Tuning endothelial monolayer adhesion: a neutron reflectivity study

Abstract

Endothelial cells, master gatekeepers of the cardiovascular system, line its inner boundary from the heart to distant capillaries constantly exposed to blood flow. Interendothelial signaling and the monolayers adhesion to the underlying collagen-rich basal lamina are key in physiology and disease. Using neutron scattering, we report the first ever interfacial structure of endothelial monolayers under dynamic flow conditions mimicking the cardiovascular system. Endothelial adhesion (defined as the separation distance ℓ between the basal cell membrane and solid boundary) is explained using developed interfacial potentials and intramembrane segregation of specific adhesion proteins. Our method provides a powerful tool for the biophysical study of cellular layer adhesion strength in living tissues.

large endothelial and epithelial surfaces control fluid and molecular flow throughout tissues (1, 10, 11, 13). These biologic membranes, although composed of individual cells, have elaborate connections both between the neighboring cells and between the cells and the underlying extracellular matrix (10). Endothelial integrity and adhesion to the basal lamina are linked to diseases such as sepsis (13), acute respiratory distress syndrome (ARDS; refs. 2–3), atherosclerosis (5, 12), and tumor progression (8). In ARDS, the pulmonary vasculature becomes leaky leading to pulmonary edema and the accumulation of albumin in the alveolar lining fluid, severely curtailing oxygenation (22–3). Endothelial barrier dysfunction is a major factor of vascular leak and fluid accumulation in the lung. Tumor vasculature is known to have highly aberrant architecture that often leads to capillary leakiness and endothelial deadhesion and promotes metastatic disease (8). Lastly, atherosclerosis is also strongly linked to endothelial barrier dysfunction (5, 12).

Mechanical stress in the form of shear flow is a powerful factor impacting endothelial adhesion and permeability and may play an important role in the diseases discussed above (2, 3, 5, 8, 12, 13, 37). Although cell-substrate interactions are considered an integral part of the monolayer barrier, dynamic changes in cell membrane-substrate distances indicative of cell adhesion and macromolecule permeation remain virtually unexplored.

While techniques like atomic force microscopy (AFM), micropipette force balance, and contrast interference microscopy (4, 32) exist to probe single cell and subcellular adhesion, no tools are available to probe multicellular tissues. Single cell experiments are useful in certain systems, such as the now classic models of leukocyte adhesion; however, in the endothelium, understanding adhesion of confluent monolayers is a key. Cell-cell signaling via adherents junctions and cell-basal lamina adhesion via focal adhesions play key roles in endothelial biology including monolayer permeability and cytoskeletal rearrangements (1–3, 37).

Neutrons probe matter in a nonperturbative manner with minimal energy transfer; furthermore, because neutrons interact with nuclei and not electrons, chemical bonding is minimally affected (7, 20, 34). Using neutron reflectivity, we developed a novel tool that probes confluent endothelial monolayer adhesion under dynamic conditions and with nanometer resolution for the first time. Reflectometry has played an immense role in soft condensed matter physics, helping to elucidate the complex bulk and interfacial structures of many polymer, lipid, and liquid crystal systems (20). Its use in the most complex of soft matter, i.e., living biological systems, is a natural extension of the technique (34). In this work, we seek not only to perform a novel dynamical type of measurement but to correlate it to biological and physical understanding of endothelial adhesion.

MATERIALS AND METHODS

Cell Experiments

Human pulmonary artery endothelial cells (Lonza, Basel, Switzerland) at passages 6-8 were seeded onto optically pure single crystal quartz blocks and cultured at 37°C in a humidified CO₂ incubator. The quartz was chemically cleaned with chromic acid before deposition assuring that there are no surface active agents on the surface (34). Endothelial cells were grown on quartz until complete confluence under standard conditions (1–3). The quartz block was then transferred into a custom-built parallel plate fluid shear cell (20). The fluid perfusion cell was flushed with endothelial culture medium reconstituted in D₂O and supplemented with 5% bovine serum. The same D₂O reconstituted culture medium served as the perfusion fluid during the neutron experiment.

