Microfluidic lung cancer models: Bridging clinical treatment strategies and tumor microenvironment recapitulation Authors: Zhiyun Yu, Arsalan A. Khan, Wara Naeem, Jeffrey A. Borgia, Michael J. Liptay, Christopher W. Seder, Jian Zhou Published: 2025 ABSTRACT Lung cancer remains the leading cause of cancer-related mortality worldwide, with non-small cell lung cancer accounting for a majority of cases. Despite advances in targeted therapies and immunotherapy, challenges such as tumor heterogeneity, resistance mechanisms, and limited preclinical models hinder treatment efficacy. Traditional cancer models, including 2D cell cultures and animal models, often fail to accurately replicate the lung's complex architecture, microenvironment, and biomechanical cues, leading to poor predictive performance in drug development. Microfluidic-based organ-on-a-chip technology offers a promising alternative by integrating human-derived cells with precisely controlled perfusion, mechanical cues, and tumor–stroma interactions in physiologically relevant 3D models. These platforms enable the study of lung cancer biology, drug responses, and patient-specific therapeutic outcomes with improved accuracy. In this review, we discuss recent advancements in microfluidic systems for recapitulating normal lung physiology and 3D lung cancer microenvironment, covering various microfluidic platforms with applications in disease modeling and drug testing. Unlike other review articles, we bring first-hand insights from clinicians about the current treatment practice for lung cancer and the clinical utilities of lung cancer-on-a-chip models, which bioengineers have been seeking. We also highlight the translational potential of these systems in personalized oncology and the need for interdisciplinary collaborations, particularly with clinicians, to enhance their clinical impact. INTRODUCTION Lung cancer is the leading cause of cancer-related deaths worldwide, with approximately 1.8 × 106 deaths annually.1,2 About 85% of cases are non-small cell lung cancer (NSCLC), primarily lung adenocarcinoma (LUAD), and lung squamous cell carcinoma (LUSC).3 About two-thirds of lung cancer deaths worldwide are attributable to smoking, particularly in regions where smoking is prevalent, with stronger associations in LUSC and small cell lung cancer (SCLC) than LUAD. However, lung cancer in never smokers is linked to environmental factors like secondhand smoke, pollution, occupational carcinogens, and genetic predisposition.1 The development of targeted therapies and the successful application of immunotherapy in select patient populations have marked significant progress in treating advanced NSCLC. However, challenges remain, including the need to identify new driver gene alterations to broaden the scope of targeted therapies and to better understand resistance mechanisms to improve their long-term effectiveness. Lung cancer exhibits significant intratumoral heterogeneity due to the presence of sub-populations of cells with distinct molecular features. An increased sub-clonal mutation fractions in localized LUAD are linked to a higher risk of post-surgical relapse, suggesting early metastatic potential in heterogeneous tumors.3 Thus, a personalized treatment regimen is necessary to identify the clonal targetable genetic alterations early in diagnosis and to timely monitor the emergence of post-therapy acquired resistant targets. Traditional models like animal systems and cancer cell lines are widely used to study anticancer drugs, offering cell-type-specific mechanistic insights. Animal cancer models, including patient-derived xenograft models, genetically engineered mouse models, and microsurgical tumor cell injection models, require extended establishment time ranging from three months to over a year.4 Despite their utility, these models often fail to fully capture the intratumoral heterogeneity observed in individual patients. While simple 2D in vitro models focus on molecular mechanisms, but they failed to replicate the human lung's complex architecture and physiological functions, including the alveolar–capillary barrier, breathing motion-induced mechanistic force, extracellular matrix (ECM), and multiorgan interactions, resulting in poor predictive performance and high drug development failure rates.5 Developing advanced, accurate, and efficient models is essential for improving anticancer drug evaluation and success rates (Box 1). BOX 1. Clinical significance of lung cancer models. From a purely clinical standpoint, the ability to simulate the true physiological environment of the lung would represent a significant advancement in managing lung cancer. In addition to evaluating the response to adjuvant therapy—delivered post-surgical resection—using cells from biopsy samples, we can also monitor the evolution of the tumor's genomic profile and its immune microenvironment. This would enable the identification of targetable mutations and the tailoring of treatments to achieve the most optimal and effective therapeutic outcomes. Furthermore, understanding how tumor cells influence the immune microenvironment and interact with the blood vessel endothelium would help assess the effectiveness and bioavailability of immunotherapy drugs, such as PD-L1 inhibitors. Clinicians face the challenging responsibility of balancing the benefits and harms of chemotherapy and immunotherapy to ensure patient safety. The most serious side effects include organ toxicities, such as hepatotoxicity, pneumonitis, cardiac toxicity, and neurological dysfunction. With this technology, the effects of chemotherapeutic drugs on other cells and organs within the body could be studied in a patient-specific manner, allowing for the prediction of potential toxicities before treatment begins. Additionally, combination therapies involving multiple drugs could be assessed rapidly on patient-derived samples, reducing trial-and-error approaches on the patient themselves and thereby enhancing patient safety. Such insights would help in preventing the development of drug resistance and would enable a more personalized and safer approach to lung cancer management. Organ-on-a-chip, or microphysiological systems, integrate human-derived cellular material with microfluidic technology to recreate the in vivo microenvironment of organs in vitro.6 These platforms mimic key physiological parameters, such as perfusion and mechanical forces, offering a controlled and biologically relevant system for drug testing and disease modeling. Microfluidic chips provide precise control of fluid flow within small channels seeded with cells in 2D and 3D formats. For lung cancer research, these chips enable the integration of 3D co-cultures, nutrient and gas supply, shear stress, tensile forces, drug concentration gradients, and real-time biosensing to monitor cellular behaviors under various conditions. 3D organoid models in microfluidic systems have been used to elucidate cancer biology and facilitate drug testing.7 Preclinical organoid models integrated with these systems are increasingly being explored for predicting patient responses to treatment.8 This review highlights advancements in microfluidic systems for modeling normal lung tissue and lung cancer (Fig. 1). It begins by outlining the current challenges in lung cancer management within clinical practice. The subsequent sections delve into microfluidic fabrication techniques and chip design for lung cancer research, followed by a discussion of recent models of lung-on-a-chip and lung cancer-on-a-chip, which excel at replicating tissue microenvironments and hold promise for clinical diagnostics and drug development. Finally, the review addresses the critical role of cross-disciplinary collaboration among scientists, bioengineers, and clinicians, along with the challenges inherent in fostering such partnerships. Microfluidic lung cancer models bridging lung physiology and cancer biology for precision lung cancer care. Modeling lung architecture and physiology in microfluidic devices enables readily recreation of key features of the lung tumor microenvironment, including complex cellular compositions and vascular networks. These lung and lung cancer-on-a-chip systems have broad applications in disease etiology, therapeutics, drug development, and personalized medicine. CURRENT LANDSCAPE IN CLINICAL MANAGEMENT OF LUNG CANCER Due to the emergence of advanced therapeutic modalities, the treatment landscape for NSCLC has advanced significantly; however, improving survival outcomes remains a challenge. About half of NSCLC cases are diagnosed at stages I–III, and even in these cases, the 5-year disease-free survival (DFS, refer to Table I for common terms) ranges from 34% to 80%, improving from stage III to stage I disease.9–11 Additionally, 5-year overall survival (OS) for resectable NSCLC ranges from 50% to 70%.12,13 TABLE I.Nomenclature in cancer research and care.TerminologyDefinition5-year overall survivalRefers to the percentage of patients who are still alive after 5 years of follow-up5-year recurrence-free survival rateRepresents the percentage of patients who have not experienced a recurrence of NSCLC within 5 yearsDisease-free survivalRefers to the length of time a cancer patient survives without any signs or symptoms of cancer after treatmentComorbiditiesComorbidities are medical conditions or diseases that coexist with a primary condition in a patientAdjuvant therapyAdditional therapy provided after primary treatment to reduce the risk of disease recurrenceNeo-adjuvant therapyAdditional therapy administered before primary treatment, such as surgical resection, to reduce tumor size and spread, facilitating more effective removalTumor downstagingReduction in tumor stage achieved through neoadjuvant therapyImmune checkpoint proteinsProteins that act as checkpoints in the immune system and downregulate immune system responses to stimuli; most cancers exploit these proteins to evade immune detectionImmune checkpoint inhibitorsMonoclonal antibodies that target checkpoint proteins to restore immune responses against cancer; widely used as immunotherapy in NSCLCPathologic complete