Device for observing biological activity in sample and method for quantifying tumor infiltration events in biological sample

Abstract

Devices for observing biological activity in a sample, and methods for quantifying tumor infiltration events in a biological sample. The apparatus comprises: a microfluidic device having a plurality of fluid channels, including a central channel adjacent to a medium channel and defining a pourable partition therebetween; wherein the central channel is used for promoting the biological cells contained in the central channel to undergo interstitial transformation (EMT) induced by the reagent supplied to the culture medium channel.

Description

Technical Field

[0001] This disclosure relates to an apparatus for observing biological activity in a sample, and a method for quantifying tumor intravasation events in a biological sample. Specifically, but not exclusively, this disclosure relates to a system and method for observing and quantifying the occurrence and characteristics of tumor intravasation events. Background Technology

[0002] Lung cancer is the leading cause of cancer-related deaths worldwide. It is often diagnosed only in its advanced metastatic stage, where treatment options are limited, resulting in a low 5-year relative survival rate of just 9%. Therefore, developing physiologically relevant human metastatic lung cancer models is of paramount importance to better study its underlying pathological processes, identify suitable drug targets, and facilitate high-throughput drug screening. Microphysiological models of cancer metastasis have the potential to bridge the gap between 2D cell culture models, animal models, and real-world clinical settings, thereby addressing the high failure rates in drug development and facilitating the transition to personalized medicine.

[0003] Lung tumors can be classified into non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC). The former accounts for approximately 80-85% of all lung cancer cases, while the latter accounts for the remaining 15%. For metastasis of epithelial lung cancer, cancer cells originating from the primary solid tumor typically undergo five key stages: local invasion, infiltration, dissemination, extravasation, and ultimately, colonization as a secondary tumor, driven by a landmark process called epithelial-mesenchymal transition (EMT). Of these five stages, EMT and infiltration are rate-limiting steps determining the number of circulating tumor cells with metastatic potential. Therefore, addressing EMT, local invasion, and infiltration can, in principle, minimize and / or prevent systemic spread of cancer cells at the initial stage, thereby improving clinical outcomes for lung cancer patients. Summary of the Invention

[0004] According to a first aspect of this disclosure, an apparatus for observing biological activity in a sample is provided. The apparatus includes a microfluidic device having a plurality of fluid channels, including a central channel adjacent to a culture medium channel and defining a perfusionable septum therebetween; wherein the central channel is configured to promote EMT (epithelial-to-mesenchymal transition) induced by a reagent supplied to the culture medium channel in biological cells contained within the central channel.

[0005] According to a first aspect of this disclosure, the central channel has the biological function of blood vessels.

[0006] According to a first aspect of this disclosure, the central channel includes a hollow lumen vascular structure.

[0007] According to a first aspect of this disclosure, the hollow lumen vascular structure exhibits a permeability coefficient of less than 1.0 × 10⁻⁶ for 40 kDa dextran. -6 cm / s.

[0008] According to a first aspect of this disclosure, the hollow lumen vascular structure is formed through angiogenesis.

[0009] According to a first aspect of this disclosure, the hollow lumen vascular structure is formed by culturing human umbilical vein endothelial cells (HUVECs) to obtain a microvascular network of human umbilical vein endothelial cells.

[0010] According to a first aspect of this disclosure, the human umbilical vein endothelial cells and biological cells contained within the central channel are co-cultured.

[0011] According to a first aspect of this disclosure, the human umbilical vein endothelial cells and the biological cells are co-cultured in a hydrogel complex injected into the central channel.

[0012] According to a first aspect of this disclosure, the reagent is an EMT-induced mixed complex.

[0013] According to a first aspect of this disclosure, the EMT-induced mixed complex comprises transforming growth factor β1 (TGFβ1) and macrophage conditioned medium.

[0014] According to a first aspect of this disclosure, each of the plurality of fluid channels has an inlet and an outlet.

[0015] According to a first aspect of this disclosure, the microfluidic device includes a plurality of partition structures for separating the central channel from the culture medium channel.

[0016] According to a first aspect of this disclosure, the partition structure includes triangular prisms arranged at regular intervals.

[0017] According to a first aspect of this disclosure, the microfluidic device includes a pair of culture medium channels that clamp the central channel.

[0018] According to a first aspect of this disclosure, the pair of culture medium channels are also connected to a common entrance.

[0019] According to a first aspect of this disclosure, the biological cells include tumor spheroids.

[0020] According to a first aspect of this disclosure, the biological cells include cancer cells selected from the group consisting of A549 cells, NCI-H1975 cells, and BEAS-2B cells.

[0021] According to a first aspect of this disclosure, the apparatus further includes: an imaging device for capturing images of biological cells in the central channel to facilitate visualization and recording of endosmosis events of biological cells in the central channel.

[0022] According to a first aspect of this disclosure, the imaging device includes a microscopic imager.

[0023] According to a first aspect of this disclosure, the apparatus further includes a processor for analyzing the captured images to identify tumor infiltration events in the biological cells based on features extracted from the images.

[0024] According to a first aspect of this disclosure, the processor is a machine learning-based processing engine.

[0025] According to a first aspect of this disclosure, the machine learning processing engine utilizes a random forest classifier, such as a fast random forest classifier, for image segmentation.

[0026] According to a second aspect of this disclosure, a method for quantifying tumor intravasation events in biological samples is provided. The method includes: injecting and co-culturing human umbilical vein endothelial cells and tumor spheroids in a central channel of a microfluidic device according to a first aspect; supplying an EMT-induced mixed complex according to a first aspect in a culture medium channel of the microfluidic device; capturing microscopic images of the central channel at predetermined time intervals; and processing the microscopic images using a machine learning processing engine according to a first aspect to quantify the occurrence and characteristics of tumor intravasation events.

[0027] According to a second aspect of this disclosure, supplying the EMT-induced mixed complex reduces the expression of laminin and VE-cadherin in the microvascular network, thereby increasing the tolerance for cancer cell infiltration.

[0028] According to a second aspect of this disclosure, the method is capable of distinguishing between biological cells with high transfer potential and biological cells with low transfer potential.

[0029] This disclosure provides a method for quantifying intravasation events in lung cancer. The method involves culturing human umbilical vein endothelial cells and lung cancer spheroids in a central channel. The human umbilical vein endothelial cells then form a microvascular network (MVN)—a 3D hollow lumen vascular network—near the tumor spheroids. After the microvascular network is formed, a macrophage conditioned medium mixture is added to the culture channel, which induces the biological cells to undergo endogenous transluminal metastasis (EMT) in the microfluidic device model. A machine learning model is provided, trained on a dataset of lung cancer MVNs in a microfluidic device. The machine learning model is used to analyze EMT-induced intravasation events in cancer cells within the microfluidic device. The method further includes quantifying the number and characteristics of the lung cancer intravasation events based on the analysis results obtained through the machine learning model.

[0030] Furthermore, this disclosure provides a 3D hollow lumen vascular system within a microfluidic device for quantifying lung cancer intravasation events. The device includes a microfluidic chip with a culture chamber (i.e., the central channel), wherein the culture chamber includes the hollow lumen vascular structure simulating a blood vessel and the lung cancer spheroids. The device also has a fluid network connected to the culture chamber, enabling the introduction of a macrophage conditioned medium mixture to induce EMT in the spheroids. The device includes inlets and outlets for controlled flow of culture medium and fluid within the microfluidic device. The device integrates imaging capabilities to visualize and record events of lung cancer spheroids infiltrating into the hollow lumen vascular structure. Furthermore, the device integrates a machine learning model for quantifying and analyzing the lung cancer intravasation events based on captured images and features extracted from the images.

[0031] Furthermore, this disclosure provides a machine learning model specifically designed for quantifying events of lung cancer infiltration into a three-dimensional (3D) hollow lumen vascular metastatic microfluidic device. The model utilizes training data comprising a dataset of lung cancer infiltration events obtained from the microfluidic device. It integrates feature extraction algorithms to extract relevant features from images or data representing the lung cancer infiltration events. The model employs a series of image processing steps, including but not limited to image decomposition, thresholding, subtraction, synthesis, and pixel-based classification using a random forest classifier, to ultimately segment individual extravasated tumor cells and vascular structures, subsequently quantifying the occurrence and characteristics of the lung cancer infiltration events. Validation and evaluation metrics are used to assess the performance and accuracy of the machine learning model. The model is integrated with a 3D hollow lumen vascular metastatic microfluidic device to quantify and analyze lung cancer infiltration events in real time based on the model's predictions and outputs.

[0032] This disclosure provides the advantages of combining and integrating macrophage-induced epithelial-mesenchymal transition (EMT), a three-dimensional (3D) hollow lumen vascular metastasis microfluidic device, and a machine learning model for quantifying and analyzing intravasation events in lung cancer. By utilizing these technologies, this disclosure provides a comprehensive solution for studying and understanding the process of lung cancer metastasis, particularly the intravasation events that occur during this process. This information is crucial for identifying predictable biomarkers, discovering novel and highly specific drug targets, and facilitating drug screening processes, thereby enabling earlier and more effective treatment of cancer patients and thus increasing the likelihood of positive therapy outcomes. Attached Figure Description

[0033] Embodiments of this disclosure will be described below by way of example with reference to the accompanying drawings.

[0034] Figure 1 This is a schematic diagram illustrating an embodiment of the present disclosure for observing biological activity in a sample.

[0035] Figure 2 This is a schematic diagram illustrating a microfluidic device for observing bioactivity in a sample according to an embodiment of the present disclosure.

[0036] Figure 3 To show in Figure 2 The flowchart shows the process of angiogenesis and co-culture of biological cells in the central channel of a microfluidic device.

[0037] Figure 4A For EGM-2 (unconditioned medium) and Images of protein arrays, where signals from various secretory components are visualized via chemiluminescence. The levels of selected cytokines were quantified.

[0038] Figure 4B The quantitative results are shown, plotted as a Log2 fold change compared to the control, and categorized according to their role as EMT inducers or inhibitors and as pro-inflammatory and / or anti-inflammatory cytokines.

[0039] Figure 5A Representative microscopic images depicting cell morphology were generated. Cell roundness was analyzed using ImageJ software, and the images were plotted after normalization to the unprocessed control group.

[0040] Figure 5B Images of live / dead cells stained with treated A549 cells, and quantification of cell viability after 2 days of incubation with various EMT inducers.

[0041] Figure 5C This is a representative micrograph of A549 cells immunostained with the epithelial marker E-cadherin. The fluorescence intensity of E-cadherin and vimentin signals was determined by measuring corrected total cell fluorescence (CTCF) and then normalized to their respective control groups.

[0042] Figure 5D These are representative micrographs of A549 cells stained with the mesenchymal marker vimentin. The fluorescence intensity of the vimentin signal was determined by measuring the corrected CTCF and then normalized to their respective control groups.

[0043] Figure 5E The relative gene expression of SNAI1, SNAI2, TWIST1, ZEB1, and ZEB2, as determined by real-time quantitative PCR, is shown. The relative mRNA level (ΔΔCT value) of each gene is normalized to its respective GAPDH (housekeeping gene) level.

