Introduction

According to global cancer statistics, in 2022 [1] there were approximately 2.48 million new lung cancer cases diagnosed worldwide. By 2025, there are projected 2,041,910 new cancer cases and 618,120 cancer deaths in the United States [2]. The statistics also reported that lung cancer was responsible for 1.82 million deaths which appears to be the leading cause of cancer deaths [3]. In patients with advanced metastatic non-small cell lung cancer (NSCLC), 20%−40% of patients will develop bone or brain metastases, and median overall survival reduces to 4 − 6 months. Metastasis is a major contributor to cancer-related deaths [4], and almost 90% of cancer deaths are due to the invasive and metastatic nature of the tumor. Current clinical treatment options for patients with metastatic lung cancer remain quite poor, and therapies only provide a minor contribution to survival and although they do improve progression free survival, they do not provide an increase in overall survival rate [5].

The growing demand for more effective therapies has driven the development of anti-metastatic agents and improved evaluation platforms. The growing demand for more effective therapies has driven the development of anti-metastatic agents [6, 7] and improved evaluation platforms [8, 9]. For instance, astragaloside IV suppresses tumor migration via the PKC-ERK1/2-NF-κB pathway [10], ginsenosides inhibit epithelial–mesenchymal transition, and resveratrol reduces proliferation while inducing apoptosis in lung cancer cells [11].

Conventional in vitro and in vivo models, such as animal models, scratch assays, and Transwell systems, are limited in their ability to recapitulate the complexity of tumor metastasis [12]. These methods lack precise control over the tumor microenvironment, fail to capture dynamic cell–cell and cell–matrix interactions, and typically rely on endpoint measurements [13, 14]. In contrast, organ-on-a-chip (OoC) systems enable the integration of multiple cell types and allow real-time monitoring under controlled microenvironmental conditions, offering higher sensitivity, throughput, and physiological relevance.

Organ-on-a-chip (OoC) technology represents an exciting alternative model system by recreating the human organ physiological microenvironment on microfluidic(a fluid flow channel with dimensions less than 100 μm) platforms [15, 16]. This system duplicates the in vivo tumor microenvironment more closely while also correcting for the “black box” disadvantages of using animal models [17]. These systems are cost-effective, scalable, and well-suited for high-throughput drug screening.

In this study, we developed a lung cancer–vascular tumor-on-a-chip model to simulate interactions between tumor cells and the microvascular environment. This platform enables precise regulation of microenvironmental conditions and real-time monitoring of tumor spheroid growth and invasion [18,19,20]. To evaluate the “parameters and stability” of our model for drug screening, we used five credible and widely used anti-metastatic lung cancer drugs Paclitaxel (PTX) [21], Cisplatin(CDDP) [22], Irinotecan (CTP-11) [23, 24], Oxaliplatin (OXA) [25], and Gemcitabine (GEM) [26]. PTX [27], CDDP [28], CTP-11 [29], and GEM [30] are all classic first-line clinical chemotherapy drugs [29, 31]. In addition to being widely used in the treatment of colorectal cancer, OXA ‘s efficacy in NSCLC has been confirmed by multiple studies: its combination with drugs such as GEM can enhance anticancer effects and improve patient survival rates [32, 33]. Moreover, OXA has demonstrated efficacy comparable to CDDP, with potential advantages in reducing certain adverse effects.

Furthermore, evaluated rocaglamide (ROC-A), we identified ROC-A from the natural family Aglaia species which is a potent cGAS-STING pathway agonist with regulatory functions on proliferation and apoptosis of tumor cells [34, 35]. With a molecular weight of 505.56 Da and good water solubility, ROC-A can readily cross the endothelial barrier and act on tumor cells. We successfully confirmed ROC-A as a dose-dependent inhibitor to tumor invasion. These results indicate that our lung cancer-vascular tumor-on-a-chip model develops an accurate model of the ECM invasion of the tumor, which proved to be suited for a platform screening compounds that are anti-metastatic in nature - including natural products, and open new avenues in cancer treatment and drug discovery.

