Endothelial Cell Selectivity to Nanoparticles Depends on Mechanical Phenotype

Endothelial cells (ECs) elongate in the direction of blood flow, are stiffer, and are considered atheroprotective in areas of the vasculature where flow‐induced shear stress is high and unidirectional and are softer, atherogenic, and polygonal in areas experiencing oscillatory multidirectional flow. To understand the precise roles of EC mechanics and morphology in the uptake of therapeutic nanoparticles (NPs) by atherogenic endothelium, human aortic ECs are induced to adopt prescribed shapes and areas imposed by microcontact patterned adhesive islands. NP uptake per cell increases with increasing spreading area and decreases with increasing cell aspect ratio at constant cell spreading area. Biomechanical analysis shows that elongated cells exhibit higher cellular stress and stiffer membranes than cells with low aspect ratios, indicating a strong correlation between morphology, mechanical phenotype, and NP uptake. Further, ECs elongated by high laminar shear endocytosed NPs to a far lesser extent than those that are nonelongated in the chaotic, lower shear areas when cocultured in the same chamber. Results indicate that conditions leading to atherogenesis, such as low, chaotic shear‐induction of EC polygonal morphology may be used to increase the uptake of therapeutic NPs as a preventative measure against atherosclerosis.


Introduction
Endothelial cells (ECs) in vivo exhibit distinctive phenotypes that depend on the prevailing blood flow patterns. Atheroprotective directly from the blood vessel lumen, in which NPs accumulate due to the disturbed flow, or via the vasa vasorum which can provide atherosclerotic lesions with a distinct blood supply. [7] Additionally, NPs can be used as a diagnostic tool for the detection of near-term atherothrombotic events to ensure ideal preventive treatment. [8] The combination of nanomedicine with new blood or imaging biomarkers [9] can enable a sensitive readout for early screening for asymptomatic atherosclerotic lesions as a preventive method. [7] To achieve improved targeting specificity, NPs are coated with targeted ligands to cell surface receptors that are upregulated on atherogenic endothelium. [10] Such a targeting strategy relies on the biased surface chemistry of cells, termed chemotargeting. Further studies have shown that areas of low flow such as in the liver and kidneys have higher accumulation of NPs than areas of high and directed flow such as in the straight parts of the aorta, and the brain. [11] While the prevailing flow is one possible reason for NP accumulation and adherence to endothelium in these regions, it is possible that the shear flow-differentiated mechanical phenotype of ECs presents another key factor for the differential endocytic kinetics of NPs. Such phenotype-specific targeting, if scientifically validated, may enable a new targeting strategy, e.g., mechanotargeting. [12] In vitro evidence of mechanophenotype-specific targeting has been accumulating. We have recently shown that when cultured on a stiff surface, ECs are stiffer, more spread, and accumulate NPs at a lower capacity as measured by NPs per unit cell surface area than cells on a soft surface. [13] However, because of the larger surface area of spread cells on stiff surfaces, these cells overall take up more NPs than rounded cells on soft surfaces. Thus, it remains unclear as to whether surface area, morphology, or surface mechanics is the dominant factor in governing NP uptake. In the present study, to elucidate the relative role of cell mechanics and spread area on NP uptake, we designed experiments in which we prescribed cell spreading area and shape independently and derived cellular stress and its role in governing NP uptake. Our results indicate the elongated endothelial cells, characteristic of atheroprotective phenotype, exhibit lower NP uptake than cells with a reduced aspect ratio, when area is controlled for. We further show that elongated ECs are mechanically more stressed, with a higher cellular stress and membrane tension than the less elongated ones. Second, using a microfluidic platform with neighboring zones for high unidirectional shear and chaotic low shear, we induced cells to adopt both atheroprotective (elongated) and atheroprone (cobblestone) phenotypes similar to ECs found in bifurcating arteries. When exposed to uniform concentrations of NPs, elongated endothelial cells grown under laminar flow exhibit lower NP uptake than nearby cobblestone cells formed under chaotic flow. Our results confirm that such a differential mechanical phenotype leads to biased NP uptake capacity, demonstrating the feasibility of mechanotargeting. In addition to providing insight into the underlying cause of non-uniform biodistribution of NPs seen in the vascular endothelium, our studies highlight the significance of cell mechanics in the design of NPbased diagnostic, preventive and therapeutic agents for enhanced targeting.

Elongated Cell Shape Suppresses Cellular Uptake of NPs
Human aortic endothelial cells display an ellipsoidal shape and align with the shear flow direction when experiencing shear force. [14] The ellipsoid-shaped cells under shear form actin fiber bundles that are aligned with the long axis of the cells, and the nuclei adopt shapes that mirror the overall cell shape. In the absence of shear force, the endothelial cells exhibit a polygonal shape. [15,16] These results suggested that elongated cells grow to be more similar in shape to cells with high aspect ratios found in atheroprotective regions (Figure 1a and Figure S1, Supporting Information), where the cells are stiffened in response to the shear force. [17] To mimic the morphology induced by high shear force and explore the roles of cell shape on the cellular uptake of NPs, we patterned extracellular matrix (ECM) molecules (fibronectin) on polydimethylsiloxane (PDMS) substrates using microcontact printing. The remaining regions were passivated with Pluronic F-127 to prevent cell adhesion, confining single cells to fibronectincoated islands (Figure 1b).