For static condition experiments, there was no flow in the chamber; for shear experiments, flow was adjusted to maintain a constant nonpulsatile 1.5-Pa shear stress. Importantly, the temperature change or shear conditions were implemented only upon transfer into the neutron experimental chamber; otherwise, the different cell systems were prepared under identical conditions. Neutron data collection was carried out from 3 to 5 h to obtain proper statistics. At the end of the experiments, we visually examined the quartz blocks under a cell-culture microscope and confirmed >90% cell surface coverage.

Cell staining experiments were carried out with standard techniques as detailed before (1–3). Actin cytoskeletal remodeling was monitored by immunostaining with Texas red phalloidin. Adherens junction remodeling was monitored by immunostaining of endothelial monolayers with β-catenin antibody. Shear stress-induced remodeling of focal adhesions was monitored by immunostaining of endothelial monolayers with paxillin antibody. Shear stress-induced activation of peripheral actin polymerization and enhancement of peripheral actin cytoskeleton were monitored by immunostaining of endothelial monolayers with cortactin antibody.

Neutron Reflectivity Analysis

Neutron measurements were performed at the Surface Profile Analysis Reflectometer (SPEAR), Los Alamos Lujan Neutron Scattering Center (7). The spallation neutron beam, after liquid hydrogen moderation, was directed onto the sample at a glancing angle (1.5 mrad), and the specular reflection was recorded by a time-of-flight (ToF), position-sensitive detector. Reflectivity is a function of the neutron momentum transfer vector, q_z, where q_z = 4π sin(θ)/λ, θ is the beam angle of incidence, and λ is the neutron wavelength (SPEAR produces a wavelength distribution in the 0.45–1.6 nm range). Figure 1 shows a schematic representation of the reflectivity setup. The reflectivity spectrum was modeled using an open source reflectivity package, Motofit, which runs in the Igor Pro environment (27). Motofit approximated the interface by a number of discrete layers with constant scattering length density and thickness. A theoretical reflectometry curve was calculated using Abeles matrix formalism (28). Both genetic optimization and Levenberg-Marquardt nonlinear least-square methods were employed to obtain the best fits for the NR data.

Fig. 1.

Schematic representation of the neutron shear cell experiment and the multiple length scales involved. Confluent layer of cells is grown on the quartz substrate that is inverted into the shear cell. A neutron beam enters through the quartz at a glancing angle and scatters from the multiple basal lamina and cell membrane interfaces. The reflectivities are modeled to generate real space images of the interfacial structure. Most importantly, the separation distance between the basal plasma membrane and quartz substrate (ℓ) is obtained; ℓ is an inherent measure of monolayer adhesion.

Three layers were used to model our cell interface within the Motofit platform: basal lamina, lipid plasma membrane, and deep cortical layer. The plasma membrane has a well-defined neutron scattering structure and is modeled using parameters from prior lipid bilayer reflectivity studies (29, 34). The cortical layer corresponded to the endothelial cytoplasm and was modeled as a diffuse water rich layer ∼10-nm thick. While the layer of greatest interest, the basal lamina was spectroscopically defined as the volume element separating the quartz/liquid interface and plasma membrane. The composition and thickness of this layer were systematically varied using a defined technique of ultrastructural refinement (see Ref. 29), generating a family of reflectivity curves from which the best fit model was selected (see Fig. 2).

Fig. 2.

Neutron reflectivity spectra (square with error bars) plotted as R·Q_z⁴ vs. Q_z at the conditions studied (A, C, E, and G). Black bold line is the best fit model from our analysis. Robustness of the fits is tested by varying the scattering light density and thickness of the basal lamina layer as shown with the corresponding plotted models. Right: calculated scattering length density profiles in real space (B, D, F, and H). The basal lamina thickness is indicated in each plot for the profile corresponding to the best-fit model.

