Engineering models of head and neck and oral cancers on-a-chip

Abstract

Head and neck cancers (HNCs) rank as the sixth most common cancer globally and result in over 450 000 deaths annually. Despite considerable advancements in diagnostics and treatment, the 5-year survival rate for most types of HNCs remains below 50%. Poor prognoses are often attributed to tumor heterogeneity, drug resistance, and immunosuppression. These characteristics are difficult to replicate using in vitro or in vivo models, culminating in few effective approaches for early detection and therapeutic drug development. Organs-on-a-chip offer a promising avenue for studying HNCs, serving as microphysiological models that closely recapitulate the complexities of biological tissues within highly controllable microfluidic platforms. Such systems have gained interest as advanced experimental tools to investigate human pathophysiology and assess therapeutic efficacy, providing a deeper understanding of cancer pathophysiology. This review outlines current challenges and opportunities in replicating HNCs within microphysiological systems, focusing on mimicking the soft, glandular, and hard tissues of the head and neck. We further delve into the major applications of organ-on-a-chip models for HNCs, including fundamental research, drug discovery, translational approaches, and personalized medicine. This review emphasizes the integration of organs-on-a-chip into the repertoire of biological model systems available to researchers. This integration enables the exploration of unique aspects of HNCs, thereby accelerating discoveries with the potential to improve outcomes for HNC patients.

I. INTRODUCTION

Head and neck cancers (HNCs) are the sixth most common cancer worldwide, resulting in more than 450 000 annual deaths globally.^1,2 Most HNCs are squamous cell carcinomas (SCCs), originating from the lining mucosa of the oral cavity, pharynx, and larynx. HNCs are strongly associated with lifestyle factors, including exposure to radiation, tobacco use, alcohol consumption, and human papillomavirus (HPV) infection.^2,3 The other subset of HNC is represented by rare aggressive tumors from soft tissues, glands, and bone, each with distinct clinical, histological, and molecular profiles. Managing HNCs often involves a multimodal approach, involving surgical intervention followed by adjuvant radio/chemotherapy or immunotherapy.⁴ Despite progress in developing treatment options, the five-year survival rate is stagnant at around 50%, with survivors frequently experiencing a compromised quality of life due to functional impairment and disfigurement following aggressive treatments.^4,5 While there has been considerable progress in understanding the mechanisms of head and neck (HN) carcinogenesis and progression, barriers persist in achieving a comprehensive understanding of the biology of these cancers and treatment responses, especially in the context of precision medicine. This persistent limitation primarily stems from the lack of suitable in vitro platforms capable of faithfully replicating the multifaceted hallmarks of HNC and its complex microenvironment.^6,7

Developing biological systems that accurately model the complex biology of HNC poses a significant challenge. Acknowledging its importance, traditional 2D cell cultures fail to faithfully recreate the intricate features of the tumor microenvironment (TME), including the extracellular matrix (ECM), immune system, vasculature, and innervation.⁸ In vivo models provide the most authentic preclinical model for studying HNCs. While xenograft and humanized mice have contributed to our understanding and treatment of HNC, animal studies remain expensive, time-consuming, and ethically controversial.⁹ Consequently, tissue-engineered 3D models have garnered attention for their ability to mimic various aspects of the TME in vitro, offering high throughput capabilities and improved fidelity for predicting clinical responses.¹⁰

Among various engineering approaches, microfluidics has emerged as a robust system to reproducibly replicate the dynamic behavior of both healthy and cancer tissues. A notable advancement in the field came with the development of organs-on-a-chip (OoCs), where miniaturized devices recreate physiological conditions closely resembling the tissue microstructure and function.¹¹ The fabrication of these platforms can be achieved through various processes, including chemical, mechanical, and energy-assisted protocols initially developed for the semiconductor industry and adapted to bioengineering.^12–14 These techniques employ various materials, such as glass, silicon, metals, ceramics, and polymers.^13,15 Polydimethylsiloxane (PDMS) is widely favored in microfluidic device fabrication for its numerous advantages, such as cost-effectiveness, ease of fabrication and sterilization, biocompatibility, and versatility, despite challenges associated with drug adsorption.¹³ The development of OoC has opened new avenues for more relevant preclinical models^16–18 with customizable designs, materials, and cells that can be combined to address specific research questions.¹⁹ These platforms can incorporate features such as spatially defined heterotypic co-cultures,¹⁷ confined cell–cell interactions,²⁰ and precise control over physiological parameters such as nutrient/hypoxia gradients,²¹ shear stress,²² and traction forces.²³ These capabilities facilitate the exploration of fundamental questions surrounding carcinogenesis, early detection, tumoral transformation, invasion, resistance, and cancer-related aspects of the human oral microbiome that are still poorly reproduced in vivo.²⁴

Here, we discuss the major advancements and challenges associated with engineering OoC models specifically tailored for HNC, as well as the exciting research possibilities unlocked by these innovative technologies. We first present a comprehensive view of current opportunities to reproduce the TME on-a-chip by mimicking soft, glandular, and hard tissues of the head and neck region (Fig. 1). Subsequently, we discuss the major applications of OoC models to investigate HNC-related questions regarding biological mechanisms, drug discovery, and personalized medicine. Last, we offer a perspective on the advancements, challenges, and future directions in employing physiologically relevant OoC models for HNC research.

FIG. 1.

Modeling soft, glandular, and hard tissues on-a-chip for HNCs. These models can be used to develop complex in vitro tissues, including (a) oral epithelium,^25,26 (b) salivary glands,²⁷ and (c) calcified bone.²⁸ Created with BioRender.com.

II. EMULATING THE HEAD AND NECK TUMOR MICROENVIRONMENT ON-A-CHIP

The HN region encompasses distinct tissue types in specific anatomic organization within the craniofacial region. For example, salivary glands are closely connected to innervation, tooth-forming tissues, and the lining mucosa adjacent to bone, as seen in the periodontium and maxillary sinuses.²⁹ HNCs can originate from virtually any cell type within these regions. Unlike most cancers in the body, HNCs often impact soft and hard tissues at early stages due to their close localization.^30,31 Such complexity introduces technical, engineering, and biological challenges in replicating the hallmarks of HNC in vitro. Therefore, this section discusses the possibilities to investigate HNC according to different tissue types: soft, glandular, and hard tissues.