These adhesive patterns were in square and rectangular geometries with independently controlled spreading areas and aspect ratios. Human aortic endothelial cells (HAECs) were cultured on the adhesive islands and induced to adopt the morphology with the prescribed spreading areas and aspect ratios ( Figure S2, Supporting Information). The HAECs were seeded on the substrate for 12 h before exposing them to fluorescent, monodispersed carboxylate polystyrene NPs (PS-COOH NPs) with a diameter of 100 nm. The encapsulated fluorescent dye in NPs has minimal photobleaching which allowed us to positively correlate fluorescence intensity with intracellular NP concentration. As previously reported, these NPs cause no apparent changes in either cell morphology or cell stress state. [12] To determine the influence of cell shape on NP uptake, cell spreading area was maintained constant at 2025 μm 2 , while aspect ratio (AR = length/width) varied from AR = 1 (square) and AR = 1.5, 2, 4, and 8 (rectangles). Figure 1b demonstrates that the cells adopted and retained prescribed ARs while maintaining a constant spreading area. With different ARs, the cells exhibited distinctive actin fiber distribution and nuclear morphology ( Figure 1b and Figure S3, Supporting Information), which are consistent with the previous observations that cytoskeletal reorganization and anisotropy regulate cell shape. [18,19] We quantified the NP uptake by the total fluorescence yield of the micropatterned cells incubated with fluorescent NPs at time points of 1, 2, 4, 8, and 12 h (Figure 1c). The NP uptake level increased with increasing incubation time and continued to rise after prolonged incubation time. Since NPs usually are cleared out of the circulation within several hours after intravenous injection, [20] the NP exposure maximum was 12 h with longer times having little clinical relevance. By 12 h, NP uptake increased nearly 13-fold over the 1 h timepoint and was strongly dependent on the AR at constant cell spreading area. The cellular uptake at 12 h was nonlinearly dependent on the AR, being the highest for the ARs of 1.5 and 2 at all time-points. In particular, the uptake increased from AR = 1 to AR = 2, then decreased by nearly two-fold from AR = 2 to AR = 8. Importantly, when cells were allowed to spread Cell-geometric constraints modulate NP uptake. a) Aortic mechanobiology: Endothelial cells sense and respond to the prevailing local wall shear stress; low-oscillatory shear causes cells to be cuboidal; high unidirectional, pulsatile shear results in elongated, ellipsoidal cells. b) Representative DIC and fluorescence images of single HAECs constrained on 2025 μm 2 protein-coated micropatterns of aspect ratios ranging from AR 1 to AR 8. Top panel: DIC and DAPI (blue), Bottom panel: phalloidin (green), DAPI (blue) and NP (red). (Scale bar = 10 μm). c) Cellular uptake of NPs after 1, 2, 4, 8, and 12 h of culture. Increasing aspect ratio (AR = length/width) from 2 to 8 caused a significant decrease in NP uptake, the intensity of the nanoparticles is increased for better presentation (Data are presented as mean ± SEM, n = 7-8 cells from 3 independent experiments per condition per time point, *p ≤ 0.05). without constraints, they adopted an aspect ratio between 1.5 and 2 ( Figure S1, Supporting Information). Thus, imposing aspect ratios less than 1.5 and greater than 2 reflect the impact of external forces on the cells. To ensure that the fluorescence arose from internalized NPs, we confirmed, using structured illumination microscopy (SIM), that NPs were internalized and localized in the cytoplasm and around the nucleus ( Figure S4, Supporting Information). The accumulation of most of the NPs at the central and ventral regions of the cell rather than on the cell surface attested to internalization of the NPs.

High Cell Spreading Area Suppresses NP Uptake
We next examined the roles of cell spreading area and shape in NP accumulation in HAECs. Cells were constrained geometrically on adhesive islands of areas ranging from 900 to 2500 μm 2 , and aspect ratios ranging from AR = 1 to 8. Cells adopted shapes consistent with the ECM adhesive patterns (Figure 2a).
Staining cells with phalloidin and 4′,6-diamidino-2phenylindole (DAPI) demonstrated the reorganization of actin cytoskeleton was accompanied by alteration in nuclear orientation and shape as cells adopted different prescribed morphologies. The dramatic change in the cell shape forced by the adhesive islands is consistent with the cytoskeletal regulation of endothelial cells by shear forces in atherosclerosis-susceptible and atherosclerosis-resistant vascular regions. [21] After incubation of the cells with NPs we quantified the fluorescence intensity arising from the NPs inside single cells. We observed that cellular uptake of NP consistently increased with increasing cell spreading area for each fixed aspect ratio of AR = 2 ( Figure 2b). However, NP uptake capacity (uptake per unit area) was reduced with increasing cell spreading area (Figure 2c).

