The spectrin cytoskeleton integrates endothelial mechanoresponses

Physiological blood flow induces the secretion of vasoactive compounds, notably nitric oxide, and promotes endothelial cell elongation and reorientation parallel to the direction of applied shear. How shear is sensed and relayed to intracellular effectors is incompletely understood. Here, we demonstrate that an apical spectrin network is essential to convey the force imposed by shear to endothelial mechanosensors. By anchoring CD44, spectrins modulate the cell surface density of hyaluronan and sense and translate shear into changes in plasma membrane tension. Spectrins also regulate the stability of apical caveolae, where the mechanosensitive PIEZO1 channels are thought to reside. Accordingly, shear-induced PIEZO1 activation and the associated calcium influx were absent in spectrin-deficient cells. As a result, cell realignment and flow-induced endothelial nitric oxide synthase stimulation were similarly dependent on spectrin. We conclude that the apical spectrin network is not only required for shear sensing but also transmits and distributes the resulting tensile forces to mechanosensors that elicit protective and vasoactive responses. Mylvaganam et al. report that an apical spectrin network in endothelial cells can transmit mechanical forces in response to shear flow-induced stress, requiring hyaluronic acid and involving PIEZO1.

P recise haemodynamic control largely relies on endothelial cells, which sense forces exerted by flowing blood 1,2 . In response to shear, endothelial cells secrete vasoactive compounds, including nitric oxide (NO), that trigger acute changes in vascular tone and/or initiate vascular remodelling to modulate vessel diameter 1,2 . NO also prevents pro-coagulant factors from interacting with the vessel wall to preserve blood fluidity 3,4 . Physiological blood flow also promotes endothelial cell elongation and reorientation parallel to the direction of shear 5 . This cellular alignment confers protection against haemodynamic stress and is associated with anti-inflammatory gene expression 6,7 .
Several mechanisms of endothelial shear-sensing and mechanotransduction have been proposed. One widely accepted model implicates a junctional mechanosensory complex involving PECAM-1 (also known as CD31), VE-cadherin and vascular endothelial growth factor receptor 2 (VEGFR2). According to this model, shear exerts strain across intercellular junctions and this activates PECAM-1, which in turn activates Src family kinases 8 . Close apposition of VE-cadherin, VEGFR2 and PECAM-1 facilitates the transactivation of VEGFR2 by PECAM-1-associated Src in response to shear 9 . Phosphorylated VEGFR2 then activates downstream signalling molecules, including PI3K and AKT 9,10 . However, PECAM-1-independent, shear-induced activation of several signalling pathways has also been observed 11 . Moreover, PECAM-1-deficient mice develop normally 12 . These observations are inconsistent with an indispensable role for PECAM-1 in endothelial responses to shear.
Other junction-independent processes probably play a role in endothelial mechanotransduction. In this regard, the glycocalyx has been implicated in shear-induced responses, and hyaluronic acid (HA) seems particularly important. HA, a non-sulfated glycosaminoglycan, is anchored to the cell surface by binding to its receptors, primarily CD44 (ref. 13 ). Enzymatic degradation of HA in vitro and in vivo disrupts endothelial responses to shear [14][15][16] . Mechanosensitive ion channels, including PIEZO1, are also impli-cated in endothelial responsiveness to shear 17 . Thus, multiple, seemingly independent, molecules cooperate to generate endothelial responses to flow. However, a coherent mechanism describing how these are coordinated is lacking.
Endothelial cells display a polarized spectrin network restricted to their apical (luminal) aspect, which most directly experiences fluid shear 18 . Spectrin networks, comprising α-subunit and β-subunit heterodimers that stabilize short actin filaments, regulate the dynamics of plasmalemmal glycoproteins and ion channels [19][20][21][22] . Importantly, in the endothelium, spectrin controls the apical retention and stabilization of CD44, the primary receptor for HA 18 . It therefore seems plausible that spectrin regulates endothelial responses to shear, conceivably through its associations with CD44 and HA.
In an effort to resolve the uncertainties regarding the relative contribution of the various endothelial mechanotransduction processes, we used a combination of imaging and biophysical approaches to systematically analyse the role of the intercellular junctional complex, HA and PIEZO1 in sensing shear. We also examined the possibility that spectrin networks serve as integrators and/or transducers of the forces imposed by flowing fluid.

Cell-cell junctions are not required for shear-induced endothelial alignment.