RESULTS

Neutron Reflectometry

In this report, using neutron reflectometry, we measured the first ever nanometer level interfacial structure of confluent endothelial monolayers under dynamic shear stress conditions (1.5 Pa). Basal lamina thickness (ℓ) is an inherent measure of monolayer adhesion and is directly derived from modeling the reflectivity spectra (see Fig. 2). Under static conditions, endothelial monolayer adhesion decreases with increasing temperature: ℓ³⁷ = 60 nm while ℓ²⁵ = 20 nm. This trend reverses with shear stress, whereby at physiologic temperature adhesion is strongest: ℓ_s³⁷ = 20 nm while ℓ_s²⁵ = 70 nm. Importantly, these data reflect an average measure of adhesion over thousands of cells composing the monolayer, since the incident neutron beam footprint is O(10 cm²). Furthermore, since all systems are prepared under identical conditions, the changes in ℓ are reflective of differential system response to imposed temperature and shear conditions. Given the interconnectedness of cell signaling within endothelial monolayers, our system is more reflective of physiologic endothelium than methods probing single cell adhesion, e.g., reflection interference contrast microscopy (RICM; Refs. 4, 32).

Figure 3 shows diffuse neutron reflectivity data for the four systems studied. The off-specular diffuse data are plotted in standard fashion with Q_z on y-axis and Q_xy on x-axis. The high intensity peak at Q_xy = 0 corresponds to the specular reflection. The off-specular (diffuse) component is circled in each plot. Standard reflectivity theory shows that the intensity of off-specular scattering (Q_xy ≠ 0) directly corresponds to the degree of interfacial thermal fluctuations (16). Our data show that the off-specular peak enlarges with increasing temperature under static conditions while it substantially decreases with shear flow for both temperatures. Temperature, as expected, increases interfacial thermal fluctuations. Shear stress acts in a reverse manner and suppresses thermal fluctuations.

Fig. 3.

Off-specular diffuse data plotted in standard fashion. In each spectrum, the diffuse peak is encircled along Q_xy ≠ 0. Temperature, as expected, increases interfacial thermal fluctuations (diffuse peak enlarges under both static and shear conditions at 37°C compared with 25°C). Shear stress acts in a reverse manner and suppresses thermal fluctuations.

Endothelial Membrane Protein Rearrangement

Shear stress induces remodeling of actin cytoskeleton and cell junction complexes (1–3, 37). Endothelial monolayers are exposed to static conditions or to laminar shear stress (1.5 Pa) for 15 min and 4 h. Even on relatively short biologic time scales (4 h) relevant to the time scale of our neutron experiments, adherens junctions (β-catenin stain), and focal adhesions (paxillin stain) undergo cortical localization and formation of enhanced adhesion zones (see Fig. 4).

Fig. 4.

Top: shear stress induces remodeling of actin cytoskeleton and cell junction complexes. Endothelial monolayers were exposed to static conditions, or to laminar shear stress (1.5 Pa) for 15 min and 4 h. Actin cytoskeletal remodeling, monitored by immunostaining with Texas red phalloidin, shows the beginnings of stress fiber formation. Adherens junction remodeling was monitored by immunostaining of endothelial monolayers with β-catenin antibody. Shear stress-induced remodeling of focal adhesions was monitored by immunostaining of endothelial monolayers with paxillin (pen) antibody showing strong cortical concentration with shear. Shear stress-induced activation of peripheral actin polymerization and enhancement of peripheral actin cytoskeleton was monitored by immunostaining of endothelial monolayers with cortactin antibody. Bottom: schematic presentation of focal adhesion protein complexes organization and their linkage to extracellular matrix and intracellular actin cytoskeleton. Pxn, paxillin; 1, talin; 2, Src kinase; 3, p130Cas; 4, focal adhesion kinase (FAK); 5, vinculin; 6, Arp2/3 complex; 7, cortactin.

DISCUSSION

Tissue adhesion plays a major role in human pathology and disease. As discussed in the introduction, endothelial adhesion is linked to difficult clinical problems like ARDS, tumor metastasis, and cardiovascular disease. A major challenge in studying biologic adhesion is the lack of tools for measuring surface energies in living systems. Concepts of surface energy and adhesion are firmly rooted in the fields of thermodynamics and engineering. However, techniques developed for soft biologic interfaces like AFM or RICM are limited largely to single cell or subcellular adhesion making the application of thermodynamic models difficult. In this article, we studied endothelial adhesion on a tissue level. The precision of our neutron method allowed us to interpret the presented data within a thermodynamic framework. Development and validation of adhesion models in endothelial monolayers enrich the understanding of the complex biophysical and biochemical control parameters for endothelial adhesion.

