate more, but the deletion of Tgfbr1 resulted in an increase in inflammation, accelerated angiogenesis, and immunosuppression.(Ishida et al. 2017)
PIK3CA-GEMM + 4NQO	Phosphatidylinositol 3-kinase (PI3K)/PDK1 pathway & TGF-β signaling	The buccal and tongue epithelium showed a distinct increase in PIK3CA expression after RU486 induction. Additionally, 39% of mice at 5–6 months and 100% of mice at 10–12 months developed SCCs after exposure to 4NQO. In addition to promoting cell growth and survival, this overexpression of PIK3CA promotes tumour invasion and metastasis to the lymph nodes and lung. Instead of AKT, molecular analysis reveals that 3-phosphoinositide-dependent protein kinase (PDK1) and increased transforming growth factor (TGF) signaling accelerate the advancement of PIK3CA-driven HNSCC. (Du et al. 2016)
Smad4 del/del	TGF-β signaling	In head and neck epithelial cells, deletion of the SMAD4 gene increases genomic instability(Tinhofer et al. 2020). As a result, the levels of the proteins Brca1 and Rad51 drop, which triggers severe inflammation marked by increased TGF-1 production and SMAD3 activation (Bornstein et al. 2009)
N-myc downregulated gene 2 (NDRG2)	PI3K/Akt signaling	Promoter methylation → NDRG2 mRNA downregulation → level of phosphorylated Akt at Serine 473 decreased → Increased cell proliferation & Metastasis → Tumorigenesis(Furuta et al. 2010; Tamura et al. 2017)

The use of CDX and PDX models in research also raises significant ethical concerns. Both require immunocompromised animals, necessitating strict adherence to institutional and national ethical guidelines that uphold the principles of Replacement, Reduction, and Refinement (3Rs) to minimize animal suffering. Additionally, PDX models involve the use of patient-derived biological materials, bringing further ethical complexities such as obtaining informed consent, ensuring donor privacy, and maintaining transparency in the use and sharing of human tissues. Furthermore, issues related to commercialization and equitable access to therapies developed using these models must be addressed through appropriate ethical oversight. As PDX models become increasingly prevalent in preclinical research, continued ethical reflection and regulatory vigilance are essential to ensure a balance between scientific advancement and responsible research practices.

Cell-based 3D models: a targeted shift in approach

For over a century, two-dimensional (2D) cell lines and animal models have served as foundational tools in biological research. They have contributed significantly to our understanding of cellular biology, signaling pathways, disease mechanisms, biomarker discovery, and therapeutic target identification. However, these models fail to replicate critical aspects of native tissue architecture, including dimensionality, cellular heterogeneity, and hierarchical organization, and cell–cell or cell–matrix interactions. As a result, they are limited in their ability to accurately model the complex physiology of human diseases (Me et al. 2021).

Ethically, the use of rodents raises concerns about animal welfare, particularly in procedures that cause pain or require euthanasia, and is subject to increasing scrutiny under the 3Rs principle (Replacement, Reduction, and Refinement). Despite extensive preclinical testing in these systems, many drug candidates ultimately fail in clinical trials due to poor efficacy or unforeseen toxicity (Arrowsmith and Miller 2013). These shortcomings highlight the urgent need for more physiologically relevant in vitro models that can better predict human responses.

To overcome the limitations of traditional tumor models, three-dimensional (3D) cell culture systems have emerged as powerful tools that bridge the gap between 2D in vitro and in vivo animal studies. Unlike monolayer cultures, 3D cell cultures recapitulate key features of native tissues, including architectural complexity, multicellularity, and functional differentiation, among these, spheroids and organoids are the most commonly used models, with growing applications in disease modeling, drug screening, and personalized medicine.

3D cultures can be established using anchorage-independent methods (e.g., hanging drop, liquid overlay, rotating wall vessels, magnetic levitation, low-attachment plates), anchorage-dependent systems that incorporate natural matrices (e.g., Matrigel, collagen, alginate, hyaluronic acid) or synthetic polymers (e.g., PEG, PGA, PLGA, PCL), and specialized microfluidic platforms (organ-on-a-chip) that simulate dynamic physiological conditions (Fig. 2) (Langhans 2018). Natural matrices more accurately mimic the tumor microenvironment (TME) due to their biocompatibility and inherent adhesive properties but are limited by batch-to-batch variability. Conversely, synthetic matrices offer tunable mechanical and chemical characteristics and are more cost-effective, though they often lack the biological cues necessary for full cellular function. Together, these advanced 3D systems offer a more physiologically relevant context for studying complex biological processes and enhancing the translatability of preclinical findings.

