Human Umbilical Vein Endothelial Cells (HUVECs, CC-2519) were obtained from Lonza (Basel, Switzerland); Endothelial Cell Growth Medium (EGM; PM211500) was purchased from Genlantis (San Diego, California). Dulbecco’s Modified Eagle Medium (DMEM, 10-013-CV), Fetal Bovine Serum (FBS, 35-016-CV), and Penicillin-Streptomycin (Pen-Strep, 15140-148), Human Plasma Fibronectin (341635), Matrigel (47743-722), CellTiter-Blue Reagent (G808A), Phosphate Buffered Saline (PBS; SH30356.01), 75 cm² Culture flasks (130190), 4 kDa Texas Red Dextran (T1037-100MG), PTX (Molecular Weight: 853 Da) (T1912-5MG), 0.25% Trypsin-EDTA (25200-056), Anti-Hu CD144 (VE-Cadherin; 14-1449-82), Goat anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody Alexa Fluor Plus 488 (A32723), Fixative solution (FB002), Triton X-100 (648463-50ML), Normal Goat Serum (31872), NucBlue Live Cell Stain Ready Probes (R37605), SlowFade Gold antifade reagent (S36936), Dulbecco’s Phosphate Buffered Saline/Modified (DPBS 1X; +Calcium, +Magnesium; SH30264.01), CellTracker CM-DiI (C7001), CellTracker Green CMFDA (C7025), Dimethyl Sulfoxide (DMSO, CAS 67-68-5), 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (M6494), and Isopropanol, 70% v/v (A459-4) were purchased from Fisher Scientific (Hampton, New Hampshire).

The TPH104c (Molecular Weight: 390.47 Da) compound was synthesized in Dr. Tiwari’s laboratory, and SUM159PTX (TNBC) cell line was developed in collaboration with Dr. Tiwari’s lab (Sridharan et al., 2019). SUM159PTX (parent cell line SUM159PT) is a highly metastatic, drug-resistant triple-negative breast cancer cell line. The microfluidic devices (Figures 1A–C) were purchased from SynVivo Inc. (SynTumor 102012, Huntsville, Alabama). HUVECs were cultured with EGM, and DMEM supplemented with 10% FBS and 1% Pen-Strep were used to culture cancer cells. All cells were grown in tissue culture flasks inside a 37°C incubator with 5% CO₂ and 95% humidity. Cells were sub-cultured when they reached 80%-90% confluency. HUVECs were used between Passages one to five, whereas SUM159PTX was utilized until passage 20.

The microfluidic device (SynTumor 102012, Figures 1A–C) was used to create the tumor microenvironment. The device constructed using soft lithography technique (Tang et al., 2017; Soroush et al., 2018; Soroush et al., 2019), comprises a polydimethylsiloxane (PDMS) based, disposable, and optically clear microfluidic chip that contains a tissue compartment for the culture of breast cancer cells and ECs cultured in the surrounding vascular channels. The two compartments communicate via microfabricated 8 μ m porous structures. The microfluidic device is bonded to a microscope slide for easy imaging and real-time monitoring of the cells.

The ECs were seeded in the two vascular channels with flow perfusion, whereas the tissue compartment was seeded with drug-resistant TNBC embedded in Matrigel for tumor formation in 3D. There is continuous media perfusion through the vascular channels in this platform, mimicking the blood flow and the exchange of nutrients, oxygen, and metabolites. The continuous medium perfusion was generated by connecting the device to an external automated syringe pump (PHD ULTRA, Harvard Apparatus, 70-3007). The induced perfusion shear stress was about 0–1.75 dyne/cm², sufficient to reflect the physiological blood flow in the microvasculature.

A flowchart detailing the creation of the model is illustrated in Figure 2. To establish the co-culture, the microfluidic devices were rinsed with deionized (DI) water and degassed using a pneumatic primer (N₂ gas at 7 psi for 15 min) to eliminate air bubbles. After degassing, the tissue compartment was pre-treated with 5% Matrigel solution (in serum-free DMEM) followed by 20 min incubation at 37°C. The devices were then perfused with human fibronectin (200 μg/mL in PBS) and incubated for 1 hour at 37°C to produce a homogenous coating to promote cell attachment. All devices were rinsed with fresh EGM before cell seeding.

