The Importance of Mechanical Forces for in vitro Endothelial Cell Biology

1Blood and lymphatic vessels are critical components of the vascular system, controlling the transport, delivery and recycling of nutrients and waste to all tissues in the body. The blood vascular system is comprised of a closed circulatory network of arteries, veins and capillaries. Arteries transport oxygenated blood with gases, nutrients, metabolites and immune cells to the organs, while veins return oxygen-poor blood to the heart. In contrast to the blood vascular system, lymphatic vessels are comprised of a blind-end, unidirectional vascular network of lymphatic collecting vessels and capillaries. Due to their specialized button-like cell junctions, lymphatic capillaries are able to take up fluid, macromolecules and immune cells. The lymph is then transported through collecting vessels that are equipped with zipper-like junctions and drained back into the venous circulation.

2As a result of their unique functions, each vessel sub-type is subjected to unique mechanical stresses. They are comprised of specialized subtypes of endothelial cells (ECs) with unique properties and genetic profiles, allowing them to perform their specific function. Not only does each vessel have unique ECs, the EC properties also differ across tissue beds. For example, blood vascular ECs are continuously aligned in most tissues, but fenestrated in tissues involved in filtration and secretion (kidney and intestinal mucosa) or discontinuous in sinusoidal vascular beds (liver and bone marrow) [reviewed in detail by ]. Lymphatic endothelial cells also display heterogeneity across tissue beds, with specialized Schlemm’s canal vessels found in the eye and meningeal lymphatics found in the brain [reviewed in detail by ].

3In addition to the heterogeneity of ECs, vessels are surrounded by a wide range of support structures with differing mechanical properties. They may be surrounded by supportive mural cells [such as pericytes and smooth muscle cells (SMCs)] and varying components of extracellular matrix, which is comprised of basement membrane (BM) and the interstitial matrix (occupying/filling the interstitial space). Large arteries and veins are characterized by a continuous lining of BM and layers of mural cells, whereas lymphatic collecting vessels only exhibit a thin BM layer and sparse SMC support. Lymphatic capillaries lack mural cell support and are characterized by a discontinuous or absent BM. These features allow each vessel subtype to maintain its integrity while performing its unique function.

4Much of the pioneering work characterizing EC structure and function was performed using cells grown in vitro. It is beginning to be appreciated that ECs within their in vivo environment differ greatly to those that are cultured in static two-dimensional or three-dimensional (2D/3D) settings. Indeed, the physical forces that ECs are subjected to in vivo, such as fluid shear stress, stiffness and stretch, greatly influence how these cells function. Here, we will briefly summarize what is known about the mechanical forces that regulate EC function and further discuss existing and emerging techniques to model these forces in vitro. By growing cells in mechanically relevant conditions, we can more accurately predict cellular behavior in physiological environments.

5Blood ECs lining the vessel wall are subjected to mechanical forces due to the friction between the ECs and the blood being pumped from the heart, known as fluid shear stress (FSS). The effects of FSS depends on a multitude of factors, such as the magnitude and direction of flow and the speed and strength of pump pulses. Shear stress can be steady laminar (where fluid moves in one direction at a steady magnitude), disturbed laminar (where flow is separated, recirculates and subsequently reattaches) oscillatory (where laminar flow fluid moves in a bidirectional manner) and turbulent (where flow is chaotic and moves in all directions). All FSS may be pulsatile, where magnitude of flow is varied. Both oscillatory and turbulent flow may also be classified as disturbed. These different stresses play a key role in regulating vascular physiology, as alterations to FSS parameters or changes in the EC response can lead to a range of vascular pathologies [reviewed in ]. Indeed, disturbed flow is known to have a wide range of effects on ECs, inducing changes in gene expression, cell shape and inflammatory profiles. This can lead to diseases such as atherosclerosis and pulmonary arterial hypertension. For a more detailed view on the effects of disturbed flow in the blood vasculature, please see the recent review by.

6As arteries transport blood away from the heart, they are subjected to high FSS (10–50 dyn/cm2) which is highly pulsatile, whereas veins are subjected to a more consistent flow with forces around 10-fold less. Unlike blood vessels which are controlled by a central pump, lymphatic capillaries are exposed to interstitial fluid flow, thus are exposed to the lowest FSS, approximately 10-fold less than blood vessels. The uptake of interstitial fluid and resulting forces are guided by hydrostatic and osmotic pressure between lymphatic vessels, blood vessels and the interstitial space [known as Starling forces]. Collecting lymphatic vessels have an intrinsic pumping capacity and smooth muscle cell coverage, thus are exposed to greater FSS than capillaries to ensure efficient transport of the lymph back to the venous circulation. Here, we will discuss how ECs sense and respond to these different forces to generate and maintain a functional vascular network.

7In order for ECs to respond to FSS, they require mechanisms to sense changes in flow conditions. This may be done via cilia, glycocalyx, ion channels, G-proteins or protein kinases. Perhaps the most well characterized mechanism of FSS sensing is through cell surface receptor ‘mechanosensory hubs’ found at cell–cell junctions [otherwise known as adherens junctions (AJs)]. Receptors within these AJ hubs include PECAM-1 (CD31), Vascular endothelial (VE)-cadherin, VEGFR2 and VEGFR3. Most central to this hub is PECAM-1, which is a direct sensor of shear stress tension. Increased tension on PECAM-1 via steady laminar FSS [12 dyn/cm2 for 16 hours (h)] triggers association of this cell surface receptor with the vimentin cytoskeleton, which permits transmission of force from myosin to PECAM-1. These tensile forces on PECAM-1 trigger the activation of Src family kinases (SFK), which in turn phosphorylate VEGF receptors in a ligand independent manner. Phosphorylated VEGFRs then activate downstream signaling cascades, including the PI(3)K and NFκB pathways. Within this complex, VE-cadherin acts to assemble VEGFRs at the junction through its transmembrane domain, facilitating their activation by SFK.

8While known to be a critical component of AJ mechanosensory hubs and a key controller of vascular integrity and sprouting, the precise role of VE-cadherin as a sensor of tension in response to shear stress remains undefined. In contrast to PECAM-1, when steady laminar FSS is applied to VE-cadherin (15 dyn/cm2), tension is reduced within 2 min. Interestingly, 2 min at 15 dyn/cm2 is sufficient to induce tension on PECAM-1. In vivo, VE-cadherin tension in vessels under high levels of FSS (arteries) is lower than that observed in low shear, immature vessels. However, VE-cadherin can be phosphorylated in response to PECAM-1-SFK on its Tyr658 site. This results in dissociation of p120-catenin and association with polarity protein LGN, which subsequently activates inflammatory pathways at sites of disturbed flow. Additional research is required to fully elucidate the mechanosensory role of VE-cadherin in response to FSS in physiological settings.