Membrane Mechanics Regulates Cellular Uptake
The endocytic entry of a NP to a cell requires membrane wrapping of the NP, driven by the cell-NP adhesion energy at the cost of the membrane mechanical energy, which includes bending and tension energies. [13] Overall, an increase in tension energy, Figure 2. Influence of cell morphology on NP uptake. a) Representative images of actin (green)-and nuclei (blue)-stained images of HAECs with various spreading area, and ARs controlled by microcontact patterning. b) HAECs cultivated on adhesive micropatterns (various size of 900 to 2500 μm 2 ) were exposed to 0.02 mg-mL −1 particles for 8 h. Cellular uptake increased with increasing cell spreading area at the same AR. (*p ≤ 0.05, comparison between AR1.5, 2 and 8, n = 5 from 3 independent experiments). c) Increasing the spread area of the cells led to increased cellular uptake. However, the cellular uptake capacity (NP per unit area) decreased with increasing cell spread area (AR = 2).
, lowers NP uptake rate,ṅ, if the membrane bending energy, , and adhesion energy, , are constant according tȯ whereṅ 0 is a rate constant and k B T is the thermal energy. Thus, we sought to determine changes in tension in the cell membrane as a function of cellular morphology. Conventional techniques, such optical traps [22] or AFM [23] have been developed to measure membrane mechanics. However, these methods require perturbation of the cell surface and do not measure the bilayer tension directly. Here we incorporated a fluorophore dye, DiI-C 12 , in the bilayer membrane of the patterned HAECs and measured its fluorescence lifetime as a membrane tension indicator. [24] The underlying principle is that the fluorescence lifetime of the DiI chromophore decreases as the area experienced by the DiI embedded in the plasma membrane increases owing to the increased non-radiative decay of fluorescence. A high membrane tension reduces membrane viscosity and lipid order, leading to increased area experienced by the DiI chromophores, more avenues for non-radiative decay of fluorescence, and hence reduced lifetime. [13,24,25] Using time-correlated single photon counting (TCSPC) we acquired the fluorescence lifetime of the embedded fluorophores in HAECs of AR 1.5 and 8 (Figure 3a,b). As shown in Figure 3c elongated cells exhibit a shorter DiI fluorescence lifetime ( AR8 = 0.955 ± 0.042 ns), than low aspect ratio cells ( AR1.5 = 1.443 ± 0.063 ns) indicating significantly higher membrane tension in elongated cells. Considering the equal spreading area of the adhesive islands on which cells were seeded, we conclude that the difference in NP uptake arises in part from increased membrane tension from cell elongation.

Cell Contractility Governs Membrane Mechanics
Though the plasma membrane is regarded as a primary barrier for NP endocytosis, the cortex, a thin network of actin filaments attached to the plasma membrane, may also play a role in cellular uptake. [26] This is because cell surface mechanics, including membrane bending modulus and membrane tension, is profoundly tied to the cortex mechanics. [27] On the one hand, the close proximity of cortex spatially confines the membrane and affects membrane organization, thus effectively modulating membrane bending modulus. [28] On the other, the cortical network is under contractile stress generated by actomyosin motors; the cortical tension can be transduced to the plasma membrane through various membrane-cortex protein linkers, [29] thereby regulating membrane tension. Since the membrane-cortex interaction changes dynamically as the cell spreads and migrates, [30] membrane mechanics also changes.
To demonstrate the physical link between the contractile stress fibers and NP uptake, we analyzed the overall NP uptake in the square and rectangular HAECs in the presence of Y-27632, a Rho-associated kinase (ROCK) inhibitor, which blocks myosin IIa light chain phosphorylation thus interfering with contraction of actomyosin. Treating HAECs with Y-27632 did not impair cell spreading on fibronectin islands ( Figure 3d). However, the pronounced reduction of actomyosin contractility abolished cell-shape-dependent differences in NP uptake ( Figure 3e). We detected a twofold higher NP fluorescence signal for elongated cells (AR = 8) treated with Y-27632 compared to the non-treated cells. This result indicates that cytoskeletal contractility associated with cellular morphology may be the mechanism by which morphology regulates NP uptake kinetics.

Cell Shape Is a Geometrical Marker of Cytoskeletal and Focal Adhesion Anisotropy
In view of the intimate correlations between membrane mechanics and contractile stress fiber organization, we characterized cell shape-dependent F-actin filament assemblies [31] and focal adhesion patterns [32] in the patterned HAECs. We hypothesize that actin network anisotropy and focal adhesion patterns impose cell shape, which modulates membrane mechanics and hence cellular uptake of NPs. We characterized the average spatial distribution of actin filaments in the central region of the cell using pixel scan of a line crossing the cell, as shown in Figure 4a.
The central region of the cell was identified based on the nuclear position inside the HAECs. We found that cells with a lower aspect ratio (such as a square shape) have slimmer central actin stress fibers preferentially distributed diagonally. We next examined the changes in actin filament orientation and isotropy with the cell shape changes on the adhesive patterns ( Figure 4b). The actin network evolved from a flat and spread shape to parallel actin bundles on either side of the cell when cells were fully elongated. A sharp peak appeared for the highest aspect ratio, indicating a unidirectional orientation of the central actin fibers. The contractile actin stress fibers exert a compressive force onto the nucleus, causing nuclear shape remodeling. [33,34] Nuclei were randomly oriented for square-shaped and low aspect ratio cells, whereas elongated cells (higher aspect ratios) exhibited diagonal orientation of the nucleus. As aspect ratio increased, nuclei became gradually aligned with respect to the cell's long axis, and finally reached an oriented and elongated state ( Figure S3, Supporting Information).
