Introduction

Across several major diseases including acute lung injury, atherosclerosis, tumor neovascularization, hypertension, diabetes mellitus, and coronary artery disease, the endothelial barrier becomes disrupted through the formation of paracellular gaps [1,2,3,4,5,6,7,8]. Since the pioneering work of Guido Majno, and George Palade in the 1960s [9, 10], paracellular gaps in response to inflammatory mediators have been considered to arise by contractile forces that act to pull intercellular junctions apart [11,12,13,14,15,16,17,18,19,20,21,22]. These forces are generated via acto-myosin interactions and manifest upon the EC substrate as tractions [12, 18,19,20,21, 23, 24].

Endothelial barrier function is routinely measured through changes in electrical resistance or macromolecular permeability across a confluent monolayer [25]. While these measurements are fast, they cannot simultaneously quantify underlying biophysical forces. As a result, further validation requires follow-up measurements of EC contraction using single cells [21], cell doublets [20, 23], micropatterned cell clusters [12, 18], and continuous cell monolayers [19, 24]. These follow-up measurements are slow, complex, and low-throughput, limiting the pace of mechanistic insights and the breadth of pharmacological investigation.

Here, we have developed “mini-XPerT”, a force-based high-throughput assay that combines monolayer traction microscopy [19, 26] to measure EC contraction, Express Permeability Testing (XPerT) [27] to measure paracellular gaps, and immunofluorescence imaging to measure structural and morphological changes. These measurements are miniaturized in 96-well plates using the approach of contractile force screening [28, 29]. Mini-XPerT utilizes elastic silicone substrates that are tunable in Young’s modulus (0.3–150 kPa) [29], thereby enabling studies spanning the entire known (patho)physiological range of substrate stiffness. Additional practical advantages of mini-XPerT include the following: it is non-invasive, non-destructive, and suitable for both short- and long-term assessment of barrier function.

Applying mini-XPerT, we have studied monolayers of primary human dermal microvascular ECs subjected to acute and progressive mediators of pathological permeability. We have compared barrier-protective mechanisms of two well-known and mechanistically distinct barrier-protective agents, the Rho-kinase inhibitor Y-26732 [30] and the canonical Tie2 receptor agonist Angiopoietin-1 (Angpt-1) [31]. The multiplexed nature of mini-XPerT unveiled evidence to suggest that Angpt-1 achieves barrier protection without diminishing contractile forces. Taken together, these results reveal that the endothelium can protect itself by: (1) reorganizing the actin cytoskeleton—a contraction-dependent mechanism—as in the case of Y-27632 or (2) by accumulating VE-cadherin to fortify cell–cell junctions—a contraction-independent mechanism—as in the case of Angpt-1.

Material and methods

While our methodologies are generalizable across adherent cell types for a wide spectrum of experimental conditions, we focused here on a representative endothelial cell type, a commonly studied range of substrate stiffness values, and a commonly used extracellular matrix ligand, collagen I.

Cell culture

All experiments were performed using primary dermal human microvascular endothelial cells (Lonza, Basel, Switzerland) at passages 4–7 in EGM2-MV medium (Lonza, Basel, Switzerland) containing 4.7% fetal bovine serum. The cells were plated for 2–3 days prior to experimentation.

Fabrication of multi-well plates

Soft, elastic (Young’s Modulus = 0.3 or 3 kPa), and deformable substrates were prepared using NuSil® 8100 (NuSil Silicone Technologies, Carpinteria, CA) in 96-well plates [29]. Embedded in the substrate surface are fluorescent beads (diameter = ~400 nm) whose displacements enable traction force calculations. Ligated to the substrate surface is biotinylated collagen I (collagen I, Advanced Biomatrix, Carlsbad, CA; 10x PBS, Corning Life Sciences, Tewksbury, MA; EZ-Link NHS-LC-LC-Biotin, Thermo Fisher Scientific, Waltham, MA) whose binding with FITC-avidin enables the identification of paracellular gaps. Detailed information of plate preparation, ligation, and cell seeding is provided in the supplemental material.

Measurement of cellular contractile forces: monolayer traction microscopy

The local contraction that an endothelial cell exerts upon  its substrate is called traction [26, 32]. To measure tractions, we utilized an inverted epi-fluorescence microscope (DMI 6000B, Leica Inc., Germany) equipped with a heated chamber (37 °C), a monochrome camera (Leica DFC365 FX), and a motorized stage. We recorded spatial images of substrate-embedded fluorescent beads at 10x magnification. Based on the bead displacements (resolution = ~15 μm) relative to a cell-free image, together with knowledge of substrate stiffness and thickness, we computed tractions using the approach of Fourier Transform Traction Cytometry [33], modified to the case of cell monolayers [26]. From each traction map, we calculated the root-mean squared traction (RMST) value and report this as a measure of average contraction in the monolayer. On a well-by-well basis, we first computed the ratio of the RMST after vs. before stimulation. We then normalized all values of a given treatment group to the mean of the corresponding vehicle control group and further calculated the average of each treatment group.