HUVECs (3.5 × 10⁷ cells/mL), suspended in EGM supplemented with 20% FBS, were introduced into the vascular channels via a precision syringe pump (Pump 11 Elite, Harvard Apparatus) at a 5 μL/min flow rate. The exact process was repeated to seed HUVECs in the other vascular channel. The microfluidic devices were placed in the cell culture incubator for 4 hr to facilitate cell attachment. Afterward, the vascular channels were manually rinsed with EGM to flush out debris and unattached cells. After initial seeding, the HUVECs were subjected to continuous media perfusion to mimic the in vivo flow shear conditions. The perfusion level in both vascular channels is controlled simultaneously by a precision syringe pump (PHD ULTRA, Harvard Apparatus) with a multirack attachment. Tygon Microbore tubing with an outside diameter of 0.06 inch (1.524 mm) and inner diameter of 0.02 inch (0.508 mm) served as the connecting ports for the fluidic and pneumatic interface. A stepwise linearly increased shear stress scheme was applied to facilitate EC physiological function and lumen formation (no shear, 4 h; expansion, 0–1.75 dyne/cm², 24 h; quiescent culture, 1.75 dyne/cm² maintained till the conclusion of the experiment).

Briefly, TNBC cells (10⁷ cells/mL) resuspended in 5% Matrigel/serum-free media was introduced into the tissue compartment using a precision syringe pump (Pump 11 Elite, Harvard Apparatus). The TNBC cells were seeded 24 h after HUVEC seeding using our established methods (Yang et al., 2021). The devices were then kept inside incubator for another 48 h for the tumor to develop, after which the device was ready for further testing described below. The cancer cells in the tissue compartment were perfused via bolus injection (10 min media perfusion at 1 μ L/min) every 8 h throughout the experiment.

Fluorescence and bright field images were acquired using an Olympus IX71 inverted microscope equipped with a Proscan XYZ 3D automated stage and a Hamamatsu ORCA Flash 4 camera. Olympus cellSens Dimension with Multi-Position Solutions software package and NIH ImageJ software were used to post-process captured images and videos. Each device was imaged daily using the stitch image feature of the CellSens software at 10X or ×4 magnification up to the point of the experiment. The detailed image acquisition procedure for each experiment is described below. This microscope system was used for on-chip drug efficacy study including permeability measurement (Figure 7), cell viability and proliferation assay (Figures 5, 6), intravasation assay (Figure 8) as well as immunofluorescence staining of VE-Cadherin and phalloidin in ECs (Supplementary Figure S1).

Confocal images were acquired using Leica TCS SP5 laser scanning confocal microscope with a 20X dry 0.7NA (Numerical Aperture) objective. Images were acquired in the XYZ plane with a tile scan of 512 × 512 pixels. Imaging was done sequentially with excitation and emission of Alexa fluor 488 and CM-DiI 553 at 488_ex/518_em and 488_ex/570_em and DAPI at 400_ex/460_em in 3 μ m steps. Images represent the projections of z stacks in xyzs mode. This microscope system was used to image the entire device with cells in different z planes (Figure 4).

The efficacy of TPH104c against drug-resistant TNBC cells was tested in the microfluidic device and compared to the most common FDA-approved chemotherapeutic for TNBC cell treatment, PTX. TPH104c (0.1, 1, 10 µM in EGM) or PTX (0.1, 1, 10 µM in EGM) were introduced into the vascular channels using a precision syringe pump (PHD Ultra with a multirack attachment, Harvard Apparatus) to mimic the intravenous route of drug delivery in vivo. The vascular channels were perfused with a drug solution for 72 h where each device was imaged daily. After 72 h of drug treatment, permeability assay and cytotoxicity assays were performed to quantify the effects of drug treatments on HUVEC barrier integrity and cytotoxicity in both cells.

To characterize the effects of drug treatments on EC barrier integrity, the permeability of the endothelial barrier was quantified by measuring the diffusion of 4 kDa TRITC dextran from vascular channels to the tissue compartment after 72 h of drug treatment. Following our already published protocol (Tang et al., 2017), the vascular channel of the microfluidic device was connected to a Hamilton syringe (GASTIGHT 1001) filled with dye solution and driven by a programmable syringe pump (Pump 11 Elite, Harvard Apparatus). The microfluidic device was then mounted onto the automated stage of the Olympus microscope system (Olympus IX71). At the time of the experiment, the dye solution was prepared using EGM to get a final concentration of 0.25 mM. The dye solution was then set to perfuse in the vascular channel with a flow rate of 1 μL/min (1.75 dyn/cm²) for 1 h. In the meantime, time-lapse images were taken every 1 min over 1 h using cellSens Dimension software and the Olympus microscope system. Region of Interests (ROIs) were drawn (Supplementary Figure S2) and the obtained intensity values were analyzed, and the permeability was calculated with the following equation: P = 1 − H C T   1 I 0 d I d t V S

