Substrate stiffness modulates migration and local intercellular membrane motion in pulmonary endothelial cell monolayers

Abstract

The pulmonary artery endothelium forms a semipermeable barrier that limits macromolecular flux through intercellular junctions. This barrier is maintained by an intrinsic forward protrusion of the interacting membranes between adjacent cells. However, the dynamic interactions of these membranes have been incompletely quantified. Here, we present a novel technique to quantify the motion of the peripheral membrane of the cells, called paracellular morphological fluctuations (PMFs), and to assess the impact of substrate stiffness on PMFs. Substrate stiffness impacted large-length scale morphological changes such as cell size and motion. Cell size was larger on stiffer substrates, whereas the speed of cell movement was decreased on hydrogels with stiffness either larger or smaller than 1.25 kPa, consistent with cells approaching a jammed state. Pulmonary artery endothelial cells moved fastest on 1.25 kPa hydrogel, a stiffness consistent with a healthy pulmonary artery. Unlike these large-length scale morphological changes, the baseline of PMFs was largely insensitive to the substrate stiffness on which the cells were cultured. Activation of store-operated calcium channels using thapsigargin treatment triggered a transient increase in PMFs beyond the control treatment. However, in hypocalcemic conditions, such an increase in PMFs was absent on 1.25 kPa hydrogel but was present on 30 kPa hydrogel—a stiffness consistent with that of a hypertensive pulmonary artery. These findings indicate that 1) PMFs occur in cultured endothelial cell clusters, irrespective of the substrate stiffness; 2) PMFs increase in response to calcium influx through store-operated calcium entry channels; and 3) stiffer substrate promotes PMFs through a mechanism that does not require calcium influx.

INTRODUCTION

The endothelium forms a semipermeable barrier that separates blood from the underlying tissue (1–3). The barrier properties are established by endothelial cells tethering to neighbors through adherens and tight junctions and to the basement membrane through focal adhesions (4–6). These structural elements are not static in space but rather remodel constantly (7, 8). This remodeling is sustained in part by the mechanical drive for endothelial cells to move forward into open space (3, 9). Motion at the cell periphery is seen during wound healing, angiogenesis, disruption, and resealing of the barrier following inflammation, and even in quiescence when mature junctions have already formed (10–14). Morphological adaptations are not shared among all mammalian cells. They are necessary for the endothelium to fulfill its essential role as a barrier and gateway controlling the bidirectional movement of water, solutes, and macromolecules.

Endothelial cells are anchored onto the surface of a complex cylindrical protein cast in vivo (5). The innermost ring of proteins in this cast represents the basement membrane, the mechanical properties of which are extensively characterized (15). The basement membrane stiffness of the pulmonary artery endothelium in vivo is 1–4 kPa, although this value increases with age and with chronic vascular disease, as seen in pulmonary arterial hypertension (PAH), where stiffness increases to 25–50 kPa (16–18). In PAH, elevated substrate stiffness is a potent, persistent, and cytoskeletal force-mediated endothelial barrier compromising stimulus (19–21). However, the impact of these physiologically relevant shifts in substrate stiffness on endothelial cell morphology, movement, and, in particular, morphological adaptations that sustain barrier integrity is unknown.

The morphological adaptation considered in our study can be grouped into two categories, based on the length scales of their impact. The first category of adaptation is at the length scale of a cell, which induces changes in cellular shape, changes in cellular size, and changes in cellular spatial position; for example, we can visualize how the location of a single cell changes over time. The second category of adaptation is at the subcellular length scale, which induces local motion of the cellular peripheral membrane; for example, we can visualize how the position of the cell’s peripheral membrane(s) changes over time. The cellular-level adaptations are most relevant in wound healing, and the subcellular-level adaptations are the most relevant in the context of barrier function.

The adaptations that cause changes in cellular size, shape, and location result from an accumulation of small length scale morphological fluctuations, and hence, represent persistence in these fluctuations. Here, we focus on the transient and small-length scale morphological fluctuations that occur at the cell periphery and define them as paracellular membrane fluctuations (PMFs). Across the vessel segment, the PMFs are heterogeneous, as evident from paracellular leak sites being localized (8, 10, 22, 23). Rac1 and reactive oxygen species-dependent local lamellipodia formation and low local VE-cadherin concentration-induced Junction Associated Intermittent Lamelliopodia (JAIL) formation are correlated with local intercellular gap formation and healing in a variety of endothelial cells (9, 24, 25). But lack of experimental data on PMFs has limited the progress in our understanding of spatiotemporal heterogeneity of barrier function.

Barrier integrity is controlled by two separate but related mechanical forces, including 1) centripetally directed tension within individual cells (26) and 2) force vectors generated by and coordinated through cell colonies (19). The cytoskeletal forces and their spatiotemporal heterogeneities have been measured in individual cells and in cell colonies (27–30). However, to reveal the molecular and mechanical signals relevant to the disruption of the endothelial cell barrier, the spatiotemporal heterogeneity of paracellular membrane fluctuations needs to be closely examined (10, 20, 22).

Interactions between adjacent cell membranes have been examined using high-resolution time-lapse imaging of the fluorescently labeled endothelial cytoskeleton or junctional proteins (9, 24). This approach illustrated the value of determining paracellular cytoskeletal reorganization as a means to visualize the dynamics of the paracellular gap between immediately adjacent cells. However, barrier properties are inhomogeneous across the cellular monolayer. Hence, paracellular reorganizations are also expected to be inhomogeneous.

To examine the spatial heterogeneity of PMFs, all intercellular interfaces across the imaged area must be assessed simultaneously. Cellular images acquired using transmitted light contain appreciable information about the paracellular region. Image pixels located in the paracellular region exhibit intensity fluctuations corresponding to local morphological fluctuations. Using these pixel intensity changes, we present a novel technique called the Paracellular Fluctuations Analyzer (PFA), which simultaneously quantifies PMFs across the entire imaged field of view.

We examined the effect of substrate stiffness as the mechanical stimulus on both small- and large-length scale morphological responses in pulmonary artery endothelial cells (PAECs). For the large-length scale properties, we quantified cellular size, shape, and speed, and for the local responses, we quantified PMFs as a function of the stiffness of the substrate on which the cells were cultured. In this study, the stiffness of in vivo basement membrane (1–4 kPa) was defined as “normal” stiffness. Our observations suggest two novel hypotheses relevant to any lung endothelial condition that involves stiffened pulmonary arteries. First, a substrate that is stiffer or softer than normal compromises the ability to repair an endothelial injury. Second, a stiffer substrate induces a novel calcium influx-independent mechanism of membrane movement.