A. Engineering soft tissues on-a-chip to study head and neck cancers

Various soft tissue components constitute the HN and oral region, such as the epithelium, mucosal membranes, and connective tissues, which collectively contribute to the physiological function of the HN region. Squamous cell carcinoma (SCC) is the predominant soft tissue malignancy of the HN region, originating from the mucosal epithelium of the upper respiratory and digestive tract.² SCC is influenced by multiple risk factors, including tobacco and alcohol use, HPV, environmental pollutants, poor oral hygiene, and radiation exposure.^2,3 SCC originating from the oral cavity and larynx is typically HPV negative (HPV−) and primarily associated with tobacco and alcohol use, while tumors arising from the pharynx are typically HPV positive (HPV+), primarily involving HPV-16 and HPV-18.² Certain risk factors for SCC exhibit geographical and cultural variations, exemplified by trends of carcinogen-containing products in Southeast Asia, high HPV prevalence in the USA, and higher risk in men due to gender-related behaviors.² Numerous initiatives have aimed to mitigate these exposures through primary prevention methods, including behavioral support programs for tobacco cessation and global campaigns promoting the implementation of Food and Drug Administration (FDA)-approved HPV vaccines.² To comprehend the evolution, progression, and treatment responses of SCC, it is essential to accurately model its associated tissues for thorough scientific investigation.

Studying HNCs using on-chip technologies offers numerous advantages to traditional cell culture, as it enables the replication of physiological conditions oversimplified in conventional in vitro models. For example, the engineering of OoC systems introduces the potential for modulating various tissue components and their properties. Factors such as stiffness, shear flow, and mechanical cues from the TME are key factors in cancer development.³² OoCs can recreate and allow further study of these aspects, including the contribution of mechanics, flow, and oxygen concentrations and how they contribute to the onset of the metastatic cascade.³³ Tumor-on-chip platforms enable the integration of heterogeneous cell populations and noncellular constituents, thereby enhancing the study of cellular interactions and multi-layer interfaces.^34,35 Studying cellular interactions is crucial for gaining insight into the intricate TME of SCC, involving interactions between the tumor and stromal cells, endothelial cells, cancer-associated fibroblasts (CAFs), and immune cells.³⁶ These cells secrete various proteins, chemokines, and cytokines that facilitate immune evasion, progression, and metastasis.² To better mimic physiology, OoCs can be leveraged for developing two- and three-dimensional architectures with precise spatiotemporal control, such as 3D tumor spheroids on-a-chip.^37,38 OoCs can further integrate features like 3D scaffolding within a vascularized matrix, thereby expanding the data interpretability and enriching the clinical relevance of these models.³⁸ The continuous advancement in this field has led to the establishment of real-time monitoring and analysis capabilities, augmenting our understanding of tumor dynamics.³⁹

Advancements in fields outside of HNC can serve as a foundation for the development and use of accurate OoC models, underscoring the importance of cross-disciplinary collaboration in pushing the boundaries of biomedical research. These platforms enable researchers to study risk factors and better understand early cancer development. For example, the pioneering “smoking airway-on-a-chip” demonstrates a powerful tool for investigating normal and disease-specific responses across molecular, cellular, and tissue levels.⁴⁰ To assess smoke-induced pathophysiology in vitro, the authors created an airway-on-a-chip lined with human bronchiolar epithelium from healthy and chronic obstructive pulmonary disease patients (Fig. 2). Cigarette smoke exposure is integrated via a connected vacuum pump, revealing insights into ciliary micropathologies, disease-specific molecular signatures, and epithelial responses to electronic cigarettes. This innovative technology could be adapted and expanded to HNC research, offering a means to replicate biomimetic exposures of tobacco-derived carcinogens affecting squamous epithelium.¹² By leveraging this established technology, we have the potential to gain a comprehensive understanding of HNC risk factors, crucial for early detection and prevention strategies. The readiness of this technology serves as a foundation for expanding its application, offering a platform to study HNC risk factors on-a-chip.

FIG. 2.

Translating insights from the “smoking airway-on-a-chip” model to advance HNC research. This technology has the potential to be adapted for HNC research, enabling the replication of biomimetic tobacco-derived carcinogen exposures on the squamous epithelium. (a) Left to right: a photograph of the smoking airway-on-a-chip setup, a schematic of a differentiated human mucociliated airway epithelium cultured on the top channel of the device, and a cross-sectional micrograph of a pseudostratified bronchiolar epithelium cultured on-chip for 4 weeks [green, β-tubulin IV; blue, DAPI (4',6-diamidino-2-phenylindole)]. (b) Methodology schematic for analyzing the impact of inhaled cigarette smoke on an airway-on-a-chip, featuring a custom cigarette smoke machine with the control of smoking parameters. (c) Cigarette smoke machine with loaded cigarettes (left) and the complete setup inside an incubator (right). (d) Time lapse of ciliary beating in the presence of absence of cigarette smoke while the color table at right represents the ciliary beating frequency (CBF) of individual cilia. Real-time polymerase chain reaction analysis. (e) Significant upregulation of antioxidant gene heme oxygenase 1 (HMOX1) expression with smoke exposure. (f) Western blot analysis illustrates the smoke-induced phosphorylation of the transcription factor nuclear factor (erythroid-derived 2)-like 2 (pNrf2) in on-chip epithelial cells. (g) Changes in the secretion of interleukin 8 (IL-8) in chips lined with bronchiolar epithelial cells isolated from normal or chronic obstructive pulmonary disease patients with or without exposure to whole cigarette smoke. (h) Heatmap comparing expression of genes associated with oxidation–reduction in bronchiolar epithelial cells. Adapted from K. H. Benam et al., Cell Syst. 3(5), 456–466.e4 (2016),⁴⁰ licensed under a Creative Commons Attribution (CC BY license, Elsevier).