[0044] Figure 5F Representative micrographs of two-dimensional (2D) scratch migration measurements. The yellow dashed line represents the initial scratch area. The percentage of the initial scratch area without cells was quantified using ImageJ software after 12 hours of treatment.

[0045] Figure 5G The 3D invasion assay of A549 spheroids embedded in type I collagen hydrogel is shown. The degree of invasion of A549 cells into their surrounding environment was defined as the total invasion area (represented by the yellow dashed line) divided by the initial spheroid area (represented by the black circle) and normalized to the control group.

[0046] Figure 5H The quantification of cell nuclei using DAPI staining is shown. A549 cells were compared with TGF-β1, Alternatively, a combination of both can be incubated for two days. Part A is a representative image showing DAPI-stained cell nuclei (blue). Cell nuclei were counted using particle analysis on ImageJ. Part B is a bar chart representing the average number of cell nuclei per field of view, normalized for the control group.

[0047] Figure 5I Representative micrographs depicting the formation process of A549 spheres during a 3-day incubation period show the formation of A549 spheres in micropores over time.

[0048] Figure 6AA photograph showing the dimensions of a three-channel microfluidic device according to an embodiment of the present disclosure.

[0049] Figure 6B The representative micrographs show the formation of MVNs by HUVECs in a microfluidic device after 24 hours of incubation with or without the introduction of EMT-IC on day 3.

[0050] Figure 6C The quantification of total microtubule length and number of vascular junctions for each field of view (FOV) is shown using the ImageJ Angiogenesis Analyzer plugin.

[0051] Figure 6D This is a confocal Z-stack image of laminin and vinculin co-stained proteins, viewed in orthogonal view. The location of the cross-sections is indicated by white lines, with orthogonal cross-sections shown on the side.

[0052] Figure 6E This is a confocal microscopy image of vascular endothelial cadherin (VE-cadherin) and F-actin (fibrous actin) co-stained, as shown in a Z-stack image.

[0053] Figure 6F Western blot and densitometric band analysis of laminin in samples collected on day 4 are shown, with their respective GAPDH levels normalized. Protein levels are shown as fold changes and compared with the control group.

[0054] Figure 6G Western blot and density band analysis of vascular endothelial cadherin (VE-cadherin) in samples collected on day 4 are shown, with normalization of their respective GAPDH levels. Protein levels are shown as fold changes and compared with the control group.

[0055] Figure 6H This is a schematic diagram of a cancer-on-a-chip model according to an embodiment of the present disclosure.

[0056] Figure 6I The quantification of (A) branch width, (B) microtubule length, and (C) number of connection points for each FOV using the ImageJ angiogenesis analyzer plugin is shown, along with the EMT-IC (TGF-β1& It did not adversely affect the stability and integrity of MVN.

[0057] Figure 7ARepresentative micrographs show the formation of MVNs in a microfluidic device after 24 hours of incubation with or without EMT-IC on day 3, in the presence of A549 spheroids. Initial tumor masses are outlined by yellow dashed lines.

[0058] Figure 7B The quantification of total microtubule length and number of vascular junctions per field of view (FOV) is shown using the ImageJ Angiogenesis Analyzer plugin.

[0059] Figure 7C This is a confocal microscopy image of immunostained laminin, optionally overlaid with a phase contrast (PhC) image. The boundaries of the initial tumor mass are indicated by a white dashed line.

[0060] Figure 7D Western blot and densitometric band analysis of laminin in samples collected on day 4 are shown, with their respective GAPDH levels normalized. Protein levels are shown as fold changes compared to the control group.

[0061] Figure 7E Confocal microscopy images co-stained with vascular endothelial cadherin (VE-cadherin) and F-actin (fibrous actin). The boundaries of the initial tumor mass are indicated by white dashed lines. Close-up areas are indicated by solid rectangles, with close-ups shown on the next line. Close interactions with well-defined boundaries between cancer cells and MVNs are indicated by white arrows. Close interactions with less distinct boundaries between cancer cells and MVNs are indicated by white asterisks.

[0062] Figure 7F Western blot and density band analysis of vascular endothelial cadherin (VE-cadherin) in samples collected on day 4 are shown, with their respective GAPDH levels normalized. Protein levels are shown as fold changes compared to the control group.

[0063] Figure 7G Live cell images taken hourly over 24 hours starting from day 3 after supplementation of EMT-IC or control medium in co-cultures of A549 spheroids and MVNs. Representative magnified close-ups of pre-labeled A549 cells (red) and MVNs (green) at selected time points, superimposed with their respective phase-difference images, are shown. Migrating and invasive cancer cells are indicated by dashed circles.

[0064] Figure 7HLive-cell fluorescence imaging is shown for A549 cancer cell spheroids co-cultured with HUVEC-derived MVNs for over 24 hours. Time-lapse live-cell imaging was initiated hourly for 24 hours after control medium was added to the co-culture of A549 spheroids and MVNs on day 3. Representative 10x magnification images show pre-labeled A549 cancer cells (red) and MVNs (green), along with corresponding phase-contrast images. White arrows within the boxed areas highlight migrating and invasive cancer cells.

[0065] Figure 7I Live-cell fluorescence imaging of A549 cancer cells migrating and infiltrating into MVNs promoted by EMT-IC over 24 hours is shown. Time-lapse live-cell imaging was initiated hourly for 24 hours after EMT-IC was added to the co-culture of A549 spheroids and MVNs on day 3. Representative 10x magnified images show pre-labeled A549 cancer cells (red) and MVNs (green), along with corresponding phase-contrast images. White arrows within the boxed areas highlight migrating and invasive cancer cells.

[0066] Figure 7J Live-cell fluorescence imaging is shown for A549 cancer cell spheroids co-cultured with HUVEC-derived MVNs for over 24 hours. Live-cell imaging was initiated hourly for 24 hours after control medium was added to the co-culture of A549 spheroids and MVNs on day 3. Representative magnified close-ups of pre-labeled A549 cells (red) and MVNs (green) superimposed with their respective phase-difference images taken hourly are shown. Migrating and invasive cancer cells are indicated by dashed circles.

[0067] Figure 7K Live-cell fluorescence imaging is shown, illustrating A549 cancer cell spheroids co-cultured with HUVEC-derived MVNs for over 24 hours. Live-cell imaging was initiated hourly for 24 hours after EMT-IC was added to the co-culture of A549 spheroids and MVNs on day 3. Representative magnified close-ups of pre-labeled A549 cells (red) and MVNs (green) superimposed with their respective phase-difference images taken hourly are shown. Migrating and invasive cancer cells are indicated by dashed circles.

[0068] Figure 7L Time-lapse live-cell imaging of A549 cancer cells migrating and infiltrating over 24 hours is shown. Representative videos capture dynamic cell behavior over 24 hours under different conditions. A and B are control conditions, and C and D are EMT-IC conditions, showing the migration and infiltration of A549 cells into the MVN. White arrows highlight migrating and invading cancer cells.

[0069] Figure 8AThis is a 20x magnification confocal microscopy image of CD31 co-stained under control co-culture and EMT-IC exposure conditions. The white dashed line represents the initial A549 tumor mass. White rectangles indicate close-up areas, shown in the next row. Migrating but non-infiltrating A549 cancer cells are indicated by white arrows, while infiltrated cancer cells are indicated by asterisks. The MVN was reconstructed and visualized using Qiber3D to observe infiltration events.

[0070] Figure 8B In order to be with Figure 8A Confocal microscopy images at 20x magnification of F-actin co-stained under the same conditions, in control co-culture and under EMT-IC exposure conditions.

[0071] Figure 8C To visualize the invasion area of ​​A549 spheroids after 24 hours of incubation in a microfluidic device, the degree of invasion of A549 cells into their surrounding environment was defined as the total invasion area divided by the initial spheroid area, normalized to the control group.

[0072] Figure 8D A graph showing the manual semi-quantitative analysis of each spherical endothelial event.

[0073] Figure 8E This is a z-stack image that includes high-resolution confocal microscopy images (63x) under control and EMT-IC conditions, displayed in orthogonal view and 3D projection. The location of the cross-sections is indicated by white lines, with orthogonal cross-sections shown on the side.

[0074] Figure 9A This is a 20x confocal microscope image of the control conditions in an experiment where EMT-IC (TGF-β1) and This study promoted the infiltration of NCI-H1975 cancer cells into MVNs to varying degrees. NCI-H1975 spheroids and MVNs were co-cultured in a microfluidic device, with optional supplemental EMT-IC incubation for 24 hours on day 3. On day 4, cells were stained for CD31 and F-actin (fibrous actin).

[0075] Figure 9B In order to be in Figure 9A 20x confocal microscopy images of the same EMT-IC treatment conditions from the same experiment. White dashed lines mark the original tumor boundaries, and white boxes highlight the areas shown in the close-up panel below. Migrated but non-infiltrating cancer cells are marked with an asterisk, while arrows indicate infiltrated cells.

[0076] Figure 9C It shows in Figure 9AIn the same experiment, the invasive area of ​​tumor spheroids after 24 hours of incubation in a microfluidic device was measured. The degree of invasion of cancer cells into their surrounding environment was defined as the total invasive area divided by the initial spheroid area, normalized to the control group.

[0077] Figure 9D It shows in Figure 9A Manual semi-quantitative analysis of each spherical endosmosis event in the same experiment.

[0078] Figure 9E High-resolution confocal microscopy images (63x) of the co-culture of NCI-H1975 and MVN processed with EMT-IC, including z-stack images displayed in orthogonal view. The location of the cross-sections is indicated by white lines, with orthogonal cross-sections shown on the side.

[0079] Figure 9F This is a 20x confocal microscope image of the control conditions in an experiment where EMT-IC (TGF-β1) and This promoted the infiltration of BEAS-2B cancer cells into MVNs to varying degrees. BEAS-2B spheroids and MVNs were co-cultured in a microfluidic device, and optionally supplemented with EMT-IC incubation for 24 hours on day 3. On day 4, cells were stained for CD31 and F-actin (fibrous actin).

[0080] Figure 9G In order to be in Figure 9F 20x confocal microscopy images of the same EMT-IC treatment conditions from the same experiment. White dashed lines mark the original tumor boundaries, and white boxes highlight the areas shown in the close-up panel below. Migrated but non-infiltrating cancer cells are marked with an asterisk, while arrows indicate infiltrated cells.

[0081] Figure 9H It shows in Figure 9F In the same experiment, the invasive area of ​​tumor spheroids after 24 hours of incubation in a microfluidic device was measured. The degree of invasion of cancer cells into their surrounding environment was defined as the total invasive area divided by the initial spheroid area, normalized to the control group.

[0082] Figure 9I It shows in Figure 9F Manual semi-quantitative analysis of each spherical endosmosis event in the same experiment.

[0083] Figure 9J High-resolution confocal microscopy images (63x) of EMT-IC-treated BEAS-2B co-cultures with MVN, including z-stack images displayed in orthogonal view. The location of cross sections is indicated by white lines, with orthogonal cross sections shown on the side.