Fig. 2.

Blueprint of different culture platforms available for 3D culture. This figure illustrates the major categories of 3D cell culture models used to replicate the complex architecture and microenvironment of in vivo tissues. These models are broadly classified into three types: (1) Anchorage-dependent models, which rely on natural or synthetic scaffolds—such as hydrogels, collagen, or Matrigel—to support cell attachment, proliferation, and differentiation by mimicking the extracellular matrix (ECM); (2) Anchorage-independent models, where cells spontaneously form multicellular spheroids without requiring a substrate, commonly via techniques like the hanging drop method, ultra-low attachment plates, or spinner flasks, thereby simulating avascular tumor conditions and promoting enhanced cell–cell interactions; and (3) Specialized 3D culture platforms, including microfluidic platforms and organ-on-chip technologies, which incorporate fluid dynamics, mechanical forces, and spatial organization to emulate tissue-level functions. Together, these models offer enhanced physiological relevance and are widely used in cancer biology, drug discovery, stem cell research, and regenerative medicine

Spheroid models

Spheroids were the first widely adopted 3D culture models to address the limitations of 2D systems (Mueller-Klieser 1997; Layer et al. 2002). Depending on their cellular composition, spheroids can be categorized as either single-cell type or multicellular tumor spheroids (MCTSs). These models rely on homotypic cell–cell adhesion and the inhibition of cell–substrate interactions, facilitating the spontaneous self-assembly of cells into 3D aggregates. Commonly used techniques to generate spheroids include the hanging drop, liquid overlay, matrix embedding, and non-adherent surface culture methods. Spheroid size and architecture are determined by the type and number of seeded cells, and these models support the co-culture of multiple cell types, extracellular matrix (ECM) deposition, and the formation of spatially heterogeneous microenvironments.

The architectural and physiological features of spheroids resemble non-vascularized tumor nodules, making spheroids valuable for cancer research (Laschke and Menger 2017). Features such as hypoxic cores and nutrient gradients, which more accurately mimic in vivo tumor conditions, lead to differential cell proliferation and drug responses not observed in 2D monolayer cultures (Laschke and Menger 2017). For instance, the EGFR inhibitor Erlotinib demonstrated promising results in 2D glioblastoma models but ultimately failed in phase II clinical trials due to lack of efficacy—highlighting the predictive limitations of 2D systems (Gomez-Roman et al. 2017). In contrast, Tirapazamine, a hypoxia-activated prodrug, showed significantly enhanced activity in 3D cultures when combined with cisplatin in melanoma and non-small cell lung cancer models, emphasizing the relevance of 3D systems in preclinical drug evaluation (Denny and Wilson 2000).

Spheroid models have become essential tools in oral squamous cell carcinoma (OSCC) research, offering a more accurate representation of the in vivo tumor microenvironment than traditional 2D cultures. Studies using OSCC-derived spheroids have demonstrated increased resistance to chemotherapeutic agents like cisplatin and 5-fluorouracil, reflecting the drug tolerance observed in clinical settings (Ishiguro et al. 2017). Spheroids have also been employed to assess the role of hypoxia-inducible factors (HIFs) in promoting OSCC invasiveness and metastasis. Furthermore, advanced 3D platforms using ultra-low attachment plates and Matrigel scaffolds have enabled the co-culture of OSCC cells with stromal or immune cells, enhancing the understanding of tumor-stroma interactions and immune evasion mechanisms (Lo et al. 2023). The findings from these and related studies underscore the growing importance of spheroid models in preclinical drug screening and in developing personalized.

Organoids models

The term"organoid"refers to a three-dimensional (3D) cellular structure that resembles an organ in terms of function and architecture (Me et al. 2021). The concept dates back to the 1960 s, when developmental biologists used the term to describe organogenesis in dissociated and re-aggregated cell systems (Tang et al. 2022). The advent of organoids in experimental research began with Sato and colleagues, who successfully generated gut organoids from adult intestinal stem cells using 3D culture in Matrigel—a basement membrane matrix derived from Engelbreth-Holm-Swarm mouse sarcoma cells (Huang et al. 2014). This was followed by generation of organoids from human pluripotent stem cells(Spence et al. 2011), including retinal (Eiraku et al. 2011; Nakano et al. 2012), brain (Lancaster et al. 2013) and kidney organoids (Takasato et al. 2014; Taguchi et al. 2014).