The clinical treatment of SCC typically involves a multimodal approach, necessitating multidisciplinary care. This often entails surgical intervention followed by adjuvant chemoradiotherapy or immunotherapy.² However, few preclinical models of SCC faithfully recapitulate the unique features of the disease, and such models are costly, time-consuming, and lack reliable predictive capabilities for preclinical drug response.^41,42 For example, only 10%–30% of HNC patients respond to chemotherapy treatments with cisplatin (anti-angiogenic), cetuximab [epidermal growth factor receptor (EGFR)-targeting drug], or immunotherapy with pembrolizumab and nivolumab [anti-programmed cell death protein 1 (PD-1)], highlighting the unmet need to broaden available treatment options.⁴³ Yet, reductionist in vitro models of HNC using plated cells can obscure the accurate clinical response of HNC patients to these drugs.⁸ In contrast, OoC technology offers a unique capability to meet the demands of high-throughput drug testing with clinically predictive reliability that informs precision therapy for HNC. Al-Samadi et al. (2019) developed a 3D microfluidic device aimed at creating a fully human in vitro microfluidic chip for testing immunotherapy drugs [PD-L1 (PD-1 ligand) antibody and indoleamine 2,3-dioxygenase 1 (IDO1) inhibitor], focusing on providing a personalized medicine approach for HNC patients. Patient-derived cancer cells, including carcinoma and stromal cells, were isolated from tumor tissue and embedded in a human tumor-based ECM known as “Myogel/fibrin.” Immune cells and serum were also collected from the patient’s blood. These components were loaded into the microfluidic chip with or without immunomodulators to assess immune cell migration toward cancer cells and their cytotoxic activity.⁴ This OoC platform utilizing patient cells demonstrated that the efficacy of the PD-L1 antibody and IDO1 inhibitor was patient-dependent, paving the way for using microfluidic devices to predict the efficacy of immunotherapeutic drugs for individual patients.

B. Glandular tissues on-a-chip to study head and neck cancers

The HN region contains various glands with distinct structures, functions, and locations. Glands share a common function for producing, secreting, or releasing substances to perform essential bodily functions.⁴⁴ Examples of glands in the HN region include salivary, lacrimal, thyroid, parathyroid, and sweat/sebaceous glands. Among these, the salivary glands stand out due to their multifaceted functions and unique properties. The major salivary glands are exocrine glands comprised of three subtypes (parotid, submandibular, and sublingual), while the minor salivary glands line the oral mucosa and upper aerodigestive tract.⁴⁵ These glands play a crucial role in the oral cavity by releasing saliva containing enzymes, including amylase, facilitating processes including mastication, swallowing, speech, and digestion.^45,46 Notably, saliva is essential for digestion and acts as a primary defense against oral infections as it contains unique antimicrobial enzymes and immunoglobulins.⁴⁵ Salivary glands are vulnerable to carcinogens, such as tobacco and radiotherapy exposure.⁴⁷ This susceptibility and its diverse functions, locations, and compositions make studying salivary gland cancers challenging.

Salivary glands are unique organs within the HN region due to their anatomy and organization.^48,49 The complex nature of salivary gland anatomy demands a dedicated approach to replicate these structures in vitro. Considering the biological complexity, advancements in tissue engineering such as organoids and OoC have provided tools to recreate some of the characteristics of salivary tissues, such as the secretion of key salivary proteins and genes related to saliva production.^27,50 Combining microbubble array technologies and engineered poly(ethylene glycol) (PEG) hydrogels, a functional salivary gland on-a-chip was developed in 2021, mimicking the main characteristics of human and murine gland structures (Fig. 3).²⁷ The chip was designed using PDMS and featured spherical cavities with a diameter of ∼400 μm, creating an excellent paracrine and autocrine conditioning niche. This system maintained high cellular viability and expression of key genes and proteins involved in the secretory process for 14 days. Additionally, intracellular calcium release in human and mice models confirmed the secretory function of the salivary glands on-a-chip.²⁷ Based on the countless possibilities of the salivary glands on-a-chip model, its application could be expanded to encompass a wide range of pathologies affecting this distinctive tissue, including malignancies.

FIG. 3.

Functional salivary gland on-a-chip. (a)–(e) Engineered PEG ECM with microbubbles was used to mimic mice and glandular tissues. (f)–(k) These engineered salivary glands express diverse gene markers related to acinar, ductal, and myoepithelial cells, such as aquaporin5 (Aqp5), sodium-potassium-chloride channel 1 (Nkcc1), Mist1, keratin 7 (K7), keratin 5 (K5), and smooth muscle actin (Sma). (l)–(o). After 14 days in culture, the salivary gland mimetic expressed specific salivary markers, such as NKCC1, K7, AQP5, PIP, and Amy1. Adapted from Y. Song, Commun. Biol. 4(1), 361 (2021), L. Piraino et al., bioRxiv:2023.07.12.548707 (2023), and Mereness et al. Acta Biomater. 166, 187–200 (2023).^27,51,52 Reproduced with permission from Nature Publish group, authors, and licensed under a Creative Commons Attribution (CC BY license, Elsevier).

Salivary gland cancers, although rare, exhibit high mortality rates and tumor variability. Early diagnosis is challenging as the disease is often asymptomatic or misdiagnosed, leading to delayed detection. At the time of diagnosis, salivary gland cancer commonly presents with metastasis, particularly invading neural tissues and bone.⁵³ Late-stage manifestation limits treatment options, resulting in a poor response to chemo-radiotherapy, high recurrence rates, and the need for radical surgeries such as complete maxillectomy, causing severe deformity and loss of function.^54–56 Surgical resections of salivary gland tumors are delicate procedures primarily due to the glands' location near vital structures, such as vasculature and nerves.⁴⁵ Consequently, these treatments significantly impact the patient’s quality of life, highlighting the clinical need for comprehensive research to better understand and address salivary gland cancers.