METHODS

Cells

Primary pulmonary artery endothelial cells (PAECs) were acquired from the Cell Culture Core of the Center for Lung Biology at the University of South Alabama. Initially, these cells were isolated from the pulmonary truncus of a rat lung as previously described (2, 31). Experiments were performed on cell passages 8 through 18. Passaging was done on plastic culture dishes.

Culture Medium

The culture medium included Dulbecco’s Modified Eagle Medium (DMEM, Santa Cruz, Cat. No. sc-224478) with 10% fetal bovine serum (Peak Serum, 100% USA origin, Cat. No. PS-FB1) and 1% penicillin-streptomycin (Invitrogen, Cat. No. 15140.122).

Preparation of Polyacrylamide Hydrogels

The hydrogels were prepared using the protocol developed by Yeung et al. (32). Briefly, 35 mm dishes with a 22-mm glass surface (FluoroDish, World Precision Instruments, Lot No. 28032018) were treated with 200 µL of 0.1 M NaOH and left to dry overnight. On the next day, these dishes were treated with aminopropyl-trimethoxysilane (Sigma Aldrich, St. Louis, MO) for 5 min at room temperature and then rinsed thoroughly with deionized (DI) water. After removing excess water, each dish was treated with 0.5% v/v glutaraldehyde in PBS for 30 min, thoroughly rinsed with DI water, and then dried for at least 30 min. Solutions of acrylamide (Ac) and bis-acrylamide (Bis) (Bio-Rad, Hercules, CA) were diluted in DI water over a range of dilutions to yield the desired gel stiffness (0.22 kPa: 7.5% Ac and 3% Bis; 1.25 kPa: 6.2% Ac and 2.8% Bis; 4 kPa: 6.2% Ac and 2.9% Bis; and 30 kPa: 10% Ac and 4.7% Bis). The gel polymerization process was catalyzed using 5-μL 0.1% w/v ammonium persulfate (Bio-Rad, Hercules, CA) and 0.5 μL of 1X N,N,N-tetramethylethylenediamine (Bio-Rad, Hercules, CA) per milliliter of the gel mixture. Then 23 μL of the resulting mixture was placed at the center of each dish and covered with RainX (to make its surface hydrophobic) treated 18-mm glass coverslip (Electron Microscopy Sciences, Hatfield, PA) to cast ∼100-μm thick hydrogels.

After removing the RainX-treated coverslip, the top surface of the hydrogels was functionalized with amine groups by adding 200 µL of N-sulfosuccinimidyl-6-(4'-azido-2'-nitrophenyl-amino) hexanoate (SANPAH, CovaChem, LLC, Cat. No. 13414) treatment, which included 10-min exposure to ultraviolet light (375 nm wavelength). The hydrogels were then washed with 0.1 M [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid] HEPES (Gibco, Cat. No. 206872) buffer at a pH of 7.3 followed by washing with Dulbecco’s phosphate-buffered saline (DPBS, Gibco, Cat. No. 14190136). Then, type I collagen from rat tail with a concentration of 3.78 mg/mL (Corning, Cat. No. 354236) was diluted in a 1:30 ratio in DPBS. SANPAH-treated hydrogels were incubated overnight at 4°C with 1.5 mL of the diluted collagen solution. The next morning excess collagen solution was removed, and the dishes were rinsed with DPBS and were ready for the culture.

Culturing and Imaging Cells

Before culturing the cells, the collagen-coated hydrogels were incubated for at least an hour in serum-free DMEM (4.5 g/L glucose). PAECs were trypsinized from the plastic culture dishes and counted using the Countess Automated Cell Counter (Thermo Fisher Scientific). Cells were seeded on two types of 35 mm dishes: glass bottom dishes with collagen-coated hydrogels and plastic bottom dishes without additional treatment. Each dish contained 10⁵ cells in 2 mL of culture media (DMEM with 4.5 g/L glucose, 10% FBS, and 1% penicillin-streptomycin). Cells were incubated at 37°C with 5% CO₂-room air for 3–4 days. Responses were measured in 55%–65% confluent cell clusters (Supplemental Figs. S1 and S2; see https://doi.org/10.6084/m9.figshare.17098742).

Cellular size, shape, and speed were assessed on five substrates (i.e., collagen-coated hydrogels) with a shear modulus of 0.22, 1.25, 4, and 30 kPa and plastic (shear modulus of ∼1 GPa). Paracellular morphological fluctuations were assessed in cells cultured on two different substrates (hydrogels with a shear modulus of 1.25 and 30 kPa). Data acquisition for cellular size, shape, speed, and PMFs was conducted in a Nikon T2-Eclipse microscope with a stage-top environmental control chamber (Okolab, UNO Stage Top Incubator, UNO-T-H-CO2, version 02.16), which maintained tissue culture conditions (37°C, 21% O₂, 5% CO₂, and 74% N₂) throughout the experiment. Subconfluent regions were then imaged using the data acquisition setup appropriate for each of the two types of experiments (Table 1).

Table 1.

Two data acquisition setups used in the study

Examined Cellular Properties	Type of Imaging	Frame Interval of Time-Lapse Imaging, s	Duration of Time-Lapse Imaging, min	Microscope Optics
Size, shape, and speed	Phase contrast	300	360	Nikon T2-Eclipse microscope, ×20 objective with 0.45 NA and 8.2–6.9 WD, ×1.5 internal optical magnification
PMFs		10	15

Cellular size, shape, and speed were examined with the setup described in the first row, and PMFs were examined with the setup described in the second row. PMFs, paracellular morphological fluctuations.

Preparation of Calcium Buffer

Two different calcium buffer solutions were prepared: 2 mM Ca²⁺ buffer that mimics the normocalcemic condition and 0.1 μM Ca²⁺ buffer that mimics a hypocalcemic condition. The 2 mM Ca²⁺ buffer was prepared by adding 0.369 mL of 1 M Ca²⁺ in 500 mL of Hank’s Balanced Salt Solution [HBSS (1×) with calcium, Gibco, Cat. No. 14025092]. The 0.1 μM Ca²⁺ buffer was prepared by adding 25 μL of 2 mM Ca²⁺ buffer solution to 500 mL of HBSS (1×, without calcium, Gibco, Cat. No. 14175095).

