Device Design and Operation
The sinusoidal blood flow was mimicked using a peristaltic pump and an external medium reservoir, as we previously reported,5 to recirculate the medium (Figs. 1c and 1d) and generate τw through the rectangular sinusoid chamber (0.8 mm width and 0.137 mm depth) on the order of 0.1 Pa. COMSOL Multiphysics 5.3 was used to compute the velocity profiles and shear stress distribution in the sinusoid chamber (Fig. 2d). The medium flow was simulated as a steady state and laminar flow using the following parameters: volumetric flow rate (8.98 µL/min), density (103 kg/m3), and dynamic viscosity (0.9 mPa·s). The simulation (Fig. 2d) shows that the shear stress distribution can be controlled to be relatively uniform along the sinusoid chamber with a wall shear stress of 0.094 Pa. A bubble trap was placed before the sinusoid chamber (Fig. 2e) to mitigate the propensity for bubble introduction due to the high-velocity flow of the culture medium. It was designed to increase fluidic resistance by narrowing the chamber width while providing an additional volume above the chamber for bubbles to be stored.31
The sinusoidal endothelium was mimicked by placing a transparent polyester (PETE) membrane between the sinusoid and stroma chambers (Figs. 1b and 2a-c) and establishing a confluent layer of ECs on the bottom surface of the membrane. In this design, the membrane provided: (1) mechanical support for the stroma chamber, (2) mechanical protection of the collagen scaffold in the stroma chamber from the high-velocity flow in the sinusoid chamber, (3) a surface for EC adhesion and growth, and (4) open pores of 10 μm to allow the migration of MM cells through the reconstructed endothelium. The culture medium ports (Figs. 2a and 2c) were used to introduce ECs into the sinusoid chamber.
The stroma chamber (i.e., the yellow compartment in Figs. 1b and 2a-c) was designed to culture BMSCs dispersed in collagen (volume = 60 mL). In this design, collagen was used to represent the viscoelastic extracellular matrix of the BM. Its elongated shape provided a large surface area and surface tension to minimize the remodeling and contraction of collagen upon culture. The medium reservoir placed above the stroma chamber (Fig. 2a and Fig. 3b) was designed to: (1) provide periodic replenishment of culture medium by pipetting during long-term culture, (2) prevent collagen overflow above the stroma chamber during the BMSC seeding step, and (3) introduce stains and dyes for convenient endpoint characterization. Nine wells were used to construct and integrate the sinusoid and stroma chambers, thus enabling six culture experiments to be conducted within each microfluidic device.
The microfluidic device (Figs. 3 and S1) was fabricated using previously reported techniques.5, 7, 30 Details of the device fabrication procedures are provided in
Methods and Supplementary Information 1. As shown in Figs. 1c and 3c, 9 wells of the well-plate were used to produce one pair of sinusoid and stroma chambers and therefore 6 pairs for each device. Therefore, each device could be used to perform 6 culture experiments. As shown in Fig. 1d, 6 sinusoid chambers in each device were connected to 6 external culture medium reservoirs and 6 out of 12 flow channels available in the peristaltic pump (IP ISM942, ISMATEC) using external tubing. Fluidic connections with the devices and external tubes were made using two polydimethylsiloxane (PDMS) pieces inserted into the inlet and outlet culture medium ports to stabilize and seal the tubes (highlighted with gray in Figs. 1c and 2a). For long-term culture, the well plate and the external medium reservoirs were placed inside a conventional CO2 incubator while the peristaltic pump was located outside. The square PDMS cover, illustrated in Fig. 1c, was used to isolate and protect the cell culture areas from the incubator during culture. Real-time spatiotemporal imaging of constitutive cells in the sinusoid and stroma chambers were achieved using an inverted microscope (Nikon Ti-E) equipped with an automated stage housed in an incubator with CO2, humidity, and temperature controls.
Tissue Construction in Sinusoid and Stroma Chambers
Detailed procedures used for cell preparation and seeding are provided in Methods. In brief, ECs were prelabelled with CellTracker Green CMFDA Dye (C2925, ThermoFisher). BMSCs were prelabelled with CellTracker Red CMTPX Dye (C34552, ThermoFisher). The sinusoid chamber was seeded with ECs at the cell density of 5x106 cells/mL. The stroma chamber was seeded with 60 μL BMSC/collagen mixture. 2 mg/mL of collagen was used as an optimum concentration that provides stiffness to support stromal cells during 3D culture. The BMSC density in the collagen mixture was 0.5x106 cells/mL.