Contractile stress fibers are physically linked to focal adhesion points, providing a mechanical force transduction pathway. [35] To determine how focal adhesions are organized in micropatterned cells, we stained HAECs adhered in protein micropatterns with cell spreading area of 2025 μm 2 for vinculin, a mechanosensitive focal adhesion-associated protein (Figure 5a).
The focal adhesions were localized at the corners of the squareshaped cells (AR1) but were found near the ends of the cells for elongated cells (AR8). This gradual increase in elongated cells, in addition to the formation of actin in the central cell layer and nuclear deformation suggests an overall increase in cellular stress.

Cell Morphological Phenotype Correlates to Mechanical Phenotype
Focal adhesion patterns and actin filament assembly are chemomechanical switches that generate and transmit mechanical forces, modulating cellular stress. [36] In particular, actomyosin motors power contractile stress fibers, and the contractile forces are transmitted through focal adhesions to generate extracellular traction. The traction force reacts back to the cell body, generating cellular stress. In view of the correlations between cell contractility and membrane mechanics and between cell contractility and cellular stress, we conceive of a direct relationship between cellular stress and NP uptake. Owing to the relatively high stiffness of the PDMS substrate (800 kPa-4 MPa), [37] traction forceinduced displacement in the substrate is very small, presenting a challenge to measure cellular stress by monolayer stress microscopy. Here we employed a recently developed biophysical model (Supplementary Materials) to simulate traction force and cellular stress. This model was validated by traction force microscopy and monolayer stress microscopy for cells seeded on soft hydrogels (<50 kPa). [12] As shown in Figure 5b,d, focal adhesions were concentrated at the elongated ends of the cells, and cell average stress increased from 726 to 1534 Pa by increasing cell elongations from AR1 to AR8, respectively. The cellular stress distribution is consistent with the presence of well-oriented central actin filaments for the elongated HAECs. For the square cell (in both the experimental results and the model, Figure 5c), the similar trend of the diagonal distribution of stress fibers and cellular stress validates both results. Interestingly, an increase in cellular stress is also consistent with nuclear elongation in high aspect ratio cells. This is understood from the compressive effect of the highly aligned, contractile stress fibers on the cell nucleus. Evolution of the lateral deformation, stress fiber formation on the side of the cells and increase in cellular stress are consistent with experimental data. Our finding agrees with results suggested by Balaban et al., showing a dependence between vinculin-containing focal adhesion area and tensional forces in endothelial cells. [35] The simulated cellular stress allowed us to directly correlate NP uptake to the cellular stress level. We found that cellular uptake was inversely dependent on cellular stress, as shown in Figure 5d. Notably, NP uptake was the highest (AR 1.5) when cellular stress was the lowest for the cells at the same spreading area (A = 2025 μm 2 ). This indicates that cellular stress is positively correlated with membrane mechanics, correlations which have not been explored previously.

Uptake of NPs Correlates with Cell Morphology in an Atherogenic Mimetic Microfluidic System
Endothelial cells exhibit morphological variability in vivo, especially at branches and regions of high curvature where the blood vessels are prone to plaque formation [15] (Figure 6a). In the previous sections, we showed that the accompanying mechanical characteristics of endothelial cells play a major role in nanoparticle uptake. We thus explored the competitive uptake of NPs by cells whose varying morphology reflects the prevailing fluid shear stress. To simulate these more realistic conditions, we adapted a microfluidic flow chamber/mixer that permitted laminar and chaotic flow, to induce cells to adopt elongated and non-elongated phenotypes, respectively. [38,39] In this microfluidic system the middle part of the chip uses a herringbone pattern on the channel ceilings that works as a mixer and creates chaotic flow profiles (Figure 6b and Figure S5a, Supporting Information) similar to the disturbed laminar flows found in atheroprone areas of the vasculature. To assess the flow patterns, we perfused 10 μm fluorescent beads and showed that the beads' trajectories were consistent with both laminar and chaotic flows (Figure 6b). This system was connected to a peristaltic pump that withdrew media from the reservoir, pumped it into the microfluidic channels connected in series, and then returned the media to the falcon tube to replenish the reservoir (Figure 6c and Figure S5b, Supporting Information).
To assess shear distributions experienced by cells, we used computational fluid dynamics simulations. Figure 6d shows the simulation of shear stress and particle trajectories in the microchannels with and without a herringbone pattern on the channel ceilings at a flow velocity that yielded a nominal shear stress on the cells of 20 dynes cm -2 . Simulations of 100 nm particles evenly distributed at the inlet where the initial position is color coded indicate significant mixing of flow streams in the area where chaotic flow is evident.
We then cultured HAEC or human umbilical vein endothelial cells (HUVEC) (supplementary data) in the chamber under dynamic conditions such that cells in the laminar flow region were exposed to a constant physiological shear flow of 20 dynes cm -2 . After culturing endothelial cells inside the microfluidic device, we observed structural adaptations to the cells induced by both laminar and chaotic flow. CD31 staining of endothelial cell borders revealed that cells in uniform laminar flows (without Streamlines generated at 20 dynes cm -2 using 10μm fluorescent bead validate the flow. c) Schematic and microfluidic set up, connected to a peristaltic pump and media reservoir. Arrows show the direction of fluid flow. d) Simulation obtained by finite element analysis software shows shear flow (ii) and particle tracing (iii). e) Cellular elongation, and nanoparticle uptake are shown via immunofluorescence imaging of CD31. The HAEC morphological adaptations are induced in cells exposed to laminar or chaotic shear flow (20 dynes cm -2 ). f) Normal distribution of cellular alignment in static, uniform flow, and chaotic flow against controls, (n = 10, 100 cells per condition). g) Quantification of cell spreading area, shows there is no difference between cell spreading under various flow conditions. h) Cellular circularity is measured to quantify elongation of the cells under laminar and chaotic flow conditions (n = 10, 100 cells per condition). i) NP uptake experimental timeline using microfluidic mixer. j) Cellular elongation, and nanoparticle uptake are shown via immunofluorescence imaging of cytoskeletal stress fibers (F-Actin staining), nuclei (DAPI) and red nanoparticles. k) HAEC uptake of nanoparticles changes in cells under different flow conditions, and the uptake is less for cells adapted to laminar flow compared to chaotic flow (n = 3). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.001. mixer) elongate in the direction of flow while cells in chaotic flow have less directionality and exhibit morphology similar to static conditions (Figure 6e). For schematics of the microfluidic device and experimental setup, see Figure S5 (Supporting Information).
We further quantified the cell orientation, cell spreading area, and cellular circularity (Figure 6f-h) and observed that, while cells maintained the same cell spreading area in all different conditions, in static and chaotic conditions HAECs maintained a cobblestone-like morphology (high circularity, ≈0.74 and 0.68), and under laminar flow conditions they decreased their circularity (≈0.25), as cells elongated in the flow direction.
After we ensured that the endothelial cells were shear-adapted and formed a physiological blood vessel structure, we perfused 100 nm nanoparticles (at flow rates 1/10 the rate used to shear the cells) and quantified the amount of uptake into endothelial cells at different time points using confocal microscopy. The timeline for the experiment and how cells form is shown in Figure 6i. As shown in Figure 6j, we observed that endothelial cells experiencing chaotic flow endocytosed a greater number of nanoparticles compared to cells experiencing laminar flow. We further quantified the amount of nanoparticle uptake in each cell type by measuring the fluorescence intensity in each condition. As shown in Figure 6k, when HAECs experienced higher shear stress, they www.advancedsciencenews.com www.advmatinterfaces.de endocytosed fewer nanoparticles. A study using human umbilical vein endothelial cells (HUVECs) resulted in similar findings ( Figure S6, Supporting Information).
Additionally, we also studied the overall NP uptake in monolayer HAECs in microfluidics in the presence of the ROCK inhibitor. We show that treating with Y-27632 did not impair cell direction or monolayer structure under different flow regimes in the chips; however, it led to an increase in NP uptake in areas under laminar flow ( Figure S7, Supporting Information). These results showed that treatment with ROCK inhibitors and reducing the actomyosin contractility abolished the morphology-and mechanics-related differences in NP uptake.

Discussion
In vivo, vascular ECs under shear flow exhibit distinctive morphologies, characteristic of chronic diseases such as atherosclerosis. [3,14,40] To mimic the shear flow-induced cell morphologies, we induced individual HAECs to adopt prescribed spreading areas and aspect ratios on adhesive islands using microcontact printing, resembling cell morphological phenotypes found in atherosclerosis-susceptible and atherosclerosis-resistant areas of the vasculature. [15,21,41] Nanoparticles have been shown to possess remarkable potential for treatment of atherosclerosis, [11,42,43] and studies have also been performed to investigate the role of NP size, shape, and elasticity on the rate of endocytosis. [44] In this study, the role of cell physical morphology, actin cytoskeleton, and membrane tension on NP uptake were studied.
Micropatterned endothelial cells were exposed to 100 nm spherical PS-COOH NPs, for which the clathrin-mediated endocytosis pathway is known to be the major endocytic route. [45][46][47][48] Actin cytoskeleton and membrane bending are the main driving forces in endocytosis [49,50] of PS-COOH NPs because of their contribution in the formation of clathrin-coated vesicles. In assessing NP uptake by the geometrically confined cells, we found that the overall NP uptake (per cell) increased with cell spreading area for fixed aspect ratio, while the uptake capacity (per area) decreased with cell aspect ratio at fixed spreading area. These results demonstrate a regulatory role of cell morphology on cellular uptake of NPs. Larger cell spreading area provides a larger membrane area for NPs to access, contact, and adhere, and thus enhances the overall uptake per cell. Although NPs internalization is driven by available net membrane area, it is penalized by an increase in membrane tension. Wang et al. showed a significant increase in the Young's modulus of cells for larger cell spreading area. [51] Therefore, higher membrane tension in larger cells can explain the demonstrated lower uptake capacity (per unit area). Notably, in simulated in vivo conditions, the EC aspect ratio is likely a more relevant driver of NP uptake differences since the area in cells aligned to laminar and chaotic flows are nearly identical (Figure 6g).