response (pCR)Refers to the presence of zero viable cancer cells in the resected tumor and regional lymph nodes after neoadjuvant treatment; assessed via pathological examination of tissue post-treatmentTNM stageStands for tumor, node, and metastasis. It is a system that describes the size and spread of a tumor in a patient's body. Nomenclature in cancer research and care. Overview of disease management plan Early-stage NSCLC (stages I–IIIA) is generally considered resectable, except in cases where stage IIIA patients have mediastinal lymph node involvement or bulky/matted lymph nodes.11 Conversely, unresectable cases are usually advanced cancers, the majority of which are bulky stage IIIB–IV tumors with metastases.14,15 In addition to advanced stage, several factors can make a tumor unresectable. These include proximity to major blood vessels like the superior vena cava, and location near the carina, where the bronchi divide. Furthermore, poor lung function from conditions like chronic obstructive pulmonary disease, severe cardiac disease, advanced age with significant comorbidities, or inability to tolerate major surgery can make surgical intervention inadvisable. This underscores the need for ongoing innovation in the multidisciplinary management of resectable NSCLC. Treatment of early-stage NSCLC Surgical resection and adjuvant therapy The current treatment guidelines for NSCLC are determined based on cancer's stage, genetic mutations, and the patient's overall health (Fig. 2). Historically, surgical resection was the primary treatment for early-stage NSCLC, though recurrence rates were high, approaching 50%.16 The advent of cisplatin-based adjuvant chemotherapy marked a critical development by significantly reducing recurrence and improving survival.17 The Lung Adjuvant Cisplatin Evaluation (LACE) meta-analysis found modest yet notable improvements in 5-year DFS and overall survival (OS), increasing by 5.8% and 5.4%, respectively.18 However, adjuvant chemotherapy has seen limited uptake due to poor compliance, as many patients struggle to complete treatment after surgery.19 Current treatment guidelines for non-small cell lung cancer (NSCLC). BAL: bronchoalveolar lavage; CT: chemotherapy; HR: high risk; CRT: chemoradiotherapy; SBRT: stereotactic body radiation therapy; RT: radiotherapy; PD-L1: programmed death-ligand 1; SSC: squamous cell carcinoma; Bev: bevacizumab pulmonary function test; PS: performance status, which is a standardized score, ranging from 0 (fully active) to 5 (deceased: PFT) used to assess fitness for surgery, with scores ≤2 generally indicating suitability for surgery. Neoadjuvant therapy Neoadjuvant chemotherapy has recently gained favor due to its potential for higher compliance, tumor downstaging, and enabling more complete resections.20 It also provides valuable opportunities for translational research, such as biomarker identification and tumor response assessment.20 Integrating immune checkpoint inhibitors (ICIs) with neoadjuvant therapy has transformed the field, with trials like CheckMate 816, KEYNOTE-671, and CheckMate 77T,21–23 which showed improved pathologic complete response (pCR) and event-free survival (EFS) when combining chemotherapy with immunotherapy. These findings support chemoimmunotherapy as the preferred induction regimen for patients without actionable gene alterations.24 Immunotherapy benefits are also evident in adjuvant settings, raising the question of whether adjuvant or neoadjuvant administration offers better outcomes.25 Table II shows the systemic therapy regimens for the perioperative management. TABLE II.Systemic therapy regimens currently in use for the perioperative management of non-small cell lung cancer.RegimenAction of mechanism Neoadjuvant or adjuvant chemotherapy Cisplatin + pemetrexed (non-squamous)Induces DNA damage; inhibits nucleotide synthesis via targeting folate metabolismCisplatin + gemcitabine (squamous)Induces DNA damage; inhibits DNA synthesis and repairCisplatin + docetaxel (squamous)Induces DNA damage; induces microtubules depolymerization; induces apoptosis; interferes angiogenesis and tumor microenvironment interactions Adjuvant targeted therapy Alectinib (ALK rearrangements)Inhibits ALK fusion proteins; inhibits PI3K-AKT, MAPK-ERK, JAK-STAT pathwayOsimertinib (EGFR mutations)Inhibit EGFR with mutations including Del19, L858R mutation, and T790M; inhibits PI3K-AKT, MAPK-ERK, JAK-STAT pathwayAtezolizumab (IV) PD-L1Monoclonal antibodies to block PD-L1 ligand on tumor cell surface; enhances T cell-mediated immune killing of cancer cellsDurvalumabPembrolizumabMonoclonal antibodies to block PD-1 receptor on T-cell surface; enhances T cell-mediated immune killing of cancer cellsNivolumab Neoadjuvant chemo-immuno therapy Nivolumab + platinum doublet chemotherapyChemotherapy reduces tumor burden and releases tumor antigens; immunotherapy enhances T cell-mediated killing of cancer cellsPembrolizumab + cisplatin-based doublet chemotherapyDurvalumab + platinum doublet chemotherapy Systemic therapy regimens currently in use for the perioperative management of non-small cell lung cancer. Targeted and immuno-therapies For patients with actionable mutations in genes such as epidermal growth factor receptor (EGFR) or anaplastic lymphoma kinase (ALK), targeted therapy has become a cornerstone of NSCLC management.26 However, the optimal timing—preoperative vs postoperative—remains under study.23 While preoperative administration may elicit a stronger immune response, further trials are needed to confirm long-term benefits. Targeted therapies, such as osimertinib for EGFR-mutant NSCLC and alectinib for ALK-rearranged tumors, have shown significant DFS gains in the adjuvant setting.23,26,27 Trials like NeoADAURA and ALNEO are now evaluating these agents in neoadjuvant phases, potentially reshaping treatment protocols.28,29 Management of advanced and metastatic NSCLC According to the National Comprehensive Cancer Network (NCCN) guidelines, management of advanced and metastatic NSCLC is complex. The mainstay of treatment involves tumor-specific targeted therapy, often with concurrent chemoradiation, depending on the tumor's TNM stage and the extent of metastases. Molecular testing is essential to identify potentially targetable mutations in genes such as EGFR, ALK, KRAS, ROS1, BRAF, NTRK1/2/3, MET (exon 14 skipping), RET, and ERBB2 (HER2), and for assessing PD-L1 expression levels, which guide further treatment decisions. For EGFR, this includes detecting specific mutations such as exon 19 deletions, exon 21 L858R mutations, S768I, L861Q, G719X mutations, and exon 20 insertions. Unmet clinical need Currently, approved targeted therapies include drugs like osimertinib, afatinib, erlotinib, larotrectinib, and selpercatinib (LOXO-292), among others.30 Recommended treatment regimens often combine chemotherapy, radiation, immunotherapy, and sometimes surgery. However, a common challenge clinicians face is that both targeted and non-targeted therapies, while initially effective, often succumb to acquired drug resistance due to factors like epigenetic mutations or tumor heterogeneity.31–33 This underscores the need to develop new targeted therapies, identify drug response, and necessary therapy adjustments in vivo for more effective management. Therefore, treating advanced-stage NSCLC is akin to addressing a moving target.34 Advancing personalized care in NSCLC requires a collaborative, interdisciplinary approach to effectively integrate emerging therapies into clinical practice. Such a collaboration is particularly manifested in establishing preclinical models for lung cancer disease management. MICROFLUIDIC PLATFORMS FOR LUNG CANCER RESEARCH Microfluidic platforms for lung cancer studies hold great promise for advancing personalized medicine by closely mimicking the physiological and pathological conditions of lung cancer in preclinical models. These systems effectively replicate respiratory mechanical motion, tumor tissue architecture, and dynamic fluid flow, thereby recreating in vivo conditions. Furthermore, the integration of patient-derived cells/microtissues and the high-throughput capabilities of these chips enable time- and cost-effective evaluations of treatment efficacy tailored to individual patients. Microfabrication for tissue engineering The microfluidic technology enables processing or manipulating small (10−9–10−18 l) amounts of fluids and biological tissues using channels and microwells with dimensions of tens to hundreds of micrometers.35 Conventional fabrication methods, such as laser micromachining, molding, and imprinting, have been foundational in microfluidic device production.36 However, the growing demands of biomedical research—requiring higher precision, flexibility, and scalability—have driven the adoption of advanced techniques like photolithography, soft lithography, and 3D printing.36,37 Photolithography employs ultraviolet (UV) light to transfer geometric patterns from a photomask to a light-sensitive photoresist on substrates like silicon or glass. Its high resolution, precision, and reproducibility make it ideal for high-volume manufacturing.38 Soft lithography, on the other hand, uses elastomeric materials, such as polydimethylsiloxane (PDMS), to create stamps, molds, or devices with fine features replicated from master molds, which are often made from photolithography, milling, or 3D printing. PDMS adhesive-based techniques further allow tuning of bonding strength to meet specific application needs.39 This approach supports the fabrication of three-dimensional, multiple layer, and complex microstructures. Its low cost, fast turnaround, and cleanroom-free process make it well-suited for rapid prototyping and small-scale production.38 While PDMS is biocompatible and optically transparent, the hydrophobicity of this material makes it difficult for direct cell culture, which requires appropriate surface coating for proper cell adherence.40 Recently, 3D printing has gained traction in biomedical tissue engineering. Techniques such as fused deposition modeling (FDM) and digital light processing (DLP) enable the layer-by-layer additive manufacturing