Developing a transwell millifluidic device for studying blood-brain barrier endothelium

Abstract

Blood-brain barrier (BBB) endothelial cell (EC) function depends on flow conditions and on supportive cells, like pericytes and astrocytes, which have been shown to be both beneficial and detrimental for brain EC function. Most studies investigating BBB EC function lack physiological relevance, using sub-physiological shear stress magnitudes and/or omitting pericytes and astrocytes. In this study, we developed a millifluidic device compatible with standard transwell inserts to investigate BBB function. In contrast to standard polydimethylsiloxane (PDMS) microfluidic devices, this model allows for easy, reproducible shear stress exposure without common limitations of PDMS devices such as inadequate nutrient diffusion and air bubble formation. In no-flow conditions, we first used the device to examine the impact of primary human pericytes and astrocytes on human brain microvascular EC (HBMEC) barrier integrity. Astrocytes, pericytes, and a 1-to-1 ratio of both cell types increased HBMEC barrier integrity via reduced 3 and 40 kDa fluorescent dextran permeability and increased claudin-5 expression. There were differing levels of low 3 kDa permeability in HBMEC-pericyte, HBMEC-astrocyte, and HBMEC-astrocyte-pericyte co-cultures, while levels of low 40 kDa permeability were consistent across co-cultures. The 3 kDa findings suggest that pericytes provide more barrier support to the BBB model compared to astrocytes, although both supportive cell types are permeability reducers. Incorporation of 24-hour 12 dynes per cm² flow significantly reduced dextran permeability in HBMEC monolayers, but not in the tri-culture model. These results indicate that tri-culture may exert more pronounced impact on overall BBB permeability than flow exposure. In both cases, monolayer and tri-culture, flow exposure interestingly reduced HBMEC expression of both claudin-5 and occludin. ZO-1 expression, and localization at cell-cell junctions increased in the tri-culture but exhibited no apparent change in the HBMEC monolayer. Under flow conditions, we also observed HBMEC alignment in the tri-culture but not in HBMEC monolayers, indicating supportive cells and flow are both essential to observe brain EC alignment in vitro. Collectively, these results support the necessity of physiologically relevant, multicellular BBB models when investigating BBB EC function. Consideration of the roles of shear stress and supportive cells within the BBB is critical for elucidating the physiology of the neurovascular unit.

Fig. 1

Blood brain barrier millifluidic device. (A) Schematic of the blood–brain barrier in vivo. (B) Expanded view of the blood–brain barrier transwell millifluidic device. (C and D) Outline of the transwell plating protocol for both (C) flow experiments and (D) experiments performed in static conditions. (E) Views of the computational model demonstrate that the HBMECs directly contact fully developed flow patterns and shear stress levels that are at 12 dynes per cm², comparable to flow experienced by ECs in vivo. In the view on the right, the concentric circles on the center of the image indicate the location of the transwell insert. (F) Live tracking of fluorescent microspheres validated the expected flow patterns predicted by the computational model. (G) Schematic of the TEER protocol used to analyze HBMEC monolayer integrity. (H) HBMEC TEER values in static conditions over a five-day period demonstrate peaked barrier integrity following the third day of culture.

Fig. 2

Human brain endothelial cells (HBMECs) do not align in the direction of flow while human aortic endothelial cells (HAECs) align parallel to flow. (A) Representative images are shown of HBMEC cells after 24 hours of flow exposure at 12 dynes per cm². HBMECs do not readily align in the flow direction and may align perpendicular to flow when cultured on the transwell in a monolayer. (B) HAEC cell alignment in the direction of flow after 24 hours of flow exposure at 12 dynes per cm² in the transwell millifluidic device demonstrates parallel alignment.

Fig. 3

Validation of transwell BBB model. En face fluorescent images showing the (A) human brain microvascular endothelial cells and (B) the human brain pericytes and astrocytes within the co-culture BBB model. Three-dimensional perspective of the transwell BBB co-culture via the (C) endothelial cell side and (D) the pericyte–astrocyte side. (E) Higher magnification imaging demonstrates (top) strong junctional PECAM-1 signal (white arrows) in HBMECs, (bottom left) GFAP expression with identifiable processes in astrocytes (dashed yellow lines), and (bottom right) NG2 expression in pericytes.

Fig. 4

Dextran permeability analysis identifies astrocyte and pericyte roles in lowering BBB. (A) Schematic of the dextran permeability assay in static-conditioned cultures. (B) In comparison to HBMEC monolayers (E), 40 kDa dextran permeability analysis demonstrates a decrease in permeability in EC/PC (EP), EC/AC (EA) and EC/PC/AC (EPA) co-cultures when normalized to the average value of the HBMEC monolayer (E). (C) While the 3 kDa dextran permeability analysis still exhibits a decrease in permeability of the three cultures compared to the HBMEC monolayer (E), there is also a significant increase in permeability of the EA culture compared to EP and EPA and a significant decrease in permeability of the EP culture compared to EA and EPA. (D) Permeability coefficients (cm s⁻¹) of 40 kDa and 3 kDa dextran for E, EP, EA, and EPA cultures. Data corresponds to Fig. 4B and C (for Fig. 4B and C, * denotes p < 0.05; ** denotes p < 0.01; and *** denotes p < 0.001).

Fig. 5

Dextran permeability shows that flow exposure improves barrier integrity in HBMEC monolayers but not co-culture BBB models. (A) Schematic of the dextran permeability assay in flow-conditioned cultures with opposite cellular orientation. (B) 40 kDa dextran permeability analysis demonstrates a statistically significant decrease in permeability in E flow conditions vs. static but no apparent difference in EC/PC/AC (EPA) co-cultures. Raw permeability values were normalized to static controls to account for differences in cell passage number and other experimental factors. (C) Similarly, 3 kDa dextran permeability analysis demonstrated the same trend of a significant reduction in BBB permeability in E flow conditions vs. static conditions. (D) Permeability coefficients (cm s⁻¹) of 40 kDa and 3 kDa dextran for E static, E flow, EPA static, EPA flow conditions. Data corresponds to Fig. 5B and C (for Fig. 5B and C, * denotes p < 0.05).

Fig. 6

HBMEC (E) occludin appears to be the most statistically significantly regulated protein in static conditions, particularly due to co-culture of the HBMECs with pericytes. HBMEC (E) ZO-1, claudin-5, and occludin are all statistically significantly regulated in flow conditions when HBMECs (E) are co-cultured with both ACs and PCs (to form the EC/PC/AC (EPA) BBB model). (A) Western blot quantification of HBMEC ZO-1, occludin, claudin-5, VE-cadherin, and caveolin-1 expression in HBMEC monolayers (E), EC/PC (EP) co-cultures, EC/AC (EA) co-cultures, and EC/PC/AC (EPA) co-cultures. Occludin expression is significantly reduced in both EC/PC (EP) and EC/PC/AC (EPA) conditions. (B) Representative western blot bands. (C) Western blot quantification of HBMEC ZO-1, occludin, and claudin-5 in both EC/PC/AC (EPA) co-cultures and HBMEC monolayers (E) after flow exposure. Dotted line represents static condition values by which the experimental data is normalized. Flow exposure increases ZO-1 expression in the tri-culture model but reduces claudin-5 and occludin in both the tri-culture and HBMEC monolayer models. (D) Representative western blot bands (for Fig. 6A and C, * denotes p < 0.05; ** denotes p < 0.01; *** denotes p < 0.001; and **** denotes p < 0.0001).

Fig. 7

Immunocytochemistry (ICC) of ZO-1, claudin-5, and WGA demonstrate strong junctional localization and varying GCX expression; cell orientation with respect to flow direction is also clarified. (A) ZO-1 and claudin-5 immunocytochemistry in HBMEC monolayers (E) and EC/PC (EP), EC/AC (EA), and EC/PC/AC (EPA) co-cultures in static and flow conditions. Strong junctional signal can be seen in all samples. ICC of HBMEC monolayers (E) and EC/PC/AC (EPA) co-cultures exposed to flow highlight reduced claudin-5 expression and increased ZO-1 junctional thickness following flow. It can also be observed that the ZO-1/claudin-5-stained HBMECs in the flow conditioned EC/PC/AC (EPA) co-culture exhibit the most prominent alignment with the direction of flow, when compared to other conditions (e.g., monoculture or static). (B) WGA staining of EC (E), EC/PC (EP), EC/AC (EA), and EC/PC/AC (EPA) in static and flow conditions demonstrates varying GCX abundance in the different conditions. Tri-culture and/or flow conditions reveal increased WGA fluorescent intensity and GCX thickness.