[0084] Figure 10A A schematic overview of ML-assisted vessel segmentation is shown (blue lines during training: background; magenta lines during training: vessels; blue entities: automatically segmented and identified vessels).

[0085] Figure 10B A schematic overview of ML-assisted quantification of tumor-vascular contact is shown (green solids: ML-segmented vessels; magenta: fluorescently labeled A549 cells).

[0086] Figure 10C The automatic identification and quantification of each spherical endosmotic event assisted by ML are shown.

[0087] Figure 10D The segmentation settings for training images using trainable Weka segmentation are shown. An overview of the parameters used for training the images with trainable Weka segmentation is provided, including the selected training features, membrane thickness, and patch size. Segmentation is performed using a fast random forest classifier. These settings are optimized to improve the accuracy and performance in distinguishing relevant structures.

[0088] Figure 10E The comparison of manual and ML-assisted quantification of endosmosis events under A-control and B-EMT-IC conditions is shown. Heatmap C on the right shows the endosmosis events per spheroid for both methods, n=15 spheroids / condition.

[0089] Figure 11A This is a confocal microscopy image of immunostained laminin, optionally overlaid with a PhC image. The boundaries of the initial tumor mass are indicated by a white dashed line.

[0090] Figure 11B Western blot and densitometric band analysis of laminin in samples collected on day 4 are shown, with their respective GAPDH levels normalized. Protein levels are shown as fold changes compared to the control group.

[0091] Figure 11C Confocal microscopy images co-stained with vascular endothelial cadherin (VE-cadherin) and F-actin (fibrous actin). The boundaries of the initial tumor mass are indicated by white dashed lines. Close-up areas are indicated by solid rectangles, and close-ups are shown on the next line. Close interactions with well-defined boundaries between cancer cells and MVNs are indicated by white asterisks. Close interactions with less distinct boundaries between cancer cells and MVNs are indicated by white arrows.

[0092] Figure 11DWestern blot and density band analysis of vascular endothelial cadherin (VE-cadherin) in samples collected on day 4 are shown, with their respective GAPDH levels normalized. Protein levels are shown as fold changes compared to the control group.

[0093] Figure 11E The following is a representative frame from live-cell imaging, showing the perfusion of MVN from one culture medium channel to another on day 4.

[0094] Figure 11F The MVN osmotic coefficients of FITC-PVP (40 kDa) perfused on day 4 under control and EMT-IC conditions are shown. Control group: n=16, EMT-IC group: n=17, collected from 3 biological replicates. Detailed Implementation

[0095] Through their trials and experiments, the inventors recognized that in vivo animal models of cancer infiltration have low conversion rates from preclinical to clinical treatment, raising growing concerns about the use of animals as tools for predicting human responses. These issues highlight the need to develop physiologically relevant in vitro human models for cancer biology research and therapy development.

[0096] On the other hand, in vitro models such as the Transwell migration / invasion assay can use porous membranes seeded with 2D endothelial cell layers to separate two compartments for observing transendothelial migration. However, endothelial monolayers cannot reproduce many functions, including intact endothelial barrier properties, thus oversimplifying the 3D tumor microvascular system.

[0097] The inventors recognized that for many years, cancer research has been conducted in in vitro 2D cell cultures and in vivo animal models. While 2D monolayer cultures have been widely adopted due to their high availability and reproducibility, these models cannot mimic the etiology of disease or facilitate comprehensive cellular and environmental manipulation. Animal models also fail to fully capture the complex biological processes occurring in the human body and inherently possess low resolution during real-time monitoring, thus limiting the acquisition of their detection results.

[0098] 3D in vitro microphysiological models offer a solution by bridging the gap between cell cultures and living tissue, enabling better control of the microenvironment while allowing the use of human cells to study human physiology. The inventors recognized that various epithelial cancer infiltration models, particularly on-chip models, can be used to more closely simulate primary and metastatic TMEs in an in vivo manner. However, some exemplary systems do not fully reproduce vascular anatomy because they focus on seeding endothelial cells onto permeable membranes or hydrogel surfaces, thus still exhibiting the characteristics of an endothelial monolayer.

[0099] Furthermore, the endothelial monolayer lacks many functional characteristics of a normal microvascular system, such as the physiologically representative barrier function. Since barrier function is one of the main obstacles that cancer cells must overcome for infiltration, the role of the functional microvascular system and the metastatic behavior of cancer cells cannot be well simulated in these models.

[0100] Examples of cancer-on-a-chip models utilize the tumor microsphere (MVN) formed around the tumor spheroid via angiogenesis and exhibit improved barrier properties. However, this model relies on the natural detachment of cancer cells from the spheroid into the MVN and does not account for the endogenous tumor metastasis (EMT) process, thus reducing the physiological relevance of the study.

[0101] The inventors have developed a method to stabilize microvessels (MVNs) in microfluidic devices by utilizing macromolecular crowding (MMC), thereby allowing the on-chip fabrication of functional microvascular systems with hollow, circular lumens, top-to-bottom polarity, and appropriate vascular barrier function. Importantly, the permeability coefficient of the MVN measured in a lung cancer endorphin chip model is significantly higher than that measured in vivo in rat microvenous vessels (1.37 ± 0.26 × 10⁻⁶). -7 The high consistency in cm / s indicates that the EMT-induced cocktail (IC) did not impair vascular barrier function and that the MVN maintained physiological permeability levels.

[0102] Although MVNs form spontaneously through angiogenesis, their density is determined by the initial cell seeding density and is reproducible across different devices and biological replicates. These MVNs were subsequently used to build an on-chip model of lung cancer cell infiltration, thereby studying the infiltration of cancer into functional microvascular systems. This method is simple and straightforward, ensuring ease of use in different laboratory settings as well as high verifiability and reproducibility. Furthermore, compared to previously reported on-chip models of infiltration, the inventors treat EMT as a major driver of infiltration and combine it with machine learning-assisted quantification. Therefore, the inventors provide a physiologically more representative new model that requires less time to observe, visualize, and quantify infiltration events, enabling more efficient and less time-consuming experiments and screenings.

[0103] The inventors recognized that on-chip models of cancer can be powerful tools for studying the tumor microenvironment and its contribution to cancer infiltration and metastasis. Preferably, they provide a controlled environment to simulate vascular-like conditions and enable visualization of events involving infiltration into artificial microvessels using live-cell imaging.

[0104] While some in vitro models have allowed for the study of how various chemical and physical aspects affect the transfer cascade reaction, these models may contain microchannels lined with a single layer of endothelial cells, which do not fully reproduce the anatomy of blood vessels and lack key features of functional microvessels, such as good, physiologically representative vascular barrier function.

[0105] Therefore, it is preferable to provide a physiologically relevant model capable of reproducing the two events of local invasion from the primary tumor via EMT and subsequent infiltration into the MVN. According to one embodiment of this disclosure, an in vitro human microphysiological model of lung cancer infiltration driven by EMT in a microfluidic device is provided, combined with machine learning (ML)-assisted quantification of infiltration events within the device. In this example, a robust EMT-inducing mixture is applied to a co-culture system comprising lung cancer spheroids (microtumor clumps) derived from A549 cells embedded in a 3D physiologically representative MVN within the microfluidic device. Furthermore, infiltration events are detected using live-cell imaging, high-resolution microscopy, and ML-assisted vessel segmentation incorporating colocalization analysis, respectively, to achieve visualization and quantification of the infiltration events. Advantageously, this disclosure opens pathways for studying physiologically relevant pathological processes of lung cancer at high spatiotemporal resolution, as well as for drug development and high-throughput screening using established platform technologies.

[0106] Reference Figure 1 An exemplary embodiment of a device 100 for observing bioactivity 102 in a sample is shown, which includes a microfluidic device 104. The microfluidic device 104 has a plurality of fluid channels, including a central channel 104A adjacent to a culture medium channel 104B and defining a perfusionable septum 104C therebetween; wherein the central channel 104A is used to promote EMT induced by a reagent supplied to the culture medium channel 104B in biological cells contained within the central channel 104A.

[0107] In this example, the microfluidic device 104 can be used to observe EMT that may occur in the central channel 104A of the microfluidic device 104, wherein the central channel 104A can accommodate biological cells (e.g., spheroids) undergoing EMT, and the EMT will occur in the central channel 104A and near the vascular-like structure.

[0108] For example, biological cells such as tumor spheroids can be injected into the central channel 104A through corresponding inlets of the same channel, and the tumor spheroids remain within the central channel 104A during the experiment. Endothelial cells forming a MVN (vascular network) can be co-cultured in the central channel, mimicking the properties and function of blood vessels. To initiate the EMT mechanism in the tumor spheroids, EM-IC can be injected into one or both culture media channels 104B to trigger EMT in cancer cells within the central channel 104A, which is spatially separated from the adjacent culture media channels 104B but can be perfused.

[0109] Also refer to Figure 2 An exemplary structure of a microfluidic device 100 is shown. In this example, the microfluidic device 100 is a three-channel microfluidic chip having a central channel 104A and a pair of culture medium channels 104B clamping the central channel 104A. The central channel 104A is connected to a cell inlet 106 and a cell outlet 108, and each culture medium channel 104B is connected to a culture medium inlet 110 and a culture medium outlet 112. Optionally, the pair of culture medium channels 104B are also connected to a common inlet 114, thus there are a total of 3 culture medium inlets and 2 culture medium outlets.

[0110] Furthermore, the microfluidic device 100 includes multiple partition structures 116 for separating the central channel 104A from one or more culture medium channels 104B. The fluid channels 104A and 104B have a height of 100 μm and a length of 14.5 mm, and are separated from the adjacent culture medium channels 104B by triangular prisms 116. The prisms 116 are arranged at regular intervals (e.g., 100 μm apart). The prisms 116 are arranged to ensure that the hydrogel injected into the central channel 104A is properly retained within the cavity sandwiched between these prism structures 116.

[0111] The widths of the central hydrogel channel 104A and the culture medium channel 104B are 1300 μm and 500 μm, respectively. The diameters of the culture medium inlet 110 and the cell inlet 106 are 1000 μm and 500 μm, respectively. Those skilled in the art will understand that the dimensions of the different features in the microfluidic device mentioned above are merely exemplary and can therefore be varied in alternative embodiments for other applications.