Adenoid cystic carcinoma (ACC), the most prevalent HNC that affects salivary glands, is characterized by poor long-term prognosis due to relentless recurrences and a high tendency to metastasize via perineural invasion or hematogenous dissemination.^57,58 Given its rarity, the pathophysiology of ACC remains understudied. Nevertheless, specific genetic alterations have been associated with this cancer, specifically the deletion of chromosome 1p35–36 and chromosomal translocations involving 6q and 9p forming a MYB:NFIB gene fusion.⁵⁷ Given its aggressive nature and tendency for late-stage diagnosis, there is a compelling need for improved methods to study ACC formation, which would greatly facilitate early detection, tumorigenesis, and the development of more targeted therapies. To illustrate the role of the stroma, specifically CAFs, in ACC progression, a microfluidic device composed of PDMS was designed with two chambers connected by a central medium channel. ACC cells formed spheroids in the chamber containing Cultrex basement membrane extract, emulating the natural ECM. Co-cultures of CAFs and ACC resulted in higher cellular migration to the chamber with Cultrex basement membrane extract compared to ACC alone.⁵⁹ The mechanism underlying the interaction between CAFs and the TME may be specific to matrix metalloproteinase (MMP), as the inhibition of this complex of enzymes significantly reduced the invasiveness of ACCs in the presence of CAFs. This study emphasizes the importance of considering CAFs and a 3D matrix when developing in vitro model systems for HNC. Furthermore, this OoC model offered a real-time migration platform under flow, which can be expanded to investigate the role of other tumor-associated cells.

Cellular and tissue heterogeneity represents one of the most challenging aspects of cancer modeling. Heterogeneity results in multidrug resistance, aggressiveness, metastasis, and recurrence.⁶⁰ The complexity of HNC models can be enhanced by replicating the cellular heterogeneity in 3D. For example, 3D co-culture models for HNC have been employed for better disease modeling and drug screening.^61–66 Enhanced tumor cell invasion into an acellular collagen layer was observed in an ECM-like microenvironment comprising a co-culture of HNC cells with CAFs.⁶⁴ Additionally, primary tongue cancer cells and immune cells were co-cultured on-a-chip to assess the immune cell migration toward tumor cells and immunotherapeutic drug efficacy. The migration of immune cells toward tumor cells was found to be cancer cell density-dependent.⁶⁷ Although most HNC cancers are known to be associated with soft and glandular tissues, some of these diseases can initiate or evolve into mineralized tissues in the craniofacial and systemic bones. Building mineralized interfaces on-a-chip is challenging, yet this can open new opportunities for bioengineers and clinical professionals.

C. Emulating hard (mineralized) tissues and their associated cancers in the head and neck region

HN bones exhibit notable distinctions from other bones within the human body, primarily attributed to exposure through the naso-oral cavities (e.g., exposure to polymicrobial infections, tobacco, and HPV) and the anatomically complex structures surrounding vital structures.^68,69 Conventional 2D in vitro models fall short in replicating the complexity of native bone due to its dynamic metabolic activity, continuous remodeling, and intricate 3D architecture spanning macro- to nano-scale hierarchies, heavily mineralized cell-laden matrix, and complex tissue interfaces.^28,70,71 To address these limitations, advanced bone-on-a-chip models have emerged, enabling the replication of key characteristics of bone to be fabricated in microfluidic devices, greatly enhancing the predictive capacity of these models for clinically translatable studies.²⁸ The remarkable potential of bone-on-a-chip stems from the inherent customizability, which allows scientists to engineer models according to their research questions. These models offer the flexibility to tailor features, including specific cell types, interfaces, vascular networks, and mineralization. This high precision control allows the replication of complex physiological conditions that closely resemble the intricate 3D microenvironment and cellular interactions in bone.^70,72 Consequently, bone-on-a-chip models hold great promise in advancing our understanding of HN bone pathophysiology and serve as instrumental tools in developing novel and targeted therapeutic interventions for HNC and other related disorders (Fig. 4).

FIG. 4.

Bone metastasis on-a-chip. (a) A biomimetic-engineered vasculature network-on-a-chip model for studying osteosarcoma metastasis. (i) Cellular morphology at various locations (1–7) in the microfluidic device after 5 days of culture. (ii) Immunofluorescent images of circulatory tumor cells as they move inside the chip. (iii) A correlation plot quantifying the adhered tumor cells. (iv) Adhesion of osteosarcoma cells to the endothelialized microchannel vascular network. (b) A bone perivascular niche on-a-chip to study breast cancer metastasis to bone. (i)–(ii) Co-culture of mesenchymal stem cells and endothelial cells on a decellularized bone matrix with cancer cells were infused on the bone-like tissue to mimic bone invasion under interstitial flow. (iii) Endothelial cells exhibited reduced growth in serum-reduced media. (iv–v) Vasculogenic genes were upregulated in the bone-like tissue, especially under flow conditions. Adapted from (a) Priyadarshani et al., ACS Biomater. Sci. Eng. 7(3), 1263–1277 (2021)⁷³ and (b) Marturano-Kruik et al., Proc. Natl. Acad. Sci. U.S.A. 115(6), 1256–1261 (2018).⁷⁴ Reproduced with permission from the Chemical American Society and Proceeding of the National Academy of Sciences.