To maintain uniformity in organoid culture and research, the Intestinal Stem Cell Consortium established guidelines in 2012 defining ‘organoids’ as 3D cultures comprising multiple cell types, including both epithelial and mesenchymal cells. In contrast, 3D cultures consisting solely of epithelial cells are referred to as ‘spheroids’ (Me et al. 2021). Organoids mimic key physiological and pathological aspects of their tissue of origin, including cell–cell interactions, lineage commitment, and disease-specific traits. Figure 3 summarizes key applications of organoids across various research domains (Watanabe et al. 2017)(Beshiri et al. 2018)(Kozuka et al. 2017)(Voges et al. 2017)(Nie and Hashino 2017)(Broutier et al. 2017)(Schutgens et al. 2019).

Fig. 3.

Key applications of organoids in Disease biology. This figure highlights the major applications of organoid technology in disease modeling and clinical research. In drug screening, organoids enable high-throughput and more predictive testing of therapeutic agents, improving the assessment of efficacy and toxicity. In the context of personalized medicine, patient-derived organoids allow individualized drug testing and treatment selection, supporting precision oncology and tailored therapeutic strategies. Additionally, organoids contribute to regenerative medicine by providing platforms for tissue repair, transplantation studies, and organ replacement research. These versatile applications bridge the gap between conventional in vitro studies and clinical outcomes, advancing both basic science and translational medicine

Organoids can be derived from either pluripotent stem cells (PSCs) or adult stem cells (ASCs). PSCs include embryonic and induced pluripotent stem cells (iPSCs), while adult stem cells (ASCs) are harvested directly from patient tissues. An organoid can be started by a single ASC type, but a cell pool of PSCs is needed (Iakobachvili and Peters 2017; Me et al. 2021). ASC-derived organoids offer a better understanding of adult tissue repair, viral infection disease, and can also be used to study personalized medicine approaches, whereas PSC-derived organoids are preferable for investigating early organogenesis and human developmental biology(Tang et al. 2022). Examples of human ASC-derived organoids include colon(Sato et al. 2011), intestine(Sato et al. 2009), liver (Huch et al. 2013b), prostate(Karthaus et al. 2014), pancreas (Huch et al. 2013a), fallopian tube(Kessler et al. 2015), stomach(Barker et al. 2010), tongue (Hisha et al. 2013), and endometrium(Turco et al. 2017). PSC-derived organoids have been developed for the intestine (Spence et al. 2011), liver (Takebe et al. 2013), lung(Dye et al. 2015), kidney(Takasato et al. 2015), pancreas(Huang et al. 2015), stomach(McCracken et al. 2014), and retina (Völkner et al. 2016).

Organoids can be cultured using several distinct methodologies. One of the most widely used method is the submerged method, where cells are embedded in basement membrane extract (BME) and cultured in media enriched with cocktail of growth factors. The air–liquid interface (ALI) method is an advance approach for organoid culture where tissue explants or cells embedded in BME are placed in Boyden chambers or cell insert, which are placed in the media containing culture dish (Gunti et al. 2021). The insert is suspended above a culture dish containing media, so the upper surface of the gel is exposed to air, while nutrients diffuse from the medium below. This method is especially beneficial for cultivating more complex or differentiated structures, including epithelial and neuronal organoids. Additionally, low-attachment plates promote spontaneous aggregation and are ideal for high-throughput screening and early-stage 3D modeling.

OSCC Organoids

Organoids derived from OSCC cells were first reported in 2018. This pioneering study used an organoid model to investigate how OSCC cells evade antitumor immune responses and to identify novel immunotherapy targets (Tam et al. 2018). OSCC organoids are typically generated by embedding primary OSCC cells in Matrigel at a density of 1000–5000 cells per 30 μL of Matrigel per well in 24-well plates (Zhao et al. 2019, 2021; Wang et al. 2022). Ultra-low attachment plates have also been employed as an alternative method to support organoid formation (Tanaka et al. 2018; Wang et al. 2021b; Chen et al. 2023). While the specific composition of organoid culture media varies widely, a commonly used base formulation includes DMEM/F12 supplemented with 1 × N2, 1 × B27, human EGF, and A83-01. This base medium is often customized with additional components to optimize growth and differentiation. The Clevers group enriched the basal medium with N-acetyl-L-cysteine, nicotinamide, human FGF10, human FGF2, prostaglandin E2, CHIR99021, forskolin, as well as 4% R-spondin and 4% Noggin—both essential for Wnt pathway activation and stem cell maintenance. In contrast, Zhao et al. included gastrin in the medium, illustrating how formulations can be tailored based on tissue-specific requirements (Zhao et al. 2019; Driehuis et al. 2019).

