Engineered three-dimensional microfluidic device for interrogating cell-cell interactions in the tumor microenvironment

A. Device fabrication and preparation

The experiments were conducted in a microfabricated device composed of a semicircular chamber for culture of tumor cells in 3D matrix, as shown in Figure 1. On the straight edge of the semicircle, an array of posts creates a permeable wall to contain the collagen and matrigel matrix within the culture well and supports the growth of endothelial cells in the perfused vascular compartment, as shown in Figure 1(a). Both the input and output flows in the vascular compartment were controlled independently by means of external syringe pumps connected to the fluidic channels by the inlet and outlet holes, which allowed for precise generation of interstitial flow into the semicircular chamber.

Fabrication of similar devices is described in detail elsewhere.^27,28,34 The microfluidic device was designed in AutoCAD, laser ablated with a femtosecond laser machining station, and etched in 10 M KOH solution at 80 °C for 1 h. The device was fabricated and assembled in an ISO 1000 class clean room at the Center for Laser Application of the University of Tennessee Space Institute. Two mm diameter inlet/outlet and culture-well holes were drilled through 500-μm thick fused silica wafers. One wafer was cleaved through the diameter of the well hole and half of that piece used for the semicircular culture well structure seen in Figures 1(a) and 1(e). The straight, right edge of the culture well was provided by the other wafer, whose edge presented an array of microchannels that serve as a barrier to separate the 3D matrix from the endothelial cells. At this interface, small openings of less than 6 μm wide and approximately 50 μm long were machined every 30 μm for a length of 2 mm in the area between the interstitial flow channel and the culture well of the device. The machined 500-μm thick pieces containing the well and the interstitial flow channels were then bonded to a 170-μm thick fused silica coverslip spin coated with 20 μm of polydimethylsiloxane (PDMS). The glass device was then bonded with silicone sealant to an acrylic manifold assembly that functions as a well for buffer solution and an interface to the interstitial flow and pump system. The acrylic manifold has silicone O-ring seals to the supply tubing. Interstitial flow is provided by a Harvard Apparatus Syringe Pump (Figure 1(c)).

For sterilization, the device was flushed with 70% ethanol and incubated 20 min at room temperature under a laminar flow hood. 1× PBS supplemented with 400 μg/ml penicillin, 400 μg/ml streptomycin, and 1 μg/ml amphotericin B was then flowed through the device overnight under UV light in the laminar flow hood using a. Harvard syringe pump at a flow rate of 20 μl/min. To facilitate cell adhesion, the device was coated with 200 μg/ml collagen I (high concentration, rat tail, BD Biosciences), diluted in Attachment Factor (Life Technologies) for a minimum of 1 h at 37 °C. The semicircular chamber was coated with 0.1% poly-L-lysine solution (Sigma) for 1 h at room temperature and rinsed with PBS before use.

B. Cell culture

All cell culture was performed using sterile technique under a Class IIA laminar flow hood. The human breast adenocarcinoma cell line, MDA-MB-231 transfected to express GFP, was cultured in Dulbecco's Modified Eagle Medium supplemented with 10% Fetal Bovine Serum (FBS) and 2 mM L-glutamine. Cancer-associated fibroblasts, bCAF33 and bCAF40T, and normal fibroblasts, hNAF112 and bNAF13, were isolated from breast cancer tissue or normal breast tissue and generously provided by the laboratory of H.L. Moses (Vanderbilt University) and the laboratory of S. Hayward (Vanderbilt University) as described by Erez.³⁵ Fibroblasts were cultured in MCDB-131 medium (Gibco), supplemented with 10% FBS, 1× insulin-transferrin selenium, 1× non-essential amino acids, 1.4 mM L-glutamine (Gibco), 13% amniomax basal medium (Gibco), and 2.1% amniomax C100 supplement (Gibco). HMVECad primary cells (Life Technologies) were cultured in MCDB-131 medium supplemented with Microvascular Growth Supplement (Life Technologies).