of structures.36 3D printing's flexibility allows for the creation of complex, customized designs without molds, facilitating the replication of intricate tissues containing multiple cell types, such as vascular grafts, skin, bone, and heart tissue.41 However, FDM-based 3D printing is limited by resolutions of hundreds of micrometers, while other methods are costly and offer limited material options.36 Despite these advancements, soft lithography remains the most widely used fabrication method for microfluidic devices in research settings due to its accessibility, versatility, and cost-effectiveness. Microfluidic systems are well suited for biomedical research. A microfluidic system for cell biology research has several key components: (1) methods of introducing fluids, gases, and cell samples; (2) methods for controlled moving and mixing these fluids, gases around on the chip; (3) methods for realizing the microanalytical activities (e.g., imaging, biosensors) on the chip.35 Compared to conventional cell biology systems, these microsystems are uniquely advantageous for cell biology studies. In the case of fluid flow, the effect of the inertial forces to frictional forces can be reduced in microsystem, leading to the formation of laminar flow in the microfluidic channels. This property enables the formation of static and dynamic gradients at subcellular resolution, which are critical factors in regulating various biological processes, such as immune response, cancer spread, and cell differentiation.42,43 The rapid diffusive heat and mass transfer at the microscale, with characteristic times ranging from approximately 10−3 to 1 s (compared to 102–104 s at the macroscale), allows for quick media and environmental changes as well as efficient temperature control.39 The utility of a hypoxic incubator along with side gas channels, gas-permeable materials, and chemical deoxygenation on chips, recreate the intricate gaseous environment of lung and the tumor by enabling the flexible control of oxygen concentration and gradients.44–46 Other advantages include low consumption of reagents and cellular materials, simultaneous, independent, high-throughput manipulation of cells, and simulation of the in vivo cellular microenvironment. The design of a microfluidic chip must meet the need of rendering the cellular behavior and physiological environment, favoring the experimental practice of revealing pathological process and drug response. Engineering principles and strategies of microfluidic lung and lung cancer models The primary principle of microfluidic design is to reproduce the complexity in cellular architecture and microenvironment. The varying geometry of lung trachea, alveoli, and tumor as well as their cellular composition need to be considered. Recreating the lung barrier, composed of epithelial and endothelial layers, typically requires channels or wells that separate air and media compartments while maintaining access to both. Intricate channel designs, combined with flow perfusion, simulate blood flow and shear stress within the lung's microvascular network. Cyclic stretching of alveoli during breathing can be mimicked by pressure-driven deformation of elastic materials integrated into the microsystem.47 The ability to precisely control gradients and spatiotemporal variations of gases and fluids positions microfluidic lung models as powerful tools for disease modeling, drug screening, and personalized medicine. The most typical microfluidic designs mimicking lung physiological components can be categorized into membrane-based platform, lumen platform, microchannel platform, and microwell platform. The design details of these various platforms have also been covered in recent reviews.45,48 The membrane-based platform is widely used to model airway epithelium and alveoli structures, featuring upper and lower channels separated by a porous membrane.49 The epithelial and vascular cells can be seeded on opposite sides of the membrane, and the pores on membrane allow the crosstalk and migration of different cell types. A defining characteristic is the air–liquid interface (ALI) created by air perfusion in the upper channel and medium supply in the lower channel, effectively mimicking airway tissue architecture. This ALI-based design underpins many lung-on-a-chip platforms, facilitating studies of infection and respiratory pathologies.50,51 A notable modification of ALI-based lung cancer-on-chip models involves implanting cancer cell lines into the airway or alveolar epithelial layer and simulating breathing movements using cyclic mechanical strain.47 This study revealed that mechanical breathing motion significantly influences NSCLC growth. Under static ALI conditions, tumor cell growth and invasion into the endothelial layer were enhanced. In contrast, simulated breathing motion markedly suppressed tumor cell migration. The lumen platform features hydrogel-filled channels and is commonly used for vascularizing tumor cell–hydrogel mixtures or tumor spheroids/organoids. Lee et al.52 developed the injection-molded plastic array 3D culture (IMPACT) platform, which simplifies and diversifies the patterning of vascular structures within 3D cellular hydrogels.49 In this system, human umbilical endothelial cells (HUVECs) are seeded into the outer channel containing fibrinogen–thrombin hydrogel, where capillary forces facilitate droplet patterning along the channel. Lung tumor cells, such as A549, are introduced into the center channel in the same hydrogel, creating spatially organized vascular and tumor cell lumens. This arrangement promotes crosstalk between cell types and enables systematic drug response studies. The choice of hydrogel is critical for ensuring healthy cell behavior and providing appropriate mechanical properties for compartmentalized cell packing. Commonly used hydrogels include collagen I, fibrinogen, fibrin, and Matrigel, as reviewed in prior studies.53,54 The microwell platform is typically well suited for high-throughput applications and is often integrated with high-content screening tools to evaluate multiple drug treatment conditions on tumor cells. Its primary use is the generation of uniform tumor spheroids or organoids with a small number of cells, which requires careful optimization of the cell culture substrate material and microwell size.55,56 The combination of microwells and fluidic channel allowed the involvement of more cell types and environmental factors, such as the perfusion of immune cells, co-culture with vascular cells, and drug administration.57,58 The microchannel platform, utilizing microscale gradient generators (uGGs), creates precise biochemical gradients with controlled spatiotemporal distribution and subcellular resolution. uGGs rely on diffusive mixing between parallel laminar streams of different compositions to generate molecular gradients. To address shear stress issues caused by laminar flow, which can alter cell phenotypes and wash away secreted factors, advanced designs incorporate flow resistive elements or flow barriers like hydrogels to achieve passive diffusion. The uGGs are the perfect platform for drug sensitivity testing by generating different concentrations and combination of drugs in a time and reagent efficient manner. Xu et al. integrate the gradient concentration generator with 3D co-culture system of lung fibroblasts and NSCLC cell line (SPCA-1) or primary tumor cells, evaluating the drug response to gefitinib, PTX, and GCB simultaneously.59 Zhang et al. incorporated the automated discrete drug logarithmic concentration generation with cell culture chamber on chip to assess potency across different concentrations.60 The integration of diverse spatial designs representing tumor microenvironment (TME) architectures within microfluidic chips has unlocked significant potential for studying various tumor mechanisms and evaluating therapeutic efficacy. Advance in microfluidic chip design has increasingly incorporated cutting-edge biological technologies, such as microbead-based immunofluorescence,61 to enable precise molecular profiling. Additionally, the inclusion of electronic automation systems62 has enhanced the functionality and efficiency of these platforms, paving the way for more sophisticated and high-throughput investigations. In the next two sections, we will discuss the state of the art on microfluidic systems mimicking the complexity of normal lung microenvironment (lung-on-a-chip) and tumor microenvironment (lung cancer-on-a-chip). LUNG-ON-A-CHIP: THE STATE OF THE ART The architecture of human lung The lung, as the primary organ of the respiratory system, comprises various tissues and cell types that enable gas exchange between the bloodstream and the external environment. It consists of two main regions: the conducting portion and the respiratory portion. The conducting portion, including the bronchi and bronchioles, channels air to the respiratory portion, which begins at the alveolar ducts and culminates in the alveolar sacs and alveoli, the primary sites of gas exchange.63 The main cell types in the conducting portion of airway include basal cells, goblet cells, ciliated cells, and pulmonary neuroendocrine cells (PNECs) that control the inhaled particles and microbes in the airways, and the basal cells are the stem cells which can differentiate into the other three cell types.64 In the respiratory portion, the alveolar type II (AT2) cells produce surfactant proteins and have the progenitor features to differentiate into alveolar type I (AT1) cells, which facilitate gas exchange with the surrounding capillaries.64 In adult lung, the homeostasis and regeneration of epithelium are supported by several mesenchymal cell types, such as airway smooth muscle cells, vascular smooth muscle cells, chondrocytes, and alveolar fibroblasts.64 Additionally, lung endothelial cells (ECs) show region-specific specialization. In the alveolar capillaries, two major endothelial subtypes coexist: aerocyte capillary (aCap) cells, which interface closely with AT1 cells to optimize gas exchange, and general capillary (gCap) cells, which serve as progenitors for aCap cells.65 These