9In addition to junctions, mechanosensory hubs are found at sites of cell-matrix adhesion [otherwise known as focal adhesions (FAs)]. FAs are comprised of extracellular matrix (ECM) binding proteins known as integrins, which are coupled to the cytoskeleton through actin binding proteins, including the mechanosensors vinculin and talin. FAs are vital components of the EC mechanosensory complex, mediating a wide range of signaling cascades in response to traction forces and FSS. However, while FAs are remodeled in response to FSS and ECs display larger FA under steady laminar flow (17 dyn/cm2) than those under disturbed flow after 14 h, whether FA proteins themselves can physically sense changes in tension in response to FSS remains unclear.

10Within FAs, integrins play a central role in transducing chemical and physical signals from their surroundings by interacting and binding to their adhesion ligands presented by the BM. For a detailed overview of integrins and their ECM interactions, please see the following reviews. Notably, integrin can mediate bi-directional signaling. “Inside-out” signaling activates the ligand binding function of integrins (for example when FSS is sensed and subsequently transduced inside the EC) and “outside-in” signaling mediates cellular responses (for example in response to changes in ECM stiffness) leading to changes in cell spreading, retraction, migration, and proliferation. Although integrins do physically respond to FSS, albeit at a significantly lower degree compared to their response to traction forces, whether they confer biologically relevant tension changes in response to FSS remains unclear. However, they are activated downstream of PECAM-1 and PI(3)K activity, revealing they may indeed involved in a FSS mechanosensory complex. The Integrin-actin linker protein talin is a well-established traction force mechanosensor. Yet similar to integrins, whether it confers physical tension sensing capacity in response to FSS remains unclear. Talin activates integrins by binding to the β-tail and loss of talin results in impaired integrin activation. EC-specific knockout of either integrin-β1 itself or talin (leading to loss of integrin-β1 activation) leads to perturbed VE-cadherin localization and loss of junctional integrity. This reveals that FAs are intimately associated with AJs, potentially through their attachment to the cytoskeleton. However, it may also suggest that integrins and talin may not be FSS tension sensors, but are simply activated downstream of AJ mechanosensory hubs though the physical link with the cytoskeleton or by downstream signaling pathways. More work is required to elucidate the FSS tension sensing capacity of integrins and their associated actin-binding proteins.

11Recently, it was reported that guidance receptor PlexinD1 is a mechanosensor, and upon loss of PlexinD1 cells fail to align after 24 h steady laminar FSS (12 dyn/cm2). This occurs through a PlexinD1-Neuropilin-1 complex with VEGFR2, which acts as a mechanosensor upstream of the PECAM-1 mediated junctional complex and integrin-based focal adhesions. If PlexinD1 is held in its ring-like conformation it cannot act as a mechanosensor, suggesting it is not only the expression, but also the structure, of each mechanosensory component which confers its role.

12As blood and lymphatic vessels sprout, mature and remodel, they become subjected to different FSS. Therefore, EC mechanosensors need to be able to adapt to each vessel’s physiological requirements. During formation of the initial blood vascular plexus in vivo, angiogenic sprouting is more likely to occur at sites of low FSS and vascular fusion is induced by steady laminar low FSS and VE-cadherin phosphorylation (induced at 1.8 dyn/cm2 but not 10 dyn/cm2 for 4 h). Later in development during vascular remodeling, in vessels with low FSS (such as capillaries) ECs are less likely to align against the direction of flow than ECs in regions of high FSS (such as arteries). This prevents polarity-induced migration into regions of high flow and subsequent vessel regression. ECs deficient in canonical Wnt signaling display increased sensitivity to FSS-induced regression in vivo or after being subjected to steady laminar FSS in culture (20 dyn/cm2 for 4 h), suggesting that a FSS ‘setpoint’ controlling EC polarity and vascular stability is modulated by Wnt. These ‘setpoints’ define the optimal FSS exposure for normal vascular function, whereby if FSS is above or below the setpoint, vascular abnormalities occur. Interestingly, loss of Wnt signaling leads to reduced sprouting capacity, yet whether this is guided through altered sensitivity to a low FSS setpoint remains unclear.

13In addition to FSS setpoints being defined during EC polarity and remodeling, they must be specified across different vessel sub-types in order for each vessel to exert its biological function. As blood vessels are exposed to higher FSS in the body than lymphatic vessels, blood ECs become misaligned and activate NFκB at much higher steady laminar FSS levels (25 dyn/cm2 and over for 16 h) than that of lymphatic ECs (10 dyn/cm2 and over for 16 h). This allows blood vessels to be exposed to higher rates of FSS without causing inflammation and disease. This is a reflection of vessel physiology – lymphatics are exposed to significantly lower FSS in vivo as their function is to transport interstitial fluid back to venous circulation. The FSS setpoint in blood EC versus lymphatic EC is mediated though expression of VEGFR3, which is enriched in lymphatic ECs compared to blood ECs. The elevated expression of VEGFR3 in lymphatic EC causes them to respond to FSS at lower levels to that of blood ECs, and reducing VEGFR3 expression in lymphatic ECs increases the FSS setpoint. Conversely, increasing VEGFR3 expression in blood EC decreases the FSS setpoint, causing induction of inflammatory profiles at FSS levels seen in lymphatic EC.

14Although FSS is lower in lymphatic compared to blood vessels, there are still physiologically relevant fluctuations in flow that drive genetic and morphological changes in lymphatics. During early development, functional lymphatic drainage can be detected as early as embryonic day (E) 11.5 suggesting that induction of FSS might contribute to the development of the primary lymphatic plexus by inducing lymphatic EC proliferation. Indeed, steady laminar FSS has been shown to induce proliferation via ORAI1 and KLF2/4-mediated upregulation of VEGF-A, VEGF-C, FGFR3 and p57 (2 dyn/cm2 for 12–24 h) and downregulation of NOTCH1 activity (2 dyn/cm2 for 12–24 h). Lymphatic valve development is predominantly initiated in areas of disturbed flow, and oscillatory flow (4 dyn/cm2 for 48 h) has been shown to reproduce a lymphatic valve EC phenotype in vitro. Oscillatory flow but not steady laminar flow (4 dyn/cm2 for 48 h) induces expression of the transcription factor GATA2 in valve-forming lymphatic ECs, which in turn upregulates expression of the lymphatic endothelial transcription factors PROX1 and FOXC2, both of which are key regulators of valve formation. In agreement, loss of lymphatic flow in a mouse model of CLEC2-deficiency (in which blood backfills the lymphatic network and thereby blocks lymph flow) results in a phenotype identical to that observed upon loss of FOXC2. In the postnatal lymphatic vasculature, FOXC2 is necessary to maintain collecting vessel quiescence and stability in valve areas by linking oscillatory flow responses to cell junction stabilization and cell-cycle arrest, revealing flow dynamics as key regulators of the formation and maintenance of a functional lymphatic system.