C. Tumor spheroid preparation

For cancer cell spheroids, MDA-MB-231-GFP cells were detached from monolayer culture with trypsin/EDTA, collected by centrifugation, and resuspended to 1 × 10⁴ cells/ml in DMEM (10% FBS, L-glutamine) with 20% methylcellulose medium. 1000 cells were loaded into each well of a 96 well sterile non-cell-culture-treated U-bottom plate and cultured for 2–3 days to allow formation of small spheroids for use in experiments.

D. Endothelial cell loading and culture

Endothelial cells (HMVECad) were stained with 8 μM Cell Tracker Blue in serum free MCD-131 medium for 1 h at 37 °C, followed by a 30-min incubation in culture medium not containing Cell Tracker. The cells were detached using trypsin (0.05%)/EDTA (0.53 mM) in HBSS without calcium, magnesium (Corning) added to culture medium, and centrifuged (800 g, 5 min). The cells were resuspended to a concentration of 5 × 10⁶ cells/ml in experiment medium (culture medium buffered with 25 mM HEPES). 500 μl of experiment medium was flowed through the channel and 100 μl were added to the semicircular chamber prior to loading. The cell suspension containing endothelial cells was loaded into the device using a 100-μl gas-tight syringe and a blunt 23-gauge dispensing needle. The device was left angled for 3 h to allow the endothelial cells to seed on the interface formed by the microchannels between the perfusion channel and cell-culture well.

E. Tumor spheroid and fibroblast loading

The GFP-MDA-231 cell spheroids were transferred to a 1.5 ml centrifuge tube using a cut pipette tip so as not to disrupt their integrity. They were washed twice with experiment medium (DMEM, 10% FBS, LG, 25 mM HEPES) and put on ice. The fibroblasts were incubated in 7 μM Cell Tracker Red for 1 h at 37 °C before detachment using TrypLE. Fibroblasts were collected by centrifugation (800 g, 2 min), suspended, counted, and aliquoted in the appropriate volume to deliver a final concentration of 2.5 × 10⁵ cells/ml in the matrix/gel suspension. The supernatant was aspirated from the GFP-MDA-231 breast tumor spheroids and the spheroids were then aliquoted into the fibroblast cell suspension to achieve a final concentration of 50–60 spheroids in 200 μl of three-dimensional reconstituted basement membrane with 1.5 mg/ml type I collagen and 10% Matrigel.

After the endothelial cells were allowed to adhere to the device interface and form a monolayer (3 h), the semicircular chamber was loaded with 7.5 μl of the prepared gel containing Cell Tracker Red-labeled fibroblasts and GFP-labeled MDA-231 breast tumor spheroids, allowed to polymerize, and the movement of the tumor spheroids was imaged continuously over 20 h.

F. Extracellular matrix gel preparation

The matrix/gel was prepared on ice using a 1:1 ratio of collagen solution and Matrigel solution, with a final concentration of 1.5 mg/ml collagen I (rat tail, BD Biosciences) and 10% Matrigel (BD Biosciences). The collagen was buffered with a solution of NaOH, HEPES, and 10× PBS in water. Matrigel was added, followed by cell suspension of spheroids with or without fibroblasts in experiment medium. Approximately 7.5 μl of gel was loaded into the semicircular well, and the device was incubated at 37 °C, 5% CO₂ for 1.5 h to allow polymerization, flipping periodically to ensure three-dimensionality of spheroid and cell distribution. 100 μl experiment medium was then added to the semicircular chamber prior to experiment. In the case of experiments with CXCL12, 100 μl of ligand diluted to 10 ng/ml in experiment medium or 100 μl vehicle solution without CXCL12, was added to the semicircular chamber after gel polymerization and prior to imaging.