Treatments to Modulate Cytosolic Calcium

To examine changes in PMFs associated with transient changes in endothelial junctional integrity, we treated the PAECs with 1 µM thapsigargin (TG; Sigma Aldrich St. Louis, MO, Cat. No. 2263454). TG inhibits the sarcoplasmic, endoplasmic reticulum ATPase, transiently increasing cytosolic calcium while depleting calcium in the endoplasmic reticulum. The resulting endoplasmic reticulum calcium depletion promotes the opening of membrane calcium channels. Such activation of store-operated calcium entry causes transient endothelial cell retraction with loss of cell-to-cell adhesion, followed by reestablishing junctional integrity (22, 33, 34).

On the imaging day, the culture medium was replaced with an HBSS buffer containing either 2 mM Ca²⁺, mimicking a normocalcemic condition, or 0.1 µM Ca²⁺, mimicking a hypocalcemic condition. After 2 min of imaging under baseline conditions, dimethyl sulfoxide (DMSO, Fisher Scientific, Cat. No. 67–68–5) or TG was added to the cells. Cells exposed to hypocalcemic medium were imaged for 2 min after TG treatment and then 10 mM Ca²⁺ was added to promote calcium influx through activated store-operated calcium entry channels.

Comparisons across Different Treatments and Conditions

We examined cellular response to three key treatments: 1) culturing cells on soft and stiff hydrogels, 2) altering extracellular calcium concentration, and 3) activating store-operated calcium entry channels. The responses from each of these three treatments were examined with respect to their corresponding references: 1) cellular response on 1.25 kPa hydrogel, 2) cellular response to normocalcemic conditions (2 mM Ca²⁺), and 3) cellular response to DMSO treatment. Cellular exposure to hypocalcemic condition was followed by replenishment of extracellular Ca²⁺, and the response to this subsequent treatment was examined with respect to the response to DMSO. The comparisons with the references were expressed as percent differences. For example, the size of the cells cultured on 30 kPa hydrogel (A₃₀) was compared with the size of the cells cultured on 1.25 kPa hydrogel (A_1.25) using expression [(A₃₀ − A_1.25)/A_1.25] × 100. The list of all the numbers obtained from such comparisons is presented in separate tables (Supplemental Tables S1–S4; see https://doi.org/10.6084/m9.figshare.16584056).

Analysis of Cellular Size, Shape, and Speed

Cellular size and shape were analyzed using a novel ImageJ-based custom software: Integrative Toolkit to Analyze Cellular Signals (iTACS, https://github.com/IntegrativeMechanobiologyLaboratory/iTACS) (29, 30). This software used distinct spatial variance in pixel intensity in the cell versus cell-free regions to identify the image regions containing cells. Both cell-free regions and small regions that contained single isolated cells were excluded from further analysis. Individual cells were detected using a series of image enhancement and object detection functions from ImageJ, including Gaussian blur and variance filters and maxima finding functions from the remaining image. This approach allowed us to detect individual cells with ∼90% accuracy; errors included missed, merged, or split cells. The size and shape of each identified cell were computed using the area (A) and circularity [C = 4$\text{π}$ (A/P²), where P is the perimeter of the cell] functions of ImageJ. The circularity property quantifies the degree to which the cellular morphology is circular. The circularity of magnitude 1 indicates perfectly circular morphology and a lower value indicates elongated or other noncircular morphologies that possess a large periphery for the cell area.

The Particle Image Velocimetry (PIV) plugin of ImageJ was used to quantify subcellular motion (35). In the PIV analysis, a progressively narrower cross-correlation window size from 512 pixels to 128 pixels was used to recover velocity patterns on uniform grid points with a spacing of 64 pixels. These grid point values of speed were used with the output of cell segmentation software to map the speed onto individual cells. This process was repeated for five separate time-lapse experiments (each from independent cultures) for the five different stiffness conditions.

The frame interval was 300 s and each time-lapse experiment generated ∼84 frames (Table 1). From each experiment, five frames were randomly selected, and from each frame up to 229 cells were considered for analysis. In total, 229 × 5 × 5 = 5,725 cells were analyzed in each experimental case. Cellular size, shape, and speed were quantified for the same cells.

Analysis of Paracellular Morphological Fluctuations

PMFs were quantified in three key steps. The first key step was the computation of time variation in pixel intensity. This involved two key operations: 1) normalizing the intensity range and 2) computing pixel intensity variability at each time instance (t). The pixel intensity range of each image was normalized to 0–255. To compute variability of pixel intensity, we collected intensities of the same pixel from five successive time instances including two preceding (${\rho }^{t-2},{\rho }^{t-1})$ and two succeeding (${\rho }^{t+1},{\rho }^{t+2})$ instances, and computed standard deviation [${\sigma }_i^t=STD({\rho }^{t-2},{\rho }^{t-1},{\rho }^t,{\rho }^{t+1},{\rho }^{t+2})$]. This standard deviation was computed using “Z-Project” function of ImageJ, and its value represented the amount of morphological activity at that pixel. The second key step was the creation of the mask of the intercellular regions. This step involved two key operations: 1) segmentation of images and 2) creation of a mask of the intercellular region. The segmentation of the phase-contrast images followed the procedure for detecting individual cells described in the first paragraph of this section (30). In these segmented images, the cell boundary pixels had a pixel value of one, and the cell interior or cell-free region pixels had a pixel value of zero. The cell boundary was also one pixel wide. This width was increased to ∼10 pixels using the “Dilate” function of ImageJ. The resulting image represented a mask ($D_i^t$) that allowed restriction of the subsequent analysis to the intercellular regions. The third key step was the computation of PMF intensity. This step involved two key operations: 1) application of the intercellular region mask and 2) computation of median pixel intensity variability at each time instance. The standard deviation across five successive images (${\sigma }_i^t)$ was multiplied by the intercellular region mask ($D_i^t).$ In the resulting image ($P_i^t=D_i^t\times {\sigma }_i^t$), the pixels corresponding to the cell free region and central region of cells had an intensity of zero and the pixels corresponding the intercellular region had an intensity (${\sigma }_i^t)$ (Supplemental Figs. S3–S6; see https://doi.org/10.6084/m9.figshare.16567749). The median value of $P_i^t$across the intercellular regions of an image was computed using the “Measure” function of ImageJ and defined as PMF intensity at the time instance $t$ (Fig. 1). These operations of the PFA technique were programmed in a custom ImageJ Macro (https://github.com/IntegrativeMechanobiologyLaboratory/PFA).

Figure 1.