Owing to the importance of endothelial mechanoresponses in the arterial circulation, we primarily used an immortalized line of human aortic endothelial cells, telo-HAECs, (tHAECs) for our analyses. Cells grown in collagen-coated microfluidic chambers were subjected to a constant shear stress (15 dynes cm -2 ) that mimicked conditions encountered in the arterial circulation 23 . Changes in endothelial morphology in response to applied shear were assessed using phalloidin to visualize F-actin (Fig. 1a). As previously reported, endothelial cells grown as confluent monolayers under static conditions had randomly oriented The spectrin cytoskeleton integrates endothelial mechanoresponses actin stress fibres 5 (Fig. 1a). After brief (30 min) exposure to flow, the cells elongated in the direction of the applied stress, which coincided with the parallel orientation of their stress fibres (Fig. 1a). To quantify alignment, individual stress fibres were segmented using a Hessian-based multiscale filter. The angle between the longest axis (maximum Feret diameter) of each segment and the direction of applied flow was then calculated (Fig. 1a). The interquartile distribution of segmented filament orientations, used to quantify the variability between conditions, revealed markedly less variation in fibre orientations (which we interpreted as alignment) in cells exposed to shear compared with those in static culture (Fig. 1a).
Notably, in subconfluent cultures, isolated endothelial cells that lacked visible intercellular contacts retained the ability to elongate and align following exposure to shear, which mimicked the behaviour observed for monolayers (Fig. 1a). This occurred despite the internalization of PECAM-1 from the surface of single   Fig. 1 | Hyaluronan is an indispensable, junction-independent regulator of endothelial responses to shear stress. a, Confluent endothelial monolayers or sparsely distributed (subconfluent) single cells were grown in collagen-coated microfluidic chambers under static conditions or subjected to a shear stress of 15 dynes cm -2 for 30 min before fixation and phalloidin staining. First column: F-actin visualization. Second column: segmented actin filaments. Third column: colour labelling of filaments according to orientation relative to the shear axis (0°). Fourth column: polar histograms of F-actin orientations from one representative experiment. Fifth column: interquartile ranges of the distribution of segmented filament orientations for the indicated conditions. Data are the mean ± s.e., n = 5 independent experiments for confluent cells and n = 3 independent experiments for sparsely distributed cells, each pooling analyses from 10 images (as in the first column) per condition. WT, wild type. b,c, Distribution of PECAM-1 in confluent endothelial monolayers (b) and single cells (c) assessed by immunofluorescence. n = 3 independent experiments, each pooling 20 images per condition. d,e, Confluent monolayers of endothelial cells were treated with vehicle (PBS) or HAase and subjected to constant fluid flow as in a. d, Representative images of cells fixed and stained with phalloidin (F-actin, left) and interquartile ranges of F-actin orientations (right). Data are the mean ± s.e., n = 4 independent experiments, each pooling analyses from 10 images per condition. e, Endothelial cells expressing the cytosolic [Ca 2+ ] indicator GCaMP6 before (white background) and after the application of shear (grey background). Images were acquired every 12 s. Left: mean GCaMP6 fluorescence over time normalized to the initial resting value. Data are the mean ± s.e. (shaded area, here and elsewhere), n = 3 independent experiments, pooling 9, 9 and 15 for vehicle-treated cells, and 8, 10 and 12 for HAase-treated cells. Right: maximum normalized GCaMP6 fluorescence induced by shear. Here and elsewhere, dots represent maximum values for each cell, colour coded by experiment. Boxplots illustrate the median (central line), 25th and 75th percentiles (lower and upper regions of the box) and full range (whiskers) of pooled data. f,g, Representative confocal extended-focus projection along with brightfield illumination (f) and orthogonal section (g) of confluent and sparsely distributed subconfluent endothelial cells incubated with fluorescent HABCs. n = 4, pooling 12 images per condition. h,i, Sparsely distributed endothelial cells were treated with vehicle (PBS) or HAase and analysed as in d (h) and e (i). P values from Student's t-test (a,e,i) or one-way analysis of variance (ANOVA) (d,h) of experimental means. Scale-bar, 5 μm (a-d,f-h).
Maintenance of luminal HA surface density depends on spectrin. It remained unclear how transmembrane glycoproteins such as CD44 activate downstream signalling cascades. Apical retention and immobilization of CD44 are regulated by the spectrin cytoskeleton 18 , which predominantly localizes to the apical surface of confluent and isolated endothelial cells ( Fig. 2f and Extended Data Fig. 1b). Notably, CD44 was not required for the accumulation of apical spectrin (Extended Data Fig. 1c), which implied that other proteins and/or lipids anchor spectrin to the plasmalemma.
We considered the possibility that spectrin might facilitate communication between the glycocalyx and other plasmalemmal molecules. The potential relationship between HA and spectrin was assessed by studying HA abundance on the surface of cells in which β-spectrin was knocked down (Fig. 2g,h). Wild-type endothelial cells grown in culture had appreciable surface HA (detectable by ELISA), which was almost eliminated by HAase (Fig. 2h). Spectrin depletion by RNAi caused a pronounced decrease in the surface density of HA (Fig. 2h). This result suggests that spectrin may regulate endothelial functions associated with HA.