1. Primary bone tumors in the oral, and head and neck regions

Primary bone cancers are rare malignancies arising from bone tissue, distinct from the more prevalent secondary bone cancer originating at distant sites and metastasizing to bone.⁷⁵ Among the rare primary bone cancers, osteosarcoma is the most common and typically arises from osteoblasts within the metaphysis of long bones in a bimodal age distribution, peaking in the first and sixth decades.^72,75 Osteosarcoma is characterized by a low 5-year survival rate (<20%), high metastatic potential (the lungs in 90% of cases), and lack of early diagnosis.⁷² However, primary bone cancer development in the HN region, especially in the jaw, has distinct biology and more aggressive clinical prognosis, requiring a complex multi-disciplinary treatment plan.⁷⁶ The low prevalence of osteosarcoma coupled with high tumor heterogeneity renders its study in humans a major challenge, further compounded by the limited models to study bone.^72,77 Furthermore, the pathophysiology of osteosarcoma is poorly understood, emphasizing the need for research models that can recreate the intricate TME to better study its genesis, progression, and treatment options.⁷²

On-a-chip systems provide a valuable opportunity to recreate the specific environment of rare cancers, such as studying cellular adhesion processes of osteosarcoma on-a-chip.⁷⁸ However, these platforms are currently limited due to the more recent introduction of bone-on-a-chip models compared to other organs.²⁸ A major challenge in recreating bone in vitro lies in mimicking the physiologically complex interfaces of mineralized tissues, prompting researchers to explore multi-component models.⁷⁹ The local microenvironment of osteosarcoma comprises bone, stromal, vascular, and immune cells embedded in a densely mineralized ECM.⁷⁷ The majority of current models oversimplify this composition and, thus, cannot recreate two vital aspects of osteosarcoma: (1) the intra- and inter-tumoral heterogeneity and (2) the dynamic and complex microenvironment of the bone cell matrix interface.⁷⁷ Despite these complexities and limitations, studying the interfaces between the local components and various cell types involved in osteosarcoma is crucial for gaining a comprehensive understanding of cancer progression and therapeutic responses. The first step is to develop a successful osteosarcoma on-a-chip model that recapitulates the unique characteristics of the HN region.

By recreating cell–cell and cell–ECM interfaces, for example, tumor extravasation from and invasion into a calcified matrix, we can gain valuable insight into progression and aggressiveness of cancer. To study circulating tumor cells in the vascular system, biomimetic-engineered vascular network-on-a-chip models have been shown to replicate the hierarchical structure and functional aspects of vasculature, facilitating a better understanding of cancer invasion and metastasis of osteosarcoma.⁷³ Embedded within a PDMS slab, a microtubular network ranging from capillaries (∼80 μm) to large arteries (∼390 μm) with a confluent endothelial layer successfully demonstrated the vascular deposition of circulatory tumor cells from an osteosarcoma cell line (MG-63) in whole blood.⁷³ Tumor cells were shown to adhere tightly to the endothelial layer of the microchannels during low shear stress.⁷³ This model emphasizes the many aspects of microvascular transport that can be studied in vitro using microfluidics techniques, such as vascular deposition of metastatic cells, shear-related inflammation, endothelial uptake, and circulating tumor cell migration.⁷³ Aligned with the concept of personalized medicine, there is a growing need for these models to faithfully recreate the specific environment for studying the lethal form of osteosarcoma that manifests in this region of the body. Specifically, the metastatic spread of osteosarcoma is a critical factor contributing to the poor survival rate associated with its diagnosis.⁷² Thus, there is a growing demand within the research community to establish a standardized platform for studying HNC metastasis.

When designing these complex bone-on-a-chip models, it is imperative to incorporate stable soft and hard tissue interfaces. Previous studies have recreated elements of the oral cavity on-a-chip and demonstrated that incorporating a hydrogel can enhance the mechanical stability of in vitro oral mucosal constructs, allowing them to successfully maintain structural geometry and barrier function for extended periods in culture.⁸⁰ There is a cascade of tumor-to-bone cellular interactions that facilitates the spread of cancer and the disrupts bone homeostasis.⁸¹ Understanding these interactions is inherently challenging due to the involvement of various factors, such as heterogenous cell populations, secreted molecular messengers, and stromal components.⁸² In vitro models serve as a valuable tool for isolating and studying these players, shedding light on their roles in tumor–bone interactions. Furthermore, studying the interface between tissues and biomaterials holds significant clinical importance for pioneering tooth-on-a-chip model.⁸³

Odontogenic tumors typically manifest in the jaw bones surrounding teeth-bearing segments, originate from tooth-forming tissues, and often remain asymptomatic until late stages.⁸⁴ While some tumors may be benign, their local aggressiveness can have severe clinical implications.⁸⁵ Notably, ameloblastoma arising from odontogenic epithelium cells possesses a high potential for malignant transformation (observed in 70% of cases), leading to substantial bone destruction.⁸⁵ Current treatment options primarily involve surgery, either through a conservative approach with a risk of recurrence or a radical approach that can significantly impact the patient's quality of life due to the removal of extensive jaw segments.⁸⁵ Therefore, the need to discover new targeted therapies and develop less invasive treatments is paramount. Recreating the unique environment of odontogenic tumors presents significant challenges, particularly when studying specific scenarios such as bone exposed to polymicrobial infections to understand the impact of periodontitis.⁸⁶ In this context, pioneering efforts to establish more physiologically relevant in vitro models, such as OoC, hold great promise. These models offer a controlled and dynamic platform for studying odontogenic tumors, enabling researchers to explore novel treatments and targeted therapies more precisely. By replicating the intricate interactions between tumor cells, bone tissues, and microbial agents, researchers can uncover crucial insights into the pathogenesis of odontogenic tumors, advancing the development of less invasive and more personalized therapeutic approaches.

2. Bone metastasis from head and neck malignancies

In HNC, bone metastases are less prevalent compared to other cancer types like breast and prostate cancers. However, 50%–80% of nasopharyngeal carcinomas metastasize to mineralized tissues.⁸⁷ Bone metastasis is associated with the patient survival of less than 50% in one year.⁸⁸ Despite advancements in diagnostic techniques for the early detection of bone metastasis, including liquid biopsies, computed tomography, and scintigraphy, clinical management remains complex due to the side effects of current treatments, which can result in osteonecrosis and infections.⁸⁹ A main challenge in addressing bone metastasis lies in the limited understanding of how malignant cells migrate, communicate, and integrate with calcified tissues, a complexity that cannot be accurately replicated in simplistic 2D in vitro models.²⁸ Hence, there is a pressing need to develop microphysiological systems that provide insight into the mechanisms underlying bone metastasis.