Interfacial Adhesion Model

The interfacial potential energy density is written as a superposition of attractive (W_ad) and repulsive terms (E_r):

$$\Delta G_{\text{adh}}\sim W_{\text{ad}}(h)+E_{\text{r}}(h)$$ (1)

The different components are functions with multiple parameters; however, the main independent variable is the separation distance h between two adhered surfaces. The equilibrium separation distance (h_o) occurs where ΔG_adh has a minimum, i.e., ∂_hΔG_adh(h_o) = 0 (4, 19, 26, 32). Adhesion strength is defined as the depth of the potential well at the equilibrium separation distance. In the subsequent analysis, we borrow from lipid vesicle physics in defining the two components of the energy density. Our goal is to calculate h_o and compare it to the neutron scattering derived ℓ that is the experimental measure of endothelial adhesion.

van der Waals attraction.

In aqueous environments, the high dielectric constant of water significantly screens any pure electrostatic interactions leaving only weaker adhesive forces such as van der Waals. With lipid bilayer vesicles and model single cell systems, the retard van der Waals potential per unit area for two parallel plates is commonly used:

$$W_{\text{ad}}=\frac{-H_{\text{a}}}{12\pi }\cdot \left(\frac{1}{h^2}-\frac{1}{{(h+d)}^2}\right)$$ (2)

where H_a is the Hamaker constant, h is the distance between the quartz and cell surfaces, and d is the cell membrane thickness. For most lipid systems in aqueous solutions H_a is in the range of 2 − 5 × 10⁻²¹ J (15, 30). Using H_a = 2.6 × 10⁻²¹ J and d ∼ 5 nm, derived directly from the reflectivity data, the attractive van der Waals portion of the endothelial/quartz interface is defined and plotted in Fig. 5 (red/blue curve labeled W_ad).

Fig. 5.

Adhesion energy density potentials (solid black curves) and their respective components: W_ad (van der Waals adhesion), V_f (membrane thermal fluctuations), and V_r (basal lamina repulsive potential). In each respective potential, the dashed blue line represents the potential calculated at 25°C and the solid red at 37°C. The fluctuation potential is plotted for 2 different values of the prefactor b: 0.1 and 0.001. The repulsive potential is also shown for 2 different sizes of the basal lamina proteins: radius of gyration 15 and 25 nm. The solid black curves (h_o²⁵ and h_o³⁷) are the total potentials calculated as the superposition of the three components at the respective temperatures and radius of gyration (R_g); the values of b do not impact the shape of the total potentials. The x-axis position of the h_o minima correspond closely to the basal lamina thickness (ℓ) under the given conditions for static experiments.

Basal lamina osmotic potential.

Endothelial cells generate their own extracellular matrix (basal lamina) making it possible to grow monolayers on inert surfaces like quartz without prior exogenous deposition of protein layers (1, 14). A key component in the structure, mechanical coupling, and stability of endothelial basement membranes is type IV collagen (14, 17, 18). Cell adhesion is impacted by the thermodynamics of basement membrane protein confinement as shown by Bruinsma, Behrisch, and Sackmann (BBS; Ref. 4). This analysis is based on modeling the confinement energy (entropy) of molecules like collagen by treating them as biopolymers. Governing parameters in the potential are basal lamina thickness, temperature, and collagen density (ρ_o). Collagen structure is taken into account by calculating its hydrodynamic radius or radius of gyration (R_g).