Spectrin is required for endothelial responses to shear. To perform a systematic analysis of the role of spectrin, we used CRISPR-Cas9 technology to delete β-II spectrin (encoded by SPTBN1) in tHAECs (Fig. 3a). SPTBN1 knockout (KO) cells did not exhibit any overt morphological defects, and the distribution of proteins that comprise the junctional complex to intercellular contacts in confluent monolayers was comparable to that in wild-type cells (Extended Data Fig. 2a). Ratios of VE-cadherin fluorescence intensity in the junctional membrane to that in the cytosol were also similar between the KO clones and the parental wild-type cells, which indicates that loss of β-II spectrin is not associated with overt defects in junctional integrity (Extended Data Fig. 2b). SPTBN1 KO cells also retained Weibel-Palade bodies that were positive for von Willebrand factor (Extended Data Fig. 2c), and the distribution and density of focal adhesion proteins was also comparable to that of wild-type cells (Extended Data Fig. 2d-f).
Following exposure to fluid flow for 30 min, stress fibre alignment failed to occur in SPTBN1 KO cells (Fig. 3b-d and Extended Data Fig. 3a). Instead, the architecture of F-actin networks in these cells more closely resembled that observed in wild-type cells in static culture ( Fig. 3b-d). Changes in cell shape were also assessed by immunostaining for VE-cadherin in cells subjected to static or shear-stress conditions. The cellular aspect ratio (the ratio of the long axis to the short axis of each cell) was quantified. As expected, aspect ratios of wild-type tHAEC monolayers following exposure to shear stress were significantly higher than those of cells maintained in static conditions, which indicated that these cells have a more elongated phenotype (Extended Data Fig. 3b). Aspect ratios of SPTBN1 KO cells exposed to shear stress were lower, which indicated that these cells failed to elongate (Extended Data Fig. 3b). Shear-induced increases in tyrosine phosphorylation of junctional proteins were comparable between wild-type and SPTBN1 KO cells (Extended Data Fig. 3c). Finally, no appreciable changes in [Ca 2+ ] cyto were detected in SPTBN1 KO cells in response to shear (Fig. 3e,f). Importantly, although HA density itself depended on spectrin (Fig. 2h), overexpression of HA synthase to increase the surface density of HA (Extended Data Fig. 4a) was insufficient to rescue defective shear responses in SPTBN1 KO cells (Extended Data Fig. 4b). Therefore, receptors such as CD44, which physically connect HA to the cortical cytoskeleton, seem essential to mediate spectrin functions in shear-stress responses.
Shear-induced changes in tension across spectrin molecules require HA. We next investigated whether the application of shear exerts tension on apical spectrin. Its spectrin-repeat domains are triple-helical coiled-coil domains that unravel and rewind in response to mechanical strain [30][31][32][33] . To assess tension, we used a fluorescence resonance energy transfer (FRET)-based probe consisting of α-spectrin with Cerulean and Venus fluorescent proteins conjugated to either end of repeat domains 10 and 11 (ref. 34 ). Increases in tension across this probe, named spectrin-cpst-FRET, cause a decrease in FRET owing to the mechanically induced unravelling of coiled-coil domains between the donor and acceptor 34 .
Similar to endogenous spectrin ( Fig. 2f and Extended Data Fig. 1b), spectrin-cpst-FRET accumulated preferentially at the apical surface of endothelial cells (Fig. 4a). Compared with cells in static culture, FRET efficiency (quantified as the ratio of Venus/Cerulean mean fluorescence intensity following Cerulean excitation) decreased following acute application of shear stress (Fig. 4b,c), which is consistent with increased tension in spectrin.
Spectrin is unlikely to directly sense the apical shear flow. More likely, HA, via its receptors, transmits forces from the vascular lumen to the spectrin cytoskeleton. Indeed, pretreatment with HAase eliminated the shear-induced change in FRET (Fig. 4c).
These experiments provide evidence that HA conveys shear signals to the spectrin network.