capillaries also encounter complex mechanical tensions due to dynamic changes in airflow, blood flow, and alveolar expansion during respiration, which can be readily modeled in microfluidic systems given their high spatiotemporal control capability. Regular exposure to harmful stimuli, such as pathogens, air pollutants, and cigarette smoke, frequently damages the lung epithelium, challenging its self-repair mechanisms and increasing the risk of carcinogenesis. To preserve homeostasis, multiple stem and progenitor cells in the lung play critical roles in cellular and tissue regeneration. Notably, it is hypothesized that lung carcinogenesis exploits these stem cell repair pathways.66 Lung cancer is believed to originate from the transformation of progenitor cells harboring genetic mutations, which proliferate and evolve in cell identity and lineage, adapting to the lung's microenvironment over time.66 Lung cancer subtypes arise from different locations in the lung, indicating the diverse lung epithelial cell origins for tumor initiation.67 SCLC, for instance, arises in the central airways and expresses markers such as calcitonin gene-related peptide (CGRP), typically associated with PNECs, the progenitors of club and ciliated cells.68,69 Lung squamous cell carcinoma (LSCC), a subtype of NSCLC, commonly develops in the ducts of submucosal glands or at the tracheal and upper airway boundaries. It is characterized by a basal cell phenotype and the expression of markers like cytokeratin 5, P63, and SOX2, indicating its basal cell origin.70 Lung adenocarcinoma (LUAD), the most prevalent lung cancer type, is frequently found in distal airways and is believed to originate from AT2 cells.63,66 These cells are highly susceptible to oncogenic insults, as evidenced by their rapid proliferation under the activation of EGFR and KRAS signaling pathways.71 Additionally, AT2-originated LUAD cells exhibit plasticity, with reports showing transformation into highly proliferative AT1-like cells following KRAS inhibition, underscoring their role in drug resistance.72 Given the heterogeneity of lung cancer subtypes and their cellular origins, accurately modeling the native physiological environment of normal lung tissue is essential for understanding the biological mechanisms driving tumorigenesis and progression. This includes replicating the composition of key cell types (e.g., lung epithelial cells, mesenchymal cells, and vascular cells) and designing appropriate culture conditions, such as extracellular matrices, cell–cell interaction, and mechanophysiological cues. This section summarizes the selection of lung cell types for designing lung-on-a-chip platforms (Fig. 3). It also explores the impact of various culture conditions, such as the choice of culture matrices and the architecture of cell–cell interactions, on effectively recapitulating the native lung microenvironment. Microfluidic lung-on-a-chip models. (a) Type I (AT1) and type II (AT2) primary human lung alveolar epithelial cells are co-cultured with lung endothelial cells on the thin collagen–elastin membrane. Reproduced with permission from Zamprogno et al., Commun. Biol. 4(1), 168 (2021). Copyright 2021 Author(s), licensed under a CC-BY 4.0.77 (b) Diagram of a cross-section through airway-on-a-chip. The differentiated airway epithelium (pink cells) and the endothelium are at opposite sides of an air–liquid interface membrane. Reproduced with permission from Benam et al., Nat. Methods 13(2), 151 (2016).80 Copyright 2016 Springer Nature America, Inc. (c) A lung-on-a-chip platform was designed to replicate the microenvironment of the lung tissue and facilitate multicellular communication between the respiratory airways and adjacent supporting stromal chambers. Reproduced with permission from Min et al., Int. J. Biol. Sci. 21(1), 17–39 (2025). Copyright 2025 Author(s), licensed under a CC-BY 4.0.81 2D lung-on-a-chip models for physiological reconstruction Most lung-on-a-chip microsystems utilize lung cell lines or primary cells (Table III) to recreate the lung microenvironment. Cell lines, such as BEAS-2B73 and 16HBE14o-,74 with their ease of acquisition and immortalized characteristics, ensure the availability of large quantities of consistent samples. However, their genetic modifications may lead to altered or diminished cellular physiological properties. To address this limitation, primary cells directly isolated from pulmonary tissue without artificial modifications are also widely incorporated into microfluidic systems. TABLE III.Cell sources for lung cancer organoid models. HPMEC: human pulmonary microvascular endothelial cells; AEC: alveolar epithelial cells; CAF: cancer-associated fibroblast; HUVEC: human umbilical vein endothelial cells; PBMC: peripheral blood mononuclear cells; TIL: tumor-infiltrating lymphocytes; CTC: circulating tumor cells; ASC: adult stem cell; PSC: pluripotent stem cell; ESC: embryonic stem cells; PDO: patient-derived organoids.Cell sourceExamplesAdvantagesLimitationCell lineA549, BEAS-2B71, 16HBE14o-, THP1Easy and long-term maintenance; high reproducibility; high throughput; ease of genetic modificationReduced physiological relevance; genetic, transcriptomic alteration; limited functional featuresPrimary cellHPMEC, AEC, CAF, HUVEC, PBMC, TIL, CTCTissue-specific physiologic and functional properties; patient-specific disease modelingLimited lifespan; donor variability; high cost; ethical and regulatory concernsStem cellASC, PSC, ESCSelf-renewal and differentiation; long-term usability; physiological relevance; patient-specific disease modelingComplex culture conditions; batch and donor variability; immature differentiation; ethical and regulatory concernsPDO⋯High physiological relevance; heterogeneous cell composition; personalized disease modeling; 3D architecture and microenvironmentHigh cost; complex culture conditions; lack of standardization; lack of vascular and immune system; longer differentiation time Cell sources for lung cancer organoid models. HPMEC: human pulmonary microvascular endothelial cells; AEC: alveolar epithelial cells; CAF: cancer-associated fibroblast; HUVEC: human umbilical vein endothelial cells; PBMC: peripheral blood mononuclear cells; TIL: tumor-infiltrating lymphocytes; CTC: circulating tumor cells; ASC: adult stem cell; PSC: pluripotent stem cell; ESC: embryonic stem cells; PDO: patient-derived organoids. The first microfluidic airway system was introduced by Huh et al. in 2007. This system employed primary human small airway epithelial cells (SAECs) to create an ALI chip, enabling precise replication of the physiological liquid plug flows observed in the respiratory system.49 This study revealed the injurious effects of liquid rupture-induced mechanical stress on lung epithelium. Notably, airflow interface in the ALI design significantly promoted SAEC differentiation and enhanced monolayer integrity under stress conditions compared to liquid culture. Later, Huh et al. developed a functional alveolar–capillary interface using human pulmonary microvascular endothelial cells (HPMECs) and primary alveolar epithelial cells (AECs) seeded on opposite sides of a porous membrane.51 This design mimicked breathing through membrane stretching and enabled studies on epithelial–endothelial interactions, neutrophil transmigration, and nanoparticle responses, advancing insights into lung physiology and pathology. Unlike the distal alveoli, the airway mucosa comprises an epithelial surface, interstitial cells, and capillaries functioning as a unit. The thickness of the interstitial compartment decreases as bronchi become bronchioles, becoming a fraction of a micrometer in the alveoli that exert gas exchange. Thus, Sellgren et al. augmented the lung airway epithelial cell ALI cultures with primary lung fibroblasts and lung microvascular endothelial cells to pursue a greater physiologic complexity of the airway mucosa.75 To accommodate the co-culture of three mucosa cell types, they tested multiple combination of collagen I or IV as the coating substrate, which supported the differentiation of airway epithelium, as indicated by the presence of mucin granules and cilia. They carefully balanced epithelial (ALI1/2) and endothelial (EGM) media, addressing challenges such as FBS's dual role in fibroblast growth and epithelial differentiation. In addition to incorporating multiple lung airway cell types, the use of novel biomaterials has enabled the recreation of native lung biomechanics, thereby supporting the maintenance of cell type-specific identities and functions. Stucki et al. advanced alveolar–capillary modeling by successfully culturing primary AT1 and AT2 cells on a cyclic stretchable PDMS membrane for at least seven days.76 The deflection of this thin membrane was regulated through a passive mechanism driven by hydrostatic and surface tension forces. Building on this, Zamprogno et al. transitioned from traditional PDMS-based single-barrier alveolar models to a bioengineered alveolar array incorporating physiological extracellular matrix (ECM) components and mechanics. The collagen–elastin membrane, with its native geometry and tunable stiffness, more accurately recapitulated the physiological lung microenvironment and offered potential for modeling various lung pathologies. Furthermore, applying cyclic physiological stretch on chip promoted cell specification, with markers like Caveolin-1 (AT1 marker), SP-C (AT2 marker), and ZO-1 (tight junction marker) confirming maturation and epithelial function76,77 [Fig. 3(a)]. Co-cultures of airway epithelial cells and smooth muscle cells (SMCs), essential for tracheal structure, were also developed using collagen–Matrigel mixtures for adhesion and maintenance.78 Further complexity was achieved by integrating immune components into an ALI lung-chip system. Perfusing human whole blood into the chip's bottom channel enabled real-time thrombus formation analysis under inflammatory stimuli like TNF-α and LPS, highlighting increased platelet–endothelial binding79,80 [Fig. 3(b)]. Most recently, a lung-on-a-chip model was developed that incorporates multiple lung cell types, including lung stromal cells, vascular endothelial cells, lung epithelial cells, and macrophages, within a microstructure designed to mimic the respiratory compartments of the lung. The spindle-shaped respiratory airway chamber is lined with human small airway epithelial cells and surrounded by adjacent stromal chambers housing human stromal cells and macrophages, effectively creating a comprehensive representation of the lung tissue microenvironment81 [Fig. 3(c)]. These advancements enhance the physiological relevance of lung-chip models for studying airway and alveolar biology. Overall, the development of primary cell and cell line-based lung chips has led to increased cellular complexity in the gas–fluid exchange unit and the incorporation of breath-mimicking mechanical forces. These lung chips have shown significant potential in modeling various respiratory diseases and pathological processes (e.g., asthma, thrombosis, viral infections),79,80,82 highlighting their translational value. By replicating the alveolar–capillary barrier, these 2D lung chips provide a physiological foundation for studying the lung cancer microenvironment, offering new insights into cancer initiation and progression. 