This permits us to compare differences across treatment conditions in each 96-well plate as well as across plates.

Measurement of paracellular gaps: Express Permeability Testing (XPerT)

At the end of drug treatment, FITC-conjugated avidin dissolved in PBS (Thermo Fisher Scientific, Waltham, MA) was added to each well and incubated for 210 s. The cellular monolayer acts as a mask; Avidin binds to biotinylated collagen I only in areas of paracellular gaps [27]. When further visualized using an inverted epi-fluorescence microscope (DMI 6000B, Leica Inc., Germany, 10x magnification), the FITC label localizes the gap and thereby permits quantitation of both its location and overall incidence in a well. The raw FITC images were converted into 8-bit binary images; the same threshold value was applied to all images in the manuscript. Gaps were computed from the areas in white and expressed as the percentage of the total area in a given image. We then normalized all values of a given treatment group to the mean of the corresponding vehicle control group and further calculated the average of each treatment group. To perform these analyses in high-throughput, a custom-macro was written using Fiji software. The macro is provided in the supplemental material.

Quantification of F-actin stress fiber expression and orientation

At the end of drug treatment, the cells were fixed with 3% Formalin (Sigma-Aldrich, St. Louis, MO) for 5 min and then washed three times with PBS. Alexa Fluor 480 phalloidin (Thermo Fisher Scientific, Waltham, MA) at a concentration of 1:300 and Hoechst dye (Thermo Fisher Scientific, Waltham, MA) at a concentration of 1:2000 was added together to each well for 30 min. The wells were washed twice with PBS and visualized using an inverted epi-fluorescence microscope (DMI 6000B, Leica Inc., Germany, 20x magnification). The Alexa 480 labels the actin stress fibers while the Hoechst dye labels the nuclei. From each image, we chose at random, six different regions. In these regions, we quantified F-actin orientation (in terms of anisotropy) using an open source plugin called Fibriltool [34] for Fiji. We also quantified corresponding F-actin expression (in terms of mean-fluorescent intensity) in those same regions. Across all images for a given treatment group, we normalized all values to the mean of the corresponding vehicle control group and further calculated the average of each treatment group.

Evaluation of VE-cadherin expression

At the end of drug treatment, the cells were fixed with 3% Formalin for 5 min, washed three times with PBS, and then permeabilized with 0.1% saponin (Sigma-Aldrich, St. Louis, MO) for 5 min. The cells were washed with PBS and a mouse-derived monoclonal anti-VE-cadherin antibody (BD Biosciences, Franklin Lakes, NJ) at a concentration of 1:250 was added to each well for 2 h. The wells were washed twice with PBS and a secondary antibody conjugated with FITC (Thermo Fisher Scientific, Waltham, MA) at a concentration of 1:500 and Hoechst dye (Thermo Fisher Scientific, Waltham, MA) at a concentration of 1:2000 was added together for an additional 30 min. The cells were then washed twice with PBS. When visualized using an inverted epi-fluorescence microscope (DMI 6000B, Leica Inc., Germany, 20x magnification), the FITC-conjugated antibody labels VE-cadherin expression at the cell–cell junction and the Hoechst dye labels nuclei.

Measurement of trans-endothelial electrical resistance (TEER)

An electric cell-substrate impedance sensing system (ECIS, Applied BioPhysics Inc., Troy, NY) was used to measure trans-endothelial electrical resistance of EC monolayers [35]. Briefly, cells were seeded on collagen I coated electrode arrays (8W10E + , Applied BioPhysics Inc., Troy, NY) and grown to confluency (24–48 h), determined by achieving electric resistance of > 1500 ohm. Cells were then incubated with indicated drugs or EBM-2 medium alone as a vehicle control with raw resistance continually measured over time at an amplitude of 4000 Hz. For every well, raw resistance values were divided by the correspondent time-zero raw resistance value to obtain the normalized TEER.

Drug treatments

EBM-2 basal medium served as vehicle for drug treatments, and as the vehicle control. Angiopoietin-1, CD14 and LPS binding protein were purchased from R&D Systems (Minneapolis, MN). Thrombin was purchased from EMD Millipore (Billerica, MA). LPS serotype O111:B4 was purchased from Sigma-Aldrich (St. Louis, MO). Y-27632 was purchased from EMD Biosciences (LaJolla, CA).

Statistics

Data are presented as mean ± SEM. For gaps, ECIS measurements, F-actin intensity and anisotropy measurements, statistical significance was tested by two-tailed unpaired t-test with Welch’s correction. For traction measurements, the more appropriate non-parametric two-tailed Mann–Whitney U-test for data deviating from Gaussian distribution was used. A p-value < 0.05 was considered significant.

