A Novel Microscopic Assay Reveals Heterogeneous Regulation of Local Endothelial Barrier Function

Abstract

Blood vessels are covered with endothelial cells on their inner surfaces, forming a selective and semipermeable barrier between the blood and the underlying tissue. Many pathological processes, such as inflammation or cancer metastasis, are accompanied by an increased vascular permeability. Progress in live cell imaging techniques has recently revealed that the structure of endothelial cell contacts is constantly reorganized and that endothelial junctions display high heterogeneities at a subcellular level even within one cell. Although it is assumed that this dynamic remodeling is associated with a local change in endothelial barrier function, a direct proof is missing mainly because of a lack of appropriate experimental techniques. Here, we describe a new assay to dynamically measure local endothelial barrier function with a lateral resolution of ∼15 μm and a temporal resolution of 1 min. In this setup, fluorescence-labeled molecules are added to the apical compartment of an endothelial monolayer, and the penetration of molecules from the apical to the basal compartment is recorded by total internal reflection fluorescence microscopy utilizing the generated evanescent field. With this technique, we found a remarkable heterogeneity in the local permeability for albumin within confluent endothelial cell layers. In regions with low permeability, stimulation with the proinflammatory agent histamine results in a transient increase in paracellular permeability. The effect showed a high variability along the contact of one individual cell, indicating a local regulation of endothelial barrier function. In regions with high basal permeability, histamine had no obvious effect. In contrast, the barrier-enhancing drug forskolin reduces the permeability for albumin and dextran uniformly along the cell junctions. Because this new approach can be readily combined with other live cell imaging techniques, it will contribute to a better understanding of the mechanisms underlying subcellular junctional reorganization during wound healing, inflammation, and angiogenesis.

Introduction

Vascular endothelial cells constitute a unique selective barrier between blood and tissue, which regulates the exchange of water, solutes, drugs, or cells, such as leukocytes and tumor cells (1). This exchange can occur via a trans- or paracellular route. There is general agreement that the paracellular way through the intercellular space, which is controlled by cell junction complexes, has a tremendous impact on inflammation, wound healing, and angiogenesis (2, 3). The endothelial cell junctions consisting of adherens, tight, and gap junctions connect the cells to each other, hence ensuring the integrity and the barrier function of the endothelial monolayer (4). Adherens junctions are stabilized by vascular endothelial cadherin (VE-cadherin) that, interdependently with actin dynamics, controls monolayer integrity involving many fine-tuned mechanisms like protein phosphorylation and Rho-GTPases (5, 6, 7, 8). Locally restricted regulation of VE-cadherin occurs, for example, during the inflammatory processes (9, 10). Consequently, the junctional patterning even along one cell might show variations in cell barrier functions at a subcellular level. These barrier-regulating processes need to be highly dynamic and locally restricted to seal the endothelial barrier and prevent plasma leakage and edema formation. Recent findings emerging from live cell imaging techniques show that junction remodeling occurs constitutively by small actin-driven junction-associated intermittent lamellipodia (JAILs) at a subcellular level. These dynamic structures with the size of only a few micrometers repair tiny local defects in the adherens junctions within a few minutes and are crucially required to maintain endothelial barrier function (11, 12, 13, 14, 15). Inhibition of the small GTPase Rac-1 or treatment with proinflammatory agents like thrombin is associated with an impaired JAIL formation and a local disruption of the VE-cadherin-mediated cell-cell adhesion (14, 16). This, to our knowledge, novel concept of subcellular barrier regulation provides a plausible mechanism for how endothelial cells can precisely modulate paracellular permeability while maintaining overall integrity of the cell layer (3). However, there is, to the best of our knowledge, currently no technique available that allows investigating the local dynamics of endothelial barrier function with the relevant resolution in time and space.

Currently, the most common experimental setups to investigate barrier functions of endothelial monolayers are either the quantification of the transendothelial electrical resistance (TER) by impedance analysis or the determination of permeability coefficients (P_E) for a soluble tracer substance, often referred to as transwell assays (17, 18). In the first approach, endothelial cells are usually cultured in measuring chambers with evaporated metal-film electrodes that allow analyzing the electrical impedance of the monolayer within a few seconds. Whereas the temporal resolution is quite high, the spatial resolution is limited by the diameters of the electrodes, which can be as small as a few hundred micrometers.

To quantify the permeability coefficient P_E, endothelial cells are cultivated on permeable filter membranes, which give access to the apical and the basal side of the cell layer. After adding a fluorescence-labeled tracer substance like albumin or dextran to the apical compartment, the permeation can be followed over time by determining the tracer concentration in samples taken from the basal compartment. The temporal resolution of this approach is given by the frequency of the sample drawing, usually several minutes. Because the diameter of the filter membranes is in the range of a few millimeters, the spatial resolution is poor.

Recently, Dubrovskyi et al. (19) have published a permeability assay with high spatial resolution. Because the method is compatible with normal immunofluorescence imaging, it has been used by several groups to correlate structural and functional parameters of endothelial and epithelial cell contacts. However, the assay is not suitable for live cell imaging and analysis of dynamic processes. Another interesting approach to resolve epithelial barrier function with lateral resolution is based on a macroporous culture substrate, whereby the pores serve as “femtoliter-sized” cuvettes (20). The filling of the pores can be followed over time by laser scanning microscopy, at least until the concentration gradient between the apical and basal side is equalized.

