Lamellipodia dynamics and microrheology in endothelial cell paracellular gap closure

Abstract

Vascular endothelial cells (ECs) form a semipermeable barrier separating vascular contents from the interstitium, thereby regulating the movement of water and molecular solutes across small intercellular gaps, which are continuously forming and closing. Under inflammatory conditions, however, larger EC gaps form resulting in increased vascular leakiness to circulating fluid, proteins, and cells, which results in organ edema and dysfunction responsible for key pathophysiologic findings in numerous inflammatory disorders. In this study, we extend our earlier work examining the biophysical properties of EC gap formation and now address the role of lamellipodia, thin sheet-like membrane projections from the leading edge, in modulating EC spatial-specific contractile properties and gap closure. Micropillars, fabricated by soft lithography, were utilized to form reproducible paracellular gaps in human lung ECs. Using time-lapse imaging via optical microscopy, rates of EC gap closure and motility were measured with and without EC stimulation with the barrier-enhancing sphingolipid, sphingosine-1-phosphate. Peripheral ruffle formation was ubiquitous during gap closure. Kymographs were generated to quantitatively compare the lamellipodia dynamics of sphingosine-1-phosphate-stimulated and -unstimulated ECs. Utilizing atomic force microscopy, we characterized the viscoelastic behavior of EC lamellipodia. Our results indicate decreased stiffness and increased liquid-like behavior of expanding lamellipodia compared with regions away from the cellular edge (lamella and cell body) during EC gap closure, results in sync with the rapid kinetics of protrusion/retraction motion. We hypothesize this dissipative EC behavior during gap closure is linked to actomyosin cytoskeletal rearrangement and decreased cross-linking during lamellipodia expansion. In summary, these studies of the kinetic and mechanical properties of EC lamellipodia and ruffles at gap boundaries yield insights into the mechanisms of vascular barrier restoration and potentially a model system for examining the druggability of lamellipodial protein targets to enhance vascular barrier integrity.

Figure 1

Decrease of effective gap diameter over time. (A) Optical phase microscopy images of gap closure for a paracellular gap prepared with microfabricated pillars using the nutrient-poor medium wash protocol. The white spots correspond to ECs detached from regions outside the gap during removal of the pillars and transferred to the gap by medium convection induced by air/CO₂ injection in the petri dish. Scale bar, 100 μm. The entire sequence of images is shown in Video S1. (B) The effective diameter (D_G) versus time (t) curve of this gap displays a steady linear decrease over most of its trajectory. The first image (t = 0) was acquired 25 min after removal of the pillars. The slope of this curve (RGC) increased by 60% after S1P injection. (C) Rates increased from 0.79 ± 0.11 μm/min pre-S1P to 1.12 ± 0.20 μm/min (41%) for the eight gaps prepared with the standard unwashed cell protocol and from 0.78 ± 0.16 to 1.15 ± 0.15 μm/min (48%) for washed ECs (N = 7). Removal of the outlier in the post-S1P distribution of unwashed gaps (Fig. S3) produced normal RGC distributions according to the Shapiro-Wilks test. The pre-S1P rates (0.71 ± 0.08 μm/min) were significantly lower than their post-S1P pairs (0.93 ± 0.07 μm/min) according to the paired sample t-test (∗p < 0.05). Inclusion of the outlier removed normality but did not change the conclusion from the t-test, as illustrated by (D) a plot of all post-S1P to pre-S1P ratios, which shows all gap closure rates were >1. Mean values ± SEs are shown in the box plots. (E) Phase images showing the area fluctuations that occurred when the cells crowded the gap for small gap diameters, where the curve often lost linearity. Scale bar, 20 μm.

Figure 2

EC motility during gap closure. (A and B) Positions of seven ECs around the gap tracked at selected time points during gap closure to create motility maps which provided EC trajectories (black and gray data points indicate pre- and post-S1P EC positions, respectively). Tracked ECs are indicated by the same numbers in (A) (t = 0) and (B). (C) Mean cell velocities (V_EC) and (D) post- to pre-S1P ratios of these velocities [$\frac{V_{EC}\left(Post-S1P\right)}{v_{EC}\left(Pre-S1P\right)}$] were obtained from this analysis. Results for 17 ECs from 3 different gaps show ECs increased their mean velocities by 27% after EC stimulation with S1P (from 0.64 μm/min pre-S1P to 0.79 post-S1P). The post-S1P EC velocities were statistically greater than the pre-S1P EC velocities at the ^∗p = 0.05 level, according to a paired sample t-test. Mean values ± SEs are shown in the box plots.

Figure 3

Kymograph analysis of fluctuating lamellipodia. (A) Bright-field images showing three stages of gap closure and the region (rectangle) where the kymographs were obtained. The entire sequence of images is shown in Video S2. This region is rotated in the kymographs, so that time is displayed in the x axis and displacements of the leading edge in the y axis. Scale bars, 20 μm. (B and C) Kymographs obtained in the same region pre- and post-S1P stimulation. An advancing ruffle is visible in dark contrast at the cell edge and fluctuating lamellipodia appear as smaller protruding peaks, which become larger post-S1P stimulation. Scale bars, 3 s/pixel along the x axis and 0.16 μm/pixel in the y axis. (D and E) Amplified images of the advancing cell fronts for the lamellipodia protrusion peaks indicated by arrows in (B and C), respectively. Scale bar, 5 μm. Protruding lamellipodia are visible in the pre- and post-S1P images but increase in size post-S1P challenge. (F–I) Boxplots of kinetic parameters pre- and post-S1P stimulation obtained from kymographs (N = 6 ECs). (F) Lamellipodia protrusion lengths, L. (G) Persistence times, T_p indicating the time lamellipodia remain at their maximum extension. (H) Protrusion and I retraction lamellipodia rates. Mean ± SE values are represented. Differences between L and T_p for pre- and post-S1P values reached statistical significance at the ^∗p = 0.05 level.