G. Microscopy and device operation

Time-lapse video microscopy was performed on a Zeiss Axiovert 200 m microscope with an incubator chamber that kept a temperature of 37 °C. The device was imaged every 15 min over 20 h under a 10× lens in bright field and for fluorescence at excitations of 488 nm and 546 nm for visualization of GFP-expressing cancer cells and fibroblasts labeled with Cell Tracker Red, respectively. Time-lapse video microscopy was performed on a Zeiss Axiovert 200 m microscope with an incubator chamber that kept a temperature of 37 °C. The device was imaged every 15 min over 20 h under a 10× lens in bright field and for fluorescence at excitations of 488 nm and 546 nm for visualization of GFP-expressing cancer cells and fibroblasts labeled with Cell Tracker Red, respectively. Individual cell migration from the spheroids was tracked using Bitplane Imaris cell tracking algorithm. Track length, track displacement, and speed were analyzed for statistical significance in Prism (GraphPad, San Diego, CA) using ANOVA.

H. qRT-PCR analysis of Slit2, Slit3, and Robo

Total RNA was extracted from cultured CAFs, NAFs, MDA-MB-231, or co-cultures of CAFs or NAFs with MDA-MB-231 cells using RNAeasy mini kit (Qaigen) according to the manufacturer's recommendations. RNA samples were subjected to reverse transcription using iSript cDNA synthesis kit (Bio-Rad) and amplified with SYBR Select Master Mix (Life Technologies) using CFX96 Touch real-time PCR detection system (Bio-Rad). Primer sequences were SLIT3 forward CGGCATCACCGATGTGAAGAA and reverse AGGCGCAGAGTTCGGATCT; SLIT2 forward GCGAAGCTATACAGGCTTGAT and reverse TGCAGTCGAAAAGTCCTAAGTTT; ROBO1 forward GACAAAACCCTTCGGATGTCA and reverse CCAGTGGAGAGCCATCTTTCT. mRNA contents were normalized to the expression of GAPDH determined with primers forward GCCTCAAGATCATCAGCAATG and reverse CTTCCACGATACCAAAGTTGTC. For the statistical analysis of fold mRNA expression change in CAFs compared to NAFs and in CAF + CC compare to NAF + CC, a one sample T-test was performed using GraphPad Prism software and p < 0.05 was taken to indicate statistically significant differences.

A. Development of bioreactor to study the tumor microenvironment

We tested a number of reagents and procedures to determine the optimum coating of the laser etched glass that allowed the growth and adherence of endothelial cells to the machined channels, thus allowing a representation of an artificial blood vessel.³⁶ These reagents included poly-L-lysine, Attachment Factor (Life Technologies) supplemented with 50 μg/ml, 100 μg/ml, or 200 μg/ml collagen I (BD Biosciences), and a solution of 50 μg/ml collagen I and 0.1% acetic acid in Millipore water. After experimentation, poly-L-lysine was used to coat the well and the collagen I with acetic acid solution was used to coat the channel to improve cellular adhesion. In addition to providing a biologically relevant environment, these coatings could also be removed readily after experiment so that the device could be reused.

Human microvascular endothelial cells (200 cells in 80 nl) were cultured in the channel and made up the walls of the artificial blood vessel that was then perfused with medium. The differential between the quantity aspirated from the outlet and that perfused through the inlet could be altered by syringe pumps to generate precise interstitial flow into or out of the semicircular chamber, making this a versatile and biologically relevant model (Fig. 1). The key feature of this design is that the magnitude and direction of the interstitial flow within the 3D matrix can be adjusted, if desired, by controlling the small differential between the rate of fluid injection into the channel by one pump and the rate of fluid withdrawal by the other. Because we used volumetric syringe pumps rather than gravity feed, the interstitial flow through the matrix was controlled independent of any small changes in the perfusate level above the matrix. Hence, with equal push-pull syringe pumps on the input and output channels, this design will insure perfusion of the endothelial cells without driving interstitial flow. This unique and innovative design also allowed in vitro visualization of cell migration to provide insight into the mechanisms of metastasis (Figure 1).

B. Cancer-associated fibroblasts increase spheroid sprouting in vitro

Since previous studies have highlighted the influence of CAFs on the TME, experiments were performed comparing spheroid sprouting in the presence of NAFs or CAFs. Endothelial cells were cultured in the channel (shown at the arrows in Figure 1(g)), spheroids with or without CAFs were suspended in reconstituted basement membrane and loaded into the semicircular well. Imaging revealed that while in the absence of CAFs there was little dispersion of the tumor cells from the spheroid (Figure 2(a), Multimedia view), when co-cultured with CAFs there was marked dispersion of tumor cells from the spheroid along with rapid cancer cell migration when tumor spheroids were co-cultured with CAFs (Figure 2(b), Multimedia view).