15Interestingly, FSS can induce changes in membrane curvature, leading to activation of a range of cell surface receptors including calcium and ion channels. The ion channel Piezo1, which is sensitive to changes in membrane tension, is also required for FSS sensing, thereby coupling FSS and membrane stretch. Piezo1 regulates vascular permeability in the lungs by inducing VE-cadherin internalization, and subsequent leakage, in response to increased hydrostatic pressure. Piezo1 is also mechanically activated in lymphatic vessels and is required for lymphatic valve formation by sensing disturbed flow to induce GATA2 and FOXC2 expression. Although not as widely studied, ion channels such as TRPV4 have also been implicated in FSS sensing to regulate vascular permeability. Not only is membrane curvature altered, but membrane fluidity changes upon the onset of shear stress, which may directly induce the difference in tension that opens up channels such as Piezo1. However, these changes occur in biologically inactive lipid membranes, thus may be a basic physical phenomena that can transduce shear stress.

16In addition to Piezo1, caveolae (invaginations on the cell membrane) are sensitive to changes in membrane tension and are known to be altered in response to FSS. ECs flatten their caveolae in response to increased osmotic pressure to allow for membrane ‘stretching,’ which may be a survival mechanism by increasing cell surface area. Interestingly, a recent study found caveolae are abundant in arterial ECs of the central nervous system and are responsible for transmitting signals from the neurovascular unit to adjacent SMC to induce vasodilation. This was proposed to be through the ability of caveolae to cluster TRPV4 ion channels rather than mediating changes in arterial membrane stretch. Precisely how physical stretch by FSS alters membrane components in ECs remains largely unknown.

17Extracellular matrix stiffness has been shown to guide the behavior of ECs to form the initial functional vascular network. Embryonic stem cells (ESCs) depend on the mechanical environment of the stem cell niche, as the fate of vascular progenitor cells (VPC) was shown to be dependent on substrate stiffness, with EC lineages favoring softer substrates (10 kPa) and SMC lineages favoring stiffer substrates (plastic, GPa range). During early lymphatic development, venous lymphatic EC progenitors laminate from the cardinal vein and encounter very soft surrounding tissue (0.27 kPa) while they migrate dorsally to form primary lymphatic structures. This value matches those measured in early chicken embryos (0.3 kPa) and adult brain (0.33 kPa) but are significantly lower compared to values from most other adult tissues, such as muscle (around 12 kPa) or bone (GPa range). Upon remodeling and maturation, ECs form the inner layer of blood and lymphatic vessels and are attached to the underlying 2D ECM environment. The composition and mechanical properties of the ECM differ along vascular trees, across tissues and during development and disease states.

18The range of venous tissue stiffness in mammals is about 3–50 kPa, which is positioned between the stiffness of the epithelium and cartilage. Unlike veins, arteries must be able to withstand higher blood pressure and are surrounded by several layers of smooth muscle cells and connective tissue. Therefore, arteries are stiffer, ranging from about 50–150 kPa. Differences in ECM stiffness experienced by endothelial progenitor cells (EPCs) have been implicated in the regulation of arterial-venous differentiation in vitro. In contrast to EPCs cultured on venous substrate stiffness (7 kPa), EPCs cultured on arterial substrate stiffness (128 kPa) showed an increase in expression of the arterial marker EphrinB2.

19The stiffness values experienced by the endothelium must be interpreted with caution as substantially different measurement techniques have been employed to generate values [reviewed by and ]. Macroscale techniques, such as Shear Wave Elastography (SWE) generally analyze the entire vessel structure as a uniform material, and do not take into account cell-scale differences of the individual vessel components, such as the endothelial or SMC layer or the surrounding ECM. Furthermore, macroscale techniques typically provide values of the stiffest component of a vessel structure. Therefore, these data underscore the importance of analyzing mechanical properties of individual vessel components using microscale techniques, such as Atomic Force Microscopy (AFM) and Scanning Ion Conductance Microscopy (SICM). It is important to attempt comprehensive measurements of in vivo ECM stiffness experienced by ECs to accurately choose the stiffness values for studies on EC-type or organ-specific vasculature.

20Adherent cells sense ECM stiffness by probing their extracellular environment, resulting in substrate displacements when substrates are compliant. The measurement of compliant ECM substrate deformations (and calculation of deformation-based metrics) via Traction Force Microscopy (TFM) or 4D displacement microscopy can provide quantitative information on cell-ECM mechanical interactions. For example, by employing 4D displacement microscopy, it has been demonstrated that ECM substrate deformations are spatio-temporally correlated with sprout-morphological dynamics during sprouting angiogenesis. Interestingly, protrusions from extending sprouts increase their pulling forces, while retracting protrusions reduce pulling.

21In vitro ECM substrate deformation analysis can provide insight into how ECs exert forces into their surroundings. However, which ECM components or EC-surrounding cells are responsible to generate the absolute in vivo ECM stiffness that counteracts those deformation forces is poorly defined. The BM underlying ECs in most mature vessel types is a very thin layer (30–500 nm) and is comprised of ECM matrix molecules such as elastin, collagen, enactin/nidogen, heparan-sulfate proteoglycans, and laminin. Interestingly, a study by showed that cells can “feel” up to several micrometers deep into a compliant substrate. Consistently, induction of EC network formation on compliant substrate (0.4 kPa) was prevented on very thin compliant substrates (<20 μm) as ECs sensed the stiffness of the underlying coverslip. These findings suggest that absolute in vivo ECM stiffness experienced by ECs might rather be generated by several different EC-surrounding structures, including the BM, the interstitial matrix and, importantly, the adjacent cells.

22Similar to their FSS mechanotransduction capacity, FAs are fundamental mechanotransducers of ECM-endothelial stiffness. FAs propagate ECM stiffness stimuli to the actin cytoskeleton, resulting in cytoskeletal rearrangements that can lead to an immediate adaptation of cells to their physiological environment, such as alignment or translocalization of proteins to the nucleus in order to modify the genetic program and thereby cellular behavior [reviewed in detail in ]. Within FAs, the expression of integrins is highly cell type- and substrate-specific, and this integrin diversity can regulate intracellular signaling cascades in response to different mechanical stimuli. In epithelial cells, an increase in matrix stiffness (0.2 kPa vs. GPa range) regulates the expression of β1 integrin and Cav1 to regulate FA assembly and turnover. Similarly, ECs matrix stiffening (3 kPa vs. 70 kPa) has been shown to induce β1 integrin activity. Strong changes in expression of β1 integrin were not observed in this study, however, analysis of ECs on softer matrices were not included.