Changes in endothelial cell glycocalyx may explain some of the variation in nanoparticle uptake. [52] Glycocalyx is a mechanosensor located on the endothelial membrane that, in addition to sensing and translating the external hemodynamic signals, plays a critical role in selective permeability, barrier function, and antiinflammatory activity at the luminal side of the endothelium. [53] It is well known that during vascular disease progression, such as atherosclerosis, there is shedding of the endothelial cell glycocalyx. [54] Cheng et al. focused on scenarios in which these changes in the glycocalyx coat greatly affected the permeability of nanoparticles in endothelial cells. They showed that glycocalyx dysfunction, both collapsed and degraded, resulted in a significant increase in nanoparticle uptake. [52] Many studies showed that membrane curvature regulates the glycocalyx and cell membrane. [55,56] however, the correlation between the glycocalyx spatial distribution, membrane curvature, and cellular nanoparticle uptake is yet to be carefully investigated. While the role of glycocalyx cannot be discounted in our study, we focused on the role of membrane mechanics on NP uptake by measuring morphology-induced changes in membrane tension using the fluorescence lifetime of a membrane-bound fluorophore. A decrease in the excited state lifetime of the fluorescent probe integrated in the cell membrane of high aspect ratio cell indicates an increase of membrane fluidity which confirms that a high aspect ratio cell is under a higher membrane tension than a low aspect ratio cell. Higher tensions in the cell membrane reduce the ability of cell membrane to wrap around the nanoparticle. [57] Our findings are in agreement with previous studies showing that elongated endothelial cells, exposed to shear stress, exhibit a significantly greater mechanical stiffness [6] and membrane rigidity compared to the control cells, due to the development of actin filaments in cells experiencing shear stress.
To confirm the role of actin cytoskeleton in the uptake we showed an increase in cellular uptake when actin was perturbed by a cell contractility inhibitor in elongated cells. This suggests that the pronounced reduction of actomyosin contractility abolished cell-shape-dependent differences in NP uptake. Furthermore, our microscopy imaging revealed that elongated cells assembled a very long and well-organized actin cytoskeleton that formed a thick parallel actin stress fibers throughout the cells. These actomyosin fibers exert compressive forces that not only cause a significant nuclear deformation but also enhance membrane tension. [58,59] Membrane bending is strongly actindependent and endocytosis under such high membrane tension requires higher force to bend the membrane, resulting in a lower amount of nanoparticle invagination. These results indicate that the cytoskeletal structure is an important factor in the regulation of membrane bending and internal energy, which contributes to nanoparticle uptake.
We further observed that elongated cell shape is supported by highly aligned actin stress fibers and concentrated focal adhesion patterns, which correlates to enhanced cellular stress. Based on the focal adhesion concentration and localization, along with internal cellular force transmitters, we constructed a thermodynamic model and showed an increase in cellular internal stress when cells are elongated. Kinetically, the increased membrane tension and cellular stress in elongated cells present an increased mechanical barrier for NP endocytosis, thereby suppressing cellular uptake. Taken together, cell shape, as a geometrical marker for mechanical phenotype of the cells, controls cellular uptake of NPs (Figure 5e).
We then studied the differences in NP uptake in ECs with intact cell junctions by culturing them in a microfluid chamber that simulated neighboring laminar shear and chaotic flow profiles. This system induced cells to adopt both atheroprone and atheroprotective phenotypes similar to those represented in human blood vessels. Staggered herringbone structures have primarily been designed and used for improved mixing in microfluidics, [38] and only a few studies have used this structure to generate different flows for cell culture. [39] A study by Levesque et al. showed that by changing vascular geometry, the morphology and shear stress of the aortic endothelial cells can be altered. [60] Although attributing the morphological changes only to flow is challenging, our microfluidic device simulating atherogenic flow conditions shows that similar morphological changes can be achieved in vitro by exposing cells to different flow regimes. Cells that were cultured in our chip maintained both elongated and non-elongated cells in a statistically distinguishable manner, like what is observedin in vivo conditions.