Methods

Cell sources

Cell lines of lung adenocarcinoma A549, H460, and HUVEC-TURBO-GFP (HUVEC-GFP) were purchased from Shanghai Zhong Qiao Xin Zhou Biotechnology Co., Ltd. A549 (Figure S1) and H460 (Figure S2) cell lines were confirmed by STR analysis and mycoplasma testing (Figure S3) to ensure genetic identity. The lung cancer cell lines (A549 and H460) were grown in RPMI 1640 medium (Gibco) that was previously supplemented with 10% fetal bovine serum (Gibco) and 1% gentamicin/amphotericin B (RPMI 1640+), and the HUVEC-GFP cells were grown in endothelial cell medium ECM (ScienCell#1001). All cells were maintained in a humidified incubator at 37 °C and 5% CO₂.

Microfluidic device fabrication

The fabrication process followed our previous reports with modification [36,37,38,39,40]. In brief, Polydimethylsiloxane (PDMS) was mixed with curing agent (DOW, SYLGARD™ 184 Silicone Elastomer Kit) at a 10:1 ratio, degassed to remove air bubbles, and poured into a mold. The mixture was cured at 90 °C for 6 h. After curing, the PDMS microfluidic device was removed and perforated using 3 mm and 5 mm biopsy punches to create inlet/outlet channels and reservoirs. The PDMS microfluidic device was then irreversibly bonded to a glass coverslip via plasma treatment. A nickel needle (diameter: 220 μm) was placed at the center of the channel, and a 1:1 mixture of SYLGARD 160 A and SYLGARD 160B was applied to secure the needle, completing the chip fabrication (Fig. 1A). Before use, the device was sterilized with alcohol and UV radiation.

Tumor spheroid preparation

A549 and H460 cell lines were placed in cell culture dishes (Corning, 430591) and incubated at 37 °C. Upon reaching 80%-90% percent confluency in a culture dish, the culture medium was removed and washed twice with 2 mL of PBS. 1 mL of 0.25% trypsin was added to the dish, which was incubated at 37 °C for 2 min. Incubation was terminated by adding 2 mL of complete culture medium to the dish. The cells were gently pipetted up and down and transferred to a sterile centrifuge tube. The tube was centrifuged for 3 min at 800 rpm. The supernatant was discarded, and the cell pellet was resuspended in complete medium at approximately 2 × 10⁶ cells/mL. A 2 mL aliquot of the cell suspension was seeded into ultra-low attachment six-well plates (Corning; 3471-ZX) to allow for spheroid formation for five days. On day 3, the cell suspension was gently pipetted to promote aggregation into tumor spheroids of moderate size. By day 5, fully formed tumor spheroids were established.

Tumor spheroid staining

On day 4 of tumor spheroid culture, spheroids were obtained by transferring the entire dish of spheroids into a sterile centrifuge tube and allowing the spheroids to settle for 10 min. The supernatant was removed; then 1 mL of basal RPMI 1640 medium was added along with 1 µL of MitoTracker® Deep Red FM (Yeasen, #40743ES50) fluorescent dye to achieve a final concentration of 10 µM. The mixture was gently pipetted and allowed to settle for 10 min in the dark. The supernatant was removed; then the spheroids were extracted and resuspended in complete culture medium at a density of approximately600 spheroids /mL. A 2 ml aliquot of the stained spheroid suspension was re-seeded into an ultra-low attachment six-well plate. On day 5 of culture, the tumor spheroid radius reached 100 micrometers (Figure S4). The tumor spheroids were then allowed to settle and were mixed with the extracellular matrix material.

Construction of the vascularized tumor-on-a-chip

Extracellular matrix material preparation and tumor embedding: Type I collagen from rat tail (Corning, 354249) was neutralized and diluted to 7 mg/mL using DI water, 1 M NaOH (S2770, Sigma), and 10× DMEM. Tumor spheroids A549 from H460 were embedded in the collagen solution at a final concentration of 5 spheroids/µL and injected into the central chamber, allowing random distribution within the matrix. Gelation and structural support: The collagen was polymerized in a 37 °C incubator for 12 min. Heated agarose was then introduced between the small and large wells to provide structural support and prevent collagen collapse. A small amount of RPMI 1640 + culture medium was added to the 5 mm large well to prevent drying, and the chip was stored at 4 °C for 6 h to remove air bubbles. Microchannel fabrication: The nickel needle was removed, forming a cylindrical channel within the gel, which was perfused with 1 RPMI 1640 + culture medium for one day. Endothelialization: HUVECs-GFP were then injected into the channel at a concentration of 10⁷ cells/mL. The device was incubated at 37 °C for one hour to allow HUVEC-GFP adhesion to the channel walls. Perfusion culture and imaging: Using previously established methods, the device was subjected to continuous perfusion at a constant flow rate of 1 mL/h (Fig. 1B). On the second day after seeding HUVEC-GFP cells, tumor spheroid and vascular morphology were imaged using the Echo Lab Revolve (Revolve Generation 2), designated as day 0. Imaging was repeated every 24 h until day 5. Drug administration: On day 1, the drug was dissolved in the culture medium and delivered through the perfusion system via the vascular channel to act on the tumor spheroids.