23Intracellularly, FAs are dynamically associated with a variety of proteins that can sense and/or transduce force, such as focal adhesion kinase (FAK), vinculin and talin. FAK is one of the first proteins recruited to FAs in response to extracellular mechanical stimuli. In fibroblasts, autophosphorylation of FAK is required for induction of adhesion and integrin-binding responses. Furthermore, FAK activity is necessary to regulate directed migration of fibroblasts toward stiffer substrates and to induce matrix stiffness-mediated translocation of the mechanotransducer Yes-associated protein (YAP) to the nucleus of hepatic stellate cells. The mechanosensor talin is a 270 kDa protein composed of an N-terminal globular head, a flexible rod domain and C-terminal helices. It binds to the cytoplasmic domain of integrin β-subunits and regulates integrin signaling in a switch-like behavior via the helix bundles of its flexible rod domains. It has been suggested that individual domains open at different tension levels, exerting positive or negative effects on different protein interactions. For example, force loading of talin leads to the exposure of cryptic hydrophobic binding sites, which can subsequently bind vinculin. As increasing force is applied to talin, more bundles are unfolded, revealing additional vinculin binding sites and thus activation of an increasing number of vinculin molecules. These FA molecules have been extensively studied as general tension sensors/transducers in non-endothelial cell types. It will be important to study the expression and function of those proteins specifically in the context of matrix stiffness-regulated EC behavior.

24After they have been engaged downstream of integrins, FAs propagate mechanical stimuli to the actin cytoskeleton. Culturing ECs on different stiffness matrices leads to a substantial remodeling of the actin cytoskeleton. With increasing matrix stiffness (3 kPa, 12 kPa, and 1.5 MPa), actin cytoskeleton remodeling becomes more organized, with an increasing amount of actin stress fibers. Interestingly, increasing matrix stiffness translates to increased EC stiffness in 2D (1.7 kPa vs. 9 kPa) and 3D (0.125 kPa vs. 0.5 kPa) environments. The actin cytoskeleton is connected to VE-cadherin via its intracellularly associated proteins β-catenin and α-catenin. Tension-induced actin remodeling tightly controls assembly and disassembly of VE-cadherin-based junctions. Cytoskeletal pulling at the VE-cadherin complex also recruits the FA tension sensor protein vinculin via α-catenin to reinforce endothelial junctions.

25Besides a direct effect on endothelial junction stability, cytoskeletal mechanotransduction can result in structural modification of membrane-bound or cytoplasmic proteins and their subsequent shuttling to the nucleus. β-catenin is not only a structural protein within the VE-cadherin adhesion complex, but can also shuttle to the nucleus upon cytoskeletal remodeling to act as a transcriptional co-activator. In chondrocytes, β-catenin expression is induced and localized to the nucleus in a β1-integrin/FAK dependent manner in cells seeded on stiff matrices (100 kPa vs. 0.5–1 kPa). Furthermore, β-catenin was shown to localize to the nucleus in valvular ECs grown on stiff matrices (50 kPa vs. 5 kPa). Another class of nuclear shuttling proteins consists of YAP and WW Domain-Containing Transcription Regulator Protein 1 (WWTR1/TAZ), which are downstream effectors of the Hippo pathway. YAP/TAZ are shuttled to the nucleus in lymphatic ECs grown on stiff substrates (25 kPa vs. 0.2 kPa) as evident by the induction of their target genes CTGF and ANKRD1. YAP/TAZ function has been extensively studied in the development of the vasculature. Precisely how changes in ECM stiffness experienced by ECs might regulate those processes remains to be elucidated.

26During angiogenesis, VEGFR2 activation and internalization induces proliferation of ECs to enable nascent blood vessels and tumor vessels to expand. In blood and lymphatic ECs, proliferation is decreased on soft 2D substrates and enhanced on stiff 2D substrates. Sub-confluent HUVECs grown on stiffer substrates (10 kPa vs. 1 kPa) increase VEGFR2 internalization, mediated via Rho activity and actin contractility. Additionally, expression of Septin9, a negative upstream effector of RhoA, is increased in ECs grown on soft 2D substrates (1.72 kPa vs. 21.5 kPa), attenuates Src/Vav2 phosphorylation and inhibits RhoA-dependent EC proliferation. However, in a confluent endothelial monolayer, stiffness-enhanced VEGF signaling is no longer observed, suggesting this mechanism is unique to actively proliferating cells and angiogenic processes.

27In contrast, maximal VEGFR2 expression and capillary blood vessel formation in an in vivo 3D Matrigel implant assay was detected in a relatively soft microenvironment of 0.8 kPa. Similarly, lymphatic ECs leave the stiffer cardinal vein (4 kPa) to migrate through a very soft 3D environment (0.27 kPa). The decrease in matrix stiffness experienced by those cells induces an increase in VEGFR3 expression that is necessary to allow efficient migration. As has been suggested by, a difference in migratory and sprouting response to matrix stiffness could be explained by differential EC requirements on a 2D ECM versus a 3D ECM environment. This hypothesis is further supported by the finding that sprouting of blood ECs is increased in synthetic 3D hydrogels with lower matrix crosslinking, which is accompanied by a decrease in matrix stiffness (1 kPa vs. 6 kPa).

28Furthermore, it is important to note that in vitro studies have been performed using different definitions of a stiff substrate. Comparing the absolute kPa values reveals that ‘stiff’ ECM values (such as 4 kPa) are used as the ‘soft’ ECM standards in other studies. This underscores the necessity to comprehensively measure in vivo stiffness of blood and lymphatic vasculature to better define the ranges that can be adapted for in vitro studies.

29More than 80% of all cardiovascular deaths in developed countries occurs in the population aged 65 or older, making aging a major risk factor for cardiovascular disease. It is well established that age-induced stiffening of the ECM in the intima alters the function of blood vessels. These alterations of ECM stiffness affect EC health, resulting in inflammation, hypertension and ultimately disease progression in disorders such as atherosclerosis and pulmonary arterial hypertension (PAH). Indeed, increased vascular ECM stiffness is induced upon aging and/or inflammation, as revealed by rigidity measurements using murine aortas of different ages and murine and human atherosclerotic plaques. Vascular calcification is often observed in human atherosclerotic plaques and is known to induce EC dysfunction, which is one of the main characteristics of arterial aging. This calcification seems to be irreversible, with a number of studies attempting to inhibit or reverse ECM stiffening being unsuccessful. Hence, current research has shifted toward investigating the translational effects of ECM stiffness on ECs, to explore the possibilities of reversing the EC response to increased ECM stiffness, instead of directly targeting ECM stiffening.