Here, we present a new optical noninvasive approach for living cells to dynamically measure local endothelial barrier function (DyMEB assay) with high spatial and temporal resolution. This approach is based on the detection of fluorescence-labeled molecules that permeate the endothelial cell sheet by TIRF microscopy. The assay is performed with cells grown on conventional coverslips and can therefore be readily combined with other live cell imaging techniques, such as differential interference contrast microscopy (DIC).

The fluorophore and the TIRF settings have been chosen in such a way that a fluorophore that has reached the basolateral space is detected and then quickly bleached irreversibly. Thus, regions of higher permeability show higher fluorescence signals and vice versa. We used the assay as a proof of principle to determine the dynamic changes in the permeability of human umbilical vein endothelial cells (HUVECs) after treatment with the well-known barrier-modulating compounds histamine or forskolin on the cellular and subcellular level.

Material and Methods

Reagents

Atto565 N-hydroxysuccinimide-ester was obtained from ATTO-TEC (Siegen, Germany), ascorbic acid from AppliChem (Darmstadt, Germany), Endothelial Growth Medium from PromoCell (Heidelberg, Germany), and Amino Dextran (10,000 MW) from Thermo Fisher Scientific (Waltham, MA). BSA, forskolin, histamine, 6-Hydroxy-2,5,7,8-tetramethylchroman-2-carbonsäure (Trolox), (3-Aminopropyl)triethoxysilane, and all other substances were purchased from Sigma-Aldrich (Munich, Germany).

Construction of the measuring chambers

The DyMEB-assay experiments were performed in self-constructed measuring chambers with a silanized glass bottom. Before silanization, the coverslips (20 mm in diameter) were incubated in 1 M NaOH (10 min, room temperature), washed two times with distilled water, incubated in 1 M HCl (1 min, room temperature), washed two times with distilled water, and finally cleaned with acetone (5 min, room temperature). For silanization, the coverslips were incubated in 3% (3-Aminopropyl)triethoxysilane solved in acetone (5 min, room temperature) and washed in acetone (5 min, room temperature). After rinsing with distilled water, the coverslips were dried overnight at 37°C and glued (ELASTOSIL; Wacker Chemie, Munich, Germany) underneath a petri dish with a 10-mm drill hole in the bottom.

Coating procedure

After silanization, measuring chambers were coated with cross-linked gelatin as described elsewhere (16). Briefly, the measuring chambers were incubated with 0.5% (w/v) gelatin from porcine skin (G-2500, type A; Sigma-Aldrich) dissolved in distilled water for 30 min at 37°C. Adsorbed gelatin was cross-linked with 2% glutaraldehyde (G-6257, grade II; Sigma-Aldrich) in distilled water for 10 min, followed by sterilization using 70% ethanol/water for 60 min at room temperature. Background fluorescence was reduced by exposure to ultraviolet light of 10,000 J/cm² (FLX-20M; Biometra, Göttingen, Germany). Subsequently, dishes were washed five times with phosphate-buffered saline (PBS), and free aldehyde groups were inactivated by 2 mM glycine/PBS (AppliChem) overnight at 4°C. Before cell seeding, chambers were rinsed five times with PBS again.

Labeling of bovine serum albumin/amino dextran with Atto565

Bovine serum albumin (BSA) or amino dextran was dissolved in labeling buffer following manufacturer’s instructions recommended for amine-reactive Atto labels. For conjugation, the amino-reactive Atto565 molecules (dissolved in dimethyl sulfoxide) were added in a fivefold molecular excessive to BSA or dextran and incubated in the dark for 1 h at room temperature while gently shaking. Purification of the Atto565-BSA/-dextran conjugate was performed by dialysis against PBS (overnight, 4°C) using dialysis cassettes (Slide-A-Lyzer, MWCO 10000 (for BSA)/2000 (for dextran); Thermo Scientific). Aliquots (final BSA/dextran concentration: 0.1 mM) were stored at −20°C until use.

Cell culture

HUVECs were isolated as described elsewhere (21), according to the principles outlined in the Declaration of Helsinki. The procedure was approved by the ethics boards of the University of Münster (2009-537-f-S). Cells were cultured on cross-linked gelatin-coated cell culture supports as described elsewhere (22) in endothelial cell growth medium (ECGM) (PromoCell) containing 100 μg/mL penicillin and 100 μg/mL streptomycin (PAA Laboratories, Pasching, Austria). Cell cultures were maintained at 37°C in a humidified 5% CO₂ incubator. Only cells of the first passage were used for experiments.

Impedance measurements

Impedance measurements were performed as described recently (21). Briefly, cells were seeded in a measuring chamber (20,000 cells/cm²), and impedance analysis was performed by an impedance spectroscopy setup (MOS Technology, Telgte, Germany). The TER was determined by analyzing the measured impedance spectra with a nonlinear least square fit (23). A stable TER was usually reached after 4 days (Fig. S1). Experiments with histamine or forskolin were performed after 6–8 days.