Fig. 1
Fig. 1
Full size image

Fabrication process of the lung cancer-vascular tumor-on-a-chip (A) Image of the device. The large hole is used for perfusing culture medium, while the small hole is for injecting type I collagen and agarose. After removing the microneedle, a complete cylindrical channel is formed inside the device, allowing for endothelial cell seeding. (B) Image of the device after connecting to the gravity-driven flow. A constant flow rate (1 mL/h) is applied to ensure that endothelial cells form and maintain a cylindrical vascular channel under shear stress

Image acquisition and processing

Images of tumor spheroid growth and ECM invasion were captured using the Echo Lab Revolve (Revolve Generation 2) in Brightfield and Texas Red channels. The growth area of tumor spheroids was calculated with ImageJ by outlining the spheroid edges using Polygon Selections. The presence of ECM invasion was determined by observing protrusions around the spheroids. The spheroid invasion was quantified by delineating invasive margins in ECM with Polygon Selections in ImageJ. The optically dense region corresponds to the compact, multilayered tumor core and was used to quantify tumor growth. In contrast, the surrounding optically sparse region represents dispersed single cells invading the ECM and was used to assess invasion. The spheroid invasion rate was calculated by dividing the invaded area by the total tumor spheroid area. Given the inherent subjectivity in judgment, the following approach was implemented to eliminate systematic bias caused by subjective assessment: Five individuals independently assessed the occurrence of ECM invasion phenomena described above and calculated the invasion rates. After excluding the highest and lowest values, the mean of the remaining three values was adopted as the final dataset.

Statistical analysis methods

All experiments were performed with at least three independent samples per group (n ≥ 3), and the exact sample size for each experiment is indicated in the corresponding figure caption. Statistical comparisons between groups were conducted using either an unpaired t-test with Welch’s correction or a two-way ANOVA, as appropriate. Data analysis was carried out using GraphPad Prism 10.1.2 software, and results are presented as mean ± standard deviation (S.D.), unless otherwise noted in the figure captions. Statistical significance was defined as p < 0.05 and is denoted as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

All experiments were performed in triplicates. Two-tailed unpaired Student’s t-test was used when only two groups were compared, and one-way ANOVA was applied with comparison of more than two groups. Differences were considered significant when p < 0.05. Statistical analyses were conducted using Graphpad Prism 10.1.2 software.

Results

Construction of the lung cancer-vascular tumor-on-a-chip model

In this study, we designed a lung cancer–vascular tumor-on-a-chip model that simulated the tumor–ECM interactions. The preparation procedure, and after some modifications for this study can be found in Fig. 2A. In brief, lung cancer spheroids were made in a non-adherent plate, placed in type I collagen solution, and injected into a polydimethylsiloxane (PDMS) device that had a 220 μm nitinol wire in the center. Spheroids were incubated at 37 °C for 12 min to allow the collagen to polymerize, and then the wire was removed creating a central channel. The GFP-labeled human umbilical vein endothelial cells (HUVEC-GFP) were placed in the channel and then perfused with ECM culture medium. This allowed for HUVEC-GFP adhesion to the ECM and the formation of stable intercellular junctions (Fig. 2A). Using this device we generated a complete cylindrical vascular structure having direct tumor–endothelial cell interactions and microvascular perfusion to simulate hemodynamic flow which then allowed for the sustained growth of the tumor spheroids.