30In the context of cancer biology, it is well appreciated that a solid tumor environment is stiffer than that of healthy tissue. How tumors overcome the mechanical barrier of a stiffer environment to promote tumor angiogenesis is likely through matrix metalloproteinase (MMP) signaling, which was found to be induced in ECs cultured in stiffer 3D collagen gels. Endothelial dysregulation results in vascular leakage due to a loss of cell junction integrity. In tumors, this dysregulation has been largely attributed to an increase in pro-angiogenic growth factors. However, tumor ECM stiffness has been shown to also play a central role in promoting a characteristically leakier tumor vasculature. Sub-confluent ECs on compliant substrates assemble spontaneously into networks reminiscent of an angiogenic process. Therefore, sub-confluent ECs grown on soft substrates (0.2 kPa) form continuous VE-cadherin-positive junctions, which are highly discontinuous and punctuated compared to those grown on stiff, tumorigenic substrates (10 kPa).

31Mechanistically, increased matrix stiffness activates FAK, resulting in Src localization to cell–cell junctions and subsequent Src-mediated VE-cadherin phosphorylation. Interestingly, loss of VE-cadherin function activates intercellular force transduction signals that increase integrin-dependent cell contractility, disrupting both cell–cell and cell-matrix adhesions. Increased substrate stiffness (40 kPa vs. 1.1 kPa) further exacerbates these effects. Tumor matrix stiffening not only impacts angiogenesis and tumor vessel leakage, but is also implicated in regulating metastasis. Matrix stiffening from 0.4 kPa to 22 kPa induced matricellular protein CCN1, which in turn upregulates N-cadherin levels on the EC surface to facilitate cancer cell binding to the endothelium and intravasation into the vessel lumen. These data suggest that identifying and targeting EC behavior in response to altered stiffness in solid tumors could provide novel therapeutic approaches to normalize tumor vasculature.

32It has become clear that changes in ECM stiffness can directly regulate EC internal stiffness and/or expression of endothelial adhesion molecules, which is correlated with altered leukocyte extravasation. Whereas differences in ECM stiffness directs EC migration during angiogenesis, in the case of leukocyte trans-endothelial migration (TEM), it is the internal EC stiffness that guides the direction of leukocyte migration. The internal stiffness of ECs is regulated by crosslinking of ICAM-1 to the actin cytoskeleton via actin binding proteins (ABPs) α-actinin and cortactin. DLC-1, a Rho GTPase Activating Protein induced by ECM stiffness, is able to stabilize this crosslinking. Loss of DLC-1 in ECs cultured on stiff substrates results in leukocyte transmigration kinetics similar to those observed for ECs cultured on soft substrates, revealing DLC-1 as a critical transmitter of ECM-induced EC stiffness required for TEM. These changes in TEM and EC internal stiffness caused by the ECM are observed in inflammatory diseases, as in areas of increased ECM stiffening (such as atherosclerotic lesions or in the lung microvasculature of PAH patients) increased leukocyte extravasation is observed.

33Besides the effects of ECM stiffness on leukocyte TEM via changes in EC internal stiffness, it has been shown that ECM stiffness regulates leukocyte and monocyte transmigration via microRNAs. Expression of adhesion molecules ICAM-1 and VCAM-1 on ECs, both essential for leukocyte and monocyte adhesion, is regulated by miR-222 and miR-126, respectively. ECs cultured on high stiffness substrates (20 kPa) showed reduced migration of monocytes, which correlated with higher levels of miR-222 and miR-126 to inhibit endothelial ICAM-1 and VCAM-1 expression compared to soft (8 kPa) and diseased (40 kPa) stiffness values. This suggest that matrix stiffness exerts a biphasic regulation of endothelial adhesion molecules ICAM-1 and VCAM-1, which promotes EC-monocyte interactions both in healthy (soft) and diseased (stiff) environments.

34While much focus has been placed on the role of FSS sensing and cell-ECM interactions, the role of cyclic stretch (CS) on endothelial cell function is far less studied. This is despite its intricate link to matrix stiffness, as a stiffer matrix will in principle result in a reduction of cyclic stretch. ECs are constantly exposed to cyclical stretch due to dilation across the vessel wall as a result of blood pumping, respiration in the lungs or uptake of interstitial fluid in lymphatics. This stretch of the EC membrane results in activation of signaling pathways and physiological responses, such as sprouting, permeability and remodeling.

35To date, much of our knowledge on how vascular ECs respond to stretch comes from cultured cells. In blood ECs, normal stretch is 5% elongation and pathological stress is 18% elongation, with acute stretch being 15 min and stretch being chronic 48 h. Stretch has been associated with a range of pathways, many of which result in loss of barrier integrity. Even after short term exposure to CS, ECs become significantly more prone to leakage in response to thrombin or VEGF, linked to an increase in Rho GTPase activity and MLC phosphorylation. However, changes in barrier integrity are not simply linked to changes in mechanical imbalance, as after cells have been re-plated to non-elastic substrates, they continue to be susceptible to leakage. Interestingly, pre-conditioning of ECs to CS provides protection against barrier breakdown, which is thought to be a result of induction of Rap1 signaling to promote junctional stability and suppress RhoA signaling. Interestingly, cyclic stretch can also induce EC proliferation in response to mechanical stimulation, in a process that requires cell–cell contacts. Indeed, stretch induced EC proliferation is lost upon VE-cadherin inhibition or knockdown and dependent on the Hippo pathway downstream effector YAP.

36Cardiovascular diseases such as hypertension and atherosclerosis are associated with CS, as an increase in blood pressure or changes in the ECM environment can affect the amount of stretch. Pathological but not physiological CS results in activation of VEGFR2 and SFKs, and subsequent phosphorylation of VE-cadherin and disassembly of junctions. Pathological and chronic exposure to CS also induces inflammation by upregulation of ICAM1 and secretion of IL-8. Therefore, increased CS can cause both pathological barrier breakdown and induce inflammation. Pulsatile stretch above 10% in brain ECs induces inflammation through ICAM1 and NO and induces amyloid precursor protein expression, which may explain why the severity of Alzheimer’s disease is increased with a higher pulsatility index and pressure. Interestingly, multi-directional stretch, analogous to the forces exerted on ECs in oscillatory FSS, results in upregulation of inflammatory stimulators such as NF-κB, Both uniaxial and equiaxial stretch (10%) have been shown to drive induction of smooth muscle genes, an effect which is abrogated upon induction of FSS (uniform at 5 dyn/cm2 or cyclic between 0 and 5 dyn/cm2), suggesting stretch alone may induce transdifferentiation of ECs toward SMCs. Thus, multidirectional stretch may be directly linked to cellular processes which drive diseases such as fibrosis and atherosclerosis.