H _CT = hematocrits in the vascular channel. In our case, H_CT = 0 (EGM does not contain blood cells). I = average intensity in the tissue compartment at a given time point, I ₀ = maximum fluorescence intensity in the vascular channel, and V/S = volume to surface area ratio of the vascular channel (Tang et al., 2017; Tang et al., 2018; Soroush et al., 2020; Yang et al., 2021).

CellTiter-Blue Cell Viability Assay was performed to determine the drug efficacy on cell viability and proliferation on-chip according to the manufacturer’s instructions. Briefly, the device was perfused with CellTiter-Blue reagent at a ratio of 1:5 with EGM and incubated for 4 h at 37 °C after 72 h of drug treatment. After incubation, the device was first placed in an orbital shaker for 15 s, then mounted to the Olympus IX71 microscope for fluorescence imaging. Finally, the images were loaded into the cellSens Dimensions software, and the intensity values were obtained, correlating with cell viability. In addition, standard MTT assay were performed in HUVECs and TNBC cells to determine their viability following TPH104c or PTX treatment. The assay was carried out on 96-well plates as per the manufacturer’s instructions.

TNBC intravasation from the tissue compartment, across the EC barrier, into the vascular channels was monitored by fluorescence imaging. For this purpose, CellTracker CM-DiI dye, which is non-toxic, well retained, and allows for multigenerational tracking of cellular movements (Krishnamurthy et al., 2008), was used. Briefly, TNBC cells were stained with CM-DiI dye before they were introduced into the tissue compartment. Meanwhile, HUVECs were stained with CellTracker Green CMFDA dye for better visualization of TNBC intravasation. Fluorescence images were acquired using the Olympus IX71 microscope system, as explained above. Images were then analyzed in cellSens Dimension software for quantifying intravasation. This was achieved using the “Count and Measure” function in cellSens Dimension software, where ROIs were drawn in different areas of the image as shown in Supplementary Figure S3. The obtained cell numbers were then analyzed and quantified (Detailed steps are given in Supplementary Material).

The formation of endothelial cell-to-cell adherens junctions was characterized using immunostaining against VE-cadherin. The devices were gently perfused with PBS. Cells were fixed using 4% formaldehyde at 4 °C for 10 min, after which it was rinsed twice with PBS and permeabilized with 0.1% Triton X-100 for 10 min at 4 °C. The device was again rinsed twice with PBS, followed by blocking with 5% normal goat serum diluted in PBS for 1 h at 37 °C. Monoclonal primary antibodies against VE-Cadherin were applied. After overnight incubation at 4°C, the corresponding fluorophore-tagged goat anti-mouse secondary antibodies were applied to the device for 1 h at 37 °C. Finally, the device was rinsed, counterstained with NucBlue (Hoechst 33342), and imaged using the Olympus system with cellSens Dimension software (Supplementary Figure S1) or using the Leica confocal microscope system (Figure 4).

Experiments were repeated in triplicates. Statistical analyses were conducted using GraphPad Prism software version 9.5.1. A factorial design was used to determine the effects of each treatment on each cell type. Results were reported as mean ± standard deviation (SD) and analyzed by Analysis of Variance (ANOVA) (Gallego-Jara et al.). Significance levels were set at α = 0.05.

A vital goal of this study is to mimic the tumor microenvironment to allow rapid signaling communication between the ECs and tumor cells while keeping them in separate compartments due to their very different culturing requirements. For this reason, a microfabricated porous barrier architecture (Figures 1B, C) based microfluidic device was used. The porous architecture (Prabhakarpandian et al., 2015; Tang et al., 2017), provides the communication area between the tissue area and the vascular channels.