The seeded ECs and BMSCs were dynamically cultured for up to 24 days while flowing culture medium through the sinusoid chamber at a volumetric flow rate of 8.98 µL/min. The culture medium above the BMSC chamber was replenished every 3 days. After culture, cells were fixed and stained with primary mouse anti-human antibody, CD31 (platelet endothelial cell adhesion molecule-1, JC70, sc-53411, Santa Cruz Biotechnology) and DAPI (D9542, Sigma-Aldrich). CD31 was used as a maker for the formation of intercellular junctions. DAPI was used to stain cell nuclei. The stained cells were analyzed using a Zeiss LSM 880 series laser scanning confocal microscope. Some tissue samples were fixed and dehydrated by serial ethanol washing and critical point drying for examination using a Zeiss Auriga Crossbeam 40 scanning electron microscope (SEM). Detailed procedures used for cell fixing, staining, and imaging are provided in Methods.
After 4 h of static culture, ECs formed a confluent layer on the sinusoid chamber walls (Fig. S2c). As shown in Figs. S2a and S2b, we used brightfield optical imaging to characterize the pore structure of the PETE membrane and their effects on the development of ECs. Fig. S2a shows ~10 μm pores unevenly distributed on the membrane surface prior to EC seeding. Fig. S2b shows that most pores were covered by ECs after static culture for 4 h.
Fig. 4b shows the overall distribution of ECs and BMSCs cultured for 12 h in the sinusoid and stroma chambers. The cross-sectional view of the sinusoid chamber (the bottom image in Fig. 4b) shows that ECs covered all the chamber walls, forming a rectangular lumen-like structure with 80-100 μm in height and 800 μm in width. The top image in Fig. 4b shows that BMSCs were uniformly distributed in the collagen matrix of the stroma chamber. These results indicate that an endothelial lumen structure could be formed in the sinusoid chamber within 12 h while BMSCs being uniformly dispersed in collagen and co-cultured in the stroma chamber.
After 24 days of culture, the endothelial lumen structure was stained with CD31 which was strongly expressed at EC junctions (Figs. 4c and 4d). The localized expression of CD31 indicated that the endothelial phenotype was maintained along with the formation of intercellular junctions.33 The effects of flow-induced shear stress were evident on the morphological alignment of ECs (Fig. 4d) and the re-arranged cortical organization of F-actin (Fig. 4e) along the direction of flow. The effects of increasing shear stress from 0.01 to 1 Pa on the increased elongation, stronger CD31 expression, and more oriented F-actin filaments of ECs were also evident as described in Supplementary Information 3 and shown in Fig. S4. Small pores of about 1 to 8 μm were observed at some EC junctions, as indicated by the white arrows in Fig. 4d and from Supplementary Video 1. Especially, the SEM image (Fig. 4f) shows that ECs were well connected to each other, despite the formation of small pores. In this image, membrane holes were not visible, as the membrane surface was mostly covered by the EC layer.
Figs. S3a and S3b show BMSCs cultured for 24 days with collagen in the stroma chamber. In comparison to BMSCs cultured for 12 h (Fig. 4b), it was apparent that BMSCs were uniformly distributed in the stroma chamber during the 24-day period. The SEM images in Figs. S3c and S3d show the detailed morphological features of BMSCs developed within the collagen matrix. BMSCs were elongated with many dendrite-like protrusions through the collagen fibers. In the absence of any detectable migration of BMSCs towards the ECs, it appeared that these cells did not physically interact.
Barrier Function of Endothelium
CD31 expression is associated with the restrictive barrier function of endothelium in vivo.33 The barrier function of the endothelium was estimated by measuring the permeability () of a fluorescein isothiocyanate-dextran solution (70 kDa, 20 μg/mL in PBS, Sigma-Aldrich) through the sinusoid chamber (detailed procedures provided in Methods and Supplementary Information 4). Briefly, as shown in Fig. S5, fluorescent images of dextran diffusing from the sinusoid to the stroma chamber were acquired at 15 s/frame for 180 s using the Nikon Ti-E microscope. Videos were taken at 0, 2, 5, 10, 15, 20, and 30 min to follow the dextran distribution. was calculated using the dextran diffusion model previously developed by William et al.34
through the PETE membrane after BMSCs were cultured for 30 h in the collagen matrix, but without ECs in the sinusoid chamber (i.e., “BMSC” in Fig. 5a) was measured to be ~ 3.5±0.5x10-5 cm/s. When ECs were placed and cultured in the sinusoid chamber (i.e., “EC+BMSC” in Fig. 5a), was significantly decreased to about 5.8±0.2x10-6 cm/s; suggesting that the EC layer significantly increased resistance to the diffusion of dextran. In separate experiments using the same culture conditions, ECs were characterized by staining with CD31 and DAPI. Fig. 5b shows that ECs formed a uniform endothelium layer on the PETE membrane surface. Fig. 5c shows well connected junctions between ECs with the occasional presence of small pores of 1.5 to 10 μm with an average pore size of 3.4 μm.