DyMEB assay

Self-manufactured chambers with silanized glass bottoms were coated with cross-linked gelatin (22). HUVECs were seeded at a density of 20,000 cells/cm² and cultured for 6–8 days. Under these conditions, cells usually reached confluency 4 days after seeding (Fig. S1).

For the DyMEB assay, the ECGM was replaced with phenol-red-free ECGM. If not otherwise indicated, the medium was supplemented with 5 mM ascorbic acid and 2 mM Trolox (pH 7.4) and incubated for at least 30 min. Either 500 nM Atto565-BSA or 50 nM Atto565-dextran were added 15 min before the assay starts.

The experiments were performed with an LSM780 ELYRA PS.1 microscope (ZEISS, Oberkochen, Germany), suitable for single molecule microscopy techniques like photoactivated localization microscopy (PALM) or stochastic optical reconstruction microscopy (STORM) with TIRF illumination. The microscope was further equipped with a DIC setup and an incubator (Tokai Hit, Fujinomiya, Japan) for live cell imaging. All images were acquired using an α-Plan-Apochromat 100×/1.46 DIC objective, a 1.6× tube lens, and an Andor iXON EMCCD Camera at 37°C in 5% CO₂ atmosphere.

To correlate structural and functional information, DIC and TIRF image series were recorded alternating for ∼40–60 s each in several cycles. It took ∼8 s to switch between the optical setups. The 561 nm Laser Line (100 mW at 60%) and the bandpass 570–650 + lowpass 750 filter cube were used for TIRF image acquisition with an angle of 77° (corresponding to a penetration depth of ∼65.2 nm) and an exposure time of 100 ms per frame. This protocol allows generating permeability data every 90–130 s.

In addition to the experiments with HUVEC, we also performed the assay with laminin- and fibronectin-coated chambers in the absence of cells to test whether alternative coatings work as well (Fig. S2).

Image analysis

Fiji (https://fiji.sc) was used to preprocess the TIRF images. To detect single, newly adsorbed fluorescence-labeled molecules in a TIRF image, difference images between two subsequent frames were calculated using the “Delta F up” function from the Cookbook plugin (https://imagej.net/Cookbook) (24, 25). The difference images were than analyzed with 3D-DAOSTORM (26) to identify the position of individual molecules. Intensity projection, application of Gaussian filters, manual segmentation of DIC images, and the generation of kymographs were also performed with Fiji using self-written macros.

Results

Principle of the DyMEB assay

To dynamically measure local endothelial barrier function with the DyMEB assay, an endothelial cell monolayer is cultivated directly on a glass-bottom dish suitable for total internal reflection microscopy (TIRF). A fluorescence-labeled tracer molecule, such as albumin or dextran, is added to the culture medium. The diffusion of the tracer molecule through the cell junctions is then followed by TIRF microscopy, which generates a thin evanescent field at the basal side of the cell layer, using high laser powers. The high laser intensities result in a bleaching of the fluorescent tracer molecules, quickly after they reach the evanescent field at the glass/medium interface. This generates and maintains a concentration gradient of unbleached tracer molecules between the apical and basal side of the cell. If the bleaching rate is fast compared to the lateral diffusion rate, it is also possible to generate lateral concentration gradients and to get information about local permeability at a subcellular level (Fig. 1 a).

Figure 1.

Assay concept. (a) Schematic drawing of the experimental setup is shown. Endothelial cells are cultivated on a glass substrate, and a fluorescence marker is added to the apical side. Marker molecules that permeate the cell layer are excited by the evanescent field of a TIRF illumination. Because of the high power of the TIRF laser, the fluorophores are quickly bleached, which results in a lateral gradient between sites of high and low permeability. Some molecules might also, in reverse, bind to the glass surface. These immobile single molecules appear as bright spots in the acquired TIRF images. (b) Shown are typical TIRF images of an endothelial cell layer that has been preincubated with Atto565-BSA at different time points. Directly after the onset of the TIRF illumination, the images show high fluorescence intensities (0 s). After a few seconds, a steady state between bleaching and diffusion from the apical to the basal side is established. In this state, there is almost no change in the overall intensity of the images (30, 60 s). Because of the quick bleaching of the tracer molecules, there are lateral gradients visible, which represent different local permeabilities (scale bars, 10 μm). (c) The corresponding course of mean fluorescence intensity illustrates the initial bleaching process, followed by a steady-state phase as mentioned above (b). To see this figure in color, go online.

In a typical experiment, tracer molecules were added at low concentrations (≤500 nM) to the culture medium. After an equilibration time of ∼15 min, TIRF images were acquired continuously at a rate of 10 frames per second and with high laser power. The initial frames show high fluorescence intensity because of the molecules that have already accumulated between the basal cell membrane and the glass surface (Fig. 1 b). The high laser power bleaches the fluorescent tracer and reduces the intensity of the images until a steady state is reached, in which the bleaching process is compensated by the diffusion of unbleached tracer molecules from the apical to the basal side. The decay phase takes ∼20 s and can be used to adjust the focal plane to the glass/medium interface if required (Fig. 1 c).