3D organoid-based lung-on-a-chip models for physiological reconstruction Although the 2D ALI microfluidic chip is a common method for modeling pulmonary gas-blood exchange, the primary cells are hard to be propagated as stable phenotypes in 2D cell culture. Thus, the 3D spherical organoid model with stem cell-derived lung epithelium is emerging as another in vitro method to study lung regeneration and cell–cell interaction. Organoids are self-organized structures composed of multiple cell types and extracellular matrix (ECM), mimicking the cellular composition and functionality of in vivo tissues. Unlike the ALI model, where primary cells are seeded into the system, organoids are derived from adult stem cells (ASCs) or pluripotent stem cells (PSCs).83 The lung ASCs are undifferentiated cells residing lung airway epithelial niche, including basal cells (NGFR+ITGA6+) in proximal airway and AT2 (HTII-280+) cells in the distal lung, giving rise to ciliate cells and AT1 cells, respectively.84,85 ASCs are quiescence during normal condition; however, when facing lung injuries, they can self-renew and differentiate into mature airway epithelial cells. These cells can be FACS-isolated from patient tissue and cultured in 3D matrices like Matrigel or collagen, allowing for the generation of organoids that maintain genetic stability and differentiate into specific epithelial cell types.85 On the other hand, PSCs, such as embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs), can self-renew and differentiate into any cell type in the body. The development of lung organoid from PSCs is a complicated process involving multiple growth factors or small molecular in differentiation medium to recapitulate the signaling pathways that regulate in vivo organ development.86 The immortalized culture of iPSCs enables large-scale organoid production and the potential for genetic modifications in specific cell types. Directed differentiation, an in vitro method to mimic early embryo development by activating or inhibiting specific signaling pathways, has been used in PSC-derived organoid studies to generate lung tissue lineages. By inducing definitive endoderm through Activin, inhibiting TGF-β/BMP pathways, and activating BMP/FGF signaling, a population of Nkx2-1+ respiratory progenitors can be generated and differentiated into lung epithelium.87 Although strategies to generate lung-specified endoderm are well-established, the induction of mature lung lineages remains under investigation. Recent advances have shown that 3D cultures of lung progenitors in hydrogel, such as Matrigel, can produce mature AT2 cells expressing SFTPC.86,88 Currently, AT2 cells can be cultured as spheroids in Matrigel, while airway progenitors form epithelial spheres containing ciliated and secretory cells.89 A key factor in efficiently generating specific lung lineages is the temporal regulation of signaling pathways. For instance, BMP signaling must transition from inhibition during AFE specification to activation to promote lung development.90 Inhibition of Gsk3b using CHIR9902 is essential for generating distal lung progenitors and AT2 cell spheroids, while its removal facilitates differentiation into both proximal and distal lung cells, including AT1 and AT2 cells and basal cells.89,91 In addition, maintaining high purity in lung progenitor populations is crucial. Reporter constructs targeting genes like NKX2.1, SFTPC, or SCGB3A2 help purify specific lung cell types via flow cytometry, though these methods are not suitable for clinical applications.91–93 Alternatively, surface markers such as carboxypeptidase M or CD47hi/CD26lo cells can effectively enrich lung endoderm progenitors for therapeutic purpose.93,94 Optimizing protocols to generate PSC-derived whole lung organoids containing both distal and proximal epithelial cells, along with mesenchymal cells, offers a more comprehensive platform for studying lung epithelial function in various pathological contexts and evaluating drug cytotoxicity. Moreover, the methods for the integration of vascular structure and immune components in lung organoid is the future direction for lung-on-a-chip.95 LUNG CANCER-ON-A-CHIP: THE STATE OF THE ART Modeling tumor microenvironment on chip Despite treatment advances in surgery, radiation, and targeted therapies, the prognosis of lung cancer remains poor due to the locally advanced and metastatic tumors in many patients at the time of diagnosis. The genomic landscape of NSCLC has been well defined in terms of the molecular subtypes. For example, mutations in EGFR, KRAS, ROS1, and ALK are considered oncogene-addicted and therefore sensitive to targeted therapies. However, the response rate and drug efficacy in most patients are limited by heterogeneous resistance mechanisms.3,96 Accurately identifying lung cancer subtypes and predicting their treatment responses can support personalized therapy design at early stages. Moreover, close monitoring of tumor progression during treatment is essential for timely adjustment of therapy, given the inevitable emergence of resistance. It is well accepted that the TME plays a critical role in the initiation and progression of primary lung carcinoma, and the enrichment of cell types in TME provides more targets for the development of anticancer agents.96 The distinct cellular heterogeneity of lung cancer spans the both tumor epithelial cells and TME, including vasculature, cancer-associated fibroblasts (CAFs), ECM, and infiltrating immune cells. At the organ level, the chronic inflammatory state in the lung displays features contributed to the carcinogenesis, indicating the lung tissue homeostasis in affecting tumor initiation.97 Enrichment of CAFs correlated with poor prognosis that associated with metastasis.98 Increased tumor-infiltrating lymphocytes (TILs) is associated with better prognosis and higher survival,99 while regulatory T cells were associated with worse recurrence-free survival.100 Given the strong influence of TME on lung tumor progression, appropriate models that mimic the in vivo TME features are fundamental to finding new therapeutic targets and testing efficacy. To this end, the microfluidic platform has been emerging as a multifunctional in vitro lung cancer modeling system. It enables the co-culturing of multiple TME cell types in a controllable spatial organization with specific ECM properties that mimic the tumor tissue geometry (Fig. 4). The incorporation of engineering factors, such as fluidic flow, air–liquid interface, interstitial pressure, oxygen gradient, and biosensors, further benefits the application of lung cancer-on-a-chip in lung tumor diagnosis, cancer stage monitoring, and drug testing. Multicellular lung cancer-on-a-chip systems. (a) Co-culture of fibroblasts with lung cancer cells in one-directional flow chip for studying paracrine signaling between cancer cells and fibroblasts. Reproduced with permission from Kim et al., Acta Biomater. 144, 258–265 (2022).108 Copyright 2022 Elsevier. (b) A vascularized lung cancer-on-a-chip platform using sacrificed vascular channels with tumor spheroids encapsulated in the hydrogel. Reproduced with permission from Park et al., Cancers 13(16), 3930 (2021). Copyright 2021 Author(s), licensed under a CC-BY 4.0.113 (c) A functional vascularized microfluidic chip with U-cup-shaped area functioning as a trap for growing organoid. Growth medium was introduced for continuous perfusion of the microfluidic chamber and the trapped organoid. Reproduced with permission from Quintard et al., Nat. Commun. 15(1), 1452 (2024). Copyright 2024 Author(s), licensed under a CC-BY 4.0.114 (d) Patient-derived lung cancer assembloids generated using microinjection-based cell-laden hydrogel manipulation to reserve multiple cell types from in vivo. Reproduced with permission from Zhang et al., Nat. Commun. 