Overview of the paracellular fluctuation analyzer technique. A: one of the two key inputs to the procedure was a sequence of phase-contrast images of the monolayer. B: the phase-contrast images were normalized to have intensity spanning the full range for an 8-bit unsigned integer image. A representative normalized image is inserted in the box. C: from these normalized image sets, we then took five sequential images as seen in the insert and computed the standard deviation intensity of each pixel over the five images. The standard deviation was the highest for pixels located near the intercellular boundaries. D: the second of the two key inputs was the segmented image of cells generated using a novel ImageJ-based custom software iTACS. This was mostly a black image where the intercellular boundary was identified as a one-pixel wide line white line. E: this segmented image was then dilated for 5 times to create a mask of the region near the intercellular boundaries. F: this mask is applied to the standard deviation image to limit the consideration of pixel intensity fluctuations to the intercellular regions. G: the median of this masked image was defined as the paracellular morphological fluctuation (PMF) intensity for that image. H: the PMF intensity for each experiment was plotted as a function of time t. I: transient changes in PMF intensity (ΔPMF) following treatments were examined by calculating the range and the duration of ΔPMF. The same steps and procedure were followed for each independent experiment. The PFA software can be obtained from the GitHub repository https://github.com/IntegrativeMechanobiologyLaboratory/PFA; and the cell segmentation software can be obtained from the GitHub repository https://github.com/integrativemechanobiologylaboratory/iTACS. iTACS, Integrative Toolkit to Analyze Cellular Signals; PFA, Paracellular Fluctuations Analyzer.

Cellular response to treatments caused transient changes in PMF intensities (ΔPMF). Such changes were quantified using two parameters: 1) range of ΔPMF defined as the maximum change in PMF intensity compared with the pretreatment value and 2) duration of ΔPMF defined as time over which PMF intensity increases from its pretreatment level and reduces back by 80% of the range of ΔPMF.

Measurement of Fluctuation at Cell Center

The procedure to measure fluctuation at the cell center was largely similar to that used to measure paracellular membrane fluctuations. The only difference was that the mask was of the central region of the cell and not the intercellular region. The altered step was implemented by inverting the mask of the intercellular region and narrowing the resulting binary image by 10 pixels using “Erode” function of ImageJ.

Statistical Analysis

GraphPad Prism v9.2.0 was used for all statistical analyses. For each condition, four or five experiments were conducted from independent cultures. The outcomes from such four or five experiments were combined to compute the mean ± standard deviation. Substrate stiffness dependence of cellular morphology and motion was analyzed for statistical significance using repeated-measures one-way ANOVA with Dunnett’s multiple-comparison test as the post hoc analysis. Substrate stiffness dependence and treatment dependence of cellular PMFs were analyzed for the statistical significance using repeated measure one-way ANOVA with Tukey’s multiple-comparison test as the post hoc analysis. A value of P < 0.05 was accepted as statistically significant. A significant difference is expressed in the figures or figure legends as ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, and ∗∗∗∗P < 0.0001.

RESULTS

PAECs Were Most Active on Hydrogels with a Substrate Stiffness Similar to a Healthy Pulmonary Artery

Cell-matrix interactions regulate cell morphology and cell movement (14, 36). However, the impact of substrate stiffness on PAEC morphology remains poorly understood. To understand how substrate stiffness affects PAECs, PAECs were grown to 55%–65% confluence either on collagen-coated hydrogels with stiffness ranging from 0.22 to 30 kPa or on plastic (Supplemental Fig. S2; see https://doi.org/10.6084/m9.figshare.17098742). Among these substrates, hydrogels with 1–4 kPa stiffness mimic the stiffness of the endothelial substrate in a healthy pulmonary artery. In contrast, hydrogels with 30 kPa stiffness mimic the stiffness of the endothelial substrate in a hypertensive pulmonary artery (17, 20, 37).

PAEC morphology was highly sensitive to substrate stiffness (Fig. 2, A–E, Supplemental Videos S1–S5; see https://doi.org/10.6084/m9.figshare.16563912). The stiffer the substrate, the larger and more variable was the cellular spread area (Fig. 3A, Supplemental Table S1; see https://doi.org/10.6084/m9.figshare.16584056). This observation is consistent with observations made using sparse bovine aortic endothelial cells (38). In contrast to the monotonic relationship between substrate stiffness and cellular spread area, the relationships between substrate stiffness and cellular circularity [defined as, Circularity = 4$\text{π\hspace{0.17em}}$(Area/Perimeter²) or speed were biphasic; Fig. 3, B and C]. The peak of this biphasic response was observed for the PAECs cultured on 1.25 kPa hydrogels. Compared with the cells cultured on softer or stiffer hydrogels, the cells cultured on 1.25 kPa hydrogels had the least circular geometry and the fastest migration speed (Fig. 3, B and C). The cells cultured on 1.25 kPa hydrogels migrated ∼66 ± 19% faster than they did on a 0.22 kPa hydrogels and ∼50 ± 13% faster than they did on a plastic surface (Supplemental Table S1; see https://doi.org/10.6084/m9.figshare.16584056). The observation that cells with the least circular geometry also moved the fastest is consistent with PAECs cultured on 1.25 kPa hydrogels being the farthest away from the jamming transition (39). Overall, these cellular morphology and migration responses showed that subconfluent PAECs were most active on the substrate stiffness consistent with that of a healthy artery (17).

Figure 2.

Subconfluent pulmonary artery endothelial cells clusters cultured on polyacrylamide hydrogels of different stiffness. PAECs were seeded from passages 8 through 18 with the density of 10⁵ cells/dish either on collagen-coated hydrogel or on an untreated plastic culture dish. Phase-contrast images of about 55%–65% confluent monolayers 3–4 days after seeding on (A) 0.22 kPa, (B) 1.25 kPa, (C) 4 kPa, and (D) 30 kPa shear modulus hydrogel and (E) on plastic (∼1 GPa shear modulus). These images were taken with a Nikon T2-Eclipse microscope under a stage-top environmental control chamber (Okolab, UNO Stage Top Incubator, UNO-T-H-CO2, Version 02.16), which maintained tissue culture conditions (37°C, 21% O₂, 5% CO₂, and 74% N₂). extra long working distance, ×20/0.45 objectives with ×1.5 internal optical magnification were used to take images. Images are representative of five separate experiments. The scale bar represents a length of 50 µm. PAECs, pulmonary artery endothelial cells.

Figure 3.