The lung cancer spheroids were embedded in collagen, and their growth and ECM invasion were observed. The spheroids exhibited stable growth within the device, increasing in size over time. ECM invasion was observed approximately three days after culture (Fig. 2B), characterized by spike-like protrusions extending from the spheroids into the surrounding ECM. Real-time imaging of the lung cancer–vascular tumor-on-a-chip model was performed to monitor tumor growth and invasion into the ECM under conditions where vasculature and tumor coexisted (Fig. 2C). Immunofluorescence staining of CD31 (an endothelial marker) further confirmed the formation of a functional vascular network. In addition, matrix metalloproteinase-2 (MMP2), a protease involved in ECM remodeling and tumor invasion, was labeled. The results indicated that endothelial cells exhibited tumor-associated phenotypic changes, thereby promoting tumor invasion (Fig. 2D). In addition, permeability assays of the vascular channel confirmed that drugs with a molecular weight below 70 kDa are able to cross the vascular barrier and act on the tumor (Figure S5).

Fig. 2
Fig. 2
Full size image

Schematic representation of lung cancer-vascular tumor-on-a-chip construction and tumor invasion into the ECM. (A) Construction of a vascularized lung cancer-vascular tumor-on-a-chip: tumor spheroids were embedded in type I collagen and seeded around a 220 μm micrometer nickel needle. After gelation, the needle was removed, and endothelial cells were seeded in the resulting channel. (B) Gradual invasion of tumor spheroids into type I collagen. (Created with bioRender.com). (C) Real-time imaging of a lung cancer–vascular organoid-on-a-chip system. Scale bar: 200 μm. (D) Immunofluorescence staining of lung cancer-vascular tumor-on-a-chip for CD31 (red), and MMP2 (yellow). Scale bar: 200 μm

ECM invasion profiles are distinctive between cell lines

A549 and H460 cell lines were inoculated into a microfluidic device that was embedded in type I collagen to visualize direct interactions between tumors and endothelial cells with the ability to monitor tumor behavior in real time. Tumor spheroids were shown to grow stably in the microvascular network, displaying substantial increases in size on day 2 post seeding. A549 and H460 derived tumors exhibited stable growth in the device (Fig. 3A). However, it was clear A549 cells exhibited greater ECM invasion, as A549 cells had a significantly higher number of newly formed invasive protrusions and higher invasion ratio than H460 cells (Fig. 3). Therefore, A549 was used as the cell line of choice for studying ECM invasion. The A549 spheroids exhibited peak frequencies of ECM invasion within the chip at day 0 and day 1 which higher than day 3 and day 5(Fig. 3B). Similarly, the invasion rate increased rapidly from day 0 to day 3, while no significant differences were observed between day 3 and day5(Fig. 3C). These data indicated that tumor spheroid invasion reached plateau within the five-day period. Therefore, this study focused on observing tumor spheroid invasion dynamics within the five-day timeframe. Collectively, these findings show flexibility of this method in developing functional microvascular lung cancer models.

Fig. 3
Fig. 3
Full size image

Comparison of tumor spheroid growth and ECM invasion between A549 and H460 cell lines in the lung cancer-vascular tumor-on-a-chip. (A) Both A549 and H460 tumor spheroids grew within the device, increasing in size over time and invading the ECM. Scale bar: 200 μm. (B) Daily newly occurring ECM invasion of A549 tumor spheroids (n = 19 tumoroids). (C) Daily ECM invasion rate of A549 tumor spheroids (*P < 0.05, ***P < 0.001, ****P < 0.0001). (D) Daily newly occurring ECM invasion of H460 and A549 tumor spheroids (n = 7 tumoroids). (E) Daily ECM invasion rate of H460 tumor spheroids (*P < 0.05, **P < 0.01). The exact P values were measured by the t-test

Cytoskeletal regulation of ECM invasion

The cytoskeleton is a key regulator in ECM invasion by tumor cells. To investigate its role, we treated the tumor-on-a-chip model with the Rho-associated coiled-coil containing protein kinase (ROCK) pathway inhibitor Y-27,632 and the actin inhibitor Cytochalasin D (CytD). ECM invasion was markedly inhibited following treatment with these inhibitors (Fig. 4A), manifested by a significant reduction in invasion growth rates (Fig. 4B) and incident rates (Fig. 4C). These findings suggest that while cytoskeletal inhibitors strongly suppress ECM invasion, confirming the cytoskeleton as a critical contributing factor in ECM invasion.