Results and discussion

A key downstream effector of EC barrier disruption is the Rho-associated protein kinase, ROCK [36,37,38]. Activation of ROCK signaling promotes actin stress fiber formation, myosin light chain phosphorylation, endothelial cell contraction and, ultimately, paracellular gap formation [15, 30, 36,37,38,39,40,41,42]. Conversely, inhibition of ROCK signaling via the pharmacological agent Y-27632 promotes endothelial barrier defense [12, 30].

We chose Y-27632 as an exemplar to evaluate mini-XPerT (Fig. 1). Based on typically used concentrations [12, 42,43,44,45,46], and supportive dose–response measurements (Supplementary Fig. S1), we picked a sub-maximal dose of 5 μM for further evaluation. Individual wells of a single 96-well plate (Young’s Modulus = 3 kPa) were pre-treated with either vehicle or Y-27632 for 20 min, and then stimulated further with the barrier disruptive agent, thrombin (1 U/ml, 30 min). Paracellular gaps and monolayer contraction were significantly enhanced by thrombin and diminished by pre-treatment with Y-27632 (Fig. 1B-E). Moreover, these force changes correlated with F-actin cytoskeletal changes, as was revealed by simultaneous measurements of the F-actin cytoskeleton (Supplementary Fig. S2). Specifically, while thrombin enhanced F-actin expression and orientation, these same quantities were significantly reduced by pre-treatment with Y-27632. Finally, the barrier-protective effects of Y-27632 was consistent with ECIS measurements (Supplementary Fig. S3).

Fig. 1
figure 1

Mini-XPerT–simultaneous measurements of barrier function and cellular contractile forces in a high-throughput format. A Silicone-based elastic substrates (Young’s Modulus = 0.3 kPa or 3 kPa) were prepared in a 96-well plate format. Embedded in the substrate are fluorescent beads (diameter = ~400 nm; shown in red) whose displacement enables the computation of monolayer contractile forces (shown as white arrows). Ligated to the substrate is biotinylated collagen I (shown as black triangles) whose binding with FITC-avidin (shown in green) enables the identification of paracellular gaps. B Representative gap images for every experimental condition, C corresponding traction force maps, and D-E averaged values. While gap formation and contraction were increased with thrombin (1 U/ml, 30 min) it was significantly reduced by pre-treatment for 20 min with the ROCK inhibitor, Y-27632 (5 µM). In D, plotted is the gap area normalized to the vehicle group. In E, on a well-by-well basis, we quantified the ratio of the average monolayer contraction after vs. before stimulation. We further normalized this ratio to the vehicle group. Each treatment group comprises of measurements performed over n = 7–16 individual wells of a 96-well plate. Plotted is the mean and standard error. *indicates p < 0.05 while ns indicates p > 0.05

Full size image

Next, we examined biophysical mechanisms of barrier protection for the Tie2 receptor agonist, Angpt-1. In cells, tissue, organs, and animals, Angpt-1 has been demonstrated to counteract permeability and weakening of the endothelial barrier [31]. While roles for actin stress fiber rearrangements, junctional remodeling, and mechanotransduction have been implicated [47,48,49,50,51,52], in the absence of direct mechanical measurements, the underlying biophysical mechanisms remain unclear. Based on typically used concentrations [35, 48, 53,54,55,56,57,58,59], and additional dose–response measurements (Supplementary Fig. S4), we picked a sub-maximal dose of 300 ng/ml for further evaluation. Individual wells of a single 96-well plate (Young’s Modulus = 3 kPa) were co-stimulated with either vehicle or Angpt-1 together with the barrier disruptive agent, thrombin (1 U/ml; 30 min) (Fig. 2). In accordance with previous findings [35], Angpt-1 co-stimulation reduced thrombin-induced gap formation (Fig. 2A). While this barrier-protective effect, including time-course and efficacy, was recapitulated by ECIS (Fig. 2C, D), it was not correlated with a reduction in cell contractile force (Fig. 2B) or in the orientation and expression of the F-actin cytoskeleton (Supplementary Fig. S2). However, it was associated with an increase in VE-cadherin expression at the cell–cell junctions (Supplementary Fig. S5).