Our computational fluid dynamics simulations and flow visualization studies showed that under an inlet flow velocity that resulted in shear on the cells in the laminar region of 20 dyne cm -2 , we obtain significant chaotic flow in the area under the grooves on the ceiling resulting in shear on the cells that is low, multidirectional, and with high shear stress gradients. We then studied the impact of shear-induced cell morphology upon nanoparticle uptake in endothelial cells. During the nanoparticle study, we used a very low flow (1 dyne cm -2 ) to avoid the effect of flow on nanoparticle distribution since it has been reported that flow has a direct effect on NP accumulation. [61] Results show a strong correlation between rounded morphology in the chronic flow adapted cells and NP uptake. This result suggests that NPs might be differentially accumulated in these areas of flow disturbance in vivo, especially at the early stages of atherogenesis. The result also points to the utility of using microfluidic in vitro models to study NP uptake in complex biological and mechanical systems. [62] Despite the demonstrated capabilities and novel insights of our models, given the complexity of blood vessel structure, new models need to be developed to replicate the structure and function of this tortuous vasculature during endocytosis. For example, it may be possible to create a model that incorporates both endothelial cells and smooth muscle cells. Another important goal for future investigation is to examine the internalization of nanomaterial with different properties including size, shape, stiffness, and surface chemistry on micropatterned cells. Cellular uptake is a complex behavior and smaller size nanoparticles may use different endocytosis pathways including caveolin-dependent, receptor-mediated, non-specific and translocation pathways. [44] In conclusion, the mechanical control of cellular uptake of NPs refocuses NP design for enhanced targeting specificity. In chemotargeting, the over-expressed receptors on the cell surface together with the high ligand-receptor binding affinity offer strong adhesion of NPs to diseased cells, but low adhesion to the normal cells, constituting the biophysical basis of chemotargeting. While chemotargeting has been widely successful, the mechanical barrier of the cell membrane presents a critical unexplored factor regulating cellular uptake of NPs. Our studies show that mechanosensitive cells, such as ECs, can sense the external mechanical cues, and respond by adopting different morphologies. Cell morphologies, in turn, are indicators of the mechanical phenotypes, which bias the cellular uptake, enabling mechanotargeting. In addition to targeting cells whose morphologies are associated with atherosclerosis as demonstrated here, mechanotargeting could also be applied to cancer diagnosis and treatment as some metastatic cells are generally softer and less stressed than non-metastatic cells. [63] The intimate correlations between cell morphological and mechanical phenotypes also open additional mechanotargeting pathways, for instance, by disrupting cell contractility and/or focal adhesions. Our findings thus pave the way for rational design of NP-based noninvasive imaging modalities for early detection, diagnostic and delivery of therapeutic agents with improved targeting efficiency for a wide range of mechanobiologically relevant diseases.

Experimental Section
Microcontact Printing and Mold Fabrication: Adhesive protein micropatterns of various shapes (square and rectangular), various areas (900, 1225, 1600, 2025, and 2500 μm 2 ) and different aspect ratios (AR = 1, 1.5, 2, 4, and 8) were fabricated on a silicon wafer by photolithography. Micropatterned substrates containing fibronectin (FN)-coated adhesive islands were produced according to a modified protocol for microcontact printing. [18] After fabricating the master mold by photolithography, stamps were made by replica casting polydimethylsiloxane (PDMS; Dow Corning) against the silicon master mold. A mixture of the PDMS base and curing agent (10:1 ratio) was poured onto the silicon master mold and cured at 70°C for approximately 2 h. The PDMS stamp bearing the negative patterns was peeled off upon solidification, followed by surface oxidization in air plasma for 7 min, and immediately used for printing the FN. Plasma-activated stamps were coated with 25 μg mL −1 FN at room temperature for 1 h. Next, PDMS stamps were dried and used to ink and pattern FN on the PDMS coated glass coverslip by 60 s conformal contact of FN-coated stamp with the substrate. Subsequently, the substrates were coated in 1% (w/v) Pluronic F127 (Sigma) for 1 h to prevent cell adhesion to non-coated adhesive islands. The cells were cultured on the surface of the plate.
Microfluidic Fabrication: The master mold was fabricated using 3Dprinting of PC-Like Advanced High Temp (Accura 5530) (Protolab, USA). The microfluidic chip is comprised of a cell culture chamber of the following dimensions: 35 mm length × 0.5 mm width × 75 μm height, with 40 μm grooves. The microfluidic culture devices were created by casting PDMS against micropatterned 3D printed molds using standard softlithography protocols. The casted PDMS was cured in a 65°C oven. The cured PDMS was stamped against a thin layer of uncured PDMS (spincoated at 1500 rpm for 3 min), and then sealed against a glass slide.
Microfluidic Fabrication-Microfluidic Chip: Human aortic endothelial cells (HAECs, CC-2535; Lonza) and human umbilical vein endothelial cells (HUVECs, C2517A; Lonza) were cultured and maintained using Endothelial Basal Medium, supplemented with all growth factors (EGM-2 BulletKit, CC-3162). Prior to seeding cells in the chip, the surface of the microfluidic devices was coated with fibronectin for 1 h, followed by priming chips with EGM-2 media for 11 h. Both HAECs and HUVECs were detached and pipetted into each channel of the microfluidic devices at concentrations of 2-3×10 6 cells mL -1 . Cells were allowed to attach to the channel walls under static conditions for 1 h at 37°C, 5% CO 2 before media was introduced to the chips, after which the attached cells were washed with fresh medium and the chip was connected to the flow setup.
On-Chip Perfusion: A closed-loop peristaltic pump (Fisherbrand Variable-Flow Peristaltic Pumps, Fisher Scientific, US) was setup in accordance with the manufacturer's protocol. The flow rates were manually determined, calibrated, and verified prior to connecting them to the microfluidic devices. All the connections and tubing were autoclaved and equilibrated with a culture medium 12 h prior to cell seeding. Endothelial cells were seeded and, after attachment, they were sheared at low flow of 1 dyne cm -2 for 24 h. Cells were exposed to 20 dyne cm -2 for 48 h to produce high shear stress for shear-adapted endothelium. And after that, at low flow yielding 1 dyne cm -2 , cells were exposed to nanoparticles.