Effect of substrate stiffness on the size, shape, and speed of subconfluent PAECs. Phase-contrast images of PAECs were acquired at a 5-min interval for 7 h by a Nikon T2-Eclipse microscope under a stage-top environmental control chamber (Okolab, UNO Stage Top Incubator, UNO-T-H-CO2, Version 02.16), which maintained tissue culture conditions (37°C, 21% O₂, 5% CO₂, and 74% N₂). These phase-contrast images were used to examine cellular size, shape, and speed as a function of substrate stiffness. These properties were quantified using a novel ImageJ-based custom software iTACS. The dot plot of (A) cellular spread area, (B) cellular circularity, and (C) cellular speed as a function of the shear modulus of the substrate on which the PAECs were cultured. The panels on the right side show the distribution of each property for each experimental condition. These frequency distribution graphs were nonlinear fitted with Gaussian least squares fit (R² > 90). The dot points in each condition for each property represent about 5,725 individual cells. These individual cell data came from 229 cells from each of five randomly selected frames from each of five independent experiments (i.e., 229 cells × 5 frames × 5 independent experiments = 5,725 individual cells). The central horizontal line that is overlapped with the dot points is a mean and the top and bottom horizontal lines that are overlapped with the dot points are standard deviation. The significance of the differences between examined conditions was analyzed using repeated measures one-way ANOVA with Dunnett’s multiple-comparison test as post hoc analysis (**P < 0.01, ****P < 0.0001). The cells examined were between passages 8 and 18, and each experiment contained cells the same passage level. iTACS, Integrative Toolkit to Analyze Cellular Signals; PAECs, pulmonary artery endothelial cells.

Signal-to-Noise Ratio of PFA

The PFA technique, developed to assess morphological movements in the intercellular regions, is based on the time variations in pixel intensity of a phase-contrast image. Such pixel intensity variations can occur not only from local cellular motion but also from noise inherent to the camera, disturbance in the culture medium, or fluctuations in the intensity of the transmitted light. As a result, even the region that does not have cells will provide a nonzero standard deviation in pixel intensities over five successive frames (i.e., the quantity we defined as PMF intensity; Fig. 4, A and B). Such PMF intensities represent the noise floor of the technique. We quantified this noise and assessed the signal-to-noise ratio for the PMFs as the ratio of the median value of ${\sigma }_i^t$ in the intercellular region when compared with the results in the cell-free region.

Figure 4.

Baseline PMF intensity under normocalcemic and hypocalcemic medium. Phase-contrast images of PAECs were acquired every 10-s interval for 15 min by a Nikon T2-Eclipse microscope. These images were used to examine baseline or PMFs on two different stiffness: 1.25 and 30 kPa. The PMF intensity was quantified using a novel ImageJ-based custom software: Paracellular Fluctuations Analyzer (PFA). The time trace of the PMF intensity for PAECs cultured on collagen-coated 1.25 kPa (solid black square dots) or 30 kPa (hollow gray square dots) hydrogels in a medium containing (A) 2 mM Ca²⁺ and (B) 0.1 µM Ca²⁺. Each dot points on the time trace represent mean PMF intensity from four different experiments and the error bars represent standard deviation. C: the dot plot of the PMF intensity of the data shown in A and B. In this plot, each dot point represents PMF intensity at different time over 15-min time interval. The central horizontal line that is overlapped with the dot points, is a mean and the top and bottom horizontal line, that is overlapped with the dot points, are standard deviation. The statistical significance was analyzed using one-way ANOVA with Kruskal–Wallis multiple-comparisons test as the post hoc analysis (***P < 0.001 and ****P < 0.0001). The cells examined were between passages 8 and 18, and each experiment contained cells at the same passage level. PAECs, pulmonary artery endothelial cells; PMF, paracellular morphological fluctuation.

The PFA had a signal-to-noise ratio of 2 ± 0.29. Following DMSO and other treatments, the noise level changed transiently. But this change was consistent and independent of the type of treatment (Supplemental Figs. S7 and S8; see https://doi.org/10.6084/m9.figshare.19878733). As such, transient changes in PMF intensity in response to change in extracellular calcium concentration or activation of store-operated calcium entry channels include a noise-associated constant shift (Figs. 5A and 6A).

Figure 5.

Transient changes in PMF intensity in normocalcemic medium. Phase-contrast images of PAECs were acquired every 10-s interval for 15 min by a Nikon T2-Eclipse microscope. These images were used to examine the transient changes in PMF intensity under normocalcemic (i.e., 2 mM Ca²⁺) medium for PAECs cultured on collagen-coated 1.25 and 30 kPa hydrogels. The PMF intensity was quantified using a novel ImageJ-based custom software: Paracellular Fluctuations Analyzer (PFA). A: the time trace of the PMF intensity for PAECs cultured on collagen-coated 1.25 kPa (solid black square dots) or 30 kPa (hollow gray square dots) hydrogels in a medium containing 2 mM Ca²⁺ and treated with DMSO at 120 s. Transient changes in PMFs (ΔPMF) were assessed using (B) the range of change of PMFs (ΔPMF) after DMSO treatment and (C) the duration of change of PMFs (ΔPMF) after DMSO treatment. D: the time trace of the PMF intensity for PAECs treated with 1 µM TG at 120 s. The transient change in PMFs was assessed using (E) the range of change of PMFs (ΔPMF) after 1 µM TG treatment and (F) the duration of change of PMFs (ΔPMF) after 1 µM TG treatment. The squares dots overlapped on the top of bars represent the range and duration from individual experiment. Each condition was examined using five independent experiments (n = 5). The height of the bar represents a mean and two horizontal lines above and below this mean represent standard deviation. The statistical significance was analyzed using repeated measure one-way ANOVA with Tukey’s multiple-comparisons test as the post hoc analysis (ns, no significance, *P < 0.05, and ***P < 0.001). PAECs, pulmonary artery endothelial cells; PMF, paracellular morphological fluctuation; TG, thapsigargin.

Figure 6.