15(1), 3382 (2024). Copyright 2024 Author(s), licensed under a CC-BY 4.0.118 Cellular complexity on lung cancer-on-a-chip 3D modeling of lung cancer-on-a-chip Cancer cell lines (Table III) have long served as the primary tool for investigating biological phenotypes due to their easy acquisition, ability to expand, and consistent passages. The US National Cancer Institute has established 60 human tumor cell lines, including 9 NSCLC lines, for drug screening.101 These cell lines are extensively characterized at various molecular levels, such as DNA methylation, mutation burden, RNA expression, and metabolic activity, providing a solid foundation for in vitro mechanistic studies. Additional lung cancer cell lines, such as NCI-H1975 (EGFRT790M) and NCI-H358 (KRASG12C), are valuable for exploring drug resistance mechanisms.102 Over recent decades, 3D tumor spheroid cultures have gained attention for more accurately replicating tumor structures, including cell shape and density. Tumor spheroids can form either through spontaneous aggregation or by embedding cells in hydrogels like Matrigel or collagen. While these spheroids may not fully mimic the histological pattern of the original tumor, they do replicate key metabolic and proliferation gradients, making them useful for high-throughput screening.103 However, cancer cell lines alone cannot completely capture the complexity of in vivo tumors. The primary cancer cells dissociated from tumor tissue serve as an alternative for in vitro modeling. It was reported that freshly dissociated lung tumor cells from patient samples could form spheroid by embedding into IrECM derived from the Engelbreth–Holm–Swarm mouse sarcoma cells.104 Patient-derived organoids (PDOs), which are directly obtained from surgical samples or biopsies, preserve the native tumor histology and are increasingly viewed as ideal models.103 Table III summarizes the common cellular sources for modeling lung cancer, and Fig. 2 shows common sample types (tissue, blood, pleural effusion, and bronchoalveolar lavage) that may be collected during the course of lung cancer disease management. Tools like hydrogel embedding and ultra-low attachment plates are commonly used for spheroid formation or PDO subculture, though they are cell-type dependent, challenging to control morphologically, and labor-intensive. Recent improvements in microwell chips with agarose coatings have enabled consistent spheroid and PDO formation, facilitating high-throughput screening and drug testing.55,105 The low cost of agarose and the efficiency of these chips have made them highly valuable. However, the limited size of tissue biopsies and PDOs presents challenges for their seeding and expansion. Tissue trapping chips, designed to efficiently trap individual biopsies106 or PDOs,57 have been developed to test chemotherapy efficacy. Recent PDO trapping chip has also been used to monitor interactions between PDOs and autologous immune cells, evaluating lymphocyte-mediated killing behavior.61 Multicellular organoid models with CAFs Given the crucial role of the TME in tumor progression, drug response, and resistance, reconstituting in vitro TME models is an emerging trend in the field. CAF, one of the most abundant and heterogeneous cell types within the TME, significantly influence tumor behavior. The presence of tumor-like CAFs is strongly associated with poor prognosis. Different CAF subtypes play different functions in TME and the associated drug response.107 The inflammatory CAFs and interferon-response CAFs are associated with inflamed TME, indicating the good prognosis, while the ECM-producing CAFs are correlated with low immune infiltration and poor prognosis.107 It is noteworthy to recapitulate the cellular interaction between CAFs and tumor cells in lung cancer-on-a-chip model. An early study mixed lung cancer cell line SPCA-1 with human lung fibroblast cell line HFL1 in Geltrex and injected in chip coupled with fluidic flow.59 The concentration gradient generator on chip enabled simultaneous testing of various drug concentration in tumor–fibroblasts chambers, screening the most sensitive treatment condition. The paracrine interaction between tumor cells and fibroblasts was investigated in the microfluidic channels, where the tumor cells and lung fibroblasts were separated by collagen I filled channel108 [Fig. 4(a)]. In another study, the head pressure between channels generated the unidirectional flow from tumor cells to fibroblasts, facilitating the delivery of secretory factors. This paracrine crosstalk reshaped the normal lung fibroblasts into CAF-like phenotype, including increased migration and invasion accompanied by upregulation of CAF activation markers (e.g., MMP-9, FAP, Vimentin). However, due to the difficulty in isolating primary CAFs and the challenges of long-term culture-induced senescence, the use of primary CAFs in cancer-on-chip studies is limited.109 Tumor cell-induced transformation of normal fibroblasts into CAF-like cells offers a promising alternative for recreating the TME. Multicellular organoid models with vascular cells The blood vessels, composed of an inner endothelial layer and outer pericytes, play a crucial role in the TME, influencing tumor aggressiveness and drug response. Tumor blood vessel is disorganized, leaky, and tortuous, with poor pericytes or smooth muscle cells coverage.110 The collapsed, narrowed lumen of tumor vessels leads to low perfusion and a hypoxic environment, hindering immune cell infiltration and drug delivery.111 Crosstalk between tumor cells and blood vessels also facilitates metastasis, where local tumor cells transition to a mesenchymal state and invade blood vessels through the endothelium. Once in the bloodstream, tumor cells can spread and seed distant organs. However, the low fraction of endothelial cells (ECs) within the TME makes isolating primary tumor vessel ECs difficult. As a result, human umbilical endothelial cells (HUVECs) are commonly used in tumor-on-chip systems, although they are not tumor-derived. Some studies have also used iPSC-derived ECs as an alternative.78 The advantages of iPSC-ECs in tumor-on-chip models include their immortality and potential to adopt TME-associated phenotypes when co-cultured with tumor cells. In one early study, A549 cells and lung epithelial cell line 16HBE were co-cultured on the upper side of a membrane, while the HUVECs and macrophages were on the opposite side. The lower blood vessel channel was perfused and connected to several portals cultured with typical cell types from brain, liver, and bone. The migration of tumor cells through endothelium toward downstream organ-like portals could be monitored at the single-cell level, mimicking the metastasis112 [Fig. 5(a)]. However, this kind of microchannel chip could not reflect the spontaneous assembly of vessel structure seen in tumor and the cell crosstalk was restricted by the membrane. Lee et al. improved this approach by seeding collagen-embedded A549 cells in microwells connected to a microfluidic channel, with HUVECs forming an endothelial layer at the junction between the channel and microwells. The fluidic flow facilitated the integrity of the endothelial layer and promoted angiogenesis of ECs into tumor cell aggregates in each well.58 A similar “lumen chip” design created endothelial channel and incorporated tumor-lung fibroblast clusters in decellularized porcine lung-derived hydrogel. The tissue-mimetic hybrid hydrogel and the existence of fibroblasts promoted the angiogenesis of tumor spheroid, and it allowed for perfusion in endothelial channel to facilitate the drug screening113 [Fig. 4(b)]. This model successfully recapitulated tumor–stroma–vasculature interactions and demonstrated drug responses, though the vascular complexity was limited to large endothelialized channels with short-term culture. More recently, a study cultured iPSC-derived vessel organoids on a chip, enabling the formation and spread of perfused microvascular networks, providing a new resource for vascularizing 3D tumor tissues114 [Fig. 4(c)]. This versatile microfluidic platform enabled self-assembly of perusable microvascular networks within hydrogel-embedded organoids, maintaining long-term perfusion and vessel functionality. Unlike primary endothelial cells such as HUVECs, iPSC-derived endothelial cells possess a more naïve cellular identity, offering the potential to be further directed or “trained” toward phenotypes that more closely resemble the microvascular characteristics of the lung TME. Lung cancer-on-a-chip application in diverse therapeutic modalities. (a) A multiorgan microfluidic chip that includes an upstream “lung organ” and three downstream “distant organs.” Reproduced with permission from Xu et al., ACS Appl. Mater. Interfaces 8(39), 25840–25847 (2016).112 Copyright (2016) American Chemical Society. (b) A symmetric flow-based lung cancer-on-chip model for studying T-cell activity in response to interactions with lung cancer cells and varying IL-6 gradient concentrations. Reproduced with permission from Sardarabadi et al., ACS Appl. Mater Interfaces 17, 4354 (2025). Copyright 2025 Author(s), licensed under a CC-BY 4.0.119 (c) Rapid EGFR mutation detection and drug response test in a dual-functional microfluidic chip. Adapted with permission from Zhang et al., Lab Chip 24(6), 1762–1774 (2024).134 Copyright 2024 Royal Society of Chemistry. Overall, fluidic flow, matrix composition, and multicellular interactions (e.g., fibroblasts) are critical in influencing vascular structure formation and growth on chips.115 Additionally, tissue clearance and imaging techniques should be integrated with microfluidic systems for comprehensive and quantitative analysis of vascular networks. Multicellular organoid models with immune cells There are various immune cells reside in TME, including lymphocytes, tumor-associated macrophages (TAMs), natural killer cells, dendritic cells, and cancer-associated neutrophils. These immune cells play complex roles in regulating the tumor response to treatment.45,116 Immunotherapy, including immune checkpoint inhibitors, has significantly improved treatment outcomes for advanced-stage NSCLC. Incorporating immune cells into microfluidic systems is essential for recreating a close-to-actual TME and studying tumor responses to immunotherapy on a chip. To recapitulate the native TME, THP1, a monocyte cell line, was mixed with tumor cells and fibroblasts as spheroids.117 It was observed that THP1 could be transformed into CD68+ TAMs after the long-term co-culture with tumors. However, since immune cell lines still cannot reveal the features of tumor-specific immune cells, the use of autologous cells from patients, such as tumor-infiltrating lymphocytes (TILs), offers a more authentic representation of the TME. A recent study used droplet microfluidic technology to assemble patient-derived TME cells, including CAFs and TILs, with PDOs in the GelMA-Matrigel microgels118 [Fig. 4(d)]. The generated tumor assembloid model demonstrated good intrabatch consistency in size, cell composition, and drug response profiling. Other than directly incorporating immune cells with tumor cells, the perfusion of lymphocytes in microchannels on chip can benefit the investigation of immune activation in the TME. One study