CXCL12-Induced Egression of Cancer Cells
In order to demonstrate that the device can be used to study the trafficking of cancer cells through the sinusoidal niche, we evaluated the effects of CXCL12 on meditating the egression of MM cells from the BM stroma chamber. It has been reported that MM cells express the chemokine receptor CXCR4 and are therefore attracted to CXCL12+ cells in the BM.16 CXCL12 can also induce cytoskeletal rearrangement, pseudopodia formation, and internalization of the CXCR4 receptor in MM cells.16, 35 CXCL12 is also known to upregulate VLA-4–mediated MM cell adhesion to fibronectin and VCAM-116, 36 and increases invasion and matrix metalloproteinases (MMP) secretion.16, 37 In this study, we used the human MM.1S cell line which has been widely used to study MM and the development of drug resistance.15, 38
Fig. 6a illustrates the experimental configuration used to induce the egression of MM1.S cells from the stroma chamber to the sinusoid chamber by adding CXCL12 (R&D Systems) to the culture medium flow in the sinusoid chamber (640 ng/mL). ECs, BMSCs, and MM.1S cells were pre-labeled with CMFDA (green), CMTPX (red), and Hoechst (blue), respectively. MM.1S cells (2x106 cells/mL) and BMSCs (0.5x106 cells/mL) were thoroughly mixed into collagen before placing the mixture in the stroma chamber. The cells in the device were cultured using the 1:1 mixture of the DMEM and RPMI complete cell culture media. CXCL12 was added in the common culture medium fed into the sinusoid chamber after ECs, BMSCs, and MM.1S were cultured in the device for 26 h. Fluorescence images and videos were taken in situ using the confocal and fluorescence microscopy for the following 4 h (Supplementary Information 4). To help maintain the CXCL12 gradient, the culture medium in the culture medium reservoir was replenished every 10 min during the 4-h period.
Fig. 6b shows the migration of MM.1S cells from the stroma chamber to the sinusoid chamber as evident from the higher density of Hoechst-stained MM.1S cells toward the endothelium. In Figs. 6d and 6e, the cross-sectional confocal fluorescence views imaged without and with adding CXCL12, respectively, are compared. MM.1S cells were evenly distributed in the stroma chamber in the absence of CXCL12 (Fig. 6d) whereas more MM.1S cells were observed at the stroma/sinusoid interface with CXCL12 (Fig. 6e). In contrast, the spatial distribution of BMSCs (red cells) was not affected by adding CXCL12. Fig. 6g shows the migration data quantified by counting the number of MM.1S cells as a function of z-axis positions specified in Fig. 6b. The cell density of MM.1S cells decreased in the axial direction of the stroma chamber toward the sinusoid chamber, as expected from their migration toward the high concentration of CXCL12 in the sinusoid chamber. In contrast, the density of BMSCs did not change along the axial direction with or without CXCL12.
The presence of migrating MM.1S cells (indicated by the red arrows in Fig. 6c) in the endothelium was observed during the z-stack confocal fluorescence analysis. From Fig. 6c, the average pore size at junction gaps was measured to be 2 to 20 µm. Fig. 6f shows a representative SEM image of the endothelium with small junction gaps as well as the presence of MM.1S cells identified by their round morphology. Furthermore, time-lapse imaging confirmed the CXCL12-induced migration of MM.1S cells toward the endothelium (Supplementary Videos 2 through 6). Fig. S6 shows BMSCs and MM.1S cultured in the stroma chamber at 4 h after adding CXCL12 in the sinusoid chamber. No close contacts between MM.1S cells and BMSCs were detected. Taken together, these results suggested that MM.1S cells migrated through the endothelium in response to CXCL12 while BMSCs did not. We further evaluated the CXCL12 production by BMSCs and ECs by ELISA. As detailed in Supplementary Information 5, BMSCs and ECs produced very low levels of CXCL12 (below 100 pg/mL), supporting that: (1) MM.1S cells did not migrate to BMSCs and ECs via CXCL12/CXCR4 axis and (2) the migration of MM.1S cells was induced by the addition of CXCL12 in the sinusoid chamber.
The effects of CXCL12-mediated MM.1S cell migration on the barrier function of ECs were evaluated by measuring and characterizing the morphological features of these cells. As described in the previous section, CXCL12 was introduced for 4 h in the sinusoid chamber after BMSCs. Fig. 7a shows that significantly increased from 5.8±0.2x10-6 to 9.3±0.1x10-6 cm/s when MM.1S cells were present and their migration was induced. As indicated by the red arrows in Figs. 7c and 7d, the presence of MM.1S cells was observed throughout the endothelium layer, confirming migration. As indicated by the white arrows in Fig. 7d, junction gaps of 2 to 20 μm formed between ECs. These gaps were substantially larger than those of 1.5 to 10 μm formed in the absence of MM.1S cells (Fig. 5c). However, the density of ECs was not significantly influenced by the presence and migration of MM.1S cells (Fig. 7b), suggesting that the viability of ECs was not compromised by MM.1S migration. These results suggested that the migration of MM.1S cells was accompanied by openings in the EC junction pores and thus an increase in the of dextran through the endothelium layer without affecting the viability of ECs.