37Lymphatic vessels are exposed to stretch as a result of their fluid uptake and pumping properties. Capillaries take up interstitial fluid via an extrinsic (passive) mechanism, where the lymph enters the lymphatic capillaries against a hydrostatic pressure gradient. It has been elegantly demonstrated that during early lymphatic development, fluid that leaks out of newly formed arteries swells the tissue that surrounds the first primary lymphatic structures and stretches these nascent structures. Stretching of the lymphatic ECs leads to activation of VEGFR3 and subsequent proliferation, resulting in enlargement of lymphatic structures so they are able to drain the excessive fluid from the embryonic tissue.

38Collecting lymphatic vessels pump lymph via cyclic compression and expansion of lymphatic SMCs that surrounds the lymphatic collecting ECs (lymphatic muscle). Increased intraluminal lymph pressure after fluid uptake leads to stretching of the collecting lymphatics and activates an intrinsic (active) lymph pump. The impact of stretch on lymphatic EC behavior exerted by SMC surrounding the collecting vessels has not been extensively addressed. In contrast to blood ECs, normal lymphatic ECs stretch is 4% elongation and pathological stretch is 8% elongation, suggesting that similar to FSS, lymphatic ECs have a lower stretch setpoint.

39To briefly summarize, blood and lymphatic vessels (and their vessel sub-types such as arteries, veins, capillaries and collectors) all display different sensitivities to mechanical stimuli, such as fluid shear stress, stiffness or stretch. These mechanical forces must be considered when performing in vitro experiments in the context of the vascular process to be studied, in order to gain biologically relevant insights into EC physiology.

40With an appreciation of the importance of the extracellular environment, the vascular field has begun to develop in vitro approaches that more closely mimic in vivo environments. Bioengineering approaches have been developed in vitro to recapitulate in vivo forces, providing more cost effective, biologically relevant experimentation. Here, we will review commonly used cell culture techniques to model forces of fluid shear stress, matrix stiffness and cell stretch, and discuss emerging technologies to combine these mechanical forces.

41There are numerous models which have been used in the vascular biology field to identify defined FSS-mediated EC outputs. Traditionally, these have been performed by seeding cells on a 2D flat and stiff substrate, allowing them to adhere and form a monolayer, then passing fluid over them (Figure 1). This can be achieved using a cone and plate system, where flow is generated by rotation of a cone around a central axis which is oriented in a perpendicular manner to the surface of a flat plate. More commonly, however, these experiments are performed using a parallel plate chamber, where fluid is pumped using a peristaltic pump, constant pressure head or syringe pump. These models have been particularly useful when modeling how ECs respond to changes in shear stress force and duration, where straight channels model laminar FSS (Figure 1B) and bifurcated channels model gradient shear stress regions (Figure 1C). Fluid can be pumped in a steady unidirectional manner or in an oscillatory manner, and in both of these cases it may be pulsatile. Examples of knowledge that has come from these models include when exposed to low levels of steady laminar FSS (10 dyn/cm2 for 5 h), EC junctions become stabilized and when exposed to disturbed shear stress, ECs acquire an inflammatory phenotype typically associated with atherosclerosis. Similar experiments have also helped to identify the mechanical contribution to formation of the lymphatic valves. When exposed to oscillatory but not steady state laminar FSS (4 dyn/cm2 or cultured on a rocking platform for 48 h), lymphatic ECs acquire features of lymphatic valves. In line with these findings, lymphatic valves form at sites of disturbed flow in vivo.

42Models of interstitial flow within lymphatic capillaries have been pioneered by the Swartz lab and are largely performed using Boyden chamber assays (Figure 1D) (i.e., transwell filter systems) or radial flow systems. Boyden chamber assays are performed by plating cells as a monolayer on a semi-porous ECM, such as collagen or fibrin, and seeded in the upper compartment of a transwell chamber. Then, a fixed hydrostatic pressure difference is maintained between the top and bottom chambers. Alternatively, radial flow chambers consist of ECM with ECs embedded and exposed to flow oriented radially, which is then placed between two glass coverslips and fixed with porous boundaries. Boyden chamber assays are also widely utilized amongst vascular biologists to study vascular permeability and chemotaxis of ECs and leukocytes.

432D FSS assays have been utilized for much of the pioneering work in understanding the mechanisms of how ECs sense flow-mediated mechanotransduction, and continue to provide valuable insights. However, these standard commercial assays are limited in that they do not fully recapitulate vessel physiology, which is significantly more complex. Of relevance to this review, these assays do not incorporate important components that can alter hydrodynamics such as mural cells, ECM stiffness, or circular stretch.

44More complex microfluidic systems to model FSS involve growing ECs in 3D microchannels using microfabrication techniques. These can be entirely poly(dimethylsiloxane) (PDMS) derived, PDMS-hydrogel hybrids, or pure hydrogel. Similar to 2D FSS models, flow can be controlled using hydrostatic pumps. PDMS-alone devices are typically linear rectangular channels that are optically transparent and permeable to gases, placed with an inlet-channel-outlet flow system (Figure 1E). These channels provide a suitable model to study dynamics of red blood cells or thrombosis. However, PDMS does not resemble native tissue environments as it is rigid, impermeable to water and not able to be remodeled by ECs.

45PDMS-hydrogel hybrids consist of a channel of hydrogel scaffold (typically collagen or fibrin) lined on both sides by PDMS channels lined with ECs (Figure 1F). These ECs then form sprouts and migrate across the hydrogel scaffolds in response to changes in pressure or growth factor gradients, forming a perfused capillary plexus with a functional vascular barrier. ECs lining the PDMS channel can also be co-cultured with mural, stromal and cancer cells, and may be derived from either normal blood, tumor or lymphatic vessels. These hybrid models have provided extensive information about how capillaries sprout, revealing that angiogenesis and lymphangiogenesis occur in response to interstitial flow, as sprouts in this model preferentially extend against the direction of flow; i.e., capillaries sprout from vessels with a low pressure gradient toward those with higher pressure. Precisely how FSS applied to the PDMS channel mediates sprouting remains under debate. It has been shown steady laminar FSS (3 dyn/cm2 for 1–3 days) inhibits VEGF-induced sprouting, whereas increasing the magnitude of FSS (to 10 dyn/cm2 for 24–48 h) can induce angiogenesis. This suggests a FSS setpoint may regulate the degree of sprouting.

