Endothelial cell behaviour within a microfluidic mimic of the flow channels of a modular tissue engineered construct

Abstract

To study the effect of disturbed flow patterns on endothelial cells, the channels found within a modular tissue engineering construct were reproduced in a microfluidic chip and lined with endothelial cells whose resulting phenotype under flow was assessed using confocal microscopy. Modular tissue engineered constructs formed by the random packing of sub-millimetre, cylindrically shaped, endothelial cell-covered modules into a larger container creates interconnected channels that permit the flow of fluids such as blood. Due to the random packing, the flow path is tortuous and has the potential to create disturbed flow, resulting in an activated endothelium. At an average shear stress of 2.8 dyn cm⁻², endothelial cells within channels of varying geometries showed higher amounts of activation, as evidenced by an increase in ICAM-1 and VCAM-1 levels with respect to static controls. VE-cadherin expression also increased, however, it appeared discontinuous around the perimeter of the cells. An increase in flow (15.6 dyn cm⁻²) was sufficient to reduce ICAM-1 and VCAM-1 expression to a level below that of static controls for many disturbed flow-prone channels that contained branches, curves, expansions and contractions. VE-cadherin expression was also reduced and became discontinuous in all channels, possibly due to paracrine signaling. Other than showing a mild correlation to VE-cadherin, which may be linked through a cAMP-initiated pathway, KLF2 was found to be largely independent of shear stress for this system. To gauge the adhesiveness of the endothelium to leukocytes, THP-1 cells were introduced into flow-conditioned channels and their attachment measured. Relative to static controls, THP-1 adhesion was reduced in straight and bifurcating channels. However, even in the presence of flow, areas where multiple channels converged were found to be the most prone to THP-1 attachment. The microfluidic system enabled a full analysis of the effect of the tortuous flow expected in a modular construct on endothelial cell phenotype.

Fig. 1

(a) The tortuous, endothelial cell-lined blood perfusion channels within a modular tissue-engineered construct were reproduced in a microfluidic device in order to quantify the effect of the resulting flow patterns on endothelium activation and integrity, as well as to provide insight into further design optimization (b) A packed module bed was scanned to produce a 3-D image using μCT. The dark grey objects are the sub-millimeter sized poloxamine modules and the lighter area around them is where fluid flows. These scans were used as the basis to create a 2-D microfluidic chamber. A section of the μCT image was converted into a photomask (inset) which was used to reproduce these channels in microfluidic chambers

Fig. 2

Experimental setup. (a) Microfluidic chamber replicating the channels found within a modular tissue-engineered construct. Red fluid fills the channels where endothelial cells were to be seeded. The image was taken prior to epoxy application and only shows the inlet tube. Scale bar=1 cm. (b) Eight flow circuits were set up in parallel using a multichannel peristaltic pump. Twelve mL of endothelial cell medium was circulated from separate reservoirs and was changed every 2 days. Flow dampeners were used to reduce the pulsation, eliminate backflow and also served to remove air bubbles. The entire system was placed in an incubator set at 37°C and 5% CO₂

Fig. 3

The network of channels found in the modular construct (Fig. 1) was reproduced in a microfluidic chamber. (a) Randomly packed modules appear black and act as obstructions to flow. The white region is open and lined with a monolayer of endothelial cells. White scale bar=1000 μm (b) Nine regions of interest, 460 μm×460 μm (boxes not to scale), were selected based on their shape and flow characteristics, the latter determined by their streamlines via computational modeling. Black scale bar= 1000 μm (c) PIV-based arrow plot of flow through the branch region of the microfluidic chamber operating at a flow rate of 11 mL min⁻¹ (τ*, the average shear stress, was 15.6 dyn cm⁻²). Scale bar=0.1 m s⁻¹

Fig. 4

The effect of flow on VE-cadherin expression (green) after 2 days of flow in four selected regions of interest. Nuclei are counter-stained in blue. Early in the experimental time course, the effect of increasing flow rates can be seen on the morphology of the endothelial cells. With no flow, VE-cadherin is continuous and well defined. As shear stress increases, expression becomes discontinuous. At the highest shear stress, the amount of VE-cadherin expression also decreased, as compared to static controls. This may be indicative of a transition into disturbed flow as the system is subject to increasingly higher average shear stresses. Scale bars are 100 μm in length

Fig. 5

VE-cadherin (green, panel (a)) and KLF2 (red, panel (b)) expression at an average shear stress (τ*) of 2.8 dyn cm⁻² after 6 days of flow, with nuclei counter-stained blue. Devices were co-stained for both markers and are presented as split images. Each region’s callout shows a confocal image (460 μm×460 μm in size) of the HUVEC lining the bottom surface. The dotted white lines define the PDMS walls that encroached into several regions of interest, beyond which cells were not present. Occasional bare patches and discontinuous VE-cadherin can be seen in the node, large contraction, obstruction and branch regions. KLF2 expression did not appear higher in any particular region. Scale bars within callout images=100 μm while white bars in the bottom right corner of central device images=3000 μm

Fig. 6

(a) VE-cadherin (green) expression at an average shear stress (τ*) of 15.6 dyn cm⁻² after 1 day of flow, with nuclei counter-stained blue. For the majority of replicates, cells did not appear well aligned, even in the straight and large contraction regions where streamlines were relatively straight. VE-cadherin expression did not appear higher in any particular region and VE-cadherin appeared discontinuous throughout the entire chamber. Scale bars are 100 μm in length (b) CFDA SE labeled THP-1 adhesion to HUVEC after 1 day at τ*= 15.6 dyn cm⁻². THP-1 cells appear green, nuclei appear blue and VCAM-1 appears red; the regional variation in THP-1 attachment (e.g., the high coverage in the node region) is described in the context of Fig. 7. The velocity field at the centre of the panel (by computational modeling) is for a total flow rate of 11 mL min⁻¹ through the circuit corresponding to the 15.6 dyn cm⁻² average shear stress. White scale bars within callouts are 100 μm and within the central device image are 3000 μm

Fig. 7

THP-1 adhesion to circuits experiencing an average shear stress of 15.6 dyn cm⁻² and corresponding static controls. The total number of cells (HUVEC and THP-1) in each location was determined by nuclear staining and the percentage of those cells that were THP-1 was determined by their CFDA-SE label; this was termed THP-1 coverage. For static controls, there were no statistically significant differences in mean coverage among any of the nine regions of interest. For flow cases, the node region, where several channels combined, showed the greatest amount of THP-1 coverage. Error bars are 1 s.e.m., n=3 or 4 and * denotes p<0.05

Fig. 8

Total (over microfluidic chamber) fold change in marker expression over time for flow cases with average shear stresses of 2.8 and 15.6 dyn cm⁻². Marker expression in flow cases (region by region) was divided by that of static controls and then averaged over all replicates and regions. The connecting lines represent a significant difference in marker expression between days for a given shear stress; † represents a significant difference in marker expression between shear stresses for a given day; and * represents a significant difference in expression between flow cases and static controls. Differences were statistically significant when p<0.05. A negative fold change represents a downregulation in marker expression while a positive change signals an upregulation. Error bars are 1 s.e.m. and n between 27 and 36 based on 9 regions for three or four sets of replicates

Fig. 9

The fold change in marker expression (normalized to that observed with a static control in the same region, time and shear stress and displayed on a logarithmic scale for different regions and times) at low and high average shear stresses. The solid connecting lines represent a significant difference in marker expression between days for a given shear stress; † represents a significant difference in marker expression between shear stresses for a given day; and * represents a significant difference in expression between flow cases and static controls (a positive value indicates expression is higher with flow). At the lower 2.8 dyn cm⁻² shear stress, virtually no statistically significant differences between flow and static marker expressions were seen in any region (p>0.05). When the average shear stress was at the higher value (15.6 dyn cm⁻²), flow reduced the ICAM-1, VCAM-1 and VE-Cadherin expression in many regions throughout the time course, as compared to static controls. With the exception of the narrow bend region, KLF2 appeared to be unaffected by the increase in average shear stress. Error bars are 1 s.e.m., n=3 or 4 and p<0.05 for statistical significance

Fig. 10

Correlation between co-stained markers. Microfluidic chambers were either stained for ICAM-1 and VCAM-1 or for KLF2 and VE-cadherin, enabling the expression levels (normalized fluorescent intensities) of each pair to be compared. The strength of the relationship was measured by the Pearson coefficient, r. The statistical significance of the correlation is determined by the corresponding p value and n is the size of the dataset. ICAM-1 and VCAM-1 both show very strong, significant correlations, confirming that activated cells express both cell adhesion molecules. A weaker, yet significant, correlation was found between the expression of the KLF2 transcription factor and VE-cadherin, suggesting a possible link between KLF2 expression and well-formed VE-cadherin junctions

Fig. 11

Comparison of channels formed in a modular construct and those reproduced in a microfluidic chamber. A cross-section of the construct is depicted with flow going into the page (y-axis). In the modular construct, endothelial cells line the entire surface of the channels and the topography (z-direction) or “thickness” of the channels along their length (y-direction) is uneven. In the microfluidic chamber, silicone analogues for the modules also block flow and force the fluid through channels; however, the channels have the same thickness throughout the length of the chamber and we examined only the cells attached to the fibronectin coated coverglass