Fig. 4
Fig. 4
Full size image

Cytoskeletal regulation of ECM invasion. (A) Schematic images illustrating ECM invasion following treatment with Y-27,632 and CytD in the lung cancer-vascular tumor-on-a-chip model. (Red: A549 tumor spheroids; white dashed line: spheroid boundary; green dashed line: ECM invasion boundary.) Scale bar: 100 μm. (B) Relative ECM invasion rate on day 5 (N = 3 devices). **P < 0.01; **P < 0.0001. (C) Incidence rate of invasion changes from day 0 to day 5. The exact P values were measured by the t-test

Drug screening versatility validation with marked anti-cancer drugs

To evaluate the effectiveness of our drug screening model, five chemotherapy drugs that have been used clinically were selected: paclitaxel (PTX), cisplatin (CDDP), irinotecan (CPT-11), oxaliplatin (OXA), and gemcitabine (GEM). The growth of tumor spheroids was then tracked, at a concentration of 100 nM, from day 0 to day 5 (Fig. 5A). GEM completely inhibited the growth of tumor spheroids, PTX and CPT-11 showed moderate inhibition of growth, CDDP and OXA showed a decrease in tumor growth until day 2 before losing effect (where the tumor growth began to exceed the control; Fig. 5B). The invasion in ECM was tracked daily (Fig. 5C). GEM inhibited ECM invasion effectively followed by OXA and PTX, and CDDP and CPT-11 had little inhibitory effect (Fig. 5D). To distinguish specific anti-invasive effects from general cytotoxicity, we analyzed the invasion rate. Notably, while OXA showed transient growth inhibition (losing effect after Day 2), it maintained a significant reduction in invasion rate. This suggests that the observed inhibition of ECM invasion is not solely secondary to growth suppression but may involve specific anti-motility mechanisms. These results serve as a proof-of-concept for the model’s sensitivity to differential drug effects.

Fig. 5
Fig. 5
Full size image

Evaluation of five positive drugs using lung cancer-vascular tumor-on-a-chip. (A) Representative images of A549 tumor spheroid growth under 100 nM treatment with five clinically used chemotherapeutic drugs in tumor-on-a-chip. Scale bar: 200 μm. (B) Relative growth rate of tumor spheroid in different treatment groups during 5-day observation. (N = 3 devices). (C) Representative ECM invasion patterns across treatment groups; blue dashed lines: tumor spheroid boundaries, green dashed lines: ECM invasion frontiers. Scale bar: 200 μm. (D) ECM invasion rates of treatment groups compared to blank control (N = 3 devices). *P < 0.05, **P < 0.01, ***P < 0.001. The exact P values were measured by the t-test

Rocaglamide inhibits tumor growth and ecm invasion

Using this model, the anti-cancer activity of the natural compound ROC-A was systematically evaluated. ROC-A was administered at concentrations of 20, 50, and 100 nM, which are effective at inhibiting tumor cells (Figure S6) in 2D cultures while exhibiting minimal toxicity to HUVECs (Figure S7), and it significantly suppressed tumor spheroid growth at all tested doses (Fig. 6A–B). Furthermore, ROC-A markedly suppressed the invasion of tumor spheroids into the extracellular matrix (ECM) at all concentrations examined (20, 50, and 100 nM), indicating that even low-dose ROC-A is sufficient to inhibit ECM invasion (Fig. 6C). Notably, the inhibitory effect observed at 50 nM was more pronounced than that at 20 nM and slightly greater than that at 100 nM; however, the difference between 50 nM and 100 nM was not substantial (Fig. 6D). Collectively, these findings suggest that ROC-A exerts a dose-dependent inhibitory effect on ECM invasion within the tested range, with a tendency to plateau at higher concentrations.

Fig. 6
Fig. 6
Full size image

ROC-A Inhibits Tumor Spheroid Growth and ECM Invasion. (A) Representative images of tumor spheroid growth under treatment with 20, 50, and 100 nM ROC-A. (Red: A549 tumor spheroids; white dashed line: boundary of the tumor spheroid; green dashed line: boundary of ECM invasion). Scale bar: 100 μm. (B) Relative growth rate of tumor spheroids in different ROC-A concentration groups over a 5-day observation period. ****P < 0.0001. (C–D) Quantitative analysis of ECM invasion at all tested concentrations of ROC-A, including the relative growth rate of ECM invasion at the endpoint (day 5) (C) and the daily progression of newly formed ECM-invading spheroids (D). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The exact P values were measured by the t-test