Figure 4

Optical phase, fluorescence, and AFM imaging of EC gaps. (A) Actin fluorescence images of closing gaps 15 min post-S1P challenge often display increased intensity at gap edges and are identified as ruffles (a few of them are indicated with white arrows). (B) Those identical regions appear in dark contrast utilizing optical-phase imaging. White arrows indicate the same regions as in (A). (C) Overlay of fluorescence (red) and phase contrast (gray) images in (A and C), respectively, shows the coincidence of actin-rich regions with those in dark phase contrast. (D) AFM height image and height cross section showing stripes exhibiting increased heights at gap edges. The same regions as above are indicated with white arrows. The measured heights are underestimates of the true EC heights due to the EC compression produced by the AFM tip during scanning. (E) Overlay of (B and E) displays the close proximity of actin-rich and elevated regions. (F and G) Results of colocalization analysis for the regions enclosed by white rectangles in (E) between the fluorescence (red) and AFM (green) channels. Costes masks (first column), Van Steensel’s cross-correlation functions (second column), and scatterplots (third column) are displayed as characterization of colocalization. Scale bars, 5 μm, 8.65 pixels/μm. (H) Values of Pearson coefficients (PCC), Mander’s coefficients (M1 and M2), and slopes of the linear fits to the scatterplots (m) for the four analyzed regions. This box plot shows the mean values ± SDs for the 4 regions. To see this figure in color, go online.

Figure 5

AFM time-lapse imaging of stiffness changes at the EC leading edge during lamellipodial protrusion and retraction. (A) Height, h and (B) Hertzian Young’s modulus, E_H. (C) h (black) and E_H (red) cross sections along the white dashed lines in (A and B). Each cross section is an average of five adjacent lines. (D) Plot of the time variation of lamellipodia area (circles) and Hertzian Young’s modulus (squares) evaluated for approximately N = 40 data points in each image within 5 μm from the cellular edge during the protrusion/retraction lamellipodial cycle shown in (A and B). Images were acquired with a precalibrated PFQNM-LC cantilever, using QI mode at 150 μm/s tip velocity, every 3 min 20 s, with optical images acquired occasionally between AFM images. Mean ± RMS values ($RMS=\sqrt{\frac{1}{N}\sum _{i=1}^N{E}_H^2}$) measured with the JPK Data Processing software are represented for E_H. Lamellipodia areas were measured using ImageJ. To see this figure in color, go online.

Figure 6

Decreased stiffness of expanding EC lamellipodia. (A) Representative force curves in R, L, and CB regions (only approach is shown). The deformation, δ, displayed in the x axis was obtained from δ = Z − Δ, where Z is the piezo displacement and Δ is the cantilever deflection. (B) Representation of the local uncompressed cellular height, h_U, obtained from h_U= δ + h, where h is the compressed cell height above the substrate measured from the AFM height image and δ is the local deformation at the maximum force obtained from the force curves. (C) h_U image of an upward-expanding EC. (D and E) Uncompressed heights in the (D) lamellipodia (L) and (E) lamella (CB) regions selected from (C). Ruffles were excluded from (D) and appear as “holes” (with h_U = 0) in the image. (F) Finite-thickness-corrected Young’s modulus calculated from fits in the 0–600 pN range. (G and H) E_Dimitriadis values in the same (G) lamellipodia and (H) lamella regions as in (D and E). Ruffles were excluded from this analysis. (I and J) Histograms of (I) h_U and (J) E_Dimitriadis for the L (gray bars) and CB (red bars) regions selected above. Gaussian fits to these histograms are shown in black and red, respectively. To see this figure in color, go online.

Figure 7

Morphology and stiffness of EC ruffles. (A) Schematic showing the relation between the uncompressed ruffle height, h_R, the ruffle deformation δ (obtained from the force curves), and the compressed height, h, extracted directly from the AFM height image. In addition to the depicted “forward” ruffle deformation, a “backward” deformation is equally possible a priori. (B) Scatterplot of the ruffle spring constant, k_R, versus h_R (N = 15 ruffles). Values are averages of the three tallest points. The ruffle deflections obtained from the force curves were added to the heights measured from the AFM image to find h_R. (C) SEM image (obtained at 5 kV) of a ruffle (∼500 nm in height) and a lamellipodium (∼250 nm length) protruding under the ruffle. The bright contrast in the ruffle indicates charging, and therefore poor electrical contact with the substrate.

Figure 8

Viscoelastic properties of lamellipodia and lamella at EC edges. (A) Variation of E₀, E′, and E″ in lamellipodia and lamella along the line cross section shown in Fig. S6. (B) Cross sections of α and Eʺ/E′ along the same lines shown above. (C) Variation of relaxation Young’s modulus, E₀, at t = 1 s with height in lamellipodia (black) and lamella (red). (D) Change of storage, E′ (full squares) and loss, E″ (open squares) moduli in lamellipodia (black) and lamella (red) regions obtained from E₀, α, and the effective frequency, f_eff, as described in materials and methods. (E) Power law exponents, α (full squares), and loss tangents, Eʺ/Eʹ (open squares), in lamellipodia (black) and lamella (red). N = 18 data points were evaluated in lamellipodia and lamella regions throughout (C–E). To see this figure in color, go online.