Endothelial‐Smooth Muscle Cell Interactions in a Shear‐Exposed Intimal Hyperplasia on‐a‐Dish Model to Evaluate Therapeutic Strategies

Abstract Intimal hyperplasia (IH) represents a major challenge following cardiovascular interventions. While mechanisms are poorly understood, the inefficient preventive methods incentivize the search for novel therapies. A vessel‐on‐a‐dish platform is presented, consisting of direct‐contact cocultures with human primary endothelial cells (ECs) and smooth muscle cells (SMCs) exposed to both laminar pulsatile and disturbed flow on an orbital shaker. With contractile SMCs sitting below a confluent EC layer, a model that successfully replicates the architecture of a quiescent vessel wall is created. In the novel IH model, ECs are seeded on synthetic SMCs at low density, mimicking reendothelization after vascular injury. Over 3 days of coculture, ECs transition from a network conformation to confluent 2D islands, as promoted by pulsatile flow, resulting in a “defected” EC monolayer. In defected regions, SMCs incorporated plasma fibronectin into fibers, increased proliferation, and formed multilayers, similarly to IH in vivo. These phenomena are inhibited under confluent EC layers, supporting therapeutic approaches that focus on endothelial regeneration rather than inhibiting proliferation, as illustrated in a proof‐of‐concept experiment with Paclitaxel. Thus, this in vitro system offers a new tool to study EC‐SMC communication in IH pathophysiology, while providing an easy‐to‐use translational disease model platform for low‐cost and high‐content therapeutic development.