Key focus points of HNC metastasis include cancer cell extravasation, migration to mineralized tissues, and establishment of a new TME within the bone. Cancer cell extravasation marks an early step in the metastatic timeline.⁹⁰ Primary cells from HNCs can undergo an epithelial-to-mesenchymal transition to invade neighboring tissues and adjacent ECMs through the lymphatic system or small vessels, respectively.⁹¹ While current microfluidic models have primarily focused on the molecular mechanisms and mechanical influence on cancer cell migration, bone-on-a-chip models have been more extensively explored in the context of other cancer types like breast and prostate cancer. Thus, analogous investigations of cellular extravasation of HNC cells to bone remain little explored.⁹² As an example, Jean et al. developed a microfluidic device with a central matrix channel and two side media channels, introducing breast cancer (MDA-MB-231) or myofibroblast (C2C12) cells into the side channel. They incorporated human umbilical vein endothelial cells, mesenchymal stem cells, and osteoblasts into the main channel to establish a 3D vascularized bone-like microenvironment. The cells encapsulated within the matrix expressed osteocalcin and alkaline phosphatase, mimicking the physiological response of bone tissue. The study revealed that breast cancer cells had a significantly higher propensity for extravasation through the vascular network (56%) compared to myofibroblasts (8.2%). Additionally, stromal cells in the bone-like environment significantly increased the extravasation of breast cancer cells to the bone by 3.8-fold.⁹³ This research, while conducted in the context of breast cancer, could inform future studies on HNC extravasation to bone, particularly by the integration of endothelial cells and pericytes given their significant role in oral SCC.

A crucial element of cellular extravasation is the lymphatic system, particularly significant in the context of HNCs, where extravasation often occurs in the cervical lymph nodes.⁹⁴ However, replicating lymph nodes in vitro presents challenges related to integrating vasculature, achieving perfusion, co-localizing immune cells, and enabling inter-tissue communication.⁹⁵ Nevertheless, recent advances in microfluidics have led to the development of more complex models suitable for studying HNC metastasis. An example is the engineered lymph node subcapsular sinus microenvironment for studying pancreatic metastasis.⁹⁶ Thomas et al. engineered a PDMS block and polystyrene culture plate with an unfunctionalized portion allowing cells to settle to the bottom before reaching the functionalized divergent portion with microfluidic perfusion and adhesion-molecule-treated regions. Adhesion molecules such as E-selectin, intercellular adhesion molecule (ICAM), and vascular cell adhesion molecule were used to mimic the cellular migration throughout the subcapsular sinus. The system successfully mimicked the shear stress radial flow in the human lymphatic vessels in both quiescent (lymphatic transit rate of 0.90–4.24) and inflammation/remodeling (lymphatic transit rate of 0.95–5.77) conditions. The system regulated pancreatic cancer cell (LS174T and PANC-1) and monocyte adhesion (THP-1), especially in the presence of E-selectin. Additionally, the adhesion of LS174T cancer cells was higher when co-perfused with THP-1 monocytes.⁹⁶ Considering this model, incorporating elements of bone, such as a mineralized matrix connected to the subcapsular sinus, could be implemented to enhance our understanding of the dynamics of HNC metastasis to bone tissue.

The migration of HNC to bone is primarily facilitated through either hematogenous or lymphatic spread to the bone marrow.⁹⁷ Subsequently, cancer cells may enter a dormant phase, evading the immune system, and eventually awakening to proliferate and establish a new TME.^98,99 These events in bone tissue are linked to an imbalance in osteoblast and osteoclast activities, orchestrated by the vasculature, lymphatic, immune, endocrine, and nervous systems.^100,101 Microfluidics allows for the isolation of these individual components to understand the role of each character in this intricate narrative. Recent microfluidic systems have explored these interactions, exemplified by a model designed to elucidate the crosstalk between sympathetic neurons, osteoclasts, and breast cancer spheroids. The PDMS device, featuring three separate compartments, allowed cell communication solely through paracrine signaling. Osteoclast differentiation was induced over a period of seven days using the receptor activator of nuclear factor kappa beta ligand and macrophage colony-stimulating factor from CD14⁺ monocytes isolated from human peripheral blood mononuclear cells. Differentiated osteoclasts were then cultivated on bone fragments within the microfluidic chip. In parallel, breast cancer cell spheroids (MDA-MB-231) were formed on Matrigel, and a neuroblastoma cell line (SH-5Y5Y) was used to model human sympathetic neurons. Notably, the tumor compartment exhibited elevated molecular signaling levels, such as Dickkopf-related protein 1, compared to cultures lacking communication with bone cells or neurons.¹⁰² While the significant role of paracrine signaling in modulating the aggressiveness of breast cancer spheroids has been highlighted, further exploration into the migration and invasion of cancer cells is warranted. Given the vast potential for enhancing HNC models, the interactions between cancer cells and the diverse cell population within bone can be achieved in a mineralized environment that closely emulates the distinctive properties of bone tissue.

III. CURRENT APPLICATIONS OF HEAD AND NECK ORGAN-ON-A-CHIP MODELS

The complex and lengthy process of drug development, coupled with its high attrition rate, incurs significant costs and delays in introducing new drugs to the market for the effective treatment of HNC. The primary cause of the elevated failure rate in drug discovery can be attributed to the reliance on current 2D cell-based drug screening platforms, which fail to accurately predict the intricate and variable therapeutic responses in humans.¹⁰³ Consequently, innovative directions are emerging for OoC models in the development of treatments for HNC. These novel models cover the crucial three-dimensional aspect of tissue development. Beyond their role in studying various facets of the TME, these platforms hold the potential to impact the pharmaceutical industry, artificial intelligence, and synthetic biology, offering new avenues for personalized medicine (Fig. 5).¹⁰⁴

FIG. 5.