46Growing microvessels in pure hydrogels is perhaps the most physiological of all FSS models, as ECs form lumenised vessels surrounded by ECM. In this model, typically a PDMS mold is produced with two media ports at each end and an ECM port in the center (Figure 1G). The hydrogel scaffold is allowed to set within this 3D ECM scaffold around a cylindrical needle, before the needle is removed leaving a circular channel within the ECM. Cells (such as EC alone or EC co-culture with pericytes) can then be seeded via the media ports, where they will subsequently adhere and form a cylindrical monolayer. Once ECs have adhered, flow can be applied by a syringe pump or using a laboratory rocker (to produce steady laminar or oscillatory FSS, respectively). As a needle is used to set the channel, it’s width can be set to a defined diameter from 15–300 μm. This is unlike vessels in PDMS-hydrogel hybrids, where form lumens spontaneously. These systems can be used to assess barrier function of the vasculature. Interestingly, in the absence of flow these vessels are leaky but when constant application of low FSS is applied (5 dyn/cm2 overnight), cell–cell adhesion is induced to prevent leakage of 70-kDa Dextran. If two cylindrical vessels are embedded within the same hydrogel these models can also be used to assess angiogenesis, with newly formed capillaries sprouting across the ECM in response to growth factor gradients between two larger vessels, a process which is dependent on EC-mediated Myosin IIA forces.

47While hydrogel-embedded microvessels are excellent models to study the effects of FSS on the vasculature, cells are embedded in a linear vessel. Therefore, they do not accurately model the effects of disturbed flow in bifurcating vessels which are found abundantly in vivo. Recently, PDMS-hydrogel hybrids have been developed to study the effects of flow at bifurcation points in vessels (Figure 1H). Interestingly both angiogenic sprouting and permeability at bifurcation points is reduced after 48 h (where the bifurcation point has stagnation pressure of approximately 38 dyn/cm2 and branched vessels are exposed to 3 dyn/cm2). This suggests bifurcation points stabilize the blood vasculature, as has been demonstrated in vivo.

48Conventional in vitro studies have long been performed using plastic or glass matrices to culture ECs. The stiffness of plastic and glass dishes (which is in the GPa range) by far exceeds the physiological stiffness that ECs experience in vivo. As discussed above, the physiological ECM stiffness experienced by ECs varies from very soft microenvironments during development to stiffer microenvironments in arterial ECs or in aged and diseased tissues. EC behavior in response to more compliant substrates has been assessed more frequently in 2D culture where ECs are seeded on top of those substrates (Figure 2A).

49Hydrogels, which are water-swollen networks of polymers, have emerged as the most promising choice for cell culture since they mimic salient elements of the ECM, have mechanics similar to those of softer tissues and can support cell adhesion and protein sequestration. The properties of hydrogels, such as mechanics, swelling, porosity and degradation, are modifiable and can be adjusted to the specific requirements of the desired experiment. For commercially available hydrogels, these properties have been experimentally characterized and can be easily adapted by the user. Diverse stiffness hydrogels can be generated from natural materials, such as collagen, fibrin or alginate, and synthetic materials, such as polyacrylamide (PA) or Polyethylene glycol (PEG). In selected settings, the use of natural materials can be advantageous as they exhibit high biocompatibility, allowing for EC interaction via integrin ligation. However, ECM stiffness provided by those natural materials is very soft. For example, the matrix stiffness of collagen hydrogels is very low (<1 kPa). Another frequently used natural hydrogel, Matrigel, which is a heterogenic mixture of various ECM proteins, was found to have a stiffness of approximately 0.45 kPa when measured with AFM in an aqueous, temperature controlled environment. Furthermore, Matrigel and other natural hydrogel materials do not allow user-defined variability of ECM components.

50In contrast, synthetic materials are more inert and usually have to be combined with chemically modified crosslinkers. This allows for integrin ligation to facilitate interaction of ECs with the substrate. Synthetic materials, such as PA, can be utilized to generate hydrogels with much wider matrix stiffness ranges, from 0.1 kPa to high kPa values. These characteristics become highly advantageous when matrix stiffness and EC-matrix interactions have to be analyzed separately. For example, PA hydrogels with different matrix stiffness values can be generated with variable collagen I concentrations. This allows for study of the balance between substrate mechanics and matrix chemistry in the process of EC network assembly.

51The impact of ECM stiffness on EC behavior has not been studied extensively in 3D in vitro models. Most 3D studies to analyze angiogenic sprouting processes have been performed using natural hydrogel materials, such as collagen type I and fibrin. For example, ECs can be coated onto cytodex microcarriers and embedded into a fibrin hydrogel (Figure 2B). Here, fibroblasts are layered on top of the gel where they provide necessary soluble factors that promote EC sprouting from the surface of the beads. These fibroblast derived factors act to minimally stiffen gels (0.03–0.05 kPa) and are required for the lumenisation of sprouting vessels. In an elegant study by, thick 3D collagen gels were demonstrated to provide a temporal hypoxic gradient (Figure 2C), allowing for the study of EC behavior in response to changes in both hypoxia and ECM stiffness. These models have simultaneously revealed encapsulation of endothelial colony-forming cells (ECFCs) leads to vasculogenic cluster formation through ROS production in hypoxic (thick) but not normoxic (thin) collagen hydrogels (Figure 2C), and that increased matrix stiffness via collagen crosslinking leads to reduced cluster formation.

52As described above, natural hydrogels provide a soft microenvironment for ECs in culture, suggesting these 3D in vitro models are most suitable to mimic the soft ECM environment that allows ECs to migrate and sprout in vivo. To generate stiffer (>1 kPa) 3D environments to study the impact of higher ECM stiffness on EC behavior, commonly used PA hydrogels are not suitable to encapsulate ECs due to the high toxicity of the hydrogel precursors. Therefore, synthetic PEG hydrogels are more widely used to analyze the effects of a stiff ECM in a 3D context.

53In addition to higher matrix stiffness ranges and the ability to independently modulate adhesive ligand presentation, tunable degradation of synthetic hydrogels offers the possibility to alter the microenvironment and stiffness experienced by cells during an experiment. This can be done using ultraviolet light or by adding engineered, MMP-degradable peptide crosslinkers to the hydrogel solution. Another advantage of using synthetic materials is that the swelling properties of these hydrogels can be modulated. An elegant study by eliminated the swelling properties of a synthetic hydrogel after polymerization, which enabled its integration into microfluidic channels (without channel closure) and subsequent channel colonization with ECs (Figure 2D).