In the steady-state phase, the images typically show locally enhanced, diffuse background fluorescence (Fig. 1 b) as well as some bright, immobile spots that are sharply in focus (Fig. 1, a and b). In each frame, new spots appear, whereas others disappear. Because the penetration depth of the evanescent field was ∼65 nm, the spots clearly derive from molecules that are adsorbed close to the glass surface at the basal side of the cell layer. The enhanced background fluorescence, however, is due to diffusing molecules (27). Because the penetration depth of the evanescent field is much smaller than the distance between the glass surface and the endothelial junctions (∼300–500 nm), we assume that the complete fluorescence signal mainly derives from molecules at the basal side and not from freely diffusing molecules above the cell layer. However, the concentration of the fluorescent tracer at the apical side is much higher and might therefore still contribute to the signal.

To test whether this might be the case, the data of the experiments (Fig. 2 a) in this study were analyzed in two alternative ways: in the first approach, the number and position of newly appearing molecules (25) attached to the glass surface is determined for every frame during the steady-state phase to generate a dot plot (Fig. 2 b1). The number of newly appearing molecules per frame is a measure for the overall permeability in the field of view (51.2 × 51.2 μm), which has the dimension of a few cells (Fig. 2 b2). Alternatively, the dot plot is smoothed with a Gaussian filter to generate a molecule density map (Fig. 2 b3), which visualizes local permeability.

Figure 2.

Visualization and quantification of local endothelial cell permeability. (a1 and a2) DIC images (a1) and the corresponding TIRF images of Atto565-BSA (a2) can be used to illustrate areas of local cell permeability in HUVECs (red areas; scale bars, 10 μm). These raw data can be further quantified and visualized by two alternative approaches. (b1) In the first approach, the number and location of newly appearing molecules, which transiently bind to the glass surface, are detected to generate a dot plot. (b2) Molecules per frame provide information about the overall endothelial permeability of the whole image (51.2 × 51.2 μm). (b3) The molecule density map is based on the generated dot plot and smoothed with a Gaussian filter, which gives information concerning endothelial cell permeability on a subcellular level and corresponding cell borders, derived from segmented DIC images. (c1) Alternatively, an average intensity projection from all frames of the steady-state phase is generated. (c2) The mean fluorescence intensity of the projection represents the overall endothelial cell permeability in the whole field of view (51.2 × 51.2 μm). (c3) Smoothing the projection with a Gaussian filter gives the intensity map, which illustrates endothelial cell permeability on a subcellular dimension. To see this figure in color, go online.

In the second approach, the fluorescence intensity of the frames, which includes both the fluorescence signal from the freely diffusing molecules as well as from molecules bound to the glass surface, is evaluated. Therefore, an average intensity projection of all frames in the steady-state phase is generated (Fig. 2 c1). The mean fluorescence intensity (Fig. 2 c2) of the projection image is a measure for the overall permeability in the field of view. To visualize local permeabilities, a Gaussian filter was applied to generate a fluorescence intensity map (Fig. 2 c3).

Optimization of the assay for long-time live cell imaging

When the DyMEB assays were performed with standard cell culture medium in the absence of antioxidative supplements, the TIRF images in the steady-state period of the first cycles, shortly after the beginning of the experiment, are relatively dark with clearly visible spots. However, over time, regions become brighter, in particular at cell contacts. After 40 min, intercellular gaps appear that are also visible in DIC images, which show considerable large regions of saturated pixel intensities (Fig. 3 a, magenta area). The effect was only observed in the illuminated areas. Although the TIRF angel was set to a high inclination angle (77°) to reduce the penetration depth of the evanescent field (approx. 65 nm), the high laser power that is required to generate the concentration gradient between the apical and basal side of the cell layer obviously still provokes a phototoxic effect.

Figure 3.

Antioxidants reduce the phototoxic effect. (a) HUVECs were cultivated, and the DyMEB assay was performed to measure local cell permeability for 40 min in the absence of antioxidants. This results in an increase of HUVEC permeability (red areas) with saturated pixels (magenta areas) because of the phototoxic effects caused by the high laser power. The arrow indicates cell gap formation visible by DIC. (b) The addition of antioxidants (5 mM ascorbic acid, 2 mM Trolox) reduces phototoxic effects (scale bars, 10 μm). (c) Saturated pixels were raised up to 54 ± 12% in the absence of antioxidants, whereas the supplementation of antioxidants causes a lowered incline of 8 ± 6% (mean ± SD). (d) In the absence of antioxidants, mean fluorescence intensity starts to increase by 100 ± 7%, whereas additions of antioxidants show a reduced increase of 45 ± 3% (mean ± SD). (e) Molecules per frame did not show these increases because of an underestimation of molecules, resulting from the saturated areas (see a and b, magenta areas) in which the detection algorithm fails (mean ± SD). (f) Examples of an intensity map and a molecule density map from the experiment shown in (a) indicate comparable local, subcellular permeabilities (scale bar = 10 μm). Line plots through the local maxima show comparable peak profiles and similar full width at half-maximum (FWHM). To see this figure in color, go online.