Fig. 2
figure 2

Angpt-1 reduces thrombin-induced gap formation but does not reduce cellular contractile force. ECs in a 96-well plate were incubated with either vehicle, vehicle + thrombin (1 U/ml), Angpt-1 (300 ng/ml), or Angpt-1 + thrombin for 30 min. A While paracellular gaps were reduced by Angpt-1 co-treatment B cellular contractile forces were not. For gaps, we report changes normalized to vehicle. For contraction, quantities have been normalized to t = 0 min values and further to the vehicle group. Each treatment group comprises of measurements performed over n = 7–16 individual wells of a 96-well plate. In all bar plots, shown is the mean and standard error. *indicates p < 0.05. C ECIS measurements confirm mini-XPerT findings. Each curve is averaged over n = 3–5 individual wells. The gray line marks the time point of 30 min post treatment; this time point corresponds to gap measurements with mini-XPerT. D Shown are corresponding resistance values. Plotted is the mean and standard error. *indicates p < 0.05 while ns indicates p > 0.05

Full size image

To demonstrate the versatility of mini-XPerT, we examined other physiologically-relevant endothelial leak provocateurs that are commonly utilized in barrier function studies: (1) Gram-negative endotoxin lipopolysaccharides (LPS; 100 ng/ml, 4 h) [48, 60] and (2) the canonical host cytokine, tumor necrosis factor alpha (TNFα; 20 ng/ml, 4 h) [48, 61]. Notably, their barrier disruptive activity manifests after several hours (e.g., [48, 62]), as opposed to thrombin, whose impact is acute, in the order of minutes (e.g. Fig. 2C). We evaluated these triggers of permeability on soft (0.3 kPa) and stiff (3 kPa) substrates with and without Y-27632 pre-treatment or Angpt-1 co-stimulation (Fig. 3). Paracellular gaps were significantly enhanced by TNFα/LPS stimulation and were largely reduced by the additional presence of Y-27632 or Angpt-1. Correspondingly, Y-27632 reduced cell contractile forces (Supplementary Fig. S6) while Angpt-1 did not (Supplementary Fig. S7). Taken together, these findings reveal the generality of the barrier-protective mechanisms of Y-27632 and Angpt-1.

Fig. 3
figure 3

Barrier protection by Y-27632 is correlated with a reduction of cellular contractile forces, whereas barrier protection by Angpt-1 is not. Plotted is gaps vs. contraction for the indicated conditions. For gaps, we report changes normalized to vehicle. For contraction, quantities have been normalized to t = 0 min values and then to the vehicle group. Each treatment group comprises of measurements performed over n = 7–16 individual wells of a 96-well plate. Plotted is the mean and standard error. Symbols in red, blue, and green correspond to the LPS/TNFα/thrombin, Y-27632 and Angpt-1 treatments, respectively. The ellipses highlight contractile force dependent (blue) or independent (green) reduction of thrombin-, LPS, and TNFα-induced gap formation. A lack of concordance for the EC response to TNFα on the soft substrate remains unclear (Supplementary Fig. S8)

Full size image

If not through contractile force reduction, how then does Angpt-1 promote barrier defense? This question is of significant interest. Given that Angpt-1 has been shown to counteract endothelial barrier disruption triggered by numerous, unrelated permeability mediators ranging from VEGF to anthrax lethal toxin to thrombin [35, 48], the most widely invoked explanation is that Angpt-1 signaling modulates one or more common downstream effectors of barrier function [31]. We evaluated the two major candidates for this conserved action: (1) F-actin cytoskeletal rearrangement, and (2) VE-cadherin accumulation at the cell–cell junction. We discovered a dominant role for the latter (Supplementary Fig. S5), consistent with the postulated role for Tie2 signaling for VE-cadherin stabilization in vivo [48, 63]. Unexpectedly, Angpt-1 did not prevent agonist-induced enhancement of F-actin (Supplementary Fig. S2) or contractile force (Fig. 2 and Supplementary Fig. S2).

In vascular biology, the dogma for EC barrier function has long held that junctional mechanisms are typically coupled to contractile force-dependent F-actin rearrangements. The present results, in contrast, are the first to our knowledge, that uncouple these two cellular processes as they relate to Angpt-1-induced barrier protection. If actin-cytoskeletal rearrangements can be uncoupled from junctional pro-barrier mechanisms in response to physiological stimuli, this could have implications for other cell types that exert barrier function such as squamous epithelium in the skin, columnar epithelium in the GI tract, and cuboidal epithelium in ducts and acinar structures of secretory organs.

Our data do not rule out the possibility that Angpt-Tie2 signaling is reorganizing forces on a much smaller scale, for example, in the peri-junctional region where the Rho family GTPase Rac1 has been shown to act in ECs [64]. Although beyond the resolution of the current technique, how such localized force reorganization could impact VE-cadherin accumulation at the cell–cell junction remains unclear.

We propose three important future directions for mini-XPerT. First, by combining mini-XPerT with monolayer stress microscopy [19], we can also determine intercellular stresses and its consequent effects on gap formation. This will require improved spatial resolution for traction measurements as well as clearly defined monolayer boundaries that can be specified through micropatterning procedures [12, 18, 19]. Second, by imaging gap formation in real-time, we can even more precisely correlate force and gap dynamics. To this end, we are currently evaluating cellular based markers to be able to delineate gaps in real-time. Finally, it would be ideal to combine mini-XPerT with concomitant measurements of molecular signaling.