Altered membrane tension in spectrin-deficient cells. Spectrin networks associate with the plasma membrane by tethering proteins and through direct interactions with phospholipids 35 . This membrane association is important as it confers a degree of flexibility to the cell surface 36,37 . We wondered whether this close association directly transmits tension changes in spectrin to the plasma membrane when shear stress is applied. To assess membrane tension, we performed fluorescence lifetime imaging microscopy (FLIM) using Flipper-TR. The fluorescence lifetime of Flipper-TR increases at higher membrane tensions 38 , which we confirmed by osmotically swelling wild-type cells (Fig. 4d). In isotonic conditions at rest, the fluorescence lifetime of Flipper-TR in the plasma membrane of SPTBN1 KO cells was significantly longer than that of their wild-type counterparts (Fig. 4d). This change occurred despite similar surface areas of the cell lines, as determined by Spectrin and HA regulate shear-induced changes in membrane tension. We next studied the effects of shear on membrane tension. Shear stress acutely induced a significant increase in Flipper-TR fluorescence lifetime in wild-type cells, thereby revealing an increase in plasma membrane tension (Fig. 4e), consistent with an increase in intramolecular spectrin tension. In SPTBN1 KO cells, while shear induced an increase in Flipper-TR fluorescence lifetime (Fig. 4e), the magnitude of this increase was markedly lower than that observed in wild-type cells (Fig. 4e). Consistent with its effects on tension across spectrin, treatment with HAase had no appreciable effect on membrane tension of cells in static culture, but greatly inhibited the shear-induced increase in Flipper-TR lifetime in wild-type cells (Fig. 4e).   24,39 ), which implicates that plasmalemmal channels that are mechanosensitive have a role. Multiple shear-sensitive Ca 2+ channels have been described. Shear triggers the release of ATP, which acts in an autocrine manner to induce Ca 2+ entry. P2X4 purinoceptors, which are ATP-operated cation channels 40 , and P2Y receptors, which are nucleotide-activated G protein-coupled receptors 41 , have both been implicated in endothelial mechanotransduction. However, we found that ATP-induced Ca 2+ transients in SPTBN1 KO cells were comparable to those in wild-type cells (Fig. 5a). The defective responses to shear in spectrin-deficient cells are therefore not attributable to impaired purinergic receptor activity.
Notably, however, responses to mechanical stimuli were impaired in SPTBN1 KO cells (Fig. 5b,c). Increasing plasma membrane tension by osmotic swelling resulted in a significant increase in [Ca 2+ ] cyto in wild-type but not in SPTBN1 KO cells (Fig. 5b,c). Importantly, swelling-induced [Ca 2+ ] cyto was regulated by stretch-activated channels, as pretreatment with GsMTx4, an amphipathic peptide that inhibits cationic mechanosensitive currents 42 , eliminated this response (Extended Data Fig. 5 and Fig. 5c).
The stretch-activated, non-selective cation channel PIEZO1 mediates (at least part of) the shear-induced [Ca 2+ ] increases in endothelial monolayers and their alignment in the direction of flow 17 . Consistent with this report, pretreatment of tHAECs with GsMTx4, which blocks PIEZO1, abrogated the shear-induced [Ca 2+ ] increase in confluent monolayers (Fig. 5d). Importantly, this treatment also blunted Ca 2+ responses in single cells (Fig. 5e). GsMTx4 also inhibited the alignment of sparsely distributed cells in response to shear (Fig. 5f). These observations indicate that the indispensable role of mechanosensitive channels in endothelial responses to flow is regulated independently of junctional complexes.
The involvement of PIEZO1 was then probed more precisely using the specific agonist Yoda1. In static culture, treatment of wild-type cells with Yoda1 induced increases in [Ca 2+ ] cyto that strongly resembled the response to flow (compare Fig. 5d to 5g). By contrast, Yoda1 did not elicit significant [Ca 2+ ] cyto changes in SPTBN1 KO cells (Fig. 5g). We conclude that the apical β-II spectrin network regulates the activity of PIEZO1.
Plasmalemmal expression of PIEZO1 is not regulated by spectrin. The unresponsiveness of SPTBN1 KO cells to Yoda1 suggested two possibilities: (1) spectrins are required for the plasma membrane retention of PIEZO1; or (2) PIEZO1 activity at the cell surface is regulated by the spectrin cytoskeleton. To test the first possibility, we examined the expression and cellular distribution of PIEZO1. Immunoblotting experiments revealed comparable concentrations of PIEZO1 in wild-type and SPTBN1 KO cells (Fig. 6a and Extended Data Fig. 6a). Using immunofluorescence microscopy, we observed that in wild-type cells, PIEZO1 appeared as punctate structures concentrated on the apical cell surface, which is consistent with its shear-sensing functions (Fig. 6b). While its expression appeared to be largely preserved in SPTBN1 KO cells, PIEZO1 no longer concentrated apically in the absence of spectrin (Fig. 6b). The altered PIEZO1 responsiveness in spectrin-deficient cells may therefore be attributable to changes in its localization.

Spectrin stabilizes endothelial caveolae that regulate PIEZO1.
Earlier studies of liposomes and membrane blebs indicated that changes in bilayer tension are sufficient to activate PIEZO1 (ref. 43 ). Importantly, PIEZO1 expression is associated with local distortion of the lipid bilayer, which results in its preferential localization to curved membrane domains 44,45 . This localization confers high sensitivity to PIEZO1 to curvature changes that result from alterations in membrane tension 44,45 .