Consistent with other studies (28, 29), this phototoxic effect could be largely reduced by the addition of antioxidative substances (5 mM ascorbic acid and 2 mM Trolox) (Fig. 3 b). Quantification of the saturated areas after 40 min showed a reduction from 54 to 8% by the addition of antioxidants (Fig. 3 c). The effect is also reflected by observing the mean fluorescence intensity, which is increasing by 100% after 40 min in the absence of antioxidants, whereas in the presence of antioxidants, this increase is only 45% (Fig. 3 d). The number of molecules per frame, however, did not increase in the absence of antioxidants because of the fact that the detection algorithm is not working in saturated regions, leading to an underestimation of this parameter (Fig. 3 e).

Although the laser-induced disruption of endothelial junction was an undesired effect and all following experiments were therefore performed in the presence of antioxidants, the experiments indicate that the DyMEB assay is able to follow cell junction regulation on a subcellular level (Fig. 3 a). In the absence of antioxidants, the formation of local spots with increased permeability usually started after 12–14 min. A comparison of the generated intensity and molecule density maps (Fig. 3 f; Video S1) showed similar permeability patterns, at least until saturated regions appear, which causes the above-mentioned problems with the detection algorithm. To estimate the lateral resolution of the DyMEB assay, the full width at half-maximum (FWHM) values of the early local spots were determined for both types of permeability maps. The average FWHM for the intensity maps was ∼13.5 ± 3.0 μm, whereas the value for the molecule density maps was 18.1 ± 4.6 μm (n = 3), indicating that both approaches have comparable lateral resolutions.

Video S1. Comparison Between Intensity and Molecule Density Maps

The DyMEB-assay was performed with confluent HUVECs in the absence of antioxidants and the intensity maps (left panel) and the molecule density maps (right panel) were generated (red channels). In the beginning permeability of the cell layer was low and both approaches gave comparable results. To the end, when the phototoxic side effects start to induce intercellular gap formation and saturated areas appear (magenta areas), the detection algorithm required to calculate the molecule density maps failed, leading to a underestimation of the permeability (red signal on the right panel).

Confluent HUVEC cultures show a large heterogeneity in permeability

Investigation of different areas within one HUVEC culture displayed surprisingly high heterogeneities in the fluorescence signal during the steady-state phase, suggesting high variations in local permeability even within the very same culture (Fig. 4 a).

Figure 4.

HUVECs show heterogeneity in cell permeability. (a) DIC images (upper row) of three regions within the same cell culture show differences in the fluorescence intensity in the corresponding TIRF frames (lower row) during the steady-state phase, indicating differences in permeability (scale bars, 10 μm). (b) The molecules per frame correlate with the mean fluorescence intensity of the images, indicating that both parameters can be used to evaluate permeability. Light gray dots represent data from different areas of the very same culture. The indicated numbers correspond to the images shown in (a). The histograms for both parameters show a great variance, indicating high variations of cell permeability.

The determined mean fluorescence intensities for 51 positions in 43 individual cultures were in the range of 13,636 and 39,581 (25,710 ± 6083), whereas the number of molecules per frame was between 55 and 435 (235 ± 100) (Fig. 4 b). The two parameters showed an approximately linear correlation (Fig. 4 b), indicating that fluorescence signals from apical molecules can be neglected and that both parameters can interchangeably be used to describe the permeability of the endothelial cell monolayer.

For the purpose of clarity, we decided to represent the permeability parameters in the following figures in terms of mean fluorescence intensities and intensity maps. The corresponding graphs for the molecules per frame and the molecule density maps will be shown in Supporting Materials and Methods.

Forskolin quickly reduces endothelial macromolecular permeability

Forskolin, an activator of the adenylate cyclase, is well known to increase endothelial barrier function (18, 30). To study the effect with the DyMEB assay, the mean fluorescence intensity was determined for three cycles (5 min) before adding PBS (as a control) or forskolin (5 μM), respectively (n = 12 each). Although the PBS controls showed only a slight increase over time, the forskolin-treated cultures displayed usually a decrease of cell permeability, indicating an improved endothelial barrier function (Fig. 5 a). However, the time course of the response between the different experiments was quite heterogeneous, with no obvious relation to the initial permeability. The averaged mean fluorescence intensity dropped from 37,318 to 29,708 (n = 12) within 40 s and basically stayed constant for the investigated time interval of 18 min (Fig. 5 b). This correlates roughly with an increase in TER (Fig. 5 c). The corresponding data for molecules per frame showed a comparable time course (Fig. S3).

Figure 5.

Barrier-enhancing effect of forskolin determined by the DyMEB assay. (a) Shown are heatmaps displaying the time course of the mean fluorescence intensity for different HUVEC cultures (n = 12) after the addition of PBS or forskolin (5 μM). Each row represents one experiment (replicate), whereas columns correspond to one time point. Regardless of the initial mean fluorescence intensity, all HUVEC cultures show a decrease in permeability upon forskolin stimulation, although the time course of the forskolin-mediated response displays variations between the single cell cultures. (b) Shown is the averaged mean fluorescence intensity for all forskolin (solid circle) or control experiments (open circle) (n = 12) depicting a decrease of cell permeability upon forskolin stimulation but not for PBS (mean ± SE). (c) Forskolin effect on transendothelial electrical resistance (TER) is shown. Confluent HUVECs were treated with 5 μM forskolin (solid circle) or PBS (open circle) as a control (mean ± SE). The TER was normalized to the initial resistance (n = 3). To see this figure in color, go online.