Given that type IV collagen is nonfibrillar (6, 17, 23, 35), the helical part of the molecule is modeled as a filament with portions assuming either an α-helical conformation or a random-free coil (like an unstructured polymer chain). In such systems, the radius of gyration is shown to follow the following scaling law (25, 36):

$${\langle R_{\text{g}}^2\rangle }^{1/2}\sim N^{1/2}\cdot {\left[b_{\text{o}}^2(1-P_{\text{h}})+l_{\text{h}}^2{P}_{\text{h}}\frac{1+P_{\text{h}}}{1-P_{\text{h}}}\right]}^{1/2}$$ (3)

where N is the total number of residues within the molecule and is set to 1,500, the number of helical amino acids in collagen IV (17). The b_o is the typical length of a residue and is set to 0.6 nm; l_h is the helical rise per residue and is set to 0.15 nm for an α-helical like structure. P_h is the probability that a given residue assumes an α-helical conformation and plays a key role in the statistical study of helix-coil unfolding transitions (25, 36). It is well established that long helical proteins undergo unfolding not by complete melting into random coil structures but rather as a step-wise transformation into filaments where segments with α-helical structures are separated by random polypeptide coils (36). P_h can be used to tune the degree of denaturation that the collagen molecule exhibits within the basal lamina. By studying the limits at zero helical content (P_h = 0) and complete helix (P_h = 1), one observes that the above formula recaptures classic polymer scaling laws:

$$\underset{P_{\text{h}}\to 0}{\lim }{\langle R_{\text{g}}^2\rangle }^{1/2}\sim N^{1/2}\cdot b_{\text{o}}$$ (4)

which recaptures the Flory type scaling for an ideal polymer chain; on the other hand

$$\underset{P_{\text{h}}\to 1}{\lim }{\langle R_{\text{g}}^2\rangle }^{1/2}\sim \infty \to N\cdot l_{\text{h}}$$ (5)

where rod-like behavior is achieved for a molecule where the entire length is α-helical, and N·l_h is the rod length. Figure 6 plots 〈R_g²〉^1/2 as a function of P_h. Interestingly, R_g has a minimum of 16 nm around P_h ∼ 0.7 or when two-thirds of the collagen molecule is helical. The radius of gyration increases upon further denaturation (decreasing P_h) to the limiting value for a random coil, R_g^rc ∼24 nm. Such nonlinear behavior is shown by Nagai (25) to depend on the extended local configuration of random coils that are forced to shrink as small portions of the chain form short helices with more confined configurations. Extensive structural work on collagen IV has shown it to be highly thermally unstable. Given its less organized network-like structure with frequent nonhelical portions, type IV collagen unfolds over a broader range of temperatures from 30–44°C, with a mid-point at 37°C, and also exhibits rapid reversibility in its helix to coli transitions (6, 31, 32, 35). The overall experimental picture of type IV collagen structure supports modeling it as a biopolymer with different degrees of helical propensity: at 37°C P_h ∼ 0 (R_g ∼ 25 nm) and at 25°C P_h ∼ 0.7 (R_g ∼ 15 nm).

Fig. 6.

Radius of gyration for a helical polypeptide N = 1,500; b_o is the typical length of a residue and is set to 0.6 nm; and l_h is the helical rise per residue and is set to 0.15 nm for an α-helical like structure.

By combining the BBS potential with the above formulation for the radius of gyration, the thermodynamic potential of the basal lamina is obtain:

$$V_{\text{r}}=\frac{{\pi }^2}{6}\cdot \frac{k_{\text{b}}T}{h^2}\cdot e^{\left\{-1.5{\mathit{\text{h}}}^2/\left[540(1-P_{\text{h}})+33P_{\text{h}}\frac{1+P_{\text{h}}}{1-P_{\text{h}}}\right]\right\}}$$ (6)

Figure 7 plots V_r for different values of P_h. It clearly shows the sensitivity that the potential has on helical content, with a complete random coil conformation (P_h = 0) having among the strongest repulsions and step-wise decrease in the repulsive strength to a minimum at P_h = 0.6 − 0.7. Figure 5 plots V_r for two different values of R_g at 25 and 37°C. R_g is the key control parameter for the strength of the repulsive interaction generated by compressing the basal lamina.

Fig. 7.

The progression of the basal lamina repulsive potential V_r as a function of collagen helical propensity P_h.

The solid black curves in Fig. 5 plot the overall adhesion potential under static conditions at 25 and 37°C by superimposing W_ad and V_r. The minimum position at 25°C is h_o ∼ 35 nm, while at 37°C it is h_o ∼ 65 nm. These values correspond closely to the measured basal lamina thicknesses ℓ (see Fig. 2) under static conditions at the two temperatures. We show for the first time that endothelial cell monolayer adhesion is sensitive to temperature and within the context of a simple physical adhesion model can be explained through the well known phenomena of collagen structural changes near physiologic temperature.