Applications of head and neck and oral tissues on-a-chip models. These models could be used for diverse applications, including drug development, metastasis via lymphatic spread, and personalized medicine. (a) Engineered salivary glands have been employed to (i) screen a drug library for radioprotection using glutathione and cell senescence assays, (ii) assess secretory function through calcium flux response and acinar marker expression, and (iii) evaluate tissue toxicity using live/dead assays. (b) Lymph nodes on-a-chip have been developed to understand the migration of monocytes through the subcapsular sinus. (i) These chips are coated with adhesin molecules such as E-selectin, intercellular adhesion molecule (ICAM), and vascular adhesion protein (VCAM), and (ii) were designed with a linearly increasing flow channel width that facilitated the adhesion of THP-1 monocytes. (c) Patient-derived HNC on-a-chip was developed to evaluate individual responses to cancer treatments. These platforms have demonstrated cellular response to administered drugs (i), which can be used to inform clinical decision making (ii) and create a personalized treatment plan (iii). Adapted from L. Piraino et al., bioRxiv:2023.07.12.548707 (2023), Mereness et al. Acta Biomater. 166, 187–200 (2023), Birmingham et al. iScience 23(11), 101751 (2020), and Kennedy et al. Sci. Rep. 9(1), 6327 (2019).^51,52,96,105 Reproduced with permission from Nature Publish group, authors, and licensed under a Creative Commons Attribution (CC BY license, Elsevier).

A. Head and neck organs-on-a-chip for drug discovery

Several innovative 3D models have been developed for tissues within the HN region. A noteworthy example is a salivary gland mimetic tissue chip platform designed to discover radioprotective drugs for HNC patients, aiming to prevent dry mouth associated with off-target radiotherapy.^106,107 Mouse and human salivary gland tissues were isolated and encapsulated within MMP-degradable PEG hydrogels into high-density microbubble arrays. These tissue chips not only support the survival of salivary gland tissues but also preserve the phenotype integrity and secretory function. Using this platform, FDA-approved drugs were screened, identifying lead compounds with high efficacy in vivo.⁵¹

B. Emulating systemic metastasis in organs-on-a-chip

In recent research, microfluidic devices have been leveraged to replicate both vascularization and lymphatics within HN OoC models to analyze systemic metastasis. Indeed, microfluidic devices have been adapted to study HN tumor-induced angiogenesis,¹⁰⁸ invasion, and metastasis for various tissue types.¹⁰⁹ Lymphatic-fibroblast crosstalk in the TME was also recapitulated on a microfluidic chip.¹¹⁰ Primary tumor-derived fibroblasts from HNC patients were co-cultured with tubular lymphatic vessels to investigate the impact of tumor-derived fibroblasts on lymphangiogenesis, a process closely associated with metastasis. It was reported that in the presence of tumor cells, lymphatic vessels can be conditioned for sprouting, vessel permeability, and pro-lymphangiogenic gene expression. The designed model was proposed to validate therapeutics prior to clinical testing as well.¹¹⁰ Additionally, a microfluidic device for HNC spheroid culture, in parallel with endothelium, was recently developed to evaluate drug efficacy and toxicity in a 3D tumor context.¹¹¹ SCC or salivary gland ACC was cultured in juxtaposition to endothelium-lined channels within a microfluidic system equipped with a concentration gradient generator. Using this approach, concentrations of chemotherapeutic drugs were identified that eliminated tumor cells and minimized toxicity to the endothelial cell layer. Interestingly, patient-specific response differences were observed for individual SCC and ACC to the drug combinations. To study tumor-induced angiogenesis, salivary gland ACC cells and oral SCC cells were evaluated in the presence of endothelial cells. Similar results were observed for the tumor-on-a-chip platform and nude mouse models regarding angiogenic capability and antiangiogenic drug testing, indicating the practicality of the microfluidic model.¹¹² In sum, better replication of complex tissues, including the incorporation of vascularization and lymphatics complete with flow, may provide greater efficacy for drug testing, thereby motivating clinical translation.¹¹¹

C. Organs-on-a-chip for personalized medicine and cancer avatars

A powerful opportunity exists to develop personalized OoC using human-derived and patient-specific cells to predict effective treatment strategies for HNC, which have been termed cancer avatars. For example, epithelial cells, Matrigel, and SCC isolated from 43 biopsies of HNC patients were treated with cisplatin and docetaxel. Not only did the resulting data support patient-specific chemoresistance, with the cisplatin and docetaxel resistant microphysiological system recapitulating HNC recurrence, but the 3D model also exhibited increased chemoresistance compared to 2D cells.¹¹³ Additionally, using 12 HNC cell lines, 3D scaffolds of Matrigel or Myogel, a hydrogel derived from human-derived leiomyoma, demonstrated low mammalian target of rapamycin inhibitor activity in contrast to EGFR and mitogen-activated protein kinase inhibitors.¹¹⁴ Moreover, a microfluidic device composed of a glass substrate and two layers of the PDMS membrane sandwiching a porous membrane was developed for the drug testing of vascularized SCC or ACC. Patient-specific differences in drug responses were evident, suggesting the application of the microfluidic device to test drug sensitivity in individual SCC and ACC clinical cases.¹¹⁵ Furthermore, with the accrual of significant multi-omics data from personalized chips like these, developing in silico OoC may be possible. Also known as digital twins, these virtual models of personalized OoC may ultimately enable the prediction of highly effective treatments based on simple tumor biopsies or even blood draws.