It is reported that junctional dynamics of highly confluent cell cultures like we used in this study is relatively low compared to subconfluent cell layers (13). Accordingly, the junctional permeability was usually relatively constant in the control experiments (Fig. 6 a). To visualize the dynamic changes in permeability along the outline of an individual cell, the fluorescence intensity along the cell border was plotted as a function of time to generate a junctional kymograph. Although the permeability of most junctions remains stable (Fig. 6 b, white arrowhead), a spontaneous and locally restricted transient increase in junctional permeability was also observed occasionally in some control experiments (Fig. 6 b, black arrow head). This demonstrates that the spatial resolution of the DyMEB-assay is able to detect and visualize such subcellular events, in which one cell regulates the paracellular barrier function at different junctions independently.

Figure 6.

Subcellular permeability dynamics in confluent HUVEC. (a) Time lapse of intensity maps illustrates the heterogeneity of HUVEC permeability toward Atto565-BSA on a subcellular level. Scale bars, 10 μm. (b) The related junctional kymograph for the selected cell further visualizes the transient, locally restricted permeability increase between two junctions (j6, j7) compared to the continuous course of another junction (j4) of the same cell. The corresponding line plot of three selected time points further illustrates this heterogeneous permeability increase. Whereas the local permeability increases in certain regions (black arrow), it remains almost constant in other parts (white arrow). (c) Time lapse of intensity maps of a forskolin experiment illustrates overall attenuation of HUVEC permeability toward Atto565-BSA upon forskolin treatment without subcellular preferences, although initial brightness shows subcellular permeability variations along cell junctions. Scale bars, 10 μm. (d) The corresponding junctional kymograph and line plot further visualize a more or less uniform permeability decrease along all junctions over time. To see this figure in color, go online.

Intensity maps of forskolin experiments show that forskolin treatment leads to a rapid and entire reduction of cellular permeability without junctional differences along the cell border of one cell (Fig. 6 c; Video S2). Junctional kymograph analyses are in line with these findings because even if there are slight differences along the cell border of one cell, the forskolin response occurs in a synchronous manner (Fig. 6 d).

Video S2. Forskolin Reduces Paracellular Permeability to Albumin

The DyMEB-assay was performed with Atto565-labeled albumin and the intensity maps were generated (red channel). Addition of forskolin (5 μM) quickly reduces the permeability of the HUVEC monolayer.

To test whether the DyMEB assay is also working with fluorescence-labeled dextran, which is another widely used tracer to investigate paracellular endothelial barrier function, equivalent forskolin experiments were performed. Compared to the PBS control experiments, the intensity maps of forskolin-stimulated cells showed reduced fluorescence intensity along the cell contacts (Fig. S4; Video S3), indicating a reduced junctional permeability. The averaged mean fluorescence intensity and detected molecules per frame of all experiments (n = 5; Fig. S4) showed also a time course comparable to the BSA experiments (Fig. 5 b).

Video S3. Forskolin Reduces Paracellular Permeability to 10 kDa-Dextran

The DyMEB-assay was performed with Atto565-labeled dextran (10 kDa) and the intensity maps were generated (red channel). Addition of forskolin (5 μM) quickly reduces the paracellular permeability of the HUVEC monolayer.

The histamine-induced transient increases in paracellular permeability show high variations on a cellular and subcellular level

Histamine is an autacoid compound that is released in allergic inflammation and is well known to transiently reduce endothelial barrier function (18, 31). Hence, we performed the DyMEB assay to study the dynamic regulation of endothelial permeability after histamine treatment. The mean fluorescence intensity of the control experiments showed only some unspecific dynamic variations over time (Fig. 7 a, left). The response to histamine, however, displayed a strong dependency on the baseline level; regions with low initial permeabilities (mean fluorescence intensity <20,000) showed a transient increase of up to 252% (130 ± 50%) within 2:20 min after the addition of histamine, whereas regions with already higher baseline permeability (mean fluorescence intensity >20,000) showed no obvious response (Fig. 7 a, right). The histamine-induced transient increase of the mean fluorescence intensity averaged over all experiments (n = 11; Fig. 7 b) correlates closely with a transient decrease in TER (n = 3; Fig. 7 c). The corresponding molecules per frame data show also a comparable time course (Fig. S5).

Figure 7.

DyMEB assay reveals heterogeneous effects of histamine on local cell permeability of HUVEC. (a) Shown are heatmaps displaying the time course of the mean fluorescence intensity for different HUVEC cultures (n = 11) after the addition of PBS or histamine (100 μM). Each row represents one experiment (replicate), whereas columns correspond to one time point. A transient increase in permeability was observed after the addition of histamine in regions with low initial permeability (mean fluorescence intensity <20,000) but not in regions with high initial permeability (mean fluorescence intensity >20,000). (b) Averaged mean fluorescence intensity for all histamine (solid circle) or control experiments (open circle) are shown (mean ± SE). (c) Histamine effect on normalized transendothelial electrical resistance (n = 3) is shown. Histamine immediately provokes reversible increases of permeability correlating to the results of the DyMEB assay (mean ± SE). To see this figure in color, go online.