designed ideal chambers and channels to form a concentration gradient generator chip, mimicking T cells interaction with cancer cells in the presence of different concentration gradients of IL6 in TME.119 This study emphasizes the dual effects of IL-6 on T-cell activation, exhaustion, and cytokine production, providing valuable insights into immune response modulation. Given the heterogeneous and phenotypic plasticity of tumor-associated immune cells, it is necessary to achieve the effective monitor and quantitative analysis of immune cells activities on chip (e.g., migration, proliferation, and activation) via imaging and secretory profiling.61 Circulating tumor cell (CTC) studies In addition to tumor cells from solid tumor biopsy, the collection and analysis of circulating tumor cells (CTCs) from liquid biopsy have emerged as an effective and non-invasive way to support early detection, monitoring response to treatment, and predicting disease recurrence. These cells hold the potential of supplying the cell source for real-time modeling of lung cancer in vitro. CTCs are shed from the primary tumor and circulate in the blood, and they are suggested to be the potential founders of distant metastasis.120 Due to their low numbers in circulation, advanced techniques are required to accurately identify and isolate CTCs from limited blood samples. The CellSearch system, an immunomagnetic enrichment based on EpCAM expression, is the only FDA-approved method for CTC detection. However, this system dismisses the potential CTC subtypes that are negative for EpCAM. An alternative method involves size-dependent microfluidic enrichment, taking advantage of the relatively larger size of CTCs compared to blood cells.121,122 During the course of treatment, the molecular landscape of tumor cells undergoes continuous evolution, ultimately leading to the development of acquired resistance to therapy.123 Lung cancer models from CTCs, including CTC-derived organoids,124 offer a minimally invasive approach to monitor tumor dynamics and can provide timely insights to guide therapeutic modifications, as CTCs can be obtained from patient blood samples without causing harm. In vitro culture and characterization of CTCs offer opportunities to explore drug susceptibility patterns in individual patients, aiding in the development of tailored therapeutic strategies. Several culture methods for CTCs were reported, including the non-adherent culture condition,125 microfluidic fibroblast co-culture model,126 and white blood cell co-culture model.127 Using microfluidic platform and proper formula of culture conditions, precise enrichment and quick expansion of CTCs become promising. The phenotype analysis and drug testing of CTC-derived tumor organoids represent a new direction for personalized cancer therapy. Nevertheless, challenges such as low success rate in deriving organoids remain to overcome.128 Commercialized lung cancer-on-a-chip platforms Built upon decades of academic research and iterative design, lung cancer-on-a-chip platforms have increasingly been commercialized, offering standardized, reproducible, and user-friendly models that accelerate translational research and preclinical drug testing. Several commercial systems now integrate features that mimic key aspects of lung physiology. For example, the Chip-S1 by Emulate and the AXLung-on-chip by AlveoliX incorporate stretchable membranes to replicate the mechanical forces stemming from breathing. Similarly, the UBeat® Stretch chip by BiomimX uses an actuation layer underneath the cell culture layer to transfer a controlled mechanical stimulation to the cell constructs. To recapitulate circulatory dynamics, platforms such as the Pod module of Chip-S1, MesoBioTech's organ-on-chip, and ComPLATE by Bi/ond include perfusable channels and flow control systems that extend tissue viability and support long-term modeling. Beyond physiological relevance, many commercial platforms are tailored for specific research applications. The OrganoPlate by MIMETAS combines microfluidic flow with multiplexed 3D culture wells, enabling high-throughput co-culture and tumor–immune interaction studies. Systems like Cherry Biotech's CUBIX and Bi/ond's ComPLATE further support real-time monitoring through live-cell imaging in controlled perfusion environments. While these commercial lung-on-a-chip platforms are far from recapitulation of human lung physiology and microenvironment, they bridge the gap between academic prototyping and scalable application, facilitating disease modeling with enhanced physiological relevance, enabling more precise and rapid drug screening, and supporting high-throughput, multiparametric readout analysis. MICROFLUIDIC APPLICATION IN DIVERSE THERAPEUTIC MODALITIES The advent of immunotherapy (e.g., PD-L1 inhibitors) and targeted therapy (e.g., TKIs) has significantly shifted the treatment paradigm for lung cancer. The development of microfluidic platforms has enhanced the modeling of TME in lung cancer, providing broad opportunities for rapid and sensitive testing of various therapeutic modalities (Fig. 5). This advancement accelerates progress in personalized medicine by enabling more tailored and effective treatment strategies. Metastasis is an important tumor progression event, and its occurrence usually accompanies tumor cell phenotype switch and requires the adjustment of the therapeutic strategies. For this point, multiorgan microfluidic chip model offers a physiologically relevant in vitro model of lung cancer metastasis112 [Fig. 5(a)]. The device used layered PDMS microchannels and microporous membranes to spatially segregate and yet fluidically connect the compartments, enabling tumor cells to migrate from the lung region into the downstream “organ” chambers under flow. Rather than simple 2D culture, the organ-on-chip design permitted more realistic evaluation of anti-metastatic drug effects, cell–cell and organ–organ interactions, and therapeutic resistance in a controllable microenvironment. To evaluate the efficacy of immunotherapy, such as immune checkpoint inhibitors (ICIs), a dynamic on-chip TME that supports interactions between immune cells and tumor cells is essential. Moore et al. developed a multiplexed microfluidic system termed EVIDENT (in vitro immuno-oncology dynamic environment for tumor biopsies), which enables real-time, high-resolution imaging of interactions between autologous TILs and tumor fragments, allowing for the assessment of various ICI treatments and tumor responses.129 The compatibility of the chip with microplate readers or high-content imaging system enhances its usability. To account for the impact of ECM barriers on immune cell-mediated cytotoxicity, an injection-molded plastic array culture device was designed to evaluate the killing efficiency of cytotoxic lymphocytes within a 3D microenvironment. This platform provides spatiotemporal analysis of lymphocyte and cancer cell behavior in ECM, revealing that while dense ECM strengthens the physical barrier, cytotoxic lymphocytes effectively kill cancer cells upon direct contact.130 In a more advanced microfluidic platform for immuno-oncology modeling, tumor cells were embedded within a 3D hydrogel matrix while microfluidic uGGs were integrated to generate cytokine gradients and sustain continuous perfusion. This configuration created a physiologically relevant dynamic environment that enabled the study of T-cell chemotaxis and differential cytokine exposure across the chip.119 The system also incorporated multiple downstream analytical modules, including real-time cell tracking, flow cytometry, and cytokine quantification, allowing comprehensive assessment of immune responses within the tumor microenvironment119 [Fig. 5(b)]. This design highlights a future direction for immuno-oncology therapeutic modeling—toward more multiparametric, high-throughput microfluidic platforms that enable integrated evaluation of complex tumor–immune interactions. As new immunotherapy modalities, including immune checkpoint blockade and adoptive T-cell transfer, continue to emerge, preclinical models capable of assessing their ability to overcome immune suppression within the tumor microenvironment become increasingly important.131 The use of microfluidic tumor immunity models offers significant potential for predicting the efficacy and clinical outcomes of personalized immunotherapies.131 The utility of microfluidic device for mutation detections (e.g., EGFR) and TKI drug testing on patient tumor samples can facilitate the personalized treatment strategy designing. Hassell et al. used H1975 cell line which carries EGFRT790M, the mutation driving resistance to TKIs, in the breathing mimicking ALI chip to model the orthotopic lung tumor growth and invasion. It was validated that the H1975 cells were more sensitive to the third generation TKI (rociletinib) than the first-generation TKI (erlotinib).47 More interestingly, the breathing motion significantly increased the drug resistance of H1975 cells to the inhibitory effects of rociletinib, providing more evidence of the drug resistance mechanism in NSCLC. Shigeto et al. utilized a combination of single-cell microarray chips and peptide nucleic acid (PNA)–DNA probes to separate H1975 cells, which harbor EGFRT790M mutations, from non-mutated cancer cells (A549), using fluorescence signals for labeling.132 Shen et al. developed a digital microfluidics-based FISH (DMF-FISH) platform that automates the traditionally labor-intensive FISH procedure for detecting EGFR amplification in NSCLC.133 The system integrates droplet handling, temperature control, and signal processing, enabling efficient cell digestion, hybridization, and staining with minimal manual intervention. Recently, a dual-functional microfluidic chip was developed to detect EGFR mutations in PDOs within 2 h using a DNA-based nanoruler. The mutation detection results guided PDO-based drug response testing for TKIs or chemotherapy. Additionally, a DNA-based nanosensor enabled real-time monitoring of ATP levels to assess PDO growth134 [Fig. 5(c)]. This integrated platform offers a personalized approach to lung cancer