Shear Stress

The above adhesion model does not take into account any active mechanical stress the endothelial cells may experience. The Fung endothelial tension-field model states that fluid shear stress τ applied to the luminal aspect of an endothelial monolayer (inside of a blood vessel) introduces an elastic tensile stress (σ_xx) in the direction of flow (x) on the upper portion of the endothelial plasma membrane (10, 11). The model assumes an equilibrium of forces on the luminal aspect of the cells:

$${\partial }_x{\sigma }_{xx}+{\partial }_y{\tau }_{xy}=0$$ (7)

where the y direction is perpendicular to the flow and equivalent to the direction of h in the above prior analysis. Integrating the equation of equilibrium across membrane thickness (d)

$$\frac{\partial }{{\partial }_x}{\displaystyle {\int }_0^d{\sigma }_{xx}\text{d}y+{\displaystyle {\int }_0^d\frac{\partial {\tau }_{xy}}{\partial y}\text{d}y=0}}$$ (8)

and using the definition of membrane tensions (or traction forces per unit length):

$${\displaystyle {\int }_0^d{\sigma }_{xx}\text{d}y=N_x}$$ (9)

the equation of equilibrium becomes

$$\frac{\partial N_x}{\partial x}+\tau =0$$ (10)

where the equivalence τ_xy ∼ τ is used. The above equation can be integrated and a relationship between the luminal membrane tension and fluid shear stress obtained:

$$N_x=-{\displaystyle {\int }_0^L\tau \text{d}x}=-\tau \cdot L$$ (11)

The choice of integration limits clearly makes a difference as to how the system distributes stress. If the endothelial monolayer was a pure continuum like a sheet of rubber, then L would be the size of the monolayer leading to very high stresses at the origin and predispose the system to failure. Endothelial monolayers, however, are not continua, rather the individual cells make tight contacts with their neighbors. These so-called cortical regions have specialized protein complexes (adherens junctions), which bring opposing cell membranes that span the thickness of the cell into close proximity (see Fig. 4) (2, 32). Fung (10) hypothesized that the cortical regions discretized the endothelial monolayer and provided a mechanism by which some of the luminal tension (N_x) could be transmitted to the basal lamina.

Balancing forces around the cortical edge gives the tension (T_y) transmitted to the basal lamina

$$T_y\sim \frac{N_x}{\cos \theta }$$ (12)

where θ is the angle between the cortical and luminal membranes (this was recently measured for endothelial monolayers to be θ ∼ 80°; Ref. 21). The discretization also defines the above limits of integration such that L becomes the size of one endothelial cell. Giving the final relationship between the shear stress and transmitted tension

$$T_y\sim \tau \cdot L$$ (13)

Using a typical cell length of 10 μm and our experimental shear stress τ ≈ 1.5 Pa, we obtain T_y ∼ 1.5 × 10⁻⁵ N/m. This force tends to lift the endothelial monolayer and do work against adhesion, a prediction made by Fung in his original paper.

To estimate the effect that shear stress has on adhesion, T_y is converted into an energy density. By dimensional arguments T_y ∼ 1.5 × 10⁻⁵ J/m²; however, the force is applied only along the cortical rim. The cortical area per cell can be calculated by assuming a circular projected adhesion area for each cell, a cortical thickness of O(10) nm, and a cell radius of 5 μm, which gives a cortical area to total cell area ratio of 0.005. T_y per cell is therefore ∼1,000 k_bT. With the use of Fig. 5, the adhesion energy strength is approximately −250 k_bT per cell. Clearly if all cortical rim tension were transmitted to the basal lamina without dissipation, the endothelial cells would de-adhere. The neutron data show that at 25°C with shear the monolayer separation from the quartz increases. The tension-field analysis presented here shows that such de-adhesion is to be expected.

