Nanoscale Imaging Reveals a Tetraspanin-CD9 Coordinated Elevation of Endothelial ICAM-1 Clusters

Abstract

Endothelial barriers have a central role in inflammation as they allow or deny the passage of leukocytes from the vasculature into the tissue. To bind leukocytes, endothelial cells form adhesive clusters containing tetraspanins and ICAM-1, so-called endothelial adhesive platforms (EAPs). Upon leukocyte binding, EAPs evolve into docking structures that emanate from the endothelial surface while engulfing the leukocyte. Here, we show that TNF-α is sufficient to induce apical protrusions in the absence of leukocytes. Using advanced quantitation of atomic force microscopy (AFM) recordings, we found these structures to protrude by 160 ± 80 nm above endothelial surface level. Confocal immunofluorescence microscopy proved them positive for ICAM-1, JAM-A, tetraspanin CD9 and f-actin. Microvilli formation was inhibited in the absence of CD9. Our findings indicate that stimulation with TNF-α induces nanoscale changes in endothelial surface architecture and that--via a tetraspanin CD9 depending mechanism--the EAPs rise above the surface to facilitate leukocyte capture.

Fig 1. Nanoscale protrusions on activated endothelial cells.

(A) AFM recordings (50 μm)² on primary human endothelial cells (HUVEC) grown on permeable filter inserts show nice spindle-like morphology and a comparably smooth cell surface (left). Upon treatment with TNF-α [20 ng/ml] for 24 h, numerous protrusions appear and the surface becomes more rough (right). (B) Detailed morphology of protrusions after TNF-α stimulation is shown in higher resolution scans: (20 μm)² (upper left) shows an almost uniform coverage with protrusions over the cell body, a (10 μm)² image (upper right) reveals single events and a (5 μm)² zoom (lower left) shows a short straight and a long curved protrusion. Another 5 μm scan demonstrates a large item of 2 μm length and >300 nm diameter/height (lower left). (C) Mechanical stiffness analysis of the white marked window of B was performed on a (3 μm)² area by recording force curves and re-plotting the calculated heights (middle left) or Young‘s moduli (middle right) in grey scale maps. Respectively, white color codes for either high or for soft areas. The height map is completely analogous to the contact mode image (far left), while the stiffness map shows dark areas at the site of the protrusion, which means mechanically soft. Accordingly, the profile plots below the maps demonstrate a protrusion height of 200 nm (red line) and a corresponding dip in mechanical stiffness to a minimum of 10E4.5 Pascal (blue line). The number values for Young‘s moduli are given as histograms (far right) for the region of interest (upper diagram) and for the total area under investigation (lower diagram). Pictures shown are representative for more than 10 independent cell preparations.

Fig 2. Immunostaining of ICAM-clusters.

HUVECs before or after treatment with TNF-α were subjected to fluorescent staining for f-actin using phalloidin and immunolabeling for ICAM-1 using antibodies, respectively. The focal planes of confocal laser scanning microscopy (LSM) were adjusted to the apical surface. (A) represents a maximum intensity projection. ICAM-1 immunoreactivity is virtually absent from control cells and the actin forms a belt at the cell periphery along the junctions. Upon activation with (TNF-α), f-actin forms stress fibers (A), which exhibit many short debranchings as seen in the zoomed image (B). The ICAM-1 adhesion protein forms a punctuate pattern of elongated or even triangular spots with diameter of around 1 μm (A) They are distributed around the perinuclear area of the cell and are often located at the debranching sites (B). The reconstructed side view (C) shows f-actin both at the basal and apical areas, and ICAM-1 only at the apical surface. (D) A closer view on the apical surface demonstrates the clustered appearance and colocalization of ICAM-1 and f-actin. (E) A highest-resolution micrograph reveals a microvilli of more than 1 μm height which are decorated with ICAM-1 (white arrows). As seen from the merged image, ICAM-1 is mostly accompanied by f-actin but not the other way round. Images shown are representative for 3 independent cell preparations. In A and B, it is maximum intensity projections, while C-E are single slices in xz-plane (x = z scale).

Fig 3. Quantitation of nanoscale surface protrusions.

For morphological phenotyping of endothelial cell surfaces (HUVEC) were recorded by atomic force microscopy (AFM) without (A-C) or (D-J) after treatment with TNF-α. (A, D) An (50 μm)² overview demonstrates the cell bodies to become (D) longer and flatter after stimulation as compared to (A) controls, which exhibit a smoother surface. (B, E) Higher resolution to a (15 μm)² scan reveals the microvilli, which have been manually marked blue in C and F, respectively. Here, contrasting examples are chosen for the sake of clarity. To achieve objectivity, computer vision was employed for the enumeration of nanostructures (G-M), referred to as nAnostic method. From raw data as in (G), profiles (H) and 3D-representations (I) were marked by an operator (in blue) to train the machine (supervised learning, for details see methods section). On the basis of identified objects, histograms of height (K) and objects‘ volume (L) were given for morphometrical profiling. Typical dimensions of the microvilli are 160 ± 80 nm height and a volume of 10–20 attoliter (10E-18 l). Finally, the increase of object count from 129 ± 14 objects / (40 μm)² without to 260 ± 40 objects / (40 μm)² with TNF-α clearly indicates cell activation(M). Shown are the mean values ± SEM of 20 images out of three independent experiments. ** p<0.01 student’s t-test.