Immunohistochemistry and Labeling: Immunostaining was performed mainly by staining F-actin and the nucleus following published methods. [64] Briefly, after washing three times with PBS buffer (to wash out the residual NPs in the solution and on the membrane surface of the cells), cells were fixed with 3.7% paraformaldehyde for 15 min. Next, cells were incubated with a permeabilization buffer, 3% bovine serum albumin and 0.1% Triton X-100 in phosphate-buffered saline (PBS), for ≈40 min. The cells were incubated with 1:1000 dilution of CF 488 conjugated phalloidin (Biotium, Hayward, CA) at room temperature for 20 min to stain the actin stress fibers and 5 min incubation with 1:5000 dilution 40-6-diamidino-2-phenylindole (DAPI; Pierce, Rockford, IL) to stain the nuclei. Subsequently, the samples were completely washed with PBS three times and were mounted using ProLong Gold antifade reagent (Invitrogen, Eugene, OR). To confirm the endocytosis of the red dye-loaded (excitation/emission wavelengths: 580/605) carboxylatemodified polystyrene 100 nm NPs, fluorescence images of the cells were taken under a super-resolution microscope, Z-stack (Nikon). To study NP uptake, samples were imaged with a Leica DM5500B microscope (Leica Microsystems, Buffalo Groove, IL) with a 10× and 40× water immersion objective (Leica Microsystems, Buffalo Groove, IL).
Microfluidic Chip Characterization: The microfluidic chip was characterized using confocal microscopy (Zeiss microscope). All devices that would be compared to each other were scanned using the same laser power settings. Cells were imaged using a 10× objective lens (10×, NA 0.45) for DAPI (405 nm laser excitation, 450/50 emission filter cube), CD31 (488 nm laser excitation, 525/50 emission filter cube), nanoparticle (561 nm laser excitation, 600/50 emission filter cube), and for phalloidin (647 nm laser excitation, 685/70 emission filter cube). NP uptake was assessed using fluorescence intensity of the red channel defined by the cell area as a mask using ImageJ (NIH, Bethesda, MD).
Disruption of Actin Polymerization Using Y-27632: Y27632 is a compound that inhibits stress fiber formation by Rho signaling pathway inactivation via inactivation of ROCK. Inhibition of actin polymerization by Y27632 was accomplished according to the manufacturer's protocol. Briefly, after cells were grown for 6 h, and prior to loading NPs, cells were treated with Y27632 (Cytoskeleton. Inc.) and diluted in cell media (final concentration of 10 × 10 -6 m) for 30 min.
Actin Directionality Analysis: The orientation distribution of cells' actin filament was quantified in OrientationJ (NIH, Bethesda, MD) based on the F-actin fluorescence images. The orientation using the structure tensor is accomplished by evaluating every pixel of the image for orientation. Software provides an orientation histogram based on considering pixels with coherency larger than min-coherency and with an energy larger than minenergy.
Staining of Cell Membranes and Fluorescence Lifetime Measurements: HAEC membranes were stained with 1,1'-didodecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate (DiI-C 12 ) (Invitrogen, Eugene, OR) at a final concentration of 1.0 × 10 -6 m of DiI-C 12 in PBS from the stock solution of 65.3 × 10 -3 m in ethanol. Cells were stained for 4 min at 37°C followed by extensive washing with DPBS. For lifetime measurements phenol-red free DMEM with 10% FBS and 2.5% HEPES was used. Time-correlated single photon counting (TCSPC) was used to measure the fluorescence lifetime of DiI-C 12 , as previously described. [24] The laser spot was located by creating a dark spot from photobleaching on a DiI-covered coverslip on the same day as the experiment. The instrument response function (IRF) was measured by collecting the data from 0.2 m Rhodamine 6G solution (Invitrogen, Eugene, OR) in methanol. With special attention and precision the peak in fluorescence on the top cell surface was found using the fluorescence images under a 60× water-immersion objective (NA = 1.20, UPlanApo; Olympus, Tokyo, Japan) and z-scan with oscilloscope module of the software. After finding the apical surface of each cell the fluorescence microtime data was collected and the fluorescence lifetime computed using a Levenberg-Marquart iterative reconvolution of a doubleexponential decay curve with the measured microtime distribution (Easy Tau 2, Picoquant, Berlin, Germany). The detailed microscopy system setup and theoretical basis can be found in ref. [25].
Numerical Simulation: A numerical simulation was conducted to investigate and predict the shear stress characteristics and particle trajectories with finite element software (COMSOL Multi-physics 6.0, Burlington, MA). A 3D model of the microfluidic chip with staggered herringbone grooves was built using the modeling function of COMSOL. A laminar flow model (Navier-Stokes, incompressible fluid) was then added to the component to calculate the flow field. An inlet velocity was chosen such that 20 dyne cm -2 was experienced on the floor of the chamber in the laminar flow region.
Statistical Analysis: A minimum of three biological replicates were used for each condition. The data are represented as mean ± standard error of the mean (SEM) except in Figure 5c, where data are presented as mean ± standard deviation (SD). Unpaired, two-tailed Student's t-tests were used to compare quantitative analyses of two groups of n = 3 or more samples, and one-way ANOVA and post-hoc tests for multiple comparisons were used to compare three or more groups. Differences were considered statistically significant at p-values < 0.05. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. GraphPad Prism (ver. 9; GraphPad Software) was used to statistically analyze the data.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.