Thermal Fluctuations

Thermal fluctuations play a large role in the physics of lipid vesicles and lipid membranes (30, 33). Vesicle adhesion is controlled by the competition between van der Waals attraction and thermal fluctuations (15, 26). The later introduce a degree of randomness (entropy) into the system as originally postulated by Helfrich and colleagues (15, 26, 30, 33). Adhesion between two vesicles or a vesicle and solid boundary suppresses fluctuations along the adhesion patch thereby lowering the overall entropy of the system (30, 33). Since thermodynamic systems approach equilibrium by maximizing entropy, fluctuations by definition drive de-adhesion. Membrane fluctuations are controlled by temperature (k_bT, where k_b is Boltzmann's constant), membrane elasticity in the form of bending stiffness (κ_b), and membrane tension (σ). A commonly used membrane fluctuation potential is

$$V_{\text{f}}=b\cdot \frac{k_{\text{b}}T\sigma }{{\kappa }_{\text{b}}}\cdot {\left(\frac{k_{\text{b}}T}{2\pi \sigma }\right)}^{1/8}\cdot e^{-h\sqrt{2\pi \sigma }/\sqrt{{\text{k}}_{\text{b}}T}}\cdot {\left(\frac{1}{h}\right)}^{1/4}$$ (14)

Rewriting the above equation to highlight the dominant scaling behavior of the different variables

$$V_{\text{f}}\sim \frac{k_{\text{b}}T}{{\kappa }_{\text{b}}}\cdot e^{-\sigma }$$ (15)

shows that increasing membrane tension and membrane stiffness decrease fluctuations, while increasing temperature increases fluctuations. Overall the neutron data support this scaling behavior for endothelial cell monolayers. Higher temperature increases off-specular scattering intensity correlating with stronger membrane fluctuations along the interface normal (see Fig. 3) (16). Furthermore, shear stress, shown in the mechanical analysis above to place the lipid membrane under tension, decreases off-specular intensity suppressing fluctuations (see Fig. 3).

To quantitate the effect of thermal fluctuations on membrane adhesion the exact potential (Eq. 14) is used. The prefactor b determines the strength of fluctuation induced interactions and is related to scattering derived measurements via $b\sim {({\xi }_{\perp }/\langle h\rangle )}^2$ (30, 33). Specular reflectivity roughness provides an indirect measurement for ${\xi }_{\perp }$ and is in the 0.5- to 1-nm range for our system. The $\langle h\rangle$ is the average thickness of the fluctuating interface, which in our system is the plasma membrane with a measured thickness ranging from 30 to 70 nm. This gives a prefactor b ∼ O(0.001). For the remaining parameters, literature values equivalent for a lipid vesicle are used: κ_b = 35 k_bT and σ = 1.7 × 10⁻⁵ J/m². Figure 5 shows V_f plotted for two values of b, the above calculated value and to demonstrate the sensitivity of the prefactor, b = 0.1. In both cases, the thermal fluctuation potential is weak compared with the van der Waals attractive potential.

From a standpoint of membrane thermal fluctuations, the diffuse scattering data show a qualitative agreement with existing scaling laws. The endothelial multi-cell monolayer is clearly sensitive to and exhibits thermal fluctuations. However, the more quantitative analysis shows that these fluctuations are likely too weak to have an impact on monolayer adhesion.

Conclusions

Adhesion is the propensity of two surfaces to remain in contact as a function of some perturbation field. The separation distance between the surfaces (ℓ) is determined by the minima in the total interaction potential (4, 15, 19, 26, 30, 32, 33). The decrease in adhesion strength with increasing temperature (endothelial monolayers under static conditions) is modeled using an interfacial potential function:

$$\Delta G_{\text{adh}}\sim W_{\text{ad}}+V_{\text{f}}+V_{\text{r}}$$ (16)