[0112] In one exemplary embodiment, the microfluidic device 100 can be fabricated by replication molding on a silicon wafer and soft photolithography using polydimethylsiloxane (PDMS). In an experiment conducted by the inventors, a 100 μm layer of SU-8 3050 negative photoresist (Kayaku Advanced Materials, Massachusetts, USA) primer was spin-coated onto the silicon wafer. It was then exposed to a photomask displaying a negative pattern of channel structures designed using computer-aided design (CAD) for photolithography. The SU8 was subsequently exposed to ultraviolet light (set to 20 mW / cm² at 365 nm) for 45 seconds, followed by pattern development. PDMS and a curing agent (Sylgard 184, Dow Corning, Michigan, USA) were mixed in a 10:1 (w / w) ratio and cast onto the SU8 master mold. After heat curing at 60°C for two hours, the positive replication molded pattern on the PDMS was separated from the wafer. The patterned PDMS was cut into individual devices, and the inlet and outlet were punched using 1 mm and 3 mm biopsy puncturists. Next, the devices and glass slides were cleaned with 100% ethanol and water, dried with a nitrogen gun, and then treated with oxygen plasma (HarrickPlasma, New York, USA) for 45 seconds to form covalent bonds between the glass slides and the PDMS devices. Immediately after the PDMS sheets and glass slides were assembled, the channels were coated with 1 mg / ml poly-L-lysine (PLL) (molecular weight: 30,000~70,000) (Meryer, Shanghai, China, Cat#.25988-63-0) dissolved in water for at least 20 minutes, followed by autoclaving. PLL coating increases the hydrophilicity of the devices, allowing for easy loading of hydrogels.

[0113] The PDMS microfluidic device 100 can be used to further form hollow luminal vascular structures in the central channel 104A, for example, through angiogenesis. Preferably, the hollow luminal vascular structures can be formed by culturing HUVECs to obtain the MVN118 of the HUVECs in the central channel 104A. More preferably, also refer to... Figure 3 The flowchart describes the co-culture of human umbilical vein endothelial cells and biological cells 120 (i.e., tumor spheroids) contained in the central channel 104A.

[0114] In one exemplary embodiment, MVNs and tumor spheroids of HUVECs can be co-cultured in the central channel of a microfluidic device, wherein both HUVECs and tumor spheroids are injected into the central channel. In an experiment conducted by the inventors, a 15 mg / ml solution of fibrinogen (Sigma-Aldrich, Cat#. F8630-5G) was freshly prepared in PBS for each experiment. A 100 U / ml thrombin solution (Sigma-Aldrich, Cat#. T4648) was prepared in PBS containing 1% (w / v) BSA (Sigma-Aldrich, Cat#. A7906) and aliquoted and stored at -20°C. HUVECs were resuspended in EGM-2 containing 6 U / ml thrombin, and the cell solution was mixed with the fibrinogen solution at a 1:1 ratio to achieve a final concentration of 7.5 mg / ml fibrinogen, 3 U / ml thrombin, and 6 × 10⁶ HUVECs / ml. As described later in the "Spheroid Formation" section, A549 spheroids containing 300 cells were formed. The spheroids were collected and resuspended in 200 μl of EGM-2, and then pipettes were applied in 2 μl / droplet amounts onto culture dishes for observation and selection. Droplets containing more than 10 spheroids were selected and rapidly introduced into the central channel along with a mixture of HUVECs, fibrinogen solution, and thrombin. The apparatus was then placed in a humidified incubator at 37°C for 1 hour to allow fibrinogen to polymerize into a fibrin hydrogel by thrombin. The hydrogel containing HUVECs and tumor spheroids remained within the central channel during incubation.

[0115] Next, EGM-2 was added to the culture medium channels. The medium was changed daily. To introduce MMCs into 3D cell culture, from day 1, medium containing Ficoll macromolecules (25 mg / ml Ficoll 400 - Cytiva, Marlborough, MA, USA, Cat#.17-0300-50; and 37.5 mg / ml Ficoll 70 - Cytiva, Cat#.17-0210-10) was introduced. The components of this Ficoll mixture had a calculated fractional volume occupancy (FVO) of 17% and were shown to enhance ECM deposition and basement membrane formation around the MVN. On day 3, the co-cultures were incubated for 24 hours with or without the EMT induction mixture (also supplemented with 25 mg / ml Ficoll 400 and 37.5 mg / ml Ficoll 70).

[0116] Further information relating to the materials and methods involved in the exemplary experiments conducted by the inventors is disclosed below: For general cell culture, the human mononuclear cell line THP-1 (ATCC TIB-202, Manassas, VA, USA) was maintained in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with GlutaMAX™ (GIBCO, Life Technologies, Grand Island, NY, USA, Cat. #61870036), while A549 cells (ATCC, Cat. #CCL-185) were cultured in Dulbecco's Modified Eagle's Medium (DMEM) (GIBCO, Cat. #10567-014) containing 1 g / L glucose and GlutaMAX™. Both media were supplemented with 1% 100 U / ml penicillin and 100 µg / ml streptomycin (P / S) (GIBCO, Cat. #15140-122), and 10% fetal bovine serum (FBS) (GIBCO, Life Technologies, Cat. #16000044). Primary human umbilical vein endothelial cells (HUVEC, mixed) (ATCC, Cat#.PCS-100-013) and GFP-expressing HUVECs (TTFLUOR HUVEC) (Innport, Primera Planta, Spain Cat#.P20201) were cultured in endothelial cell growth medium (EGM-2) (Lonza, Walkersville, MD, USA, Cat#.CC3162) until passage 8. All cells were passaged after reaching 80% confluence at 37°C and 5% CO2. HUVEC and A549 cells were cultured in polystyrene flasks coated with 0.1% gelatin (Sigma-Aldrich, Saint Louis, MO, USA, Cat#.G1890). HUVECs and PMA-treated THP-1 cells were digested with TrypLE™ Express (GIBCO, Life Technologies, Cat#.12605-010) at 37°C for 3 minutes and then resuspended in EGM-2, DMEM containing 10% FBS and 1% P / S, and RPMI containing 10% FBS and 1% P / S, respectively.

[0117] Regarding macrophage conditioned medium ( To prepare the macrophages, one million THP-1 cells were seeded into each well of a 6-well plate containing 3 ml of RPMI-1640 medium containing 10% FBS, 1% P / S, and 100 ng / ml phorbol-12-myristate-13-acetate (PMA) (STEMCELL Technologies, Vancouver, Canada, Cat. #74042) to induce macrophage differentiation. After 24 hours, adherent cells were washed with phosphate-buffered saline (PBS), and then replaced with 3 ml of EGM-2 medium and incubated for 48 hours. Subsequently, the supernatant was extracted, filtered through a 0.22 µm filter, and then diluted 1:1 with fresh EGM-2 medium. Store at -80°C and thaw immediately before use.

[0118] Regarding the cytokine array of the proteome analyzer, analysis was performed using the Proteome Profiler Human XLCytokine Array Kit (R&D Systems, Minneapolis, USA, Cat.#ARY022B). The kit contains 105 cytokines. Following the manufacturer's instructions, the antibodies and reagents in this kit were used to simultaneously detect 105 cytokines. The ChemiDoc™ MP Imaging System (Bio-Rad Laboratories, CA, USA, Cat. #17001402) was used. A quantification control (unconditioned EGM-2 medium) and... Differences in cytokine levels between replicates were assessed using ImageJv1.54f software, measuring the average pixel intensity at each replicate. Cytokines were classified as EMT inducers or inhibitors, and as pro-inflammatory or anti-inflammatory cytokines, and then plotted using GraphPad Prism 9.4.1 software. A graph showing the log2 factor change relative to the control.

[0119] Regarding spheroid formation, A549 cell spheroids containing 500 cells were formed using an Aggrewell™ 400 culture plate (STEMCELL Technologies, Cat.#34411) according to the manufacturer's instructions. Briefly, 500 µL of anti-adhesion wash buffer (STEMCELL Technologies, Cat.#07010) was added to each well, and centrifuged at 1,300 xg for 5 minutes to prevent cell adhesion. After removing the wash buffer, 600,000 A549 cells suspended in 2 ml of DMEM containing 10% FBS and 1% P / S were added to each well to achieve the target number of spheroids containing 500 cells per well. Cell aggregation was achieved by centrifuging at 100 xg for 3 minutes and incubating for 3 days, after which the cells were flushed out for subsequent experiments. Those skilled in the art will understand that, according to alternative embodiments of this disclosure, other target biological samples (e.g., spheroids of other cancerous diseases) can be cultured for observation using this system.

[0120] For 2D cell seeding, trypsin-digested A549 cells were seeded at a density of 15,000 cells / cm² in 48-well tissue culture-treated plates and incubated overnight to allow for adhesion. After overnight incubation, the culture medium was replaced with either the control (EGM-2 medium) or the following experimental media: (i) EGM-2 medium containing 5 ng / ml recombinant human TGF-β1 (PeproTech, New Jersey, USA, Cat. #100-21); (ii) containing... EGM-2; and (iii) containing EGM-2 and TGF-β1 were then incubated for 2 days before further analysis.

[0121] Preferably, the system may also include imaging equipment, such as a microscopic imager, for capturing images of biological cells in the central channel to facilitate visualization and recording of endosmotic events in the biological cells within the central channel. To analyze A549 cell morphology, phase-contrast images of A549 cells were captured using a 100x objective lens on a Nikon ECLIPSE Ti2-A inverted fluorescence microscope (Nikon Instrument Inc.). Cell roundness was assessed using ImageJ software. A roundness of 1.0 indicates perfect roundness, while values ​​closer to 0.0 indicate less round or more elongated cell morphology.

[0122] Regarding viability assays, cell viability of A549 cells subjected to different conditions was investigated using two methods. First, cell counts were performed in each field of view based on nuclear staining with 4',6-diamidinyl-2-phenylindole (DAPI). Additionally, live / dead cell assays were performed using a Live / Dead Viability / Cytotoxicity Kit (Invitrogen, Waltham, Massachusetts, USA, Cat. #L3224) according to the manufacturer's instructions. Dead and live cells were stained with 4 μM ethidium bromide homodimer-1 and 2 μM calcein acetoxymethyl ester, respectively, at 37°C for 15 min before imaging with a 10x objective lens on a Nikon ECLIPSE Ti2-A inverted fluorescence microscope.

[0123] For the immunocytochemistry of 2D cultures, the sources of antibodies and reagents used in this study are summarized in Table 1. Cells were first fixed with 4% paraformaldehyde (PFA), then permeabilized with 0.1% Triton X-100 for 15 min, and analyzed with 3% bovine serum albumin (BSA) (Sigma-Aldrich, Saint Louis, USA, Cat. #A7906). The cells were then incubated overnight with primary antibodies against E-cadherin and vimentin. After overnight incubation, the cells were washed three times with PBS and treated with Alexa Fluor 555-conjugated secondary antibody for 1 hour, followed by incubation with DAPI for 10 minutes. Fluorescently stained E-cadherin and vimentin were visualized using a 63x objective lens on a Nikon ECLIPSE Ti2-A, and fluorescence intensity was calculated using ImageJ v1.54f software using the following formula: Corrected total cellular fluorescence (CTCF) = Integrated density – (Area of ​​selected cells × Mean background fluorescence intensity) Table 1. List of antibodies and reagents used in 2D immunocytochemistry.