Co-culturing of tumor cells and ECs was achieved by first establishing ECs in the vascular compartment (Figures 3A–C), followed by culturing tumor cells in the tissue compartment (Figures 3D–F). Variable perfusion rates, as described previously, mimicking the blood flow pattern, were applied to vascular channels after HUVEC seeding to ensure dynamic shear stimulation of ECs to recapitulate the in vivo state. After 24 h of flow perfusion, HUVECs aligned along the direction of the fluid shear (Figures 3C–F; Supplementary Figure S1), which mimics the normal endothelium observed in vivo and is widely understood to be a defining morphological feature of the vascular homeostasis (Steward et al., 2015). After 72 h of flow perfusion, HUVECs formed a complete 3D lumen (Figure 4), covering all four sides of the vascular channels. This complete lumen is organized through strong intercellular contact as indicated by a prominent, continuous adherent junction signal via VE-cadherin staining (Figure 4; Supplementary Figure S1). VE-cadherin is vital for maintaining and controlling EC contacts, regulating vascular permeability, leukocyte extravasation, and various other cellular processes such as proliferation, apoptosis, growth factor receptor functions, and angiogenesis (Hynes, 1992; Kemler, 1992; Lampugnani et al., 1992; Gumbiner, 1996; Hendrix et al., 2001). Overall, these morphological features suggest the formation of a functional EC barrier under flow stimulation. Due to the unique side-by-side architecture, each cell type and its interactions were accessed simultaneously. After seeding (Figure 3D), TNBC cells quickly occupied the entire tissue compartment within 24 h (Figure 3E). Confocal imaging showed a homogeneous distribution of TNBC cells along the z-axis of the tissue compartment (Figures 4A–D ), suggesting 3D tumor formation.

A key feature of the tumor microenvironment model is its ability to mimic drug transport from the delivery site to the target tissue. In our microfluidic co-culture model, drug treatments successfully reached the tumor (central tissue compartment) from their initial delivery site (the vascular channels), resulting in concentration-dependent cancer cell killing and varying levels of EC damage (Figure 5). TPH104c treatment resulted in elongation and swelling in TNBC cells (Supplementary Figure S4). In contrast, cellular shrinkage and condensation after PTX treatment were observed (Supplementary Figure S4), which are typical hallmarks of apoptotic cell death (Elmore, 2007). This observation follows our previous findings, which suggest an necroptotic cell death induced by TPH104 (Tukaramrao et al., 2021), the parent compound of TPH104c.

Image analysis suggests significantly less EC damage by TPH104c treatment than PTX. Morphologically, the endothelial monolayer was largely intact after THP104c treatment at all concentrations tested, whereas, after PTX treatment, empty spots in the endothelial monolayer appeared at 0.1 µM concentration (Figure 5A). This loss of endothelial monolayer after PTX treatment was even more prominent at higher concentration treatments, where the endothelial monolayer became intermittent (1 µM) and completely disconnected (10 µM).

Quantitative cell viability assessment using the CellTiter-Blue assay revealed a similar trend (Figure 5B). TPH104c treatment at 0.1, 1, and 10 µM concentrations resulted in 15%, 33%, and 55% less HUVEC viability than control (co-culture, no drug treatment), respectively. In contrast, PTX treatments resulted in 22%, 40%, and 69% less HUVEC viability than the control. These findings suggest that TPH104c could be more favorable for intravenous drug delivery compared to PTX, owing to its capacity to cause minimal damage to the EC barrier. In contrast to the outcomes derived from the microfluidic model, MTT assays performed in HUVECs in well plate culture reveal a distinct disparity in the dose-response curves (Supplementary Figure S5). While PTX exhibited notably higher toxicity than TPH104c at lower concentrations (IC₅₀ values of 0.075 and 5.34 µM respectively), there is no discernible difference in efficacy between the two treatments at higher concentrations.

In TNBC cells, both treatments exhibited a concentration-dependent drug action in the microfluidic model, with PTX exhibiting slightly higher cytotoxicity in TNBC cells (Figure 5C). MTT assays conducted in well plate culture indicate comparable cytotoxicity levels between PTX and TPH104c, with IC₅₀ values of 5.12 µM and 7.41 µM respectively (Supplementary Figure S6). This consistency in response corresponds with our findings from the microfluidic platform (Figure 5C).