Endothelial-Smooth Muscle Cell Interactions in a Shear-Exposed Intimal Hyperplasia on-a-Dish Model to Evaluate Therapeutic Strategies
Andreia Fernandes, Arnaud Miéville, Franziska Grob, Tadahiro Yamashita, Julia Mehl, Vahid Hosseini, Maximilian Y. Emmert, Volkmar Falk, Viola Vogel* [43] for 100 rpm and 200 rpm and for 120 rpm [44] . The shaded area represents the estimated range of mean WSS at the conditions used in our experiments, i.e. 135 rpm. All conditions assume that fluid volume is 2 ml per well. (B) Schematics describing the protocol to create EC monocultures exposed to shear as in (C). Human Primary Umbilical Vein ECs (HUVECs) were grown for two days in Endothelial Growth Medium (EGM-2, Table 1) in static conditions until reaching confluency. Cells were transferred to the orbital shaker and grown for 3 days in 2 ml of coculture medium (CCM, Table 1) and 135 rpm rotation speed. For static cocultures, the protocol was identical except that cells were not placed on the shaker. (C) 3 HUVEC monoculture after 3 days in CCM either on the orbital shaker or in static conditions, according to protocol described in (B). Images were taken in both center and periphery of the well and show cell-cell junctions through VE-cadherin immunostaining (green). As a reference, the well wall is on the left side. Straight white arrow indicates the direction of the flow, while curved arrow represents disturbed flow. Scale bar 200µm. (D) Graph showing the distribution of EC nuclei orientation in blue (angle ⍺, degrees°) from center to periphery of the well.
Orientation of nuclei stained with DAPI were analyzed using a Matlab script. Angle ⍺ is defined as indicated in (E). Red dots represent the angle ⍺ of shear vectors that were previously published [43] and kindly provided by the author. Blue dots that co-colocalize with red dots represent cells that are aligned with the flow. The vertical white patterns in the blue dots are an artifact derived from image stitching. (E) Schematics explaining definition of angle ⍺ used in graph (D). Angle ⍺ is defined as in Salek et. al. [43] for better comparison with computer simulations.
4 Figure S2 -Characterization of the effect of different culture media on the SMC phenotype. Human aortic SMC monocultures were cultured under static conditions, for 3 days with different culture media ( Table 1) Table 1) for manipulation of phenotype, as in (A). The western blot shows protein levels of ⍺SMA (42KDa) and S100A4 (12 KDa), using GAPDH (35 KDa) as a loading control in a 12% gel. Calponin (34 KDa) was blotted using alpha-Tubulin (50 KDa) as a loading control in a 10% gel. (D) Since EGF and FGFb in CCM and shear are known to induce the synthetic phenotype, we tested whether TGFβ-induced contractile SMCs shifted back to synthetic after 3 days on the orbital shaker in CCM. This experiment served as a control for the stability of the SMC phenotype, since cocultures with ECs were performed under the same conditions. Thus, SMC monocultures were cultured in TGFβM for 3 days in either static conditions (St) or exposed to shear (Sh). Medium was then exchanged to coculture medium (CCM) and cells were cultured for additional 3 days in either shear (St/Sh) or static conditions (St/St). As a control, SMC monocultures were grown in SmGM medium and then transferred to CCM for additional 3 days in either static (St/St) or shear (St/Sh) conditions. Protein levels of ⍺SMA (42KDa) and S100A4 (12 KDa) were blotted, using GAPDH (35 KDa) as a loading control in a 12% gel. Calponin (34 KDa) was blotted using alpha-Tubulin (50 KDa) as a loading control in a 10% gel. (E) Human aortic SMC monocultures were cultured under 6 static conditions, for 3 days with different culture media, as in (A). Graph shows quantification of extracellular fibronectin, through immunostaining (intensity of signal/number of nuclei). (F) As example of images used for quantification in (E), the immunofluorescence confocal images show fibronectin immunostaining (gray) and DAPI staining (blue) in SMC monocultures after 3 days in either TGFβM or SmGM under static conditions. Scale Bar 200µm.  , Table 1), as described in Figure 1A. Total area covered with ECs was calculated using the large mosaic images showing VE-cadherin staining in the whole well or only half of the well (images not shown). VE-cadherin fluorescence signal was blurred and binarized for area calculation using Fiji software. Percentage of tissue covered with ECs was calculated based on the total area of the tissue, using Excel. Periphery and center regions of the well were compared. Two samples (n=2) or five samples (n=5) per condition were analyzed.
(B) Percentage of proliferating ECs in the IH model, Quiescent Vasculature (QV) model and EC monoculture (EC MC) at day 1, 2 and 3 of coculture on the orbital shaker and comparison between center and periphery of the well. For all conditions CCM medium was used (see Table   1). To distinguish from SMCs, all EC nuclei were stained with ERG1 antibody. Proliferative cells were immunostained with anti-ki-67 antibody, and all nuclei were stained with DAPI.
Nuclei were counted using a script in Fiji/ImageJ. For statistical analysis, one-way ANOVA followed by Tukey's test was performed to compare between different models within the same time point and region of the well. **** p<0.001, *** p<0.01, ** p<0.1, *p<0.5).
8 Figure S4 -Growth of ECs in monoculture exposed to shear. ECs were seeded at 8x10 4 cells/cm 2 and exposed to shear for 3 days, such as in IH and Quiescent Vasculature models. shaker. To investigate the phenotype of SMCs using different co-culturing conditions, ECs were seeded on top of SMCs and cultured for 3 days on the orbital shaker in CCM (Table 1) as described in Figure 1A and 4. Quiescent vasculature model: ECs seeded at low density on top of c-SMCs. Intimal hyperplasia model: ECs seeded at low density on top of s-SMCs. Highdensity coculture: ECs seeded at high density on top of s-SMCs. SMC monocultures were also cultured in CCM for 3 days on the orbital shaker. The immunofluorescent images, acquired with an Olympus FV1000 laser scanning confocal microscope, show the expression of contractile phenotype markers, i.e. ⍺SMA and calponin and, of synthetic marker S100A4. The three different stainings were performed on the same sample but at different sites.
Compartmentalization of the well for these 3 independent stainings was possible with our multi-10 staining platform, described in Fernandes et al. 2022. [104] Images were acquired at the periphery of the well. Scale bar 200µm.  Figure 1A and 4. Graph showing the distribution of EC nuclei orientation in blue (angle ⍺, degrees°) from center to periphery of the well (radial position). Orientation of EC nuclei stained with ERG1 (EC specific nuclear marker) as function of the radial position were analyzed using a Matlab script. Angle ⍺ is defined as indicated in Figure S1E. Red dots represent the angle ⍺ of shear vectors that were previously published [40] and kindly provided by the author. Blue dots that co-colocalize with red dots represent cells that are aligned with the flow. The vertical white patterns in the blue dots are an artifact derived from image stitching. in SmGM medium and imaged with phase contrast in a Zeiss light microscope before addition of ECs. Scale Bar 200 µm. (Below) Thereafter, ECs were seeded at either 8x10 4 cells/cm 2 (low density, LD CC) or 1.5x10 5 cells/cm 2 (high density, HD CC). For seeding, MCDB131 medium (Invitrogen) containing 3.3% human serum, 1M Hepes, 1x Glutamax and 1x ITS (insulinselenium-transferrin) was used. Cocultures (CC) and EC monocultures (EC MC) were imaged 13 5h after seeding with phase contrast using a Zeiss light microscope with a 10x objective. Scale bar 100µm.  , Table 1) as described in Figure 1A and 4. Total area covered with ECs was calculated using the large mosaic images showing VE-cadherin staining in the whole well or only half of the well (as in Figure 1A and C). VE-cadherin fluorescence signal was blurred and binarized for area calculation using Fiji software. Percentage of tissue covered with ECs was calculated based on the total area of the tissue, using Excel. Periphery and center regions of the well were compared.