[0124] RNA extraction and cDNA synthesis were performed using RNAiso Plus (Takara Bio Inc., Cat. #9109) and PrimeScript™ RT Master Mix (Takara Bio Inc., Cat. #RR036A), respectively, according to the manufacturer's protocols. For RT-qPCR analysis, 50 ng of cDNA and target-specific primers were amplified using ChamQ SYBR Color qPCR Master Mix (Vazyme, Cat. #Q411-03) on a QuantStudio™ 7 Pro real-time quantitative PCR system (AppliedBiosystems™, Carlsbad, CA, USA) (Table 2). Gene expression levels were quantified using a cycle threshold (ΔΔCT) value, normalized relative to the housekeeping gene GAPDH, and the relative fold change was compared to the A549 control condition (EGM2 only) collected 24 hours after medium replacement. Another alternative housekeeping gene, B2M, was used to ensure the validity of GAPDH, and it produced similar results.

[0125] Table 2. Primer sequences used for RT-qPCR For the cell migration assay, A549 cells were cultured in the wells of a 48-well plate until they confluenced. First, the medium in each well was replaced with PBS. Then, using a P1000 pipette tip, incisions were made in the confluenced A549 cell layer to create gaps. The cells were washed with PBS to remove any floating cells, and then the initial medium was added back.

[0126] Microscopic images of the central channel can be captured at predetermined time intervals. For example, phase contrast imaging was subsequently performed at 0, 6, and 12-hour time points using a 100x objective lens from a Nikon ECLIPSE Ti2-A inverted fluorescence microscope. The scratch area at specific time points was quantized using ImageJv1.54f software.

[0127] For the spheroid invasion assay, A549 spheroids were centrifuged and resuspended at a density of 1,200 spheroids / ml in media under different conditions. For each well of a 48-well plate, 75 µL of the spheroid suspension (estimated to contain 15 spheroids) was mixed with 75 µL of 2 mg / ml neutralized type I collagen TeloCol®-6 hydrogel (Advanced BioMatrix, Carlsbad, USA, Cat. #5225). The type I collagen hydrogel containing the spheroids was incubated at 37°C for two hours to allow polymerization, and then topped with 300 µL of medium. EGM-2 medium was used as a control. The spheroids were incubated for 24 hours to investigate their invasive ability under different conditions. After fixation with 4% PFA for 30 min, the spheroids were imaged using a 100x objective lens on a Nikon ECLIPSE Ti2-A inverted fluorescence microscope. To quantify the invasiveness of the spheres, the area of ​​the invasive region was manually tracked and measured using ImageJ v1.54f software, and then divided by the initial sphere area.

[0128] For the immunocytochemistry of 3D cultures, after 4 days of culture, the apparatus was washed with PBS, and the cultures were fixed for 15 minutes through the culture medium channel with 4% paraformaldehyde (PFA) (Thermo Scientific, Cat#.5735) in PBS, followed by permeabilization with 0.25% Triton X-100 in PBS for 10 minutes. After blocking with 5% BSA for 1 hour, the samples were incubated with the corresponding primary antibody (see Table 3) in PBS containing 0.5% BSA at 4°C for 16 hours. The samples were then washed three times with PBS for 5 minutes each time, followed by incubation with secondary antibody (see Table 3) at room temperature for at least 3 hours. Before imaging, the samples were washed and stored in PBS at 4°C.

[0129] Table 3. List of antibodies and reagents used in 3D immunocytochemistry.

[0130] Confocal imaging was performed using 20x and 63x objectives on an inverted confocal microscope (Leica SP8, Leica Microsystems, Germany). Images of laminin, VE-cadherin, and vinculin were acquired and visualized. Orthogonal projection reconstruction of z-stack images was performed to visualize tumor intravasation events. All images were analyzed using ImageJ v1.54f software (https: / / imagej.nih.gov / ij / ). Furthermore, the MVN near the tumor spheroid was visualized using Qiber3D.

[0131] The system may also include a processor, preferably a machine learning processing engine, for analyzing the captured images to identify tumor endosmosis events in biological cells based on features extracted from the images. Quantification of MVN vascular junctions and microtubule lengths for each FOV was performed using ImageJ v1.54f software and the Angiogenesis Analyzer plugin 1.0 (https: / / imagej.nih.gov / ij / macros / toolsets / Angiogenesis%20Analyzer.txt), analyzing phase-contrast images of the MVN (10x). Briefly, the raw images were converted to binary images using automatic thresholding for binary tree analysis in the Angiogenesis Analyzer plugin. The number of junctions and total branch length were measured and expressed as the number of vascular junctions and microtubule length for each FOV, respectively. A junction was represented as a point with at least three neighboring points, while microtubule length refers to the length of an element defined by two junctions or a junction with an endpoint.

[0132] For Western blotting, the PDMS device was detached from the glass slide using a cutter and cultured in polystyrene flasks containing a 1:1 mixture of 2x Laemmli buffer and 2x protease inhibitor (Sigma-Aldrich, Cat#. P8340). The samples were then cultured in polystyrene flasks using the BCA Protein Assay Kit (Thermo Fisher, Cat#. A53226). The samples were denatured at 95°C for 5 min and loaded onto an 8% SDS-polyacrylamide gel (Life Technologies, Cat#. HC2040) at an equal protein loading volume for electrophoresis at 120 V. After protein separation, the samples were electrotransferred onto polyvinylidene fluoride membranes (Thermo Scientific, Cat#. 88,518) using a Power Blotter XL SYS (Life Technologies, Cat#. 0.34580). For membrane staining, before incubating the membrane overnight at 4°C in TBST containing 1% skim milk powder (see Table 1), the membrane was incubated in TBS-Tween 20 (TBST) containing 5% skim milk powder (Phygene Biotechnology Co Ltd., Fuzhou, China, Cat#. PH1519) (containing 50 mM Tris, 150 mM NaCl, and 0.5% Tween 20 (Sigma-Aldrich, Cat#. P2287)) to block nonspecific antibody binding. After washing three times with TBST, the secondary antibody (see Table 4) resuspended in TBST containing 1% skim milk powder was added to the blot membrane, and the reaction was carried out at room temperature for 1 hour. Protein bands were then detected using the ChemiDoc™ MP Imaging System (Bio-Rad) via ECL Super Signal West Pico Plus (Life Technologies, Cat#34,580) and quantified using Image Lab 6.1 software (Bio-Rad).

[0133] Table 4. List of antibodies and reagents used in 3D immunocytochemistry.

[0134] For live-cell imaging, resuspended A549 and HUVEC cells were pre-labeled with universal cell membranes using the PKH26 red fluorescence and PKH67 green fluorescence cell linker Midi kits, according to the manufacturer's protocols (Sigma-Aldrich, Cat. #MIDI26, MIDI67). 24-hour live-cell imaging began on day 3 of culture and was performed after the optional addition of the EMT-inducing mixture (i.e., the reagent added to the culture medium channels), using a 20x objective lens on a Leica Mica confocal microscope (Leica Mica, Leica Microsystems, Germany) to track any tumor intravasation events. To quantify the intravasation rate from the fluorescence images, the number of cancer cells colocalized with the green fluorescence of the MVN was manually counted. Preferably, the EMT-inducing mixture complex comprised transforming growth factor β1 (TGFb1) and macrophage conditioned medium.

[0135] Regarding machine learning-assisted vessel segmentation and quantification of infiltration events, the inventors defined the localization of fluorescently labeled A549 cancer cells within the MVN as a representative indicator of cancer cell infiltration. All analyses were performed using Fiji (ImageJ v1.54f) at 24-hour time points on fluorescent live-cell images.

[0136] Images can be processed through the following modules, including blood vessel segmentation, tumor spheroid extraction, and segmentation workflows.

[0137] Vessel segmentation was achieved using a machine learning-assisted approach with image stacks of phase contrast and GFP channels. This dual-channel combination enabled reliable segmentation even when GFP expression in HUVECs decreased during culture. Initially, images were decomposed into separate stacks for thresholding. Default thresholds were applied to both GFP and Cy3 channels to segment HUVECs and tumor-like structures. Subsequently, the tumor-like structure signal was subtracted, and the images were recombined with the phase contrast channel. A trainable Weka segmentation (TWS) plugin was then trained using 28 images to distinguish between vascular and non-vascular regions. Segmentation parameter settings included trainable features such as Gaussian blur, Hessian matrix, membrane projection, Sobel filter, and difference of Gaussians. The parameter settings were: membrane thickness of 1, block size of 19, minimum sigma of 1.0, and maximum sigma of 16.0. The selected classifier option was "Fast Random Forest," using 200 trees and a batch size of 100. These settings determined how TWS learned and applied machine learning for image segmentation. Binary semantic segmentation was performed on images containing blood vessels and spheroids, classifying blood vessels into one category and spheroids and background into another. Subsequently, the binarized tumor-like fluorescence signal was intersected with the positive regions of blood vessels to identify infiltrated cells originating from the tumor-like structures.

[0138] In tumor spheroid extraction, each microscopic image used for quantification contains multiple tumor spheroids present on a microfluidic chip. A square region of interest (tROI) is drawn around each tumor region to extract each spheroid individually. A particle analysis plugin is used to locate each spheroid, and after elliptical fitting of the spheroid, the size of the tROI is set to 115% of its major axis to ensure that only the signal from its immediate neighboring region is included in the analysis.

[0139] In the segmentation workflow, a default threshold is applied to the GFP channels, and then a Weka classifier is applied to the phase difference and binarized GFP channels for each tROI to segment the MVN ( ). Figure 6C Morphological manipulations (e.g., erosion and dilation) are used to reduce imaging artifacts. A region of interest (vROI) of segmented vessels is then generated using a particle analysis plugin. This vROI is then overlaid onto a channel containing pre-fluorescently labeled A549 cancer cells. Fluorescence signals outside the vROI are excluded. Figure 6C The Otsu threshold was applied to segment cancer particles within the vROI. Fluorescent particles of A549 cells within the segmented blood vessels were counted (minimum size: 50 µm², roundness 0.5–1.0).

[0140] An ImageJ script was developed to facilitate high-throughput automated image analysis. It can be obtained at: https: / / github.com / anna-jaey / FijiScriptToolbox?tab=readme-ov-file#particle-counting-and-area-ratio-quantification-in-multi-channel-fluorescence-images. The inventors also performed statistical analyses. Each experiment was conducted with at least three independent biological runs, each run with at least two replicates. Levene's test was performed to test the homogeneity of variance between samples. For parametric samples, unpaired t-tests and one-way ANOVA were performed for samples with two or more conditions, followed by Tukey's post-hoc multiple comparison test. On the other hand, for nonparametric samples with two or more conditions, Mann-Whitney and Kruskal-Wallis tests were used, respectively. Data are presented as mean ± standard deviation, with each data point containing at least three replicates, and statistical significance was set as follows: p < 0.05; p < 0.01; p < 0.001; p < 0.0001. All photomicrographs were quantized using ImageJ v1.53t software (https: / / imagej.nih.gov / ij / ), and all analyses were performed using GraphPad Prism v9.3.1 (GraphPad Software, San Diego, CA, USA, www.graphpad.com).

[0141] Figure 4A and Figure 4B This demonstrates the use of a proteomics analyzer array for... A cytokine profiling analysis revealed a large amount of EMT inducers secreted by THP-1-derived macrophages. (Refer to...) Figure 4A and Figure 4B Observed TGF-β1 synergistically promotes EMT in A549 cells. EMT is characterized by loss of apical-basal polarity, disruption of cell adhesion, and increased ability of cancer cells to migrate and invade surrounding tissues. Standard EMT induction protocols utilize TGF-β1 because TGF-β1 / SMAD is a hallmark pathway of EMT, leading to the upregulation of EMT-related transcription factors. However, inducing EMT using a single growth factor often results in incomplete transformation, therefore other EMT drivers have been considered. Physiologically, cellular components of the tumor microenvironment, such as cancer-associated fibroblasts (CAFs) and tumor-associated macrophages (TAMs), have been shown to play a decisive role in cancer cell EMT. In fact, studies have shown that the secretome of TAMs contains multiple cytokines involved in EMT induction, and macrophages or their conditioned media have been shown to promote EMT in epithelial cancer cells. Both pro-inflammatory (M1) and immunomodulatory (M2) polarized macrophages have been shown to drive EMT.

[0142] Adenocarcinoma human alveolar basal epithelial cells (A549) can be exposed to TGF-β1 and macrophage conditioned medium ( (or a combination of both) and study their ability to undergo EMT. Derived from THP-1-derived macrophages. Proteomic analysis was performed using an array of proteomic analyzers. The composition was analyzed. Compared with unconditioned EGM-2 medium, It is rich in a variety of known EMT inducers, exceeding the number of EMT inhibitors. Meanwhile, as shown in Figure 4, the presence of pro-inflammatory and anti-inflammatory factors is relatively balanced, indicating... It contains a mixture suitable for inducing EMT in cancer cells.

[0143] Reference Figures 5A to 5G A549 cells were compared with standard EMT-inducing cytokines TGF-β1, Incubation can be performed using either a combination of methods to determine the optimal formulation of EMT-IC. It can be observed that... It synergistically drives EMT in A549 cells with TGF-β1. In these figures, p < 0.05; p < 0.01; p < 0.001; p < 0.0001. n = 3 biological replicates.

[0144] From a morphological perspective, refer to Figure 5A When exposed to TGF-β1 or At that time, A549 cells exhibited a significantly reduced cell roundness, while the combination of both induced the most pronounced change in morphology from cuboidal to spindle-shaped. (See reference) Figure 5B Although none of the treatments affected cell viability, but as Figure 5H As shown, the number of cells per field of view decreased, attributed to the aforementioned change in cell morphology towards a larger spindle shape. Meanwhile, referring to... Figure 5C A549 cells exposed to TGF-β1 did not show a visible reduction in the epithelial marker E-cadherin, while those exposed to TGF-β1 did not show a visible reduction. A549 cells treated with either treatment or a combination of both experienced the most significant reduction in E-cadherin signaling. (See reference...) Figure 5D In exposure to TGF-β1 or Upregulation of the mesenchymal marker vimentin was detected in the combined A549 cells, although it was not significant.

[0145] Gene expression analysis showed that when A549 cells were exposed to When combined with TGF-β1, key genes involved in EMT are generally upregulated, such as Figure 5E As shown. Although SNAI1 gene expression showed relatively high variability among treatment groups, it also showed significant variation when TGF-β1 was used alone or in combination with... In A549 cells treated with the combination therapy, SNAI2 was significantly upregulated. As shown in the figure, TWIST1 was only upregulated when treated with the combination therapy. The expression of ZEB1 and ZEB2 was significantly upregulated in A549 cells treated with a combination of both.

[0146] Since marker expression and morphological changes are insufficient to confirm successful EMT, the function of transformed A549 cells in terms of migration and invasion potential was investigated. (Refer to...) Figure 5F When used alone, TGF-β1 and The similarity enhanced the migration potential of A549, and the combination of the two more than doubled the effect. Next, A549 spheroids (tumor microclusters formed in micropores, refer to...) Figure 5I (The cells) were embedded in a 3D type I collagen hydrogel to measure the ability of tumor cells to invade their local environment. For example... Figure 5G As shown, treatment with TGF-β1 increased their invasiveness, while treatment with... Treatment with either TGF-β1 or a combination of both resulted in a significant and profound invasion of the surrounding environment. In summary, TGF-β1 and The combination induced the most effective EMT in A549 cells and was used as EMT-IC in all subsequent experiments.

[0147] It was observed that EMT-IC did not adversely affect the stability of the MVN in the microfluidic device. Using an established three-channel microfluidic device, HUVECs were seeded in fibrin hydrogel in the central channel, where they formed a stable MVN through angiogenesis. (See reference...) Figure 6A HUVECs were seeded in fibrin hydrogel in the central channel using an established three-channel microfluidic device 100', where they formed stable MVNs through angiogenesis. Figures 6B to 6H EMT-IC (TGF-β1&) was shown. The impact on the stability and integrity of MVN. In these figures, p < 0.01; p < 0.0001; n = 3 biological replicates.

[0148] and Figure 2 and Figure 3 Similar to the illustrated embodiment, the microfluidic device 100' in Figure 6 is also a three-channel microfluidic chip, having a central channel 104A and a pair of culture medium channels 104B clamping the central channel 104A. Furthermore, the central channel 104A is connected to a cell inlet 106 and a cell outlet 108, and each culture medium channel 104B is connected to a culture medium inlet 110 and a culture medium outlet 112. Additionally, the microfluidic device 100 includes multiple partition structures 116 for separating the central channel 104A from one or more culture medium channels 104B. As shown, these partition structures are triangular.

[0149] In this exemplary experiment, on day 3, the culture medium was replaced with EMT-IC or control medium (unconditioned EGM-2 without TGF-β1), and MVN was incubated for another 24 hours. Figure 6B As shown, the MVN forms consistently and reproducibly in different devices and biological replicates, thus ensuring the reproducibility and robustness of the model, such as... Figure 6I As shown, EMT-IC has no significant impact on the stability of MVN, as illustrated by the quantification of the total length of the microtubule structure and the number of vascular connection points, referencing... Figure 6C Co-staining of the main basement membrane components laminin and the focal adhesion marker vinculin showed that the endothelial structure is enveloped by a tight basement membrane sheath. This indicates that EMT-IC has no adverse effect on the apical-basal polarity of microvessels, and that cell adhesion focals interact closely with the vascular basement membrane, such as... Figure 6D As shown. Co-staining with vascular-specific endothelial adhesion molecules vascular endothelial cadherin (VE-cadherin) and F-actin (fibrous actin) allowed observation of intercellular connections relative to the overall MVN structure. Clear cell boundary outlines were observed under both conditions, although these boundaries appeared more taut in MVNs incubated with EMT-IC, as shown in the reference. Figure 6E Semi-quantitative analysis of laminin levels, determined by Western blotting, showed a decrease in basement membrane levels in MVNs exposed to EMT-IC, such as... Figure 6F As shown, Western blot analysis of vascular endothelial cadherin (VE-cadherin) revealed a 50% decrease in protein levels in MVNs exposed to EMT-IC, such as... Figure 6G As shown.

[0150] Reference Figure 6H To visualize endosmosis events in the microfluidic device 100, A549 spheroids and HUVECs were co-inoculated into the central channel of the hydrogel in the microfluidic device, allowing MVN to form during the first 3 days of culture. Subsequently, the culture medium was replaced with EMT-IC or unconditioned EGM-2 (control), and the cultures were incubated for another 24 hours.

[0151] Referring to Figure 7, it was observed that when the EMT-IC was co-seeded with A549 spheroids in the microfluidic device, it did not adversely affect the stability of the MVN. The phase-contrast images clearly show that the MVN formed tightly around the A549 spheroids and was uniformly distributed throughout the chip, while the tumor mass remained intact. Figure 7A As shown. (Refer to...) Figure 7B Regardless of the culture medium chosen, MVN showed considerable stability on day 4. Figures 7A to 7G It shows that within 24 hours, the EMT-IC (TGF-β1 and) Live-cell fluorescence imaging of A549 cancer cells induced to migrate and infiltrate into the MVN. In these images, n=11 biological replicates.

[0152] Reference Figure 7C and Figure 7DImmunostaining revealed that both the MVN and tumor masses synthesized laminin, while semi-quantitative assessment by Western blot analysis of the co-cultures showed a slight decrease in total laminin protein levels. As expected, immunocytochemical analysis of vascular endothelial cadherin (VE-cadherin) showed that this cell adhesion molecule was limited to the MVN, while F-actin (fibrous actin) was labeled in all co-cultured cells. This allowed for clear observation of cancer cells migrating from the initial tumor masses and moving towards the MVN, referencing... Figure 7E Notably, in unconditioned culture media, a clear boundary could be observed between migrating tumor cells and microvascular structures. Figure 7E (arrows), while under EMT-IC conditions, these boundaries are not obvious at sites of cancer cell-MVN interaction. In fact, at these specific locations, vascular endothelial cadherin (VE-cadherin) appears to be strongly downregulated ( Figure 7E (asterisk). Semi-quantitative analysis of Western blot results for vascular endothelial cadherin (VE-cadherin) showed that its level decreased by more than 50% in the EMT-IC culture system, such as... Figure 7F As shown.

[0153] In addition, live-cell imaging of fluorescently labeled A549 spheroids and MVNs was observed within 24 hours to track events of infiltration into the MVN, such as... Figures 7G to 7I As shown. Because the cells are labeled with a non-permanent membrane dye, it does not delineate the entire cell body after several days of culture. Therefore, fluorescent labeling is used as an indicator of cell identity, and PhC imaging is necessary for reliably tracking cell movement. Live-cell imaging was performed at 10x magnification to track the movement of all cells from the spheroid, as shown. Figure 7H and 7I As shown; while the time frames of the observed endorphin events are presented in magnified close-up form, such as... Figure 7G , 7J As shown in 7K and 7L, by superimposing fluorescence and phase-contrast images, it was possible to observe cancer cells leaving the tumor spheroid and moving toward and along the vessel wall. At the endorphin site, cancer cells extend membrane protrusions through the microvascular wall, and then allow their cell bodies to pass through the opening into the MVN (microvascular nucleus). Figure 7G (dashed circle). This was observed in both the control and EMT-IC conditions.

[0154] Reference Figures 8A to 8EEMT-IC was observed to promote the infiltration of cancer cells into the MVN. To visualize the infiltration event, co-staining of the A549 spheroids and MVN co-cultures on day 4 with the endothelial markers CD31 and F-actin (fibrous actin) was performed. In these figures, p<0.05; p < 0.0001. Control group = 19 spheroids; EMT-IC group = 21 spheroids, collected from 3 biological replicates.