Due to the fast permeation of CellTiter-Blue dye from the tissue compartment to the vascular channels, anti-TNBC efficacy assessment becomes challenging (as the dye enters the vascular channels and stains ECs, the number of ECs present can affect the final reading. In this case, since more ECs survived treatment with TPH104c compared to PTX, the TPH104c treatment group may produce a more false-positive signal due to the increased number of stained ECs). To solve this problem, a secondary fluorescence-based cell assay was performed. TNBC cells were stained with CM-DiI (Red) while HUVECs were stained with CMFDA (green) (Figure 6A). The efficacy of both drug treatments was then directly assessed by quantifying the number of surviving cells in the tissue compartment using CM-DiI fluorescence. As shown in Figure 6, both treatments exhibited concentration and time-dependent drug action. Meanwhile, no significant efficacy difference was detected between the two treatments (Figures 6B, C). Overall, TPH104c is equally potent in killing TNBC cells (Figures 5, 6) and, at the same time less toxic to ECs (Figure 5).

Endothelial barrier function is important in maintaining the proper functioning of the circulatory system and preventing the infiltration of harmful substances into tissues (Mehta and Malik, 2006; Stan, 2007). It also plays a critical role in preventing cancer metastasis (Chen et al., 2013; Wettschureck et al., 2019; Liang et al., 2021). EC barrier function is maintained by cellular junctions between adjacent ECs, which limit the movement of substances through the intercellular spaces (Dejana et al., 2008). The barrier’s permeability can be altered by various factors, such as inflammation, injury, or disease, leading to the breakdown of the barrier and the leakage of fluids and substances into surrounding tissues (Mehta and Malik, 2006; Lampugnani, 2010). Permeability assays (Figure 7) were performed using 4 kDa dextran dye in EGM media and perfused for 1 h at 1 μ L/min flow rate (1.75 dyne/cm²), taking a time-lapse video every min. We measured permeability values of (2.07 ± 0.1042) × 10^–5 cm/s in the empty device (no cells), (1.05 ± 0.0195) × 10^–6 cm/s in the EC only control group (no TNBC, no drug treatment), and (2.3567 ± 0.1002) × 10^–6 cm/s in the co-culture control (no drug treatment). This value is similar to reported in vivo values (Yuan et al., 2009; Shi et al., 2014) and is tighter than other reported in vitro models (Kazakoff et al., 1995; Lee et al., 2014; Frost et al., 2019).

Compared to co-culture control, a concentration-dependent permeability increase was observed after PTX or TPH104c treatment (Figures 7B, C). However, the endothelial barrier integrity was better preserved after the TPH104c treatment than PTX. The permeability values after 0.1, 1, or 10 μ M TPH104c treatment increased by 4%, 30%, and 55%, respectively, compared to co-culture control. However, the permeability values after 0.1, 1, or 10 μ M PTX treatment increased by 26%, 65%, and 101%, respectively, compared to co-culture control (Figure 7D). This result is consistent with our previous cell morphological analysis (Figure 5A) and viability assay results (Figure 5B).

Tumor cell intravasation is the process by which cancer cells escape from the primary tumor and enter the bloodstream (Sigdel et al., 2021). This allows cancer cells to travel to distant sites, forming metastases. Intravasation is a critical step in cancer progression and the formation of distant metastasis. Intravasation is regulated by complex interactions between tumor cells and the surrounding microenvironment, including the ECs forming blood vessel walls. Using CM-DiI fluorescent dye and time-lapse imaging, we monitored TNBC cells’ intravasation at various time points during the drug treatment in the tumor microenvironment. TNBC cell intravasation was first observed 48 h post-seeding (the day treatment starts). The number of intravasated cells increased steadily in the next 72 h (Figure 8A). A negative correlation was observed between drug treatment concentration and the number of intravasated TNBC cells.

The impact of drug treatments on TNBC intravasation is complicated by the decreased number of TNBC cells in the tissue compartment following drug treatment. To isolate the effect of drug treatments on TNBC intravasation, we analyzed the data by normalizing the number of intravasated cells to the number of cells remaining in the tissue compartment. This allowed us to obtain “intravasation percentage” data that revealed the portion of TNBC cells that were intravasated out of the total TNBC cell population at a given time. TPH104c treatment exhibited a concentration-dependent inhibition effect on TNBC intravasation (Figure 8B). In contrast, interestingly, the intravasation percentage of TNBC cells increased after PTX treatment at 0.1 or 1 µM. The exact mechanism for this increase is yet to be elucidated. However, it is reasonable to assume that the higher percentage of TNBC cells that were able to intravasate was due to the weakening of the endothelial barrier by PTX treatment, which reduces EC viability. At this point, significant intravasation into the vascular channels had occurred, as indicated by Figure 8B.