The intensity maps of cells with a low baseline value (mean fluorescence intensity <20,000) showed a time-dependent transient increase in permeability after stimulation with histamine (Fig. 8 a). The increase was largest in the regions between the cells. A line plot further confirms that the maximal intensity was found close to the position of cell-cell junctions as estimated from the DIC images (Fig. 8 b; Video S4). Analysis of the local permeability changes along the border of one cell typically showed high subcellular variations, indicating that some junctions contribute more to the permeability increase than others (Fig. 8 c; Fig. S6). In contrast to the phototoxic effect described above, the histamine-induced permeability increase was not accompanied by an obvious gap formation observable by DIC. This demonstrates that the DyMEB assay is also capable of analyzing more subtle junctional rearrangements.

Figure 8.

DyMEB assay detects histamine-induced leakage of HUVEC at cell junctions. (a) Time lapse images of the intensity maps illustrate a transient increase of local cell permeability at cell junctions after stimulation with histamine (100 μM). Scale bars, 10 μm. (b) Shown are line plot diagrams depicting local maximum of molecule density at cell junctions upon histamine stimulation. The circle marks the position of the cell junction. (c) The junctional kymograph visualizes subcellular heterogeneities in permeability along the cell border of one cell, revealing that one cell junction (j3) shows a higher contribution to the histamine-induced increase of permeability. The corresponding line plot for three selected time points further illustrates the local differences in permeability increase. Although histamine induced an overall increase in junctional permeability, the magnitude was higher in certain regions (j3, black arrow) compared to other sites (j1, white arrow) of the same cell. To see this figure in color, go online.

Video S4. Histamine Transiently Increases Local Permeability

The DyMEB-assay was performed with Atto565-labeled albumin and the intensity maps were generated (red channel). Addition of histamine (100 μM) transiently increases the local permeability of the HUVEC monolayer.

Discussion

Here, we described a new approach to study dynamical changes in local endothelial permeability at a cellular and subcellular level. The required hardware consists of a TIRF microscope equipped with a high-power laser and a fast camera, which is usually used for STORM/PALM superresolution imaging. Comparable setups nowadays exist in many cell biology labs and image facilities. Because the software used to perform the image analysis is also freely available, we believe that the assay is a useful and broadly available tool for many researchers in the field.

We tested two different approaches to analyze the acquired fluorescence images. The first approach is based on detecting fluorophores that are transiently bound to the glass surface. This ensures that only molecules that are underneath the cell are included in the generation of the molecule density maps and the calculation of molecules per frame. However, this approach is quite sensitive to saturation effects and a correct focal plane. The second approach integrates the complete fluorescence signal of all bound and freely diffusing molecules in the evanescent field. The analysis is less time consuming, more robust against small variations in the focal plane, and gives, in the end, comparable results. We therefore displayed most of our results in terms of mean fluorescence intensity and intensity maps, respectively.

Using this assay, we found to our surprise that the permeability of confluent HUVEC to BSA was quite heterogeneous even within the very same cell culture. A similar heterogeneous permeability distribution has been described for confluent bovine aortic endothelial cells, particularly at higher passages (32). A possible explanation might be that the permeability changes during different phases of the cell cycle. However, in both studies, the investigated cell cultures had already reached confluency for several days, so that the cellular turnover should be low. Alternatively, the cell cultures might consist of different subpopulations with a different expression pattern. The existence of such subpopulations in HUVEC at low passage has recently been suggested (33). This is also consistent with our observation that the expression of VEGF-receptor-2 is heterogeneous in HUVEC cultures of passage 1 (14). Further, we found that only regions with low permeability showed a transient increase in permeability within a few minutes after histamine stimulation, which also might be related to differences in the expression patterns. From correlations between DIC images with permeability maps, we could show that the increase occurs close to the cell junctions, without the formation of intercellular gaps, indicating that subtle changes in cell-cell interaction were detectable with the DyMEB assay. Although several studies using confluent HUVEC cultures showed a moderate, transient and reversible reduction of the TER after histamine stimulation, the effect on macromolecular permeability is less clear (18, 34). In an earlier study, the effect of histamine on endothelial junction organization was compared between HUVECs and human dermal microvascular endothelial cells (34). In contrast to human dermal microvascular endothelial cell cultures, HUVECs showed a much higher variability in the response to histamine, which is consistent with our observations that only a subpopulation of HUVECs responds to histamine stimulation.

Interestingly, we observed also a subcellular heterogeneity in the response to histamine, indicating locally restricted junctional regulation. Little is known about the signaling mechanisms underlying such a subcellular regulation of endothelial barrier function, but a local activation of the small GTPases rhoA and rac1 has been demonstrated in endothelial cells (10, 14). Interestingly, these GTPases have been suggested to play a role in histamine-induced permeability changes (6). Compared to histamine, other proinflammatory mediators like thrombin or tumor necrosis factor-α showed an even more pronounced effect on VE-cadherin distribution and the formation of intercellular gaps (16, 34, 35). Because the DyMEB assay is sensitive enough to follow the moderate histamine-induced changes, it is most likely applicable to investigate also the local changes of endothelial permeability in other inflammatory processes. Locally restricted fluctuations in permeability were also observed in some control experiments. It is tempting to speculate that these subcellular changes in permeability are a result of locally appearing JAILs at a particular cell contact (13). However, this requires further studies, including a more specific analysis of actin dynamics. Besides histamine, we also used forskolin, which activates the adenylate cyclase and leads to an increase of cortactin and actin filament along the cell junctions via an activation of the cAMP-Epac1-Rap1 pathway (36, 37). Obviously, this signaling mechanism seems to reduce the permeability relatively synchronous along the whole cell perimeter.

In addition to BSA, we also tested fluorescence-labeled dextran as a tracer molecule. As expected and in agreement with other reports (18), forskolin transiently reduced the permeability for this tracer as well. Because dextrans are available with different molecular weights, it will be possible to investigate the regulation of local permeability to molecules of different size with the DyMEB assay in future studies. It should be noted that because of the high laser power and the small basal compartment, the required tracer concentrations for the DyMEB assay were considerably lower compared to conventional transwell filter setups, which is usually in the range of milligram/milliliter. For example in this study, we used BSA (66 kDa) in a concentration of 500 nM (33 μg/mL), whereas for the smaller 10-kDa dextran tracer, a concentration of only 50 nM (0.5 μg/mL) was already sufficient to get acceptable fluorescence signals. To perform the DyMEB assay with tracer molecules of higher molecular weight or with cell layers that show a lower permeability, such as endothelial cell layers of the blood brain barrier, the concentration of the tracer molecule could simply be increased to enhance the sensitivity.

In principle, the assay can also be performed with epithelial cell monolayers. However, the different epithelial cell types display a large variability in permeability and cell dimensions. Whether the DyMEB assay allows subcellular resolution depends of course on the cell diameter but also on the cell height. Although the barrier-forming cell contacts in squamous epithelial cells (e.g., lung epithelium) are close to the substrate, the tight junctions in most cuboidal epithelial cell types are usually found at the apical part of the cells. Therefore, molecules that pass the tight junctions will diffuse in the rather long lateral intercellular space (38) before reaching the evanescent field, which will reduce the spatial resolution. However, the DyMEB assay might still be used to investigate the dynamic regulation during local events like cell division or apoptosis with a cellular resolution.

Recently, permeability of endothelial and epithelial cell layers for avidin (which has nearly the same molecular weight of albumin) was investigated with the XPerT assay (19, 39, 40, 41, 42). In this assay, fluorescence-labeled avidin is added to the apical side of the cells and allowed to diffuse through the cell layer for a few minutes, where it binds to the biotinylated extracellular matrix. Because the binding between biotin and avidin is very strong and almost irreversible, the assay has a very high lateral resolution but does not allow dynamic measurements because the binding sites for avidin are eventually saturated. In our assay, the interaction between albumin and the glass surface or the extracellular matrix is weak and transient. This reduces the lateral resolution because molecules can diffuse away from the site of passage but allows dynamic measurements because there is no saturation effect.

There are several options to further increase the lateral resolution of our assay. An increase in laser power, for example, would reduce the lifetime of the fluorescence molecules and increase lateral concentration gradients of unbleached molecules. However, as for all live cell imaging techniques, this increases also the phototoxic effects, which limits this possibility. Vice versa, a fluorophore with low photostability but high quantum yield would have the same effect. In our study, we choose Atto565 because it was reported that it shows these features at least for the yellow absorbing fluorophores (43). We also tested fluorescein isothiocyanate (data not shown), but in our hands, the phototoxic effects were much higher. Because most techniques require photostable fluorophores, there are only very few options under the commercially available markers. Also, a reduction of the lateral diffusion by the modification of either the extracellular matrix or the tracer molecule, like in the XPerT assay, would be a possibility.

A big advantage of our assay is the possibility to follow changes in permeability over time. In the presented experiments in which we alternatively acquired DIC and TIRF images, the temporal resolution was around 90 s, which could be easily reduced below 1 min if the DIC acquisition is omitted. This hightime resolution allowed, for example, to follow the histamine-induced increase after 3 min, which is usually only detectable by TER measurements but hardly in transwell filter assays (18).

To investigate the dynamics of local cell-cell interaction is, in our opinion, essential to get a deeper understanding of the underlying molecular processes. For example, it was not before we observed JAILs in live cell imaging experiments that the fundamental meaning of these structures for the dynamic remodeling of endothelial cell junctions became obvious (44). Therefore, we believe that the assay might contribute to a better understanding of locally restricted dynamic events, such as cell division, apoptosis, and transmigration of leukocytes or tumor cells. These cellular processes are all involved in many vascular-related diseases like inflammation, tumor metastasis, wound healing, or the development of arteriosclerosis. Because the area of the investigated region is very small, the assay is also suitable for high-throughput applications or integration in a “Lab on a chip” system.

Author Contributions

H.-J.S. and J.S. designed research. N.K. and J.S. performed experiments and analyzed data. N.K., H.-J.S. and J.S. wrote the manuscript.

Editor: Claudia Steinem.