[0155] To visualize the endothelial events in more detail, co-cultures of A549 spheroids and MVNs were fixed after 24 hours of exposure to control medium or EMT-IC and co-stained with endothelial markers CD31 and F-actin. A549 showed stronger fluorescence intensity for F-actin staining, allowing it to be distinguished from CD31-positive MVNs with weaker F-actin staining. As expected, cancer cells migrated from the A549 spheroids under both conditions (initial tumor masses are indicated by white dashed lines), as... Figure 8A and 8B As shown. Nevertheless, semi-quantitative analysis of the invasive regions of migrating cancer cells indicated that cancer cells exposed to EMT-IC exhibited enhanced invasive potential within the fibrin hydrogel of the microfluidic device, such as Figure 8C As shown. This is comparable to previous results observed in type I collagen hydrogels, and can also be found in [reference needed]. Figure 5G .

[0156] A close-up of a confocal microscope image under control conditions. Figure 8A and 8B The white solid-line rectangle in the image shows that although cancer cells are located near microvessels, they are usually still clearly defined. Figure 8A (White asterisk), indicating that no endovascular events occurred. Conversely, under EMT-IC conditions, migrating cancer cells extend membrane protrusions into the microvascular structures ( Figure 8B (arrow), thus initiating infiltration. For example... Figure 8B Manual quantification of the defined endoosmosis events indicates that, on average, two endoosmosis events occur per spheroid when exposed to EMT-IC.

[0157] Conversely, refer to Figure 8D In control conditions, infiltration events were rare (<1 infiltration event per spheroid). This was further confirmed by high-resolution confocal microscopy, where a clear separation between the invading cancer cells and the MVN was primarily observed in control conditions. Figure 8E(White asterisk). Under EMT-IC conditions, the membrane protrusions of cancer cells invade the microvascular structure, and subsequently their cell bodies migrate into the MVN lumen ( Figure 8E The latter case is particularly evident because, in orthogonal cross-sections and 3D projections, the cancer cell body appears to be surrounded by CD31-stained lumens (arrows). Figure 8E (The last line).

[0158] Reference Figures 9A to 9J To verify the universality of the established lung cancer endosmosis microarray model, two other lung cancer cell lines—non-metastatic BEAS-2B and metastatic NCI-H1975—were introduced into the model. In these figures, p < 0.01; and p < 0.0001. NCI-H1975: Control group = 10 spheroids; EMT-IC group = 8 spheroids, collected from 3 biological replicates. BEAS-2B: Control group = 22 spheroids; EMT-IC group = 17 spheroids, collected from 3 biological replicates.

[0159] Reference Figures 9A to 9C Even under control conditions, the more invasive metastatic NCI-H1975 cell line exhibited strong invasive potential, which was significantly enhanced (10-fold) by EMT-IC. Correspondingly, frequent infiltration of NCI-H1975 cells into the surrounding microvascular system was observed under control conditions (0.5 events / globulus), while EMT-IC supplementation significantly increased this phenomenon (3 events / globulus). Figure 9D As shown. High-resolution imaging confirmed the successful infiltration of NCI-H1975 cells under EMT-IC conditions, with cancer cells identified within the microvascular lumen, such as... Figure 9E As shown.

[0160] In contrast, the response of BEAS-2B cells was significantly weakened because the areas they invaded were smaller in nature and could only be slightly enhanced (by a factor of 2) by EMT-IC, such as Figures 9F to 9H As shown. Furthermore, BEAS-2B cells did not exhibit any significant endosmotic potential under control conditions; this potential was only slightly increased by EMT-IC, thus maintaining a baseline endosmotic level below 1 event / spheroid, as... Figure 9I As shown, this is consistent with its non-metastatic nature. High-resolution microscopy confirmed that even under EMT-IC conditions, BEAS-2B cancer cells generally maintained a clear boundary with the MVN, as shown. Figure 9J As shown.

[0161] Also refer to Figures 7G to 7LLive-cell imaging of fluorescently labeled A549 spheroids and MVNs was observed over 24 hours to track infiltration events into the MVNs. Any colocalization of red-labeled A549 cells within green-labeled MVNs was defined as an infiltration event. Live-cell imaging confirmed enhanced cancer cell shedding and migration in the presence of EMT-IC, and an increased rate at which A549 cells approached adjacent MVNs by extending their pseudopodia.

[0162] An image analysis workflow was developed to segment blood vessels and cancer cells and analyze their colocalization, quantifying endothelial events in an unbiased manner. Figure 10D As shown, the inventors reliably segmented the MVN surrounding the tumor spheroid using a TWS plugin. Combined with threshold-based segmentation of the tumor spheroid, fluorescently labeled cancer particles within the segmented vascular region were counted as a representative indicator of endorphin events, referencing... Figure 10A and Figure 10B . Reference Figure 8D and Figure 10E The automated method exhibited a similar trend to manual counting. It was observed that, under EMT-induced exposure conditions, an average of three endosmosis events occurred per spheroid, compared to one endosmosis event per spheroid under control conditions. Figure 10C Using automated workflows, the total number of events per group increased slightly.

[0163] Reference Figures 11A to 11F This shows that EMT-IC (TGF-β1 and) The effect of co-culturing A549 spheroids on the stability and integrity of MVNs. In these figures, p < 0.05; p < 0.001; n = 3 biological replicates.

[0164] Immunostaining of MVN and A549 spheroids on day 4 showed that both MVN and tumor masses (identified as dense cellular clusters in PhC images) synthesized laminin, such as Figure 11A As shown, semi-quantitative assessment of Western blot analysis of the co-cultures revealed a slight decrease in total laminin protein levels under EMT-IC conditions, such as... Figure 11BAs shown. As expected, immunocytochemical analysis of vascular endothelial cadherin (VE-cadherin) showed that this cell adhesion molecule was limited to the MVN, while F-actin (fibrous actin) was labeled in all co-cultured cells. Nevertheless, F-actin in A549 exhibited stronger fluorescence intensity, allowing it to be distinguished from the MVN. Thus, the migration of cancer cells from the initial tumor mass towards the MVN could be clearly observed, as... Figure 11C As shown. Notably, in unconditioned culture medium, a clear boundary can be observed between migrating tumor cells and microvascular structures (…). Figure 11C (asterisk); however, under EMT-IC conditions, these boundaries are not apparent at sites of cancer cell-MVN interaction. In fact, vascular endothelial cadherin (VE-cadherin) appears to be downregulated at these specific locations ( Figure 11C (arrow). Semi-quantitative analysis of Western blot results for vascular endothelial cadherin (VE-cadherin) showed that its level decreased by more than 50% in the EMT-IC culture system, such as... Figure 11D As shown.

[0165] Since vascular barrier function, characterized by low vascular permeability, depends on tightly connected intercellular junctions and a well-structured basement membrane, vascular wall permeability was subsequently assessed. To this end, on day 2 of culture, HUVECs were seeded into culture medium channels of a cancer endosmosis chip model to form an additional continuous endothelial monolayer at the culture medium-hydrogel interface. By day 4, these endothelial cells had anastomosed with the MVN in the central channel, forming vascular openings connected to the culture medium channel.

[0166] To assess perfusion capacity, FITC-PVP (40 kDa) was introduced into one of the culture medium channels along with EMT-IC. In vivo fluorescence microscopy showed that MVN could be perfused from the culture medium channel under both conditions, such as… Figure 11E As shown. Next, vascular permeability was assessed by measuring the diffusion flux of the solute across the vessel wall. Specifically, the transport of FITC-PVP across the microvessel wall was quantified by tracking changes in fluorescence intensity over time within a defined perivascular (hydrogel) region (imaged every 15 seconds over 15 minutes). Assuming the microvessels form circular tubular structures with a diameter less than 50 μm, the permeability coefficient was calculated using established methods.

[0167] Upon initial exposure to EMT-IC, the MVN exhibited a permeability coefficient of 7.79 ± 2.7 × 10⁻⁶. -7 cm / s, which is different from the control group's 5.41±5.37×10 cm / s. -7The permeability coefficients are comparable to those in cm / s, such as Figure 11F As shown. Measurement of vascular permeability 24 hours after exposure to EMT-IC was not possible because at this point in the experiment, the openings of blood vessels to the culture medium channels under EMT-IC conditions had closed, while MVN under control conditions remained permeable.

[0168] These embodiments provide an in vitro microphysiological model of lung cancer infiltration driven by EMT, combined with machine learning (ML)-assisted image processing, enabling automated and unbiased quantification of infiltration events. This platform technology allows for visualization and study of underlying biological processes at high spatiotemporal resolution and has potential applications as a drug screening tool. The universality and sensitivity of this on-chip model of lung cancer infiltration were demonstrated by functionally integrating multiple metastatic and non-metastatic lung cancer cell lines. Importantly, the baseline metastatic potential of cancer cell lines was clearly reflected in the device's measurements of invasive and infiltration potential, indicating that the established model allows for the prediction of cell-specific metastatic potential and EMT-induced responsiveness, paving the way for testing patient-specific samples.

[0169] Advantageously, the model's benefit lies in its accelerated endosmosis timeline, as live-cell imaging data confirm that endosmosis occurs frequently in the initial hours following EMT induction. Therefore, the 24-hour endpoint provides a snapshot of actively endosmotic cells, while cells that have already inseminated are unlikely to have migrated out of the imaging field of view via the MVN. Importantly, the established system maintains excellent reproducibility during these initial 24 hours. This compressed timeline represents a significant advantage over other alternative in vitro models and in vivo systems, particularly for high-throughput applications.

[0170] Furthermore, a novel EMT-inducing mixture based on the synergistic effect of macrophage conditioned medium and TGF-β was developed, which induced robust migration and invasion behavior in lung cancer cells, superior to standard EMT induction methods. Since both M1 and M2 polarized macrophages have been shown to participate in promoting EMT in cancer cells, the inventors selected unpolarized macrophages capable of secreting pro-inflammatory and anti-inflammatory factors, and chose this conditioned medium based on its rich content of EMT inducers. It contains multiple EMT inducers, including CHI3L1, IL-10, IL-6, IL-8, and TNF-α, which are known to play roles in cancer metastasis, migration, invasion, chemotaxis, and endothelial cell junction retraction. The complexity of the cytokines in the inventors' model more closely mimics the heterogeneous tumor microenvironment in advanced non-small cell lung cancer (NSCLC), making it suitable for studying EMT-induced cancer infiltration mechanisms.

[0171] The inventors used multiple biomarkers to study successful EMT at the protein and gene expression levels. In fact, the EMT process is controlled by a variety of transcription factors, including SNAI1, SNAI2, ZEB1, TWIST, CarB-box binding factor, Mesenchyme Forkhead 1, and Kruppel-like factors. These transcriptional regulators are regulated by a complex signaling network in the tumor microenvironment, particularly through pathways involving TGF-β, Notch, and Wnt signaling. Notably, the expression of these factors follows a temporal hierarchy in the EMT process: activation of SNAI1 initiates the transformation, while subsequent induction of SNAI2, ZEB1, and TWIST is used to maintain the final migratory phenotype. In particular, ZEB1 and ZEB2 showed the strongest upregulation under EMT-IC conditions, exceeding the effect of TGF-β1 alone. This is especially important because ZEB1 and 2 are associated with invasive, stem cell-like phenotypes and immune escape. The synergistic effect of TGF-β1 on ZEB1 and 2 indicates that cancer cells exhibit a more aggressive metastatic phenotype.