treatment by combining mutation detection and drug sensitivity testing. Currently, most lung cancer drug resistance studies use mutation-carrying tumor cell lines and focus on the detection of drug-resistant cells in vitro. However, the mechanisms by which new mutations are acquired after drug treatment are elusive and134 individually relevant. Microfluidic platforms offer an exciting opportunity to replicate the mutation acquisition process using patient-derived tumor cells cultured on-chip, enabling more personalized and effective drug resistance research and the identification of new therapeutic targets. CONCLUDING REMARKS AND OUTLOOK The microfluidic-based lung cancer-on-a-chip technology is advancing rapidly, demonstrating broad applications in the diagnosis and treatment of lung cancer. Compared to traditional 2D cell culture and 3D organoid models, lung cancer-on-a-chip systems recreate a miniature microenvironment that closely mimics the physiology of lung tissue and cancer. Features like ALI co-culture of lung epithelial and capillary endothelial cells emulate the gas-blood exchange unit, while 3D co-culture of tumor cells with stromal cells and perfusion channels replicates the TME for disease modeling and drug testing. These systems have significantly enhanced our understanding of lung tumor initiation, progression, and metastasis in a resource-efficient manner. However, challenges remain in fully replicating the lung cancer microenvironment at the tissue and organ levels, such as collection and subculturing of native tumor tissue, normal lung tissue–tumor interaction, and multiorgan microphysiological systems for metastasis and drug metabolism studies. Integrating lung cancer-on-a-chip platforms with technologies like high-throughput imaging, biosensors, and artificial intelligence (AI) could unlock new opportunities for clinical diagnostics and therapeutic development. The adoption of PDOs over conventional cell lines is gaining traction in the organ-on-chip field due to their superior ability to recapitulate the in vivo tumor environment and enhance personalized medicine. However, significant challenges persist, particularly the variable PDO culture success rates, which range from 7% to 87% based on sample type and processing methods.135 Surgical resection specimens generally yield higher success rates compared to smaller volume samples, such as percutaneous needle or lymph node biopsies.135 Another limitation lies in the contamination of PDO cultures with normal lung mesenchymal cells, necessitating the optimization and standardization of key processes, including sample collection, digestion, medium formulation, and biophysical microenvironment design. Once established, the validation of lung cancer PDOs to ensure faithful representation of the original tumor is critical. Basic parameters, such as morphology and histological markers (e.g., thyroid transcription factor-1, cytokeratin 5/7, and p63), can be assessed through H&E staining and immunostaining.136 Advanced genetic and transcriptomic approaches, including whole-exome sequencing, copy number variation analysis, single-cell RNA sequencing, and spatial multiomics,137,138 provide deeper insights into tumor identity. Additionally, periodic validation of PDOs during long-term culture is essential to monitor potential cell population shifts across passages, ensuring the consistency and reliability of the model for downstream applications. CTCs have emerged as an alternative cell source for advancing disease monitoring and metastasis research. Key challenges in CTC analysis include achieving high purity during isolation, enabling efficient single-cell analysis, and optimizing the rapid expansion of CTC clones. Improvements in microfluidic-based CTC enrichment techniques, along with advancements in culture media composition and substrate design, hold significant potential to enhance the efficiency and applicability of CTC-based lung cancer research.125,126 Despite advancements in reconstructing the TME with cancer and stromal cells, current lung cancer organoid models still fall short of replicating the lung-specific physiological environment. Integrating tumor organoids with lung epithelial cells, capillary endothelial cells, and resident immune cells remains a key challenge. Primary lung epithelial cells and endothelial cells that used in most lung-on-a-chip model may not reveal the specific characteristics of AT1 and AT2 cells and their interaction with lung capillaries. Co-culturing stem cell-derived alveolar organoids with tumor organoids offers a promising approach to recreating the lung microenvironment, though efforts are needed to promote the maturation of alveolar structures, particularly alveolar type 1 (AT1) cells, which experience higher mechanical tension.139 Microfluidic platforms applying controlled mechanical forces could aid this maturation process. Another critical aspect is replicating the oxygen gradient between alveoli (∼13.3 kPa) and the hypoxic tumor microenvironment (<2.03–3.04 kPa).140 Well-designed gas channels on chips could facilitate hypoxia-targeted therapy studies. Furthermore, when co-culturing tumor and normal lung organoids, the functional identity of alveolar cells—such as marker expression, surfactant production, inflammatory signaling, and viability—must be closely monitored. This necessitates microfluidic chip designs compatible with staining, imaging, and immunoassay platforms to effectively evaluate tumor–lung cell interactions. While most studies using patient-derived tumor materials focus on advanced-stage cancer, modeling early tumorigenesis on-chip represents a critical yet underexplored area. Genetic-engineered mouse models have demonstrated that the loss of Rb1 and p53 induces SCLC,141 while the mutation in KrasG12D and p53 loss in AT2 cells could generate LUAD model in both mouse and human iPSC-derived lung organoids.142 Incorporating inducible mutations into cancer-prone lung cell types, such as basal cells for SCLC or AT2 cells for NSCLC, could help elucidate the transformation processes in early tumorigenesis. Microfluidic devices provide a platform for high-throughput and biophysically controllable propagation of mutation-carrying lung organoids, enabling long-term monitoring of early tumor cell transformation through cell-type-specific marker expression. At later stages, these systems can facilitate drug screening and response evaluation on emerging tumor cells, helping to identify valuable early therapeutic targets. The integration of microfluidic platforms with other up-to-date technologies such as biosensors and artificial intelligence (AI) also holds immense potential for advancing lung cancer-on-a-chip research. Biosensors, capable of real-time monitoring of biochemical and biophysical signals, can be incorporated into microfluidic systems to dynamically measure critical parameters, such as oxygen gradients, cytokine secretion, and drug response biomarkers. While AI excels at analyzing complex and large volume datasets to identify patterns, predict outcomes, which help the optimization of experimental conditions, AI-driven image analysis could improve the detection of subtle morphological changes in 3D cultures and improve the output efficiency of personalized drug testing. By combining the high-throughput capabilities of microfluidic platforms with the precision of biosensors and the analytical power of AI, sophisticated lung cancer-on-a-chip models can be created to better replicate the complexity of lung cancer in vivo, accelerating drug discovery and improving therapeutic outcomes. The integration of microfluidic devices into lung cancer research presents several challenges, particularly for biologists and clinicians who may lack engineering expertise. A major barrier is the reliance on specialized fabrication techniques, such as soft lithography and 3D printing, which are not typically available in standard cancer biology laboratories. While commercial availability could help bridge this gap, it requires standardized device components and operational workflows—both of which are currently lacking. Such standardization is crucial not only for reproducibility but also for enabling scalability and broader dissemination. To improve accessibility, future designs should focus on minimizing assembly complexity, providing clear, user-friendly manuals, and ensuring compatibility with conventional laboratory equipment such as microscopes, multiwell plates, and high-throughput screening platforms. By overcoming these barriers, lung/cancer-on-a-chip technologies can become more user-friendly, promoting wider adoption in cancer research and biomedical applications. Finally, achieving complex lung cancer models for preclinical drug discovery and individualized therapy requires seamless collaboration among scientists, clinicians, and bioengineers.143 Current challenges and solutions in this multidiscipline cooperation include the following: (1) Divergent priorities: Scientists focus on understanding disease mechanisms and preclinical data, while clinicians prioritize immediate, evidence-based patient care.144 (2) Communication gaps: Discipline-specific knowledge hinders mutual understanding. (3) Logistical barriers: Limited access to clinical samples, funding competition, and physical separation of labs and clinics hinder integration. Shared biorepositories and data-sharing platforms, along with clinical research associates, may streamline collaboration. To overcome these obstacles, scientists are encouraged to become well-versed in clinical workflows and gain an understanding of clinical practice to accurately identify existing deficits that research can address. Similarly, clinicians must be educated in basic science to facilitate meaningful collaboration. Once a mutual understanding is established, the next step involves regular cross-disciplinary meetings to identify relevant research questions, exchange ideas, and monitor project progress.144 Addressing these barriers with efficient communication, shared infrastructure, and aligned priorities will accelerate translational research and its impact on clinical outcomes.