Discussion

A microfluidic lung cancer tumor model was created in this work where we could examine tumor behavior with ECM and cell interactions. Combining two technologies: vascular-organ-on-a-chip and lung cancer spheroids, we created lung cancer-vascular tumor-on-a-chip [41]. This work provides a more physiologically relevant microenvironment compared to conventional in vitro models [42]. The addition of fluid dynamics and shear forces to lung cancer spheroids [43], along with functionalized vasculature [44], led to the elimination of metabolite buildup caused by perfusion lack [45] and enabled the graphical representation of lung tumor invasion and expansion into the bloodstream [46]. This device forms a cylindrical vascular structure which affords tumor spheroids and endothelial cells direct interaction, simulates PRF under physiological conditions, perfusion to the tumor, and potentially most important to unlimited growth of the tumor spheroids.

ECM invasion is a key step in tumor metastasis and involves coordinated processes such as cell migration and matrix remodeling. Using the lung cancer-vascular tumor-on-a-chip, we were able to visualize the entire tumor spheroid invasion process into the ECM at a single-cell resolution. Under an inverted microscope, the tumor spheroids exhibited outward protrusions and progressive expansion of the invasion area over time, which are consistent with invasive phenotypes. By distinguishing between optically sparse and optically dense regions in the images, the overall proliferation of the tumor and the invasion of individual tumor cells into the extracellular matrix can be analyzed [47, 48]. Upon drug treatment of the tumor spheroids, we observed suppression of the spiky protrusions and punctate dissemination which indicates the tumor spheroids ability to invade the ECM was inhibited as an assessment of the drug’s anti-invasion activity.

This study investigated whether A549 and H460 cell lines were equally likely and equally invasive within the device to the ECM component. In this experiment, A549 was seen to exhibit more ECM invasion than H460, therefore using the ROCK pathway inhibitor Y-27,632 [49] and the actin inhibitor CytD [50] only on A549 cells. Although neither cytoskeletal inhibitors had any effect on tumor growth or proliferation, the both inhibitors showed the ability to attenuate ECM invasion, with both inhibitors to suggest anti invasion due to inhibition of ECM penetration. These findings suggest that the cytoskeleton plays an important role in tumor invasion within this model and support the utility of the device for evaluating invasion-related behavior.

Upon an evaluation of the model, we then selected five clinically used anti-lung cancer invasion and migration drugs: PTX, CDDP, CTP-11, OXA and GEM. Of these, albumin-bound paclitaxel is approved for NSCLC in hopes of more favorable drug formulation to allow less hypersensitivity reactions and better targeting of the tumor [21]. CDDP-based combination chemotherapies are first-line treatments for NSCLC stage IV disease and have shown to be effective [22]. CTP-11 is indicated as a monotherapy when recurrent small-cell lung cancer develops resistance to platinum-based drug therapy [23, 24]. OXA is approved for patients who are intolerant to cisplatin [25], and GEM has been demonstrated to be well-tolerated as a monotherapy for older or frail patients with NSCLC [26]. GEM directly interferes with DNA synthesis and regulates invasion-related pathways [51,52,53]. When administered to the lung cancer-vascular tumor-on-a-chip, GEM showed the strongest inhibitory effect on tumor growth, while PTX and CTP-11 demonstrated moderate effects. CDDP and OXA exhibited comparatively limited effects on tumor growth under the tested conditions. With regards to ECM invasion, GEM made the most significant inhibition effect, followed by OXA and PTX, and CDDP and CTP-11 demonstrated only limited enhanced suppression. Overall, we speculate that GEM and PTX inhibit tumor invasion through their inhibition of tumor growth and subsequent ECM invasion. Lastly, PTX is primarily inhibitory to ECM invasion, while Irinotecan is primarily influential on ECM invasion, and CDDP is non-influential. These data support our model for use in screening a small molecule drug that is specifically used for tumor invasion. We think the model could be of interest to others and future considerations would be for use of small-molecule drugs for treatment of tumor invasion and metastasis to the lung tissue. The accuracy and stability of the model were validated by evaluating the effects of five classical clinical chemotherapeutic agents on the lung cancer–vascular tumor-on-a-chip model.