Transient changes in PMF intensity in hypocalcemic medium. Phase-contrast images of PAECs were acquired every 10-s interval for 15 min by a Nikon T2-Eclipse microscope. These images were used to examine transient changes in PMF intensity under hypocalcemic (i.e., 0.1 µM Ca²⁺) medium for PAECs cultured on collagen-coated 1.25 and 30 kPa hydrogels. The PMF intensity was quantified using a novel ImageJ-based custom software: Paracellular Fluctuations Analyzer (PFA). A: the time trace of the PMF intensity for PAECs cultured on collagen-coated 1.25 kPa (solid black square dots) or 30 kPa (hollow gray square dots) hydrogels in a medium containing 0.1 µM Ca²⁺ and treated with DMSO at 120 s. The transient change in PMFs (ΔPMF) after DMSO treatment was assessed using (B) the range (ΔPMF) and (C) the duration of ΔPMF. D: the time trace of the PMF intensity for PAECs treated with 1 µM TG at 120 s followed by elevation of extracellular Ca²⁺ to 10 mM at 240 s. The transient change in PMFs following each of the two treatments was assessed using (E) the range of change of PMFs (ΔPMF) after 1 µM TG and 10 mM Ca²⁺ treatment and (F) the duration of change of PMFs (ΔPMF) after 1 µM TG and 10 mM Ca²⁺ treatment. The squares dots overlapped on the top of bars represent the range and duration from individual experiment. Each condition was examined using five independent experiments (n = 5). The height of the bar represents a mean and two horizontal lines above and below this mean represent standard deviation. The statistical significance was analyzed using repeated-measure one-way ANOVA with Tukey’s multiple-comparisons test as the post hoc analysis (ns, no significance, *P < 0.05, and ***P < 0.001). PAECs, pulmonary artery endothelial cells; PMF, paracellular morphological fluctuation; TG, thapsigargin.

Changes in Morphological Fluctuations Were Not Detected at the Cell Center

In this method, the pixel intensity fluctuations observable from the phase contrast image are commonly more intense at the cell periphery than they are at the cell center, since at the cell periphery, the cell membrane frequently moves across adjacent pixels. To closely examine this qualitative observation, we quantified pixel intensity fluctuations at the cell center in a manner similar to PMF intensity calculations. The PMF intensity at the cell center was not sensitive to the substrate stiffness (Supplemental Figs. S7 and S8; see https://doi.org/10.6084/m9.figshare.19878733). Thus, changes in the morphological fluctuations at the cell center were of the same level as the noise in the technique.

Baseline PMF Intensity Was Decreased by Exposure to Low Extracellular Calcium

As discussed earlier, the large-length scale motion represented by cellular speed of PAECs was sensitive to the substrate stiffness, but how it affects the local intercellular motions of PAECs is poorly understood. To explore how both substrate stiffness and extracellular calcium concentration impact the local intercellular motions (i.e., PMFs) of PAECs, we seeded cells on 1.25 or 30 kPa stiffness hydrogels and maintained extracellular calcium at either 0.1 or 2 mM. Under normocalcemic conditions (i.e., 2 mM extracellular Ca²⁺), the PAECs cultured on 1.25 kPa hydrogels exhibited 12 ± 8% higher PMF intensity compared with those cultured on 30 kPa hydrogels (Fig. 4, A and C and Supplemental Table S3; see https://doi.org/10.6084/m9.figshare.16584056). Reducing extracellular calcium inhibited PMFs when cells were grown on collagen-coated 1.25 and 30 kPa hydrogels (Fig. 4C and Supplemental Table S3). Under hypocalcemic conditions (i.e., 0.1 µM extracellular Ca²⁺), the PAECs cultured on 1.25 kPa hydrogels exhibited 8 ± 5% lower PMF intensity compared with those cultured on 30 kPa hydrogels (Fig. 4, B and C and Supplemental Table S3). Thus, under normocalcemic conditions, baseline membrane fluctuations were higher when PAECs were cultured on substrates with stiffness of a healthy artery as opposed to stiffness of a hypertensive artery. In contrast, under hypocalcemic conditions, baseline membrane fluctuations were higher when PAECs were cultured on substrates with stiffness of a hypertensive artery, revealing a novel calcium influx independent mechanism of membrane fluctuations. At present, the reason for this difference is not understood.

Thapsigargin Transiently Increased PMF Intensity

Since either a disturbance in the culture medium or fluctuations in the intensity of the transmitted light can increase the PMF signal, we first tested whether PMF intensity is affected by the addition of a fluid. To quantify this response, we added DMSO (TG solvent) in solution to 60%–65% confluent PAECs cultured on collagen-coated 1.25 or 30 kPa hydrogels. DMSO treatment triggered a transient increase in PMF intensity in cells grown on both types of hydrogels (Fig. 5A).

The transient increase in PMF intensity (ΔPMF) was analyzed using two properties: 1) range of ΔPMF (Fig. 5B) and 2) duration of ΔPMF (Fig. 5C). Under normocalcemic conditions, the increase in the ΔPMF duration was 37 ± 9% higher in cells cultured on 1.25 kPa hydrogels than it was in cells cultured on 30 kPa hydrogels (Fig. 5C and Supplemental Table S4; see https://doi.org/10.6084/m9.figshare.16584056). This difference was expected as a similar distinction existed in the PMF intensity without any fluid addition (Fig. 4, A and B). The precise source of these DMSO-induced changes in PMF intensity remains unclear; yet we anticipate that it is due to fluid disturbance-induced transient changes in image brightness, or potentially, due to shear stress-sensitive endothelial responses that are most pronounced when cells are grown on substrates with a 1.25 kPa stiffness.

Thapsigargin leads to the activation of store-operated calcium channels, which induce interendothelial cell gaps and increase macromolecular permeability (40). TG was compared with the DMSO treatment-induced changes in PMF intensity. TG increased PMF intensity and this increase was sensitive to substrate stiffness (Figs. 5, A and D and 7). Under the normocalcemic condition, the range of ΔPMF following TG treatment was 31 ± 7% larger in the cells cultured on 1.25 kPa hydrogels than it was in cells cultured on 30 kPa hydrogels (Fig. 5E). The duration of the ΔPMF following TG treatment was 45 ± 2% larger in the cells cultured on 1.25 kPa hydrogels than it was in the cells cultured on 30 kPa hydrogels (Fig. 5F). TG treatment-induced changes vanished about twofold slower than the DMSO treatment-induced changes (Figs. 5, C and F and 7). Thus, TG treatment increased activity of the cell membrane, and this increase in activity was long lasting, especially when the cells were grown on a substrate with healthy artery stiffness.

Figure 7.