C. NAFs alter the morphology of spheroid cells but do not increase the dispersion of the MDA-231 cells from the spheroid

When experiments were performed with MDA-MB-231 spheroids in the presence of normal fibroblasts (NAFs) (Figure 2(c), Multimedia view), the cancer cells did migrate, but this movement was only in close proximity to the spheroid and there was considerably less invasion of the surrounding matrix. Of note, the spheroid cells cultured with NAFs tended to form “Single file” strings of tumor cells that did not move away from the group of GFP-tagged tumor cells.

D. Replacement of cancer-associated fibroblasts with CXCL12 (CXCL12)

Cancer associated fibroblasts have been reported to secrete substantially elevated levels of the chemokine CXCL12 (SDF-1).³⁷ To evaluate whether the addition of CXCL12 could replicate the dispersion of tumor cells from the spheroid induced in response to CAFs, spheroids were incubated in the device with CXCL12 (10 ng/ml) added to the semicircular chamber to mimic the ligand in a stromally derived manner. The microfluidic device was incubated at 37 °C in the presence of 5% CO₂ for 20 h during which the migration of the cells was recorded using time-lapse video microscopy. Results showed that addition of CXCL12 to the spheroid cultures stimulated the migration of the CXCR4-expressing MDA-MB-231 cells in a manner very similar to that of the co-culture with CAFs (Figure 3(a), Multimedia view).

To evaluate whether addition of CXCL12 to spheroids cultured with NAFs would allow the NAFs to stimulate cancer cell dispersion from the spheroid in a manner equivalent to that of CAFs, CXCL12 (10 ng/ml) was added to the co-culture of spheroids and NAFs and the migration of the cells was followed using time-lapse video-microscopy. Results showed that addition of CXCL12 to NAF/MDA-MB-231 co-cultures induced little diffusion of tumor cells from the spheroid. These data suggest that NAFs produce factors that inhibit the migration of the tumor cells in response to CXCL12 (Figure 3(b), Multimedia view).

Quantification of time-lapse imaging data: We performed quantitative analysis of the time-lapse imaging data using Bitplane Imaris software and also by individually tracking the migration of individual cells over time. Since some of the cells move into and out of the plane of focus, the Imaris software data were backed up with individual tracking of cells. The distance cells move varies considerably depending upon their position in the spheroid with the cells near the edge often moving a greater distance than the cells in the middle of the spheroid, resulting in rather wide standard deviations (Table I). We did observe a trend toward greater migration of cells in the spheroid in the presence of CAFs or SDF-1 as compared to the absence of either of these or in combination with NAFs. The data show there is dispersion of the spheroid in the presence of CAFs or SDF-1 (CXCL12), but not in the absence of CAFs or SDF-1, or with NAFs.

E. Analysis of Slit2, Slit3, and Robo1 in CAFs, NAFs, and MDA-MB-231 cells and co-cultures of CAFs or NAFs with MDA-231 cells

The expression of Robo1 mRNA was not different between NAFs and CAFs or between co-cultures of NAFs and MDA-MB-231 cells versus CAFs and MDA-MB-231 cells. Interestingly, while Slit2 mRNA levels were very similar between NAFs and CAFs, when co-cultured with the MDA-MB-231 cells, the Slit2 levels in NAF co-cultures were about 2-fold higher than the CAF co-cultures. Moreover, we observed much higher levels of Slit3 mRNA in NAFs as compared to CAFs and with co-cultures of NAFs and MDA-MB-231 cells as compared to CAFs and MDA-MB-231 cells. These data suggest that when NAFs are co-cultured with MDA-MB-231 cells the inhibitory ligands, Slit2 and Slit3 are more available bind Robo1 and inhibit the migratory response to CXCL12 than in the cultures of CAFs co-cultured with MDA-MB-231 cells (Figure 4).