The literature shows that many natural compounds could have specific properties that are anti-tumor in that they can prevent tumor growth, induce apoptosis, modulate the tumor microenvironment, inhibit tumor angiogenesis, or even develop drug resistant properties [54, 55]. ROC-A is a natural bioactive compound derived from the Aglaia species that has shown multi-targeted anti-tumor activity against lung cancer. ROC-A seems to have a specific target for the ATM/ATR-Chk1/2 DNA damage checkpoint pathway leading to a G1/S-phase arrest, providing a way to inhibit proliferation and promote apoptosis [56]. ROC-A can also inhibit NF-κB and STAT3, by downregulating the resulting pro-inflammatory factors from the inflammation within the tumor microenvironment, leading to tumor invasion [57]. ROC-A has exhibited the ability to increase chemotherapy with sensitive Cisplatin in vivo studies decreasing the occurrence of drug resistance. While the study presented the anti-invasion effect of ROC-A, we were able to show that the ROC-A inhibitor limited tumors invasion into the ECM. While various concentrations of ROC-A do show inhibition, ROC-A appeared to show the strongest incidence rate of invasion at 50nM, while increasing ROC-A to 100 nM it slightly decreased inhibition of invasion. We provide evidence that ROC-A has a significant anti-invasion/migration effect by limiting tumor infiltration into the ECM.

Animal models are hindered by limitations such as poor visualization, prolonged modeling times, and substantial interspecies differences from humans. As a result, many drug candidates that show promising efficacy in preclinical animal studies ultimately fail to demonstrate therapeutic benefit in clinical trials [19, 58]. In contrast, traditional in vitro models typically rely on single cell types, whereas organ-on-a-chip systems integrate multiple cell types to better recapitulate complex cellular interactions within native tissues. Overall, tumor-on-a-chip models provide a potential high-throughput and cost-effective platform for the precise evaluation of drug efficacy and toxicity. Moreover, they reduce the time and financial burden associated with animal experiments, improve the predictive accuracy of preclinical studies, and enhance the likelihood of clinical success. Despite challenges such as operational complexity, their potential in precision medicine, drug development, and disease mechanism research far exceeds that of conventional models, positioning them as a key tool in translational medicine.

Compared with previous studies, the lung cancer–vascular tumor-on-a-chip model developed in this work further advances physiological relevance. By integrating tumor spheroids, extracellular matrix (ECM), endothelialized vascular channels, and continuous perfusion, our system enables real-time observation of tumor–ECM interactions and invasion processes under more biomimetic conditions. This design allows not only the evaluation of cell migration but also the investigation of tumor growth, cytoskeletal remodeling, and drug responses within a multicellular context. Therefore, our platform complements existing migration assays by offering a more physiologically relevant, albeit moderately scalable, system that is particularly well suited for mechanistic studies and the precise evaluation of anti-invasion therapeutics.

Limitations of this study include the need for further validation methods. The current analysis relies on 2D imaging. Although this approach allows long-term monitoring under low phototoxicity conditions, it does not fully capture the spatial morphology of tumor spheroids. Therefore, the implementation of 3D imaging techniques will be considered in future work to provide a more comprehensive evaluation. In addition, in vivo experiments, to predict the accuracy of pharmacological efficacy evaluations, as well as the incorporation of patient-derived samples to enhance the biomimetic performance of the chip [59, 60]. In the future, patient-derived cells are expected to be increasingly integrated into the tumor-on-a-chip model platforms, enabling automated high-throughput drug screening and advancing precision medicine applications.

Conclusions

In the current study, we built a lung cancer-vascular tumor-on-a-chip, and used basic comparative analysis for tumor spheroid growth and ECM invasion with A549 and H460 cell lines but have also used cytoskeletal inhibitors to develop a functional model that effectively simulated tumors invading the ECM, which is a key measure of tumor invasiveness. We also used the model to screen five known anti-invasion drugs (Paclitaxel, Cisplatin, Irinotecan, Oxaliplatin, and Gemcitabine) which confirmed the models functionality, reliability, and stability in drug screening. Rocaglamide was identified in this study as a potent inhibitor of tumor cell invasion within this microphysiological system. Our findings further extend its functional characterization by demonstrating its efficacy in a physiologically relevant tumor-on-a-chip platform. This highlights the utility of our model for evaluating natural compounds targeting tumor invasion. We will continue using this model to assess additional natural products with potential anti-invasion properties, using invasion into the ECM as a key assessment platform.