Transient changes in PMF intensity with respect to DMSO treatment. The transient changes in PMF intensity following thapsigargin (TG) and calcium replenishment treatments were analyzed as a response relative to that of DMSO treatment. For this, the data used for Figs. 5 and 6 were further analyzed in which the range and duration of ΔPMF following DMSO treatment was subtracted from the range and duration of ΔPMF following TG and calcium replenishment treatments. The relative values of (A) range and (B) duration of ΔPMF following TG and calcium replenishment treatments are shown for PAECs in extracellular calcium concentration of 2 mM (bars on left of each panel) and 0.1 µM (bars on right of each panel). Each condition was examined using five independent experiments (n = 5). The height of the bar represents a mean and two horizontal lines above and below this mean represent standard deviation. The statistical significance was analyzed using repeated-measure one-way ANOVA with Tukey’s multiple-comparisons test as the post hoc analysis (ns, no significance, *P < 0.05, **P < 0.01, and ***P < 0.001). PAECs, pulmonary artery endothelial cells; PMF, paracellular morphological fluctuation.

Low Extracellular Calcium Revealed a Calcium-Influx Independent PMF Regulation Mechanism

In low extracellular Ca²⁺ (i.e., 0.1 µM Ca²⁺ instead of 2 mM Ca²⁺), TG triggers Ca²⁺ release from intracellular stores that transiently increase cytosolic Ca²⁺; yet Ca²⁺ influx through store-operated calcium entry channels is absent and the transient increase in cytosolic Ca²⁺ is not sustained (41). Here, we examined whether Ca²⁺ release and/or Ca²⁺ influx is necessary for TG to increase PMFs. To test this idea, a standard recalcification protocol was utilized, where TG was added to cells incubated in 0.1 µM Ca²⁺, followed by addition of 10 mM Ca²⁺ (41, 42). In comparison with the DMSO control, TG treatment dramatically increased PMFs by 917 ± 71%, when cells were grown on a substrate with 30 kPa stiffness (Figs. 6 and 7). TG treatment induced only a modest increase in PMFs by 49 ± 14% when the cells were grown on a substrate with 1.25 kPa stiffness. Thus, high substrate stiffness reveals a calcium influx-independent mechanism that controls the paracellular membrane movement.

Subsequent addition of Ca²⁺ to the medium induced a second elevation in PMFs (Figs. 6, D–F and 7). However, in contrast to the initial addition of TG in low Ca²⁺, Ca²⁺ entry through open store-operated calcium entry channels increased PMFs by 160 ± 20% in cells grown on 1.25 kPa and by 615 ± 45% in cells grown on 30 kPa substrate stiffnesses in comparison with DMSO control. Thus, calcium influx through store-operated calcium entry channels rapidly stimulates paracellular cell movement. It is notable that whereas the TG-induced elevation in cytosolic Ca²⁺ is sustained for 15 min following the addition of 10 mM Ca²⁺ to the medium (40, 42), PMFs increased for only 3–4 min.

DISCUSSION

Pulmonary endothelial cells adhere to their substrate on the inner surface of the vessel wall. These cell-substrate adhesions are involved in generating the mechanical drive for large-length-scale morphological changes, such as migration (43). In addition to these cell-substrate adhesions, each cell adheres to its neighbors. These cell-to-cell adhesions are involved in the tight control of macromolecular and cell transport across the endothelium. A mechanism for such a controlled transport process is thought to include intercellular-force-induced fluctuation in the overlap of adjacent cell membranes (44). Indeed, endothelial cells possess an autonomous drive of their cell membranes to move forward to establish contacts at cell junctions. Using rat PAECs, we found that both migration and PMFs were strongly influenced by the stiffness of the substrate on which the cells adhered (45–48). These data suggested two novel hypotheses for substrate stiffness-associated changes in the ability of PAECs to repair after injury and maintain barrier function. First, substrates that are stiffer or softer than normal compromise the ability of endothelial cells to migrate. Second, a stiffer substrate under a hypocalcemic condition sensitizes the endothelium, so that release of calcium from the endoplasmic reticulum is sufficient to increase PMFs. The mechanisms underlying this phenomenon remain unknown.

The Need and Opportunities in Studying Subconfluent Endothelial Cell Clusters

Individual cells of a PAEC monolayer cultured from PAH subjects are morphologically similar to those in the monolayer cultured from control subjects (49). This observation raises the possibility that the effect of substrate stiffness on PAEC morphology is small when the cell is part of a confluent monolayer. But in subconfluent endothelial monolayers cultured from the bovine aorta, the substrate stiffness impacts cellular morphology (11). The availability of open space makes the motion of individual cells in a subconfluent monolayer less constrained than in a confluent monolayer. Specifically, the type of motion that is restricted in confluent monolayers and less so in subconfluent monolayers is cellular migration, which occurs over the timescale of minutes (21). However, the type of motion that is relatively less restricted in either state of confluence is the small-scale morphological changes in the intercellular region, which occur over the timescale of seconds and regulate barrier function (27, 36, 50, 51). Studying subconfluent cells, therefore, presents an opportunity to examine the effect of substrate stiffness on both large- and small-length-scale morphological changes, specifically, cellular size, shape and speed, and local motion in the PMFs. Nevertheless, to generate expectations for arteries with noninjured endothelium, the findings reported here must be examined in the confluent endothelium and then translated into the intact circulation.

PAECs Cultured on a Stiffer or Softer Substrate Exhibited Slower Motion

PAECs cultured on a stiffer substrate exhibited a larger spread area (Fig. 3A). The cellular spread area indicates the ability of the cell to form stable focal adhesions and exert forces on them and extend lamellipodia to form newer focal adhesions (52). The cellular shape indicates a complex set of cues that the cell receives from its environment (53). However, in the presence of neighbors over long distances, the cellular shape is determined by the jamming transition phenomenon (39). From the motion perspective, as the cellular monolayer approaches jamming transition, the motion of individual cells becomes increasingly correlated with that of their neighbors (54). As a two-dimensional sheet of cells approaches a jamming transition, mechanical properties such as shear modulus of the cell sheet diverge and cellular circularity approach a critical value of 0.87 (54). A lower value of circularity indicates cells farther away from the jamming transition. Consistent with being closer to the jamming transition, the PAECs on both softer (0.22 kPa) and stiffer (4 kPa, 30 kPa, and plastic) surfaces showed larger circularity (Fig. 3B). This observation coincided with the cells on both softer and stiffer substrates migrating slower (Fig. 3C). Overall, softer substrates limited cellular spreading, shape anisotropy, and migration but stiffer substrates promoted cellular spreading while still limiting their shape anisotropy and migration (see Supplemental Videos S1–S5; see https://doi.org/10.6084/m9.figshare.16563912). These findings are consistent with a recent Cellular Potts model-based simulation and emphasize the involvement of physical forces in regulating cellular shape, size, speed, and PMFs (55, 56).