Indeed, endothelial cells are uniquely enriched in curved membrane microdomains, termed caveolae 46,47 , that unfurl or flatten in response to applications of force 48 . Localization of PIEZO1 to caveolar structures has previously been reported in other cells 44,49 . Accordingly, in wild-type endothelial cells, immunofluorescence revealed significant colocalization of PIEZO1 and caveolin-1, an essential component of caveolae (Fig. 6c). Because caveolae are stabilized at the cell surface via interactions with cytoskeletal proteins 50 , we investigated the effects of spectrin on caveolae as these might indirectly influence PIEZO1 activity.
Similar amounts of total caveolin-1 were detected in wild-type and SPTBN1 KO endothelial cells (Fig. 6d). In wild-type endothelial cells, caveolin-1 formed punctate arrays that were enriched on the apical surface, which was consistent with the reported abundance of caveolae 51,52 (Fig. 6e and Extended Data Fig. 6c). This polarized distribution was lost in SPTBN1 KO cells, in which a greater fraction of caveolin-1 was found in the basal and internal membranes ( Fig. 6e and Extended Data Fig. 6c). The number of caveolin-1 monomers in each apical cluster was also decreased in SPTBN1 KO cells (Fig. 6f).
Spectrin can regulate the mobility of caveolin-1 in the plane of the membrane 18 . However, it remained unclear whether the behaviour of caveolin-1 reflects the number or appearance of the caveolae themselves. We therefore studied the integrity of curved caveolar domains by examining the degree of colocalization between cavin-2 (a coat protein that maintains caveola invagination) and caveolin-1. As previously reported 48 , colocalization between these proteins decreased following flattening of caveolae with hypotonic stress (Extended Data Fig. 6e). We also found significantly lower colocalization of caveolin-1 and cavin-2 in SPTBN1 KO cells compared with their wild-type counterparts, which suggests that there was decreased abundance of caveolar bulbs (Fig. 6g). This possibility was specifically addressed by performing transmission electron microscopy (TEM), analysing sagittal sections of endothelial cells. In wild-type cells, TEM revealed a high apical density of uncoated, bulb-shaped plasmalemmal structures, which we identified as caveolae (Fig. 6h,i). While still present, markedly fewer caveolae were observed on the surface of SPTBN1 KO cells (Fig. 6h,i). Increased surface tension at rest in SPTBN1 KO cells could account for the disappearance of caveolae, thereby causing their flattening.
Shear stress induces flattening of caveolae. Since caveolae play an important role in calibrating plasma membrane tension 48 , and given the relationship between spectrin expression, membrane tension and caveolar density and curvature, we wondered whether shear stress results in flattening of caveolae. Conceivably, changes in tension-induced caveolar flattening could underlie the defects in PIEZO1 function presented in Fig. 5. We therefore analysed the colocalization of caveolin-1 and cavin-2 before and after acute application of shear stress. Similar to hypotonic stress (Extended Data Fig. 6e), shear stress reduced the association between cavin-2 and caveolin-1 (Fig. 6j), indicative of caveolar flattening. HAase treatment eliminated this response (Fig. 6j). Together, these observations suggest that spectrin conveys forces sensed by HA to induce changes in plasma membrane tension that flatten caveolae, which in turn activates the tension-sensing PIEZO1.
Shear-induced production of NO is regulated by spectrin. We next examined whether spectrin is also implicated in the shear-induced production of NO. In the vasculature, NO is primarily produced by endothelial NO synthase (eNOS), which interacts with caveolin-1 under basal conditions. Caveolin-1 maintains eNOS in an inactive state by sequestering it on caveolae. Increases in [Ca 2+ ] cyto result in the activation of calmodulin which, in its Ca 2+ -complexed form, binds eNOS, disrupting its inhibitory interaction with caveolin-1 (ref. 53 ). Once dissociated from caveolae, eNOS can be activated through phosphorylation [54][55][56][57] . We first tested whether the increased [Ca 2+ ] cyto induced by PIEZO1 was sufficient to activate eNOS. To this end, we immunoblotted cell lysates with antibodies that detect phosphorylation of Ser1177 of eNOS. Indeed, treatment of static endothelial cells with the PIEZO1 agonist Yoda1 induced a distinct increase in eNOS phosphorylation (Fig. 7a). Application of shear (applied by rotary agitation) also induced a large increase in eNOS phosphorylation, which was eliminated by pretreatment with the PIEZO1 inhibitor GsMTx4 (Fig. 7a). PIEZO1 activation was therefore sufficient and necessary for NO production by endothelial cells. Additionally, the shear-induced increase in eNOS phosphorylation was lost following enzymatic degradation of HA (Fig. 7b), which is consistent with its role in mediating shear-induced changes in membrane tension that activate PIEZO1.
Given that the spectrin network regulates shear-induced [Ca 2+ ] cyto increases, the stability of HA and the stability of caveolin-1 and caveolae, it was probable that NO production might also be regulated by spectrin. As shown in Fig. 7c, unlike wild-type cells, SPTBN1 KO cells did not show significant phosphorylation changes following exposure to shear stress.
To examine whether impaired phosphorylation of eNOS translated into a functionally significant reduction in NO production, we examined the interaction between platelets and endothelial cells under flow. Very few platelets adhered to wild-type endothelial cells, which is consistent with the ability of NO to inhibit the attachment of platelets to the vessel wall. However, the number of platelets that adhered firmly to SPTBN1 KO endothelial monolayers markedly increased (Fig. 7d). Surface expression of platelet adhesion receptors such as von Willebrand Factor or P-selectin was similar in wild-type and KO cells (Extended Data Fig. 7). Thus, spectrin contributes to the anti-thrombotic properties of the endothelium through its effects on NO production.
Previous studies have revealed a relationship between decreased expression of spectrins and atherosclerotic plaque severity 58 . Increased activity of calpain, which cleaves spectrin and consequently network disassembly, is also associated with early stages of atherosclerosis 59 . However, it remains unclear whether these changes in spectrin expression were reflective of the endothelium or other cells in the artery. We examined endothelial expression of β-II spectrin in en face preparations of descending thoracic aortas. We immunostained aortas of wild-type and Ldlr −/− mice that had been fed a high-fat diet for 4 weeks to promote atheroma formation. Consistent with previous observations 2,60 , endothelial cell alignment was lost in diseased descending aortas. Notably, the aortic endothelium of wild-type mice expressed high levels of β-II spectrin (Fig. 7e,f). However, in Ldlr −/− aortas in which plaques had formed, loss of alignment was coincident with a significantly lowered expression of β-II spectrin (Fig. 7e,f).

Discussion
Although it has long been appreciated that the endothelium plays a central role in the regulation of vascular responses to shear stress, a comprehensive understanding of how these cells sense and transduce changes in blood flow has remained elusive. Here, we revisited the contribution of multiple proposed mechanotransducers to the shear-sensing functions of the endothelium and investigated mechanisms that coordinate their activities. We focused specifically on the functions of HA and PIEZO1, which are indispensable for shear-induced responses of the endothelium.
It has been proposed that the cytoskeleton provides an integrative mechanism to distribute forces from mechanosensors to mechanoresponsive molecular complexes in cells 1 . The majority of studies have focused on the actin cytoskeleton because it plays a fundamental role in regulating the activity of integrins and junctional protein complexes often associated with mechanotransduction 2 . However, actin filaments cannot withstand large strains and are generally highly dynamic 61 . This suggests that actin-rich structures alone are not well suited to provide a framework for mechanosensitive cells.
We propose that the spectrin cytoskeleton, which contains only short, relatively stable filaments of actin, is a fundamental integrator of mechanical signalling on the apical surface of the endothelium. Spectrins, which are widely expressed and highly conserved in metazoan tissues, stabilize cell membranes and modulate the activity of ion channels 21 . Although the underlying mechanism has remained obscure, spectrins seem particularly important for vascular function. For example, global deletion of α-II spectrin or β-II spectrin, the dominant isoforms expressed in non-erythroid cells, is embryonic lethal in mice 62 and zebrafish 63 . Moreover, deletion of β-II spectrin impairs cardiovascular development 62 .
Spectrins are extremely flexible and associate with short actin filaments that become highly stable, and the resulting spectrinactin lattice is particularly pliable and long-lived 64 . Such flexibility confers increased compliance and deformability to the membrane and is ideal for cells exposed to the dynamic environment of the vasculature. Indeed, the spectrin cytoskeleton is required to maintain the integrity of erythrocytes, which experience haemodynamic stresses 19,37 and must squeeze through narrow capillaries [65][66][67] . Endothelial cells are similarly subjected to haemodynamic stress, especially in the aorta, and in narrow capillaries, the endothelium is deformed by passing erythrocytes. Here, we found that shear stress deforms endothelial spectrin (Fig. 4b,c), a feature that probably plays a central role in blood vessel homeostasis.
The vasculature is not the only environment where cells regularly experience shear and fluctuations in pressure. Cells in the lung are subjected to cyclic fluctuations in pressure. Compressions that arise from load-bearing activities also result in fluid shear, both in the synovium of joints and in the lacunae of bones. Whether these forces deform spectrin remains to be studied.