IV. FUTURE PERSPECTIVES

The development of OoC platforms tailored for HN tumors can expedite the understanding of mechanisms related to malignant transformation, invasion, metastasis, tumor heterogeneity, biological sex differences, and drug resistance. Considering that patient samples are limited but necessary to improve preclinical research, integrating OoC with patient specimens provides an optimal fusion of discovery and mechanistic research. This fusion could help address the high attrition rates of promising compounds in preclinical research, often failing in clinical trials due to a lack of efficacy. Multi-organ systems, though complex and costly, show promise for minimizing animal studies and parallel use in early phase clinical trials.¹¹⁶ For modeling complex tissue interactions, low to medium-throughput OoC platforms are well-suited for preclinical single-organ or double-organ toxicity and efficacy studies. Conversely, high-throughput plate-based OoCs with simpler tissue constructs can aid in target identification, lead selection, and optimization.¹¹⁷ A forward-thinking approach for multi-organ-on-a-chip that could be adapted for HNC was developed by McAleer et al.²⁰ This system included a pumpless, reconfigurable, multi-organ-on-a-chip with an interconnected chamber for hepatocytes, cardiomyocytes, and tumor cells. Using a recirculating serum-free medium, this platform was used to predict preclinical on-target efficacy, metabolic conversion, and off-target drug toxicity through functional microelectromechanical systems of multidrug-resistant and non-multidrug-resistant types of cancers. The system demonstrated that tamoxifen, a selective estrogen receptor modulator, reduced the viability of breast cancer cells only after metabolite generation in the liver compartment. However, it did not affect vulva cancer cells except when co-administered with verapamil, a permeability glycoprotein inhibitor. Both tamoxifen alone and in co-administration with verapamil produced off-target cardiac effects, emulating clinical findings.²⁰ Similar platforms could be designed to dissect mechanisms of drug resistance in HNC and develop markers to distinguish indolent from lethal tumors in early stages.

Radio- and chemo-resistances remain the primary causes of poor survival rates for HN SCC.² Thus, using OoC to understand the molecular landscape of treatment resistance can significantly impact patient survival outcomes and propel advancements in cancer cell therapy. Currently, clinical applications of OoC for personalized medicine are in use for other types of cancers. For instance, Horowitz et al., developed a digitally manufactured microfluidic platform for multiplexed drug testing using intact cancer slice cultures and patient biopsies of glioblastoma.¹¹⁸ This innovative approach demonstrated the sufficient preservation of the tissue microenvironment and provided results within days of surgery to guide the selection of effective initial therapies. Additionally, patient-derived organoids-on-a-chip from colorectal cancer are part of a registered clinical trial to screen drugs via microfluidics-based assays, comparing the results with patients' responses.¹¹⁹ OoC technologies have already demonstrated their effectiveness in addressing treatment resistance in various cancer types, yet their potential remains largely unexplored in the context of HNC. Therefore, OoC platforms offer a promising avenue for future research investigating treatment resistance in HNC.

Apart from applying OoC in basic research, drug screening, and personalized medicine, increasing evidence points to the translational potential of OoC and microtechnologies for disease detection, risk stratification, and progress monitoring.¹²⁰ Particularly in tobacco users or HPV-positive individuals, seemingly healthy tissues may harbor mutated cells that may develop into cancer. Presently, identifying such alterations before the development of a clinically detectable tumor is an immense challenge. In addition, the precise process underlying this development remains incompletely understood in the context of HNC. In an effort to replicate tumor progression in the oral epithelium, a model of full-thickness tissue-engineered oral mucosa was developed using oral cancer spheroids to mimic dysplasia and tumor invasion.¹²¹ This represents a foundational step toward increasing the complexity of the engineered system and addressing fundamental questions related to identifying single or small clusters of cells that, under specific conditions, may become malignant. The OoC field has been rapidly integrated with advanced techniques such as automated sensors, gene editing, patient-derived organoids, and omics- and single-cell-based analysis.^122,123 These applications typically generate substantial data outputs, resulting in a burdensome amount of data commonly referred to as “big data.”¹²⁴ Traditionally, data have been accumulated and processed manually, often inefficiently. The application of artificial intelligence and deep learning to OoCs presents a powerful tool for the in-depth analysis of massive image-based data generated by OoC approaches, enhancing the automated level of OoCs and, thereby, increasing its throughput.¹²⁵ In addition to analyzing images, secretome, and gene changes over time, machine learning applications tailored for HNC can be explored to improve device design for enhanced modeling of fluid flow, interstitial forces, and masticatory mechanics, leading to a new generation of OoC platforms with extensive functionalities.¹²⁵

Furthermore, we envision that OoC technology will contribute significantly to mitigating global health disparities. For example, the development of high-fidelity cancer avatar models from patient cells could be analyzed under different stressors and drug combinations, enabling patients to undergo multi-drug trials without physically leaving their local region. This application marks a significant stride in extending the benefits of high-end technology to remote/rural regions and low-income populations by indirectly providing access to facilities and tools that are typically available only in large metropolitan centers.

V. CONCLUSIONS

Remarkable advances have been achieved in the development of OoC models for studying various types of cancers in recent decades. However, the unique and heterogenic nature of HNCs poses challenges in faithfully reproducing the diverse cell population, complex tissue environment, and intricate biological processes. These challenges create opportunities for interdisciplinary collaborations in an increasingly multi-disciplinary era. The immense potential of integrating microphysiological systems into the daily practices of clinicians marks just the beginning of how these tools can improve patient quality of life, especially in predicting risk factors, enhancing diagnostics, and guiding treatments through personalized medicine. As technologies advance and interdisciplinary efforts expand, the future holds great promise for further progress in cancer research and patient care through the application of OoC models.

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Mauricio Goncalves da Costa Sousa and Sofia M. Vignolo contributed equally to this work.

Mauricio Goncalves da Costa Sousa: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Sofia M. Vignolo: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Cristiane Miranda Franca: Conceptualization (supporting); Supervision (supporting); Writing – review & editing (supporting). Jared Mereness: Visualization (supporting); Writing – original draft (supporting); Writing – review & editing (supporting). May Anny Alves Fraga: Visualization (supporting); Writing – original draft (supporting); Writing – review & editing (supporting). Alice Corrêa Silva-Sousa: Visualization (supporting); Writing – original draft (supporting); Writing – review & editing (supporting). Danielle S. W. Benoit: Conceptualization (lead); Resources (lead); Supervision (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Luiz Eduardo Bertassoni: Conceptualization (lead); Funding acquisition (lead); Resources (lead); Supervision (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead).

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding authors upon reasonable request.