It is notable that in most cases, endothelial isolates are subcultured and passaged on plastic dishes. The standard protocol of isolation and multiple passaging on plastic dishes might condition the PAECs to survive on a plastic surface. As a result, the reported findings represent the cellular characteristics retained after cells were subcultured on plastic dishes and then transiently placed on elastic hydrogel substrates (32, 57). Although the responses assessed here are over a few hours, it will be important to reevaluate the responses using the PAECs that have been passaged on a soft substrate multiple times. A similar reevaluation of the responses in freshly isolated cells and under conditions native to pulmonary endothelial cells, including extracellular matrix proteins and hemodynamic forces, would also be insightful.

A Novel Technique to Quantify Paracellular Morphological Fluctuations

The structural basis of a strong barrier includes the persistently overlapping intercellular boundaries (25, 58, 59). By contrast, a weak barrier involves processes ranging from the reduced persistence of membrane overlap with occasional creation of paracellular transport pathways to persistent paracellular transport passages that coalesce to form a large separation between adjacent cells. The structural basis of a strong barrier described earlier provides the clearest, albeit a static, picture of the barrier property (10, 60). However, the opening and closing rate of transient paracellular transport passages alone does not provide a complete structural description of a weaker barrier (61). Complementary processes that need to be examined for a more complete description include 1) forward and retractive motion of the paracellular membrane and 2) cytoskeletal reorganization that maintains or alters the intercellular overlap. When viewed through an optical microscope as a transmitted light image, these processes amount to changes in local pixel intensity. PFA is designed to quantify such pixel intensity changes. In principle, changes in PMF intensity analyzed with our current approach do not differentiate between strengthening or weakening alterations of barrier function. Although the association of PMFs with barrier function was not the focus of the current work, our findings motivate such studies. In such studies, the nature of barrier alteration can be incorporated into the PMF intensity change by computing the local velocity vectors of morphological fluctuations and multiplying the PMF intensity with the sign of the velocity component perpendicular to the intercellular boundary. This can be achieved by appropriately combining the PMFs assessed using the PFA program (GitHub repository, https://github.com/IntegrativeMechanobiologyLaboratory/PFA) with the relevant velocity vector analyzed using iTACS (GitHub repository, https://github.com/IntegrativeMechanobiologyLaboratory/iTACS) (30). Incorporation of such an ability to resolve the contribution of forward versus retractive membrane motion could be a useful advancement of the PFA program. Another useful advancement would be integrating a fluorescence-based assay for detecting intercellular leak sites and superimposing the PMF intensity map (9, 20, 24, 50).

Extracellular and Intracellular Calcium Induce Contrasting Paracellular Morphological Fluctuation Intensity Responses to Substrate Stiffness

In the response to contractility-inducing treatments, an increase in PMF intensity would include the contributions from neighbors retracting away from each other (Supplemental Videos S6–S13; see https://doi.org/10.6084/m9.figshare.16563915) (33). But DMSO treatment also elicited a small increase in PMF intensity (Figs. 5, A–C and 6, A–C). Such an increase in PMF intensity could potentially originate from two effects of fluid disturbance: 1) altered fluid shear stress inducing paracellular membrane fluctuations and 2) altered phase-contrast causing small and transient changes in the image brightness. However, these effects would also be present in the measured responses to TG and calcium replenishment treatments. Hence, the response to TG treatment and/or altered extracellular Ca²⁺ were analyzed with respect to DMSO (Fig. 7, A and B).

Under a hypocalcemic condition, TG treatment does not increase PMFs in PAECs cultured on 1.25 kPa hydrogels but it does increase PMFs in PAECs cultured on 30 kPa hydrogels (Fig. 6, D–F). Although 30 kPa stiffness is similar to that of a hypertensive artery, the cells used in this study were derived from a normotensive artery. Nevertheless, the reported finding and the cellular tendency to adapt to the stiffness of its substrate present the rationale for examining PMF intensity change in subconfluent PAECs cultured from hypertensive rat lungs (62). These studies may reveal a novel mechanism for enhanced endothelial permeability in a hypertensive artery with injured endothelium (62, 63).

Contribution of Intrinsic Properties of Individual Cells

In the data analysis, all cells that were part of a cluster were treated equally. In that, cells located at the margin or the interior of a cluster were treated the same. However, the behavior of an individual cell is not independent of its neighboring cells. In addition, the cells located at the margins (as opposed to the interior) of a cluster do not have neighbors on all sides and exhibit unique properties (63). The contribution of such unique cells to the observations reported here remains unknown and presents an opportunity for future research.

In summary, we have presented a novel technique that enables quantitative assessment of PMFs from a sequence of phase-contrast images. The key findings include PMFs occur in endothelium, irrespective of the substrate stiffness; PMFs increase in response to calcium influx through store-operated calcium entry channels; and stiffer substrate promotes PMFs through a calcium influx-independent mechanism. In the area of pulmonary arterial hypertension research, these findings motivate closer examination of 1) calcium influx-independent mechanisms of pulmonary artery endothelial cell barrier disintegrity driven by increased substrate stiffness and 2) the mechanisms for repairing endothelial injury.

SUPPLEMENTAL DATA

Supplemental Figs. S1 and S2: https://doi.org/10.6084/m9.figshare.17098742.

Supplemental Figs. S3–S6: https://doi.org/10.6084/m9.figshare.16567749.

Supplemental Figs. S7 and S8: https://doi.org/10.6084/m9.figshare.19878733.

Supplemental Tables S1–S4: https://doi.org/10.6084/m9.figshare.16584056.

Supplemental Videos S1–S5: https://doi.org/10.6084/m9.figshare.16563912.

Supplemental Videos S6–S13: https://doi.org/10.6084/m9.figshare.16563915.

GRANTS

This work was supported by Research and Scholarly Development Grant, University of South Alabama (to D.T.T.) and National Heart, Lung and Blood Institution Grants HL-66299 (to T.S.), HL-60024 (to T.S.), and HL-148069 (to T.S.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.S.P., T.S., and D.T.T. conceived and designed research; S.S.P., A.d., and S.S. performed experiments; S.S.P. and D.T.T. analyzed data; S.S.P., T.S., and D.T.T. interpreted results of experiments; S.S.P. and D.T.T. prepared figures; S.S.P. drafted manuscript; S.S.P., S.S., T.S., and D.T.T. edited and revised manuscript; S.S.P., T.S., and D.T.T. approved final version of manuscript.