Figure S11 -Examples of SMC layers forming corrugated structures that resemble
intimal thickening. The IH model was produced as described in Figure 1A and exposed to shear for 3 days on the orbital shaker using CCM (Table 1) (Table 1). Thereafter, s-SMCs were transferred to m200 medium (Table 1)   ECs were seeded at low density on s-SMCs, according to the protocol sketched in Figure 1A.
Black and white images acquired by widefield fluorescence microscopy show the whole well as a large mosaic. On the left side: s-SMCs stained with cell tracker. On the right side: ECs through VE-cadherin immunostaining. Both culture media are commercially available endothelial growth media: EGM-2 (Lonza) and m200 (Invitrogen). See Table 1 for complete media formulas. Scale bar 5 mm.

Figure S18 -Number of proliferating SMCs is reduced in the absence of EGF and FGFb.
Graph showing the percentage of proliferating SMCs after 3 days of coculture with ECs at lowdensity under shear (EC-SMC CC = IH Model). Either complete coculture medium (CCM) or CCM without both EGF and FGFb was used (CCM-EGF-FGFb) (see Table 1). The different regions of the tissue (EC islands and Gaps) in both center and periphery regions of the well were compared. As a control, s-SMC monocultures (SMC MC) were cultured for 3 days under shear in CCM. Proliferative cells were immunostained with anti-ki-67 antibody, a proliferative marker, and all nuclei were stained with DAPI. To selectively count SMC nuclei, ERG+ cells (EC nuclei) were excluded. Nuclei were counted using a script in Fiji/ImageJ.   , Table 1). Additionally, ECs were seeded at high density on top of s-SMCs and (D, H) cultured in static conditions for 4h and on the orbital shaker for additional 3 days, or (E, I) 24h in static conditions followed by 2 days on the orbital shaker, both in m200 medium (Table 1)

. (B-E) Stitched mosaics of whole wells depicting SMC
labelled with cell tracker on the left and ECs immunostained with VE-cadherin on the right.
Scale bar 5 mm. (F-I) Zoomed-in images of both cell tracker (top) and VE-cadherin channels marked with an orange square in the respective whole-well overviews. Scale bar 0.5mm.

Comparison of VE-cadherin, Fibronectin and SMC cross-sectional profiles
Stitched images with 3 channels were used (channel 1: cell tracker stained SMCs; channel 2: ERG1 and VE-cadherin double staining, i.e. two different primary antibodies targeted with the same secondary antibody; channel 3: Fibronectin). ERG1/VE-cadherin images were blurred and binarized by thresholding. The positive and negative regions in the binarized ERG1/VEcadherin image were identified as "islands" and "gaps", respectively. In the image from the IH model, regions of EC networks and island regions were manually identified. These ERG1/VEcadherin binarized images indicated gaps, islands and EC networks, and the corresponding raw images of Fibronectin (Fn) and SMC were compressed to low-resolution (around 240 x 1960 pixels). The Fn and SMC images were then processed using a mean filter kernel of 15 x 15 pixels window size to remove signal fluctuations in 10 µm scales and to better visualize the large-scale trends. The contrast of each image was linearly adjusted so that its maximum pixel intensity becomes 255 in the end. Using these classified and processed images of the three channels, the cross-sectional profiles at arbitrary positions were created. Note that these processed images were only used to create cross-sectional profiles with the normalized range and were not used for quantitative analysis. See these results in Figure 3F, 4F, S19B and S20B.