OSCC organoids have been used for a variety of applications, including investigating tumour biology, drug screening, predicting treatment response, and understanding tumor–stromal interactions (Tanaka et al. 2018) (Zhao et al. 2019, 2021; Wang et al. 2022; Chen et al. 2023). Tanaka et al. established OSCC patient-derived organoids with a 30.2% success rate and characterized them using histological, molecular, and functional assays to inform precision therapy development (Tanaka et al. 2018). Driehuis et al. further advanced the field by establishing long-term organoid cultures from oral mucosa tissues. They developed a panel of 31 organoid lines derived from the oral mucosa of head and neck squamous cell carcinoma (HNSCC) patients and revealed selective drug sensitivities due to inter-patient variability, highlighting the potential of organoids for personalized therapeutic strategies (Driehuis et al. 2019).

Several studies have used OSCC organoids to uncover new therapeutic vulnerabilities. Zhao et al. demonstrated that lactate enhanced cancer stem cell (CSC)-like features in organoids via upregulation of CD133 and that this effect could be mitigated by inhibiting MCT1-mediated lactate transport (Zhao et al. 2019). Wang et al. (2021b) demonstrated that disulfiram (DSF) disrupts OPA1-dependent mitochondrial metabolism, leading to ATP depletion and AMPK-driven autophagy. Rescue experiments confirmed ATP depletion as the mechanism underlying DSF-induced cytotoxicity. Tam et al. (2018) showed that CEACAM1 expression in OSCC organoids contributes to NK cell resistance and that CEACAM1 knockdown enhances NK cell-mediated cytotoxicity, identifying CEACAM1 as a potential immunotherapy target. Moreover, Wang et al. (2022) found that knockdown of the chromatin modifier KMT2D impaired OSCC organoid formation, highlighting its role in tumor progression.

To investigate tumor–stromal interactions Zhao et al. showed that co-culturing CD44 + cells derived from OSCC organoids with CAFs enhanced the stemness and organoid formation ability of CD44 + cells (Zhao et al. 2021). Chen et al. expanded this approach by generating fibroblast-attached organoids (FAOs) from OSCC organoids and fibroblasts of adjacent normal tissues to create models that recapitulate tumor–stroma crosstalk and revealed Notch activation as a key mediator in fibroblast reprogramming to CAFs in tumour microenvironment (Chen et al. 2023). Collectively, these studies highlight the importance of tumor-stroma interactions in OSCC progression and suggest that lactate metabolism and Notch signaling could serve as potential therapeutic targets.

Limitations of organoids models

Despite the growing utility of organoids in OSCC research, several limitations persist. Most organoids consist predominantly of epithelial cells and lack immune, stromal, and vascular components, limiting their ability to model the tumor microenvironment (TME) and immune–tumor dynamics (Dijkstra et al. 2018) (Neal et al. 2018). Moreover, reproducibility remains a significant challenge due to variability in tissue source, Matrigel composition, culture media, dissociation protocols, and the absence of standardized quality control parameters, such as morphological benchmarks, passage limits, and genetic stability, further complicates inter-laboratory comparisons and clinical translation (Drost and Clevers 2018).

To address these challenges, several innovative approaches are emerging. Co-culture systems incorporating autologous immune cells, such as peripheral blood mononuclear cells (PBMCs) or tumor-infiltrating lymphocytes (TILs), are enhancing immune representation in organoid models (Sachs et al. 2018). Advanced culture platforms, including air–liquid interface (ALI) systems and organoid-on-a-chip technologies, are also contributing to a more accurate recapitulation of the tumor microenvironment (TME) (Neal et al. 2018) (Skardal et al. 2020). Concurrently, standardization efforts are underway, including the development of defined, serum-free media, commercially available organoid culture kits, and the establishment of organoid biobanks to minimize variability in culture protocols. Moreover, global initiatives such as the Human Cancer Models Initiative (HCMI) are promoting standardized methodologies and centralized biobanking to improve reproducibility and facilitate data sharing. Collectively, these advancements are progressively overcoming the limitations of traditional systems and enhancing the clinical relevance and translational potential of organoid models.