Spectrin is responsible for the maintenance of apical caveolae in the endothelium (Fig. 6h,i). As an important membrane reserve, caveolae contribute to the deformability of cells 48 and were proposed to be involved in shear-sensing 47 . However, it has been unclear whether shear causes caveolae to flatten or whether they serve primarily as organizers of mechanosensitive molecular complexes. A previous study reported that osmotically induced swelling of endothelial cells results in the flattening and remodelling of their caveolae 48 . A similar hypo-osmotic stress resulted in a roughly twofold increase in the fluorescence lifetime of Flipper-TR (Fig. 4d). This increase was comparable to that induced by application of shear stress (Fig. 4e). The similarity of the changes in membrane tension support the hypothesis that (a fraction of) caveolae flatten in response to shear. Accordingly, experiments investigating colocalization between cavin-2 and caveolin-1 provided evidence that shear stress acutely decreases the integrity of curved caveolar domains (Fig. 6j).
Our data establish a correlation between caveolar stability and retention and the activity of PIEZO1. Localization of PIEZO1 to curved microdomains is largely credited for its mechanosensitivity 68 , and resting membrane tension tunes channel sensitivity 69 . We therefore posit that spectrin regulates the sensitivity and activation of PIEZO1 in two ways: by influencing the density of caveolae and by maintaining a comparatively low apical membrane tension.
Our observations provide a tentative unifying framework to explain how responses to shear are integrated and communicated to intracellular effectors. We consider that components of the glycocalyx, notably HA, are the primary sensors of fluid shear and that CD44 is essential to convey the information to the underpinning spectrin meshwork (Fig. 4b,c). The latter is also linked to caveolae, dictating their stability and distribution. Displacement of the spectrin network could readily alter the curvature or strain applied to caveolae and hence activate PIEZO1. The resulting calcium influx, perhaps along with mechanical distortion of the caveolae, stimulates the release and activation of eNOS. In this model spectrin plays a dual role: (1) it regulates the stability of glycocalyx components (that most directly experience fluid flow) and senses and distributes mechanical disturbances experienced by the glycocalyx; and (2) it modulates plasma membrane tension by stabilizing curved membrane microdomains, such as caveolae, to influence the activation of mechanosensitive ion channels.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41556-022-00953-5.
Mice. Wild-type (C57BL/6J) and Ldlr −/− mice were bred and housed under pathogen-free conditions at The Hospital for Sick Children animal facility. In vivo experiment protocols (Animal User Protocol 1000043631) were reviewed and approved by the Animal Care Committee of The Hospital for Sick Children in accordance with guidelines established by the Canadian Council on Animal Care and federal and provincial legislation.
HAase from bovine testes (Sigma, H3506) was used at 20 units ml -1 for 10-30 min at 37 °C in DMEM. The Quantikine ELISA Hyaluronan Immunoassay (R&D Systems, DHYAL0) was used per the manufacturer's instructions. Spectrin-cpst-FRET was used as previously described 34 . Caveolin-1-RFP, caveolin-1-GFP and cavin-2-RFP have been previously described 70 . GCaMP6 was used as previously described 71 . Stock solutions of Flipper-TR (Spirochrome, SC020) were prepared in dimethylsulfoxide (DMSO) and used according to the manufacturer's instructions. Mouse high-fat diet D12108C (Clinton/Cybulsky high-fat rodent diet with regular casein and 1.25% added cholesterol) was used as supplied by Research Diets.
CRISPR-Cas9 and RNAi. SPTBN1 KO CRISPR cell lines were generated as follows: a predesigned vector containing a human SPTBN1-targeting guide RNA (gRNA) and eCas9-2A-tGFP was purchased from Millipore Sigma (MISSION gRNA ID: HSP00000399630). After 24 h of transfection of this plasmid, single GFP-positive cells were sorted into 96-well plates and expanded for 2-3 weeks to generate clonal populations. β-II spectrin deletion was confirmed in clones by immunoblotting. Stealth RNAi CD44 siRNAs (ThermoFisher, 1299001) and SMARTpool: ON-TARGETplus SPTBN1 siRNA (Dharmacon, L-018149-01-0005) were electroporated at 100 pM using a Neon transfection system. Shear stress. Endothelial cell monolayers were grown in glass-bottom microfluidic channels (Ibidi) coated with 50 µg ml -1 rat tail collagen (Sigma, C-7661). Inlets of the microfluidic channels were connected with tubing to syringes filled with complete medium in a syringe pump (Harvard Apparatus), while the outlet tubing was connected to a waste container. The syringe pump was used to apply a fluid flow rate of 3.9 ml min -1 , which yields an estimated shear rate of 15 dynes cm -2 (ref. 72 ).
Immunofluorescence. Endothelial cell monolayers were grown on coverslips or in microfluidic channels coated with 50 μg ml -1 rat tail collagen. Cells were fixed with 4% paraformaldehyde (PFA) (Electron Microscopy Sciences), permeabilized with 0.1% Triton X-100 in PBS, blocked with 2% BSA in PBS and stained with the indicated antibodies or fluorescent phalloidin in blocking solution. After staining, coverslips were mounted onto glass slides using ProLong Diamond (Life Technologies) or imaged directly in PBS.