Orientation of Cell Nuclei
For SMC and EC monocultures, DAPI images were utilized. For cocultures, images with ERG1 and VE-cadherin double staining (i.e. two different primary antibodies targeted with the same secondary antibody) and DAPI images were utilized. DAPI and ERG1/VE-cadherin images were blurred and binarized by thresholding. The binarized images in pairs were then processed using a custom script written in MATLAB (2016b, Mathworks, Natick, MA, USA).
Briefly, the connectivity of the signals in the DAPI image was analyzed to distinguish each nucleus. To select ERG-1-positive nuclei and to remove the VE-cadherin signal, with each DAPI-positive nucleus the number of positive pixels in the corresponding area of the binarized ERG1 image were counted. Then, the ERG1-positive ratio was evaluated by dividing the number of ERG1-positive pixels by the total number of DAPI-positive pixels. In the end, a nucleus with an ERG-positive ratio higher and lower than 50% was recognized as that of ECs and SMCs, respectively. Thereafter, nuclear orientations were evaluated using a custom script written in MATLAB. First, the position of the dish center was detected by fitting a circle to the edge of the cell culture dish in the binarized nuclei images. Next, connectivity analysis was performed to distinguish each nucleus. Each nucleus was then fit to an ellipse. The angle between the long axis of the nucleus and the radial direction was evaluated as the nuclear angle ⍺ (as defined in Figure S1E). This angle was defined according to Salek et al. [40] , so that the angle of shear vectors simulated in this publication can be compared with the orientation of cells in this study. See the corresponding results in Figure S1D, S6 and S23.

Immunostainings and Microscopy
Tissues/cells were fixed with 4% paraformaldehyde for 15 min at room temperature (RT). If necessary, cells were permeabilized for 10 min with 0.1% Triton. Thereafter, cells were blocked with 1% BSA for 1 hour at RT and stained with primary antibodies in 1% BSA overnight at 4°C. After 3 washes with 1x PBS, samples were incubated with secondary antibody for 1h at 32 RT, followed by another 3 washes with 1x PBS and stored in 1x PBS at 4°C before imaging. Table S1. When indicated, cell nuclei were stained with DAPI (Invitrogen, D3571) for 10 minutes, 1:500 dilution in PBS.

Primary Antibody
Dilution Primary Antibody

Secondary Antibody Dilution Secondary Antibody
Rabbit anti-VEcadherin (Cell Signaling, 2500) Simone Schürle-Finke, ETH Zurich). Only images within the same figure were imaged and edited with the same settings and can be directly compared.
After centrifugation at 10600g for 10min, the supernatant was recovered and aliquoted for Micro BCA measurement and western blotting. The Micro BCATM Protein Assay Kit (ThermoScientific, #23235) was used to measure the total protein content of the sample. For protein separation by size, an equal amount of protein was loaded in each gel pocket.
Electrophoresis was performed with 60V for 25min, then 1h 30min at 110V. The Mini-Protean Tetra Cell System from BioRad was used to transfer the protein to a membrane with pore size 0.2µm (Amersham Protran ,10600015), at 90V for 90 minutes. Afterwards, the membranes were blocked with 5% milk in 1xTBST (0.1% Tween) for 1 hour at room temperature. Primary antibodies were incubated overnight at 4°C, diluted in 5% milk at their respective dilutions (see antibody details on Table S2). After three washes with 1x TBST, membranes were incubated with secondary antibodies (1:10'000 in 5% milk) for 1h at room temperature followed by three washing steps with 1x TBST and one wash with 1x TBS.