Ethical considerations for organoids models

The expansion of human-derived organoid models also raises critical ethical concerns. These include ensuring transparent informed consent for tissue donation, especially regarding potential future uses in research, commercialization, and regenerative medicine. There is a need for clarity on donor rights, the scope of consent (broad vs. specific), and third-party data access. Concerns about genetic privacy and re-identifiability are particularly acute given the increasing use of genomic data in organoid research. Additionally, the commercialization of organoid-based therapies may disproportionately benefit certain populations, exacerbating disparities in access to cutting-edge treatments. Harmonized international policies for consent, biobanking, and data governance are urgently needed to ensure socially responsible innovation in this space (Munsie et al. 2017).

Conclusion and prospects

The landscape of oral squamous cell carcinoma (OSCC) research has evolved significantly with the advancement of preclinical model systems. Traditional 2D cell lines and animal models have laid the groundwork for understanding OSCC biology and therapeutic responses, yet their limitations in recapitulating human tissue complexity and tumor microenvironment (TME) have become increasingly evident. The introduction of three-dimensional (3D) models, particularly spheroids and organoids, marks a pivotal shift toward more physiologically relevant in vitro systems. These models more closely mimic in vivo tumor architecture, heterogeneity, and cell–cell or cell–matrix interactions, thereby enhancing the translational value of preclinical findings. Table 2 summarizes the key features of the major preclinical models discussed in this review.

Table 2.

Comparative advantages and disadvantages of the described models to aid in selection of an optimal research approach

Model	Structural complexity	Cost & scalability	Tumor microenvironment mimicry	Relevance to drug screening & personalized medicine	Timeframe to establish
2D Cell Cultures	Low – lacks tissue architecture	Low cost, highly scalable	Poor – lacks stroma, ECM, and heterogeneity	Moderate – good for high throughput screening. Low personalization	1–2 weeks
3D Spheroids	Moderate – simple 3D architecture High–tissue–like with self-organization	Moderate cost, scalable	Moderate – lacks stromal/immune cells	Improved over 2D, limited personalization	2–3 weeks
Patient-Derived Organoids (PDOs)	High–tissue–like with self-organization	High cost, moderate scalability	High –retains genetic and phenotypic heterogeneity	High – suitable for personalized therapy testing	2–4 weeks
Cell Line-Derived Xenografts (CDX)	Moderate	High cost, limited scalability	Moderate – lacks full TME context	Moderate – useful for in vivo efficacy studies	1–2 months
Patient-Derived Xenografts (PDX)	High – close to native tissue	Very high cost, low scalability	Very high – preserves original tumor architecture	Very high – strong predictive value for clinical response	2–6 months
Genetically Engineered Mouse Models (GEMMs)	High	Very high cost, low scalability	High – includes native stroma and immune system	High – models genetic drivers, less useful for drug screening	3–6 months

Among these, patient-derived organoids (PDOs) represent a promising platform for personalized oncology, offering applications in drug screening, biomarker discovery, and mechanistic studies. Co-culture systems, organoid-on-a-chip platforms, and air–liquid interface cultures are addressing key limitations by incorporating stromal and immune components, thereby more accurately simulating the OSCC tumor microenvironment. However, challenges such as batch-to-batch variability, lack of standardized protocols, and limited integration of vasculature and immune compartments continue to hinder reproducibility and broader clinical translation.

Future research should prioritize the standardization and scalability of OSCC organoid models through defined culture media, commercial kits, and biobanking initiatives. Expanding multi-lineage co-cultures that include stromal, endothelial, and immune cells will be essential to better mimic the native TME and to evaluate immunotherapeutic responses more effectively. Additionally, coupling organoid technology with high-throughput drug screening, single-cell multi-omics, and CRISPR-based functional genomics holds immense potential to identify patient-specific vulnerabilities and resistance mechanisms. To ensure the ethical advancement of organoid-based research, clear policies surrounding informed consent, genomic data use, and commercial equity must be established. International collaboration will be critical in developing harmonized guidelines for biobanking, data sharing, and regulatory oversight.

In conclusion, each preclinical model of oral squamous cell carcinoma (OSCC) has its own advantages and limitations in OSCC research. The combined use of 2D, 3D, and in vivo models, along with cutting-edge technologies, will be instrumental in driving the development of targeted therapies, enhancing clinical outcomes, and advancing personalized treatment strategies. As efforts continue to address the current limitations of organoid systems, these models are poised to play an increasingly central role in the development of more effective, personalized, and ethically sound cancer therapies.

Data availability

The manuscript has no data produced during the study; hence no data availability is included.

Declarations

Conflict of interest

The authors declare no conflict of interest pertaining to the manuscript.