[0172] In addition to the marker expression, the inventors also confirmed the success of EMT at the functional level, among which It synergistically promotes the migration and invasion potential of cancer cells with TGF-β1. In fact, Its performance was even better than that of TGF-β1, which indicates that The inherent biological complexity is favorable for inducing EMT in A549 cancer cells. It is noteworthy that, although macrophage supernatant has been reported to induce EMT in colon cancer cells, TGF-β1 and... The synergistic effect of TGF-β1 in EMT induction has been studied. Due to its biomarker expression and function, TGF-β1 and... The combination appears to work synergistically to facilitate EMT in A549 cells, therefore this combination was chosen as the EMT-IC for this microphysiological model.

[0173] Furthermore, EMT-IC directly affects the MVN, reducing its basement membrane and intercellular junction proteins, potentially making it more permissive to infiltrating cells. However, within the initial time range studied, it did not affect its overall stability or vascular barrier function against small solutes. Nevertheless, the effects of longer incubation times with EMT-IC were not tested.

[0174] The enhanced cancer cell invasive potential observed under EMT-IC conditions in type I collagen hydrogels was also reproduced in co-cultures with MVNs in fibrin hydrogels within microfluidic devices. This favorably demonstrates that A549 cells, as well as other lung cancer cell lines, undergo functional EMT in this co-culture setting, despite their metastatic predisposition. Advantageously, initiation of endosmosis was observed primarily in co-cultures exposed to EMT-ICs in metastatic lung cancer cell lines, where endosmosis events are defined as cancer cells inserting membrane protrusions into microvascular structures. In addition to the systemic effects of EMT-ICs on vascular connectivity, a reduction in vascular endothelial cadherin (VE-cadherin) was also observed, specifically at the cancer cell entry site. This suggests direct communication between EM-transformed cancer cells and endothelial cells, leading to the formation of permissible entry points on the microvascular walls. Using established microphysiological models, these interactions can be visualized and studied at high spatiotemporal resolution, and potential drug targets for anticancer therapy can be investigated.

[0175] There are various microphysiological systems involved in cancer metastasis. These primarily focus on later stages of cancer metastasis, such as extravasation events. It should be noted that physiologically representative models of cancer infiltration are noteworthy because EMT-driven cancer infiltration is the rate-limiting step in circulating cancer cells. Therefore, targeting the early stages of the metastatic cascade may be more effective in preventing cancer cell spread.

[0176] The inventors recognized that a limitation of in vitro cancer metastasis research is that quantifying cancer infiltration rates is time-consuming and requires manual intervention. For example, a similar study used a "thin-plate spline" approach to define tissue boundaries and compartments in multiple organ-on-a-chip models to quantify tumor infiltration events. Such mathematical modeling is necessary due to variations arising from organ-on-a-chip fabrication and experimental procedures. Automated image quantification using ML-based methods offers a solution, enabling unbiased analysis of large-scale data in various formats used for tumor staging, cancer susceptibility, recurrence, and patient survival prediction.

[0177] Advantageously, an image analysis workflow using readily available, non-commercial, open-source tools has been developed to reliably and efficiently quantify endosmosis events. By defining endosmosis events as the localization of fluorescently labeled cancer particles within segmented MVN regions, the inventors' automated method exhibits a trend consistent with manual counting, highlighting its reliability. Notably, according to embodiments of this disclosure, the machine learning-assisted workflow detects, on average, one more endosmosis event per sample than manual counting. This difference is likely due to the enhanced analytical capabilities of machine learning-based methods, enabling the detection of endosmosis events missed by the human eye. Furthermore, the image analysis workflow according to embodiments of this disclosure is universal because it utilizes the overlap of fluorescence signals between tumor spheroids and endothelial cells to characterize endosmosis events. The inventors have observed that the quantification of endosmosis events is similar to manual analysis and remains robust across multiple biological runs.

[0178] Furthermore, machine learning-based tools can be used for cancer research on staging, prognosis, extravasation, and lymph node metastasis. Future advancements combining physiologically relevant in vitro models with 3D microscopy, 3D image analysis, and ML-assisted analysis will benefit patient-information-based drug development, early cancer detection, and personalized treatment planning.

[0179] Furthermore, this platform technology has a significant impact by allowing for detailed study and visualization of the endosmosis process over time. It enables cell tracking and provides high-throughput capabilities for drug development. Currently, over 90% of drugs fail in clinical trials after years of costly development, primarily due to overly simplified in vitro models or non-translational animal responses. With this technology, candidate drugs and treatments can "fail quickly, fail early," thereby reducing costs and reliance on animal testing. This capability not only improves the efficiency of drug development but also aligns with ethical responsibilities.

[0180] Therefore, combining machine learning with the inventors' EMT-driven on-chip model of lung cancer endorphinization provides a physiologically relevant platform to simulate the initial process of cancer metastasis to the microvascular system. This approach holds promise for improving drug development, personalized patient treatment planning, and prognosis of cancer progression.

[0181] Machine learning-based processing engines can be implemented as computer servers or processors within external computers that process sampled data. In this embodiment, the system includes a server. The server includes suitable components required to receive, store, and execute appropriate computer instructions. These components may include: processing units, including a Central Processing Unit (CPU), a Math Co-processing Unit (Math Processor), a Graphics Processing Unit (GPU), or a Tensor Processing Unit (TPU) for tensor or multidimensional array computation or manipulation operations; read-only memory (ROM); random-access memory (RAM); input / output (I / O) devices (e.g., disk drives); input devices (e.g., Ethernet ports, USB ports, etc.); displays, such as liquid crystal displays, light-emitting displays, or any other suitable displays; and communication links. The server may include instructions that can be stored in ROM, RAM, or a disk drive and executed by the processing units. Multiple communication links can be provided, which can be connected differently to one or more computing devices, such as servers, personal computers, terminals, wireless or handheld computing devices, IoT devices, smart devices, edge computing devices, and cloud devices. At least one of the multiple communication links can be connected to an external computing network via a telephone line or other types of communication links.

[0182] The server may include storage devices such as disk drives 208, which may include solid-state drives, hard disk drives, optical drives, tape drives, or remote or cloud-based storage devices. The server may use a single disk drive or multiple disk drives, or remote storage services. The server may also have a suitable operating system residing on the disk drives or in the server's ROM.

[0183] Computers or computing devices can also provide the necessary computing power to operate or interface with machine learning networks of neural networks to provide various functions and outputs. Neural networks can be implemented locally, or they can be accessed or partially accessed through servers or cloud-based services. Machine learning networks can also be untrained, partially trained, or fully trained, and / or can be retrained, modified, or updated over time.

[0184] While not strictly necessary, the embodiments described with reference to the accompanying drawings can be implemented as an Application Programming Interface (API) or a set of libraries used by the developer, or can be included in another software application, such as a terminal or personal computer operating system or a portable computing device operating system. Typically, since program modules include routines, programs, objects, components, and data files that help perform specific functions, those skilled in the art will understand that the functionality of a software application can be distributed among multiple routines, objects, or components to achieve the same functionality required herein.

[0185] It should also be understood that any suitable computing system architecture can be used where the methods and systems of this disclosure are implemented wholly or partially by a computing system. This will include stand-alone computers, network computers, and dedicated hardware devices. When using the terms "computing system" and "computing device," these terms are intended to include any suitable configuration of computer hardware capable of implementing the described functions.

[0186] Those skilled in the art will understand that various changes and / or modifications can be made to this disclosure as illustrated in the specific embodiments without departing from the spirit or scope of this disclosure as broadly described. Therefore, this embodiment should be considered illustrative rather than restrictive in all respects.

[0187] Unless otherwise stated, any references to prior art contained herein should not be construed as an admission that the information is common knowledge.

Claims

1. A device for observing biological activity in a sample, characterized in that, include: Microfluidic devices; The microfluidic device has multiple fluid channels, including a central channel adjacent to a culture medium channel and defining a perfusionable septum therebetween; wherein the central channel is used to promote interstitial transition (EMT) induced by reagents supplied to the culture medium channel in biological cells contained within the central channel.

2. The apparatus according to claim 1, characterized in that, in, The biological cells in the central channel undergo epithelial-mesenchymal transition.

3. The apparatus according to claim 2, characterized in that, in, The central channel includes a hollow lumen vascular structure.

4. The apparatus according to claim 3, characterized in that, in, The hollow lumen vascular structure is formed through angiogenesis.

5. The apparatus according to claim 4, characterized in that, in, The hollow lumen vascular structure is formed by culturing human umbilical vein endothelial cells (HUVECs) to obtain a microvascular network of human umbilical vein endothelial cells.

6. The apparatus according to claim 5, characterized in that, in, The human umbilical vein endothelial cells and the biological cells contained in the central channel were co-cultured.

7. The apparatus according to claim 6, characterized in that, in, The human umbilical vein endothelial cells and the biological cells were co-cultured in a hydrogel complex injected into the central channel.

8. The apparatus according to claim 1, characterized in that, in, The reagent is an EMT-induced mixed complex.

9. The apparatus according to claim 8, characterized in that, in, The EMT-induced mixed complex includes transforming growth factor β1 (TGFb1) and / or macrophage conditioned medium.

10. The apparatus according to claim 1, characterized in that, in, Each of the multiple fluid channels has an inlet and an outlet.

11. The apparatus according to claim 1, characterized in that, in, The microfluidic device includes multiple partition structures for separating the central channel from the culture medium channel.

12. The apparatus according to claim 11, characterized in that, in, The partition structure includes triangular prisms arranged at regular intervals.

13. The apparatus according to claim 10, characterized in that, in, The microfluidic device includes a pair of culture medium channels that clamp the central channel.

14. The apparatus according to claim 13, characterized in that, in, The pair of culture medium channels are also connected to a common entrance.

15. The apparatus according to claim 6, characterized in that, in, The biological cells include cancer cells selected from a group consisting of A549 cells, NCI-H1975 cells, and BEAS-2B cells.

16. The apparatus according to claim 15, characterized in that, in, The cancer cells were cultured as tumor spheroids.

17. The apparatus according to claim 1, characterized in that, Also includes: An imaging device for capturing images of biological cells in the central channel to facilitate visualization and recording of endosmosis events of biological cells in the central channel.

18. The apparatus according to claim 17, characterized in that, in, The imaging device includes a microscopic imager.

19. The apparatus according to claim 18, characterized in that, Also includes: A machine learning processing engine is used to analyze the captured images to identify tumor infiltration events in the biological cells based on features extracted from the images.

20. A method for quantifying tumor endosmosis events in biological samples, characterized in that, include: Human umbilical vein endothelial cells (HUVECs) and tumor spheroids are injected and co-cultured in the central channel of the microfluidic device according to claim 7; The EMT-induced mixed complex according to claim 9 is supplied in the culture medium channel of the microfluidic device; Microscopic images of the central channel are captured at predetermined time intervals; as well as The microscopic images are processed using the machine learning processing engine of claim 19 to quantify the occurrence and characteristics of tumor intravasation events.

Images