Cytokine Array
Supernatants from 72h coculture experiments were analyzed using the Human Cytokine array C5 (RayBiotech, Inc). Previously blocked membranes were incubated with supernatants for 3.5h, followed by an overnight incubation with the Biotinylated Antibody Cocktail. On the second day, HRP-conjugated Streptavidin antibodies were added for 2h. For chemiluminescence detection, Fusion FX chemiluminescence imaging system was used (Vilber), and finally, signal intensities were analyzed using Fiji software.

Quantitative assessment of island and network regions in EC images
The ECs visualized by ERG1 and VE-cadherin staining was segmented to island and network regions via texture analysis, highlighting the local features of the brightness pattern.

35
The following procedure was all carried out using a custom MATLAB script. The horizontal and vertical resolutions of ERG1/VE-cadherin image were first reduced to 10% of the original ones. The image contrast was then tuned using "imadjust" function of MATLAB so that the bottom 1% and the top 1% of the resulting image pixels turned zero and saturate, respectively, followed by conversion to an 8-level grayscale image. Next, an 8 x 8 gray-level co-occurrence matrix (GLCM) was created by counting how often a pixel with grayscale value i occured 3 pixels horizontally and 3 pixels vertically adjacent to a pixel with the value j. The pixels counted in the central elements around the principal diagonal of GLCM (i.e., considerably bright pixels having a neighbor of the similar brightness 3 x 3 pixels away) were categorized to EC island, while those counted in the elements locating around the right-top and the left-bottom corners of the GLCM (i.e., pixels having a considerable contrast to the neighbor 3 x 3 pixels away) were categorized to EC network. The pixels counted in the (1, 1) element (top-left) of the GLCM (i.e., dark pixels having a dark neighbor 3 x 3 pix away) was the background. The obtained binary masks representing island and network regions were finalized by filling gaps spanning smaller than 3 pixels by sequentially applying dilating and eroding morphological operations. The outcomes were used to calculate the coverage of island and network in an arbitrary area.
The GLCM-based analysis captured the difference in the morphological features of the ECs in islands and networks well. It should be, however, noted that the analysis inevitably causes a segmentation error with ECs locally having an irregular contrast or those under inhomogeneous illumination, because it purely focuses on the local feature of the image and does not implement any biological context for segmentation. For quantification, we set several continuous subregions spanning from the periphery to the center of the entire image as a region of interest, keeping the error rate less than 1%. In principle, the GLCM-based analysis detects gaps between networks and holes within an island more strictly than a simple threshold-based area detection.

SUPPLEMENTARY VIDEOS
Video S1 -Orbital Shaker. 6-well plate on top of an orbital shaker rotating at 134 rpm. The travelling wave that circulates through the well creating a pulsatile flow can be observed.
Video S2 -High-resolution z-stack video of the IH model at the periphery and center of the well. Low-density coculture with s-SMC (IH-model, prepared as described in Figure   1A) Video S4 -3D reconstruction of corrugated structure in IH the model. Low-density coculture with s-SMC (IH-model, prepared as described in Figure 1A) after 3 days on the orbital shaker. Same sample as in Figure 3C, but image acquired in another location. Video shows a 3D reconstruction of a z-stack acquired with a Zeiss confocal microscope. A 40x water 37 objective was used. Tiles of 3x3 images were stitched together using microscope software (Zen). Imaris software was used to create animation. ECs stained with VE-cadherin antibody by immunostaining (red) and SMC with cell tracker (green). Scale bar 100 µm.
Video S5 -High-resolution z-stack video of the high-density model at the periphery and center of the well. High-density coculture with s-SMC (high-density model, prepared as described in Figure 4) after 3 days on the orbital shaker. Same sample as in Figure 4A  Video S6-High-resolution z-stack video of the high-density model at the periphery of the well. High-density coculture with s-SMC (high-density model, prepared as described in Figure 4) after 3 days on the orbital shaker. Videos show a z-stack acquired with a Zeiss confocal microscope corresponding to the orthogonal view in Figure 4C. The video starts at the bottom of the well and finishes at the top. A 40x water objective was used. Tiles of 3x3 images were stitched together using microscope software (Zen). ECs stained with VE-cadherin antibody by immunostaining (green) and SMC with cell tracker (red). Scale bar 200 µm.