where W_ad is the van der Waals adhesive potential, V_f is a Helfrich type membrane fluctuation energy, and V_r takes into account the presence of biopolymers existing within the adhesion zone as described above. Endothelial cells generate their own basal lamina making it possible to grow monolayers on inert surfaces like quartz without prior exogenous deposition of protein layers (14). While proteins such as fibronectin and laminin are part of endothelial extra cellular matrix, a key component in the structure, mechanical coupling, and stability of endothelial basement membranes is type IV collagen (14, 17, 18). Our model uses the thermal stability of collagen, as a control parameter in the interfacial potential. With higher temperature, collagen unfolds, the effective osmotic pressure of the basal lamina increases with R_g leading to de-adhesion as shown in Fig. 5. We chose to focus our analysis on collagen because of the rich existing literature on its thermal properties. However, any network of loosely organized proteins should undergo a volume expansion (increase in R_g) with temperature given the increase in entropy with heating. While our theoretical calculation shows that the contribution of thermal fluctuations is small, diffuse neutron scattering (Fig. 3) clearly shows an increase in thermally excited interfacial fluctuations with increasing temperature (and their suppression with flow). Intricate theoretical models of single cell adhesion in the literature predict the existence of thermal unbinding driven by membrane fluctuations alone (32), a phenomena that may be at work for endothelial monolayers studied here. For endothelial adhesion under static conditions, we show that temperature is a strong control parameter leading to weaker adhesion at higher temperatures, in agreement with thermodynamic models and intuition about adhesion in soft-matter systems.

Mechanical shear stress plays an important and growingly recognized role in endothelial biology (2, 3, 8, 10, 11, 37). Our study is the first to provide a direct measure of endothelial monolayer adhesion under physiologic shear stress conditions. The above biomechanical analysis of endothelial stress distributions predict that to avoid catastrophic endothelial delamination, part of τ is transmitted to the basal lamina (10, 11). Thermodynamically, luminal flow does work (proportional to the product of shear stress and cell size) against the static adhesion energy. The de-adhesion observed for endothelial monolayers under flow at 25°C is in agreement with this thermodynamic argument. At 25°C, most active biology is suppressed and the endothelial monolayer responds as a soft matter physical system as seen above. At physiologic temperatures, however, the full richness of cell biology is available to the system. Shear stress causes major redistribution of proteins involved in cell-cell (adherens junction complexes) and cell-substrate adhesion (focal adhesions). Figure 4 shows that even on relatively short biologic time scales (4 h) relevant to the time scale of our neutron experiments, adherens junctions (β-catenin stain) and focal adhesions (paxillin stain) undergo localization to the cortical rim of endothelial cells. This localization of adhesion proteins is postulated to promote linkage between the actin skeleton and extracellular matrix, forming enhanced adhesion zones (2) (see Fig. 4, bottom). Bruinsma et al. (4) and Sackmann and Bruinsma (32) explored the theoretical effect of specific adhesion binding proteins on single cell adhesion potentials, concluding that adhesion molecules change the interfacial energy landscape allowing access to strong adhesion zones rich in specific adhesion molecules. Our measurement shows that this biologic effect may indeed be at work in endothelial monolayers as well. Shear stress at 37°C causes localization of specific proteins which may alter the adhesion potential in favor of overall increased adhesion as measured with neutrons. This biologic response in effect competes against the purely physical effects of shear flow observed at 25°C and the increased repulsive potential of the basal lamina at higher temperatures.

The work presented measures living tissue adhesion strength under dynamic conditions. The strength of neutron reflectometry is its nonpertubative nature and ability to probe large surface areas with nanometer resolution. Using endothelial monolayers, we show that physical models of soft-adhesion can describe large scale tissue adhesion, which broadens their scope of application far beyond the single cell realm. Lastly, we show a richness exists in biologically active adhesion under physiologic conditions that we have only begun to understand.

GRANTS

This work benefited from the use of the Lujan Neutron Scattering Center at Los Alamos Neutron Science Center funded by the U. S. Department of Energy (DOE) Office of Basic Energy Sciences and Los Alamos National Laboratory under DOE Contract DE-AC52–06NA25396.

DISCLOSURES

No conflicts of interest, financial or otherwise are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: L.P., K.G.B., and J.M. conception and design of research; L.P., N.Z., K.G.B., and J.M. performed experiments; L.P., A.J., and J.M. analyzed data; L.P., A.J., K.G.B., and J.M. interpreted results of experiments; L.P., A.J., and K.G.B. prepared figures; L.P. drafted manuscript; L.P., A.J., K.G.B., and J.M. edited and revised manuscript; L.P., A.J., K.G.B., and J.M. approved final version of manuscript.