54In addition to alterations in stiffness, growing cells on flexible substrates to allow for cyclic stretch has wide implications for EC biology. Most of the studies on stretch are performed using commercial systems, such as the Flexcell Tension Plus system (Flexercell, United States). Here, cells are plated on flexible silicon membranes coated with matrix and grown in standard media before being subjected to stretch. These systems induce cellular elongation in response to stretching and deformation of the elastomere base of the plates, achieved using vacuum pressure which exposes cells to radial, but not uniaxial stretch. High throughput techniques to induce CS in a 96 well place format have been developed using PDMS devices, where cells are exposed to uniaxial stretch. However, these can only reach 10% stretch. Advances in models for measuring cyclic stretch are found in lung-on-a-chip models, discussed in further detail in the following section.

55Förster resonance energy transfer (FRET) reporter constructs have been widely used in the field of cell biology to study protein activity, interactions and conformational changes. FRET biosensors rely on transferring energy from the excited state of the donor fluorescent protein to the unexcited acceptor fluorescent protein, which then emits free photons. FRET efficiency is sensitive to the distance and orientation of the separation of the two fluorophores which make up the FRET pair, with distance changes as small as 1 nm substantially affecting the amount of energy transfer. Biosensors may consist of an intracellular construct that is sensitive to kinase activity, such as Src and FAK-Lyn (Figure 3A) or conformational changes in proteins, such as Src, non-muscle Myosin II and α-catenin. Biosensors have also been utilized in vivo to measure changes in protein activity, such as Src, AKT, NO, and RhoA.

56The vascular biology field has utilized FRET biosensors in EC culture to measure tension across vinculin and talin at FAs and PECAM-1 and VE-cadherin at cell–cell junctions. These tension sensor constructs consist of a donor fluorophore connected to an acceptor through a nanospring derived from the elastic spider silk protein flagelliform (Figure 3B). Where tension is low or absent, the nanospring is compact and FRET is high, and conversely, application of tension stretches the spring and decreases FRET. These tension sensors are now being utilized in vivo in ECs, with a VE-cadherin-tension sensor zebrafish recently described by. This zebrafish line allows for live, intravital fluorescence lifetime imaging microscopy–fluorescence resonance energy transfer (FLIM- FRET) imaging of tension across junctions. These models and imaging techniques are predicted to be more widely utilized in the vascular biology field to measure changes in tension or subcellular signaling dynamics in tissue culture or in vivo models.

57Recent scRNA-seq approaches have revealed high levels of EC heterogeneity not only between EC types, but also between organ tissues. To investigate organ-specific vasculature functions, organ-on-a-chip models have been developed to study tumor, gut, liver, pancreas and muscle function. These are also being applied in the field of personalized medicine by incorporating patient samples. These models aim to culture cells in conditions that closely mimic the body, taking into account factors such as tissue geometry and composition, dynamics and flow gradients. Generally, they consist of microfluidic devices integrated with microchip manufacturing methods, which allow for continuously perfused chambers with live cells that have been arranged to simulate tissue- and organ-level physiology.

58Due to the fact that the lung vasculature is exposed to multiple mechanical forces, here we will focus on lung-on-a-chip models. Pioneered by the Ingber laboratory, these devices are comprised of alveolar epithelial cells plated on the apical side of an ECM-coated porous membrane, with EC coated on the basal side (Figure 3C). Alveolar cells are exposed to air and ECs are exposed to FSS at a range between 1 and 15 dyn/cm2, similar to forces applied in vivo. To model cyclic stretch, suction is applied to two external chambers flanking the cell channel, causing it to extend and retract at a rate and degree to that found in the body (approximately 10% cyclic stretch). This allows for modeling of multiple mechanical forces on the cells, including stretch, ECM components and fluid flow, making them superior to traditional microfluidic devices. Such devices can also can be modified to model other tissues with polarized EC interfaces, such as kidney (EC-podocyte), liver (EC-hepatocyte), brain (EC-astrocyte), and tumor (tumor-epithelium-EC).

59Microcirculation-on-a-chip models have been developed to model organizational layers within the vasculature. One such model allows for the co-culture of blood and lymphatic vessels within the same device, separated by a porous polyethylene terephthalate (PET) membrane. This chip can be utilized to measure vascular permeability and lymphatic absorption. Other devices with an organized composition of ECs and SMCs separated by a porous membrane have been developed to better model the composition of the arterial wall. These allow for a more accurate examination of the interactions and signaling between these cellular components. Finally, models of interconnected, perfused vessels with a high flow arterial input, capillary plexus and low flow venous output have also been developed, to model multiple stages of vascular development including vasculogenesis, angiogenesis to anastomosis. While advantageous for their complex vascular organization, these chips lack the integration of other organ-specific cells.

60In addition to organ-on-a-chip models, the use of organoids for in vitro investigation of organ-specific phenotypes is emerging in the fields of biology and drug discovery. A major road block in organoid research is how to incorporate a functional vasculature network. Not only is the field faced with the challenge of EC heterogeneity and matching the vasculature to the type of tissue organoid, organoids in vitro remain small and immature due to our inability to grow a perfused, functional vascular network within them. However, when transplanted in vivo, organoids can become vascularized due to ECs derived from transplanted iPSCs or the host. For a detailed review on the challenges in vascularization of organoids, see the review by.

61The complexity and relevance of mechanical forces to generate and maintain a functional vascular system is well appreciated by vascular biologists. Sophisticated methods have been developed to model and measure forces in 2D cultured monolayers, but are now being adopted into more complex 3D models. The choice of a suitable 2D or 3D model is dependent on the specific endothelial process to be studied. While 3D models may be superior to study FSS within a model of a circular, lumenised vessel, or to model the ECM stiffness microenvironment surrounding sprouting and migrating ECs, stiffness experienced by ECs of a mature vessel network may be more accurately studied in 2D settings. This is due to the fact mature vessels are equipped with a monolayer of ECs which are naturally attached to the underlying ECM in a 2D manner.

62Given the transport of fluid exposes ECs to the mechanical forces of FSS and cyclic stretch simultaneously, and the stiffness of the surrounding matrix will affect the degree of cyclic stretch and FSS, techniques to incorporate modeling of all these forces will yield maximal physiological relevance. We believe that further improvements to in vitro and in vivo force modeling will begin to unravel the precise impact of FSS, stiffness and stretch on the development and function of a heterogeneous vessel network across different tissues.

63MF, EG, and LS wrote and edited the manuscript.

64The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling editