A high-throughput microfluidic bilayer co-culture platform to study endothelial-pericyte interactions

Abstract

Microphysiological organ-on-chip models offer the potential to improve the prediction of drug safety and efficacy through recapitulation of human physiological responses. The importance of including multiple cell types within tissue models has been well documented. However, the study of cell interactions in vitro can be limited by complexity of the tissue model and throughput of current culture systems. Here, we describe the development of a co-culture microvascular model and relevant assays in a high-throughput thermoplastic organ-on-chip platform, PREDICT96. The system consists of 96 arrayed bilayer microfluidic devices containing retinal microvascular endothelial cells and pericytes cultured on opposing sides of a microporous membrane. Compatibility of the PREDICT96 platform with a variety of quantifiable and scalable assays, including macromolecular permeability, image-based screening, Luminex, and qPCR, is demonstrated. In addition, the bilayer design of the devices allows for channel- or cell type-specific readouts, such as cytokine profiles and gene expression. The microvascular model was responsive to perturbations including barrier disruption, inflammatory stimulation, and fluid shear stress, and our results corroborated the improved robustness of co-culture over endothelial mono-cultures. We anticipate the PREDICT96 platform and adapted assays will be suitable for other complex tissues, including applications to disease models and drug discovery.

Figure 1

Development of the microfluidic microvascular co-culture model in the PREDICT96 platform. (A) Side-view schematic of the microvasculature. Capillaries are surrounded by pericytes which, under healthy conditions, help stabilize and mature the endothelium. (B) Cross-section schematic showing interaction of endothelial cells (ECs) and pericytes (PCs) through basement membrane. (C) Side-view cross-section schematic of the vascular model in the bilayer microfluidic device. ECs and PCs are cultured on either side of a microporous membrane coated with extracellular matrix, which allows interaction between the two cell types. (D) Top-view of the PREDICT96 plate, containing an array of 96 bilayer microfluidic devices that interfaces with a 384 well plate top. (E) Schematic of PREDICT96 custom pneumatic pump lid, containing 192 individual pumps that control fluid flow in each channel of the 96 bilayer devices. (F) A single PREDICT96 device corresponds to 4 wells of the 384 well plate with architecture allowing for culture and fluid flow in separate top and bottom channels, which overlap in the device center. (G) Top-view bright field image of a single device, with channel overlap area indicated as the region of interest (dotted rectangle). (H) Representative image of ECs stained for PECAM-1 (green) and Hoechst (blue) in channel overlap area.

Figure 2

Long-term co-culture viability is high in PREDICT96 devices. Representative co-culture devices assessed for viability at (A) day 7 and (B) day 14 with split channels in descending order: red, green, blue, overlay. (C) Representative co-culture device stained for actin fibers (Phalloidin) and ECs (PECAM-1) on day 14 shows good coverage for both cell types. (D) Viability was greater than 95% at days 7 and 14. (E) Total cell numbers as determined by nuclear counts in channel overlap at day 7 and 14 did not differ significantly. N = 2 at day 7; N = 4 at day 14.

Figure 3

Permeability assays in PREDICT96 vascular models. FITC-dextran was loaded at a concentration of 50 μg/ml into the top channel of devices and transfer to the bottom channel was tracked over time. Note that equilibrium between both channels would be at 25 μg/mL (indicated by black dotted line). Both EC mono-culture (Mono) and co-culture with PC (Co) were assessed. Treatment with cytochalasin B (CB) was used to disrupt the barrier (added at t = 0). Transfer of (A) 20 kDa FITC-dextran and (B) 70 kDa FITC-dextran across the vascular barrier was measured at various intervals over 260 min. (C) Permeability coefficients were calculated at the 260 min time point for each condition assayed in the PREDICT96 plate. N = 8–10 technical replicates per condition. Asterisks in A and B indicate significant difference in + CB treated conditions compared to -CB controls at corresponding time points.

Figure 4

Quantification of EC monolayer coverage in co-culture media screening. (A) Representative outputs from custom code for conditions with “good” endothelial cell coverage and “poor” coverage. (B) Plate map for conditions assessed: 10 media formulations with decreasing serum concentrations (red gradient) or 0.2% serum and decreasing supplements (blue gradient). The top half of the plate assessed co-culture (rows A–D) with corresponding mono-culture conditions in the bottom half of the plate (rows E–H). (C) Automated coverage and nuclear count measurements for channel overlap area in devices across the PREDICT96 plate. Note improved coverage in co-culture (rows A–D) vs. mono-culture (rows E–H) conditions. Note that column 12 was left blank for permeability assay and therefore no cells were detected (ND). (D) EC coverage as a function of media formulation (N = 4 per condition). The ten unique medias are indicated as 1 being the most rich and 10 as the most stripped down. Co-culture with PC maintained EC monolayers even in severe starvation formulations.

Figure 5

Detection of channel- and cell-type specific responses in PREDICT96 co-cultures. (A) Select data shown from 16-plex Luminex kit run on media collected from EC and PC channels after 24 h. Note that some secreted factors are expressed in both channels while others are differentially expressed in EC or PC channels. (B) Gene expression for cells isolated from EC and PC channels of devices show that EC markers are significantly enriched in EC channels, while PC markers are significantly enriched in PC channels. Note that IL-6 is expressed by both cell types (secreted and gene). Enrichment of EC markers in the EC channel was significantly decreased in co-culture compared to mono-culture, indicated potential inter-channel crosstalk.

Figure 6

Co-culture model response to inflammatory insult—cytokine profiles from EC channels. Select secreted cytokines in media isolated from the EC channel were analyzed by Luminex, with a comparison of EC mono- vs. EC co-cultures in acute and chronic TNFα stimulation conditions. N = 3 per condition.

Figure 7

Vascular model response to low fluid shear stress (FSS). (A) Representative images of devices 24 h after exposure to low FSS. (B) Higher magnification shows some alignment of EC in response to fluid flow. (C) Gene expression of ECs from mono- and co-cultures showed that IL-6 was significantly increased in flow, while VE-CAD expression was significantly decreased. IL-6 was also significantly increased in co-culture compared to mono-culture under flow.