Fig 4. Association, clustering and redistribution of membrane proteins.

HUVEC with or without treatment by TNF-α and CD9 siRNA were subjected to immunostaining for ICAM-1, tetraspanin CD9 and junctional adhesion molecule JAM-A, respectively. Images are taken by laser confocal microscopy (LSM) and projections in z are presented. (A) ICAM-1 is virtually absent from control cells, but is induced by TNF-α and forms a punctuate pattern of elongated or even kinked spots with a diameter of around 1 μm. They are distributed all over the cell surface with a preference for the perinuclear area. CD9, which is located at the cellular junctions in control cells becomes redistributed over the whole apical cell membrane upon cell activation—similar to the pattern of ICAM-1. JAM-A, which is often associated to CD9, exhibits the same behavior. In the activated cell state, there is a high degree of colocalization with both CD9 and JAM-A in the apical membrane clusters of ICAM-1—seen in the merged image as whitely spots. (B) Western blotting (WB) of CD9 in a siRNA-treated cell preparation as described in detail under methods section. Efficiency of siRNA-transfection was routinely controlled by WB for each umbilical vein preparation. (C) JAM-A immunoreactivity is hardly affected by CD9 siRNA-transfection; the expression level remains high, and the distribution still is mainly spread over the apical surface and not relocated to the cell junctions. CD9 protein level is vastly reduced as already confirmed in (B) by WB. The merged image demonstrates the reduced degree of clustering. Pictures shown are representative for more than 5 independent cell preparations.

Fig 5. Inhibition by CD9-siRNA of microvilli formation.

HUVECs treated with CD9 siRNA or non-targeting siRNA either stimulated with TNF-α or without TNF-α, respectively, were subjected to quantitative AFM-analysis.(A) Surface topography is recorded at 10 arbitrary areas of the sample (one representative is shown here). (B) Raw data were subjected to computer vision as demonstrated in Fig 3 to identify microvilli. The result is presented as a black and white mask. (C) For better visualization, an overlay of A and B is given. (D) The object count rises from 290 ± 40 objects / (40 μm)² without to 437 ± 16 objects / (40 μm)² with TNF-α at non-targeting siRNA. This morphometric change is abrogated by pretreatment of the cells with CD9 siRNA. Here, the object count without TNF-α 280 ± 30 objects / (40 μm)² or with TNF-α 240 ± 40 objects / (40 μm)² is not significantly different. (E) Similar is for the size, measured as „local differential volumes”(LDV) of the microvilli: The increase from 7.6 ± 1.4 f l / (40 μm)² to 14.2 ± 0.5 f l / (40 μm)² (by 87% after TNF-α) is completely abolished in CD9 knock-down cells; without TNF-α 8.4 ± 0.9 f l / (40 μm)² is not significantly different to TNF-α 7.2 ± 1.1 f l / (40 μm)². Shown are the mean values ± SEM of 30 images out of three independent experiments. **p<0.01 post-hoc Bonferroni analysis compared to each bar.

Fig 6. Schematic model of adhesive microvilli formation.

(A) Upon activation of endothelial cells, ICAM-1 is upregulated and processed into the plasma membrane; it associates with tetraspanin CD9 to form clusters 2-dimensional. (B) The clusters (or: endothelial adhesive platforms EAPs also contain JAM-A, which is not depicted here) recruit f-actin, potentially through a RhoG dependent mechanism. (C) The clustered adhesion platforms are propelled upwards by typically 160 ± 80 nm, thereby increasing the interaction probability with leukocytes. (D) Upon leukocyte contact, the microvilli are further elongated to develop a full docking structure with long filopodia engulfing the leukocyte.

Fig 7. Sketch on physiological relevance of adhesive microvilli.

Geometrical considerations illustrate the impact of microvilli on binding probability of leukocytes. General assumptions are: ICAM-1 clusters (adhesive platforms, EAP) cover 10% of the endothelial surface and are equally distributed at mean distance of 2b = 2.5 μm, and leukocytes have smooth, spherical shape, radius a = 5 μm. (A) In a simple model of a flat endothelial surface, the chance of a leukocyte to meet an EAP upon first contact just equals the ICAM-1 surface coverage (10%). (B) With an arbitrary convolvement of membrane surface as in real cells (cytoskeletal roughness), many EAPs become inaccessible for leukocytes. The binding probability decreases close to zero, depending on the degree of surface roughness. (C) This sterical shielding effect is avoided, when EAPs are elevated above ground level. With the geometrical assumptions made above, EAPs reach a binding probability of 95% when lifted upwards by 160 nm.