Confocal microscopy and image analysis. Unless otherwise indicated, imaging was performed using a Quorum spinning disc system mounted on a Zeiss Axiovert 200M microscope, using ×10 air, ×25 water, ×63 oil or ×100 oil objectives, a ×1.5 magnification lens and a back-thinned EM-CCD camera (C9100-13, Hamamatsu). Acquisitions were controlled using Volocity software (Perkin-Elmer), exported and processed with Matlab (MathWorks) for filament alignment analysis or analysed and quantified using Volocity or Fiji/ImageJ (NIH) software.

Analysis of F-actin orientations.
Confocal images of F-actin were analysed using custom Matlab scripts in Matlab_R2020a. In brief, images were skeletonized using a Hessian-based multiscale filter and converted to a binary label matrix. Tubular structures, or 'filaments, ' were then segmented by fitting structural elements to the skeleton. Segmented areas smaller than 3 pixels were excluded. The maximum Feret diameter of each segmented filament was identified as the longest possible distance between two parallel tangents at opposing borders of the structural element. The angle between the maximum Feret diameter relative to the direction of applied flow was then calculated.
FRET. Cells expressing spectrin-cpst-FRET were imaged using a Leica TCS SP8 STED 3X microscope. Following excitation at 433 nm (the peak excitation wavelength for Cerulean), images were simultaneously acquired at Cerulean and Venus emissions using HyD detectors and a ×63/1.4 oil objective. FRET efficiency was quantified as the ratio of Venus/Cerulean mean fluorescence intensity.
FLIM. Cells were incubated with 1 µM Flipper-TR for 15 min before imaging. Frequency-domain FLIM measurements were collected using an Olympus IX81 inverted microscope equipped with a Lambert-FLIM attachment, using a ×60/1.49 NA oil immersion objective and a Li2CAM iCCD camera. The frequency of modulation was 400 MHz and the instrument was calibrated based on the fluorescence lifetime of AlexaFluor 546 assuming a monoexponential lifetime of 4.1 ns (ref. 73 ). Lifetime analysis was performed using FLIM software from Lambert Instruments.
Electronic cell size measurements. Wild-type and SPTBN1 KO cells were grown to confluent monolayers or as sparsely distributed cells. Cells were then suspended and cell diameters were measured using a Beckman coulter counter. These measurements (and corresponding standard deviations) were used to approximate the average cellular surface areas as follows: radius (r) = diameter/2; σ radius = σ diameter /2; surface area = 4πr 2 ; σ surface area = 8πr × σ radius .
TEM and analysis. TEM was performed according to standard methods. In brief, tHAECs were fixed in a mixture of 2% PFA, 2.5% glutaraldehyde and 0.1 M cacodylate buffer. Subsequently, the cells were post-fixed with 1% osmium tetroxide, dehydrated in a graded series of ethanol before embedding in Quetol-Spurr resin. Thin sagittal sections were cut and stained with uranyl acetate and lead citrate. Grids were examined using a JEOLJEM-1011 electron microscope at 80 kV with a side-mounted Advantage HR CCD camera (Advanced Microscopy Techniques).

Platelet isolation.
Whole blood was collected from three healthy donors (25-year-old male, 25-year-old female and 31-year-old male) using 0.6% acid-citrate-dextrose solution A (ACD-A) as the anticoagulant. Washed platelets were prepared using a light-spin/hard-spin method and 0.6% ACD-A/PBS solution. Blood sample collection was performed according to the guidelines of the Research Ethics Board at the Hospital for Sick Children. Donors were recruited as volunteers and received no remuneration. Platelet preparations were not mixed and were subsequently used for the n = 3 biological replicates of experiments investigating platelet adhesion to endothelial monolayers.
Aortic arch en face immunostaining. En face aortic samples were prepared as previously described 74 . In brief, following intracardiac puncture, mice were perfused through the left ventricle with ice-cold PBS for 5 min followed by 4% PFA in PBS for 5 min. Once collected, aortas were further fixed in 4% PFA for 24 h at 4 °C. Next, adipose tissue was dissected while immersed in cold PBS. Permeabilization was performed with 0.1% Triton-X in PBS for 30 min. Aortas were incubated overnight with primary antibodies at 4 °C. The aorta was opened in a reproducible manner by cutting along the greater curvature then mounted on glass slides with mounting medium (Dako Fluorescent, DakoCytomation).
Quantification and statistical analysis. The number of experiments and cells quantified are indicated in the individual figure legends. No statistical methods were used to predetermine sample sizes, but our sample sizes were similar to those reported in previous publications 18,75 . For statistical analyses, data distribution was assumed to be normal, but this was not formally tested. Statistical analysis was performed using R statistical software as indicated.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding author on reasonable request.

Code availability
The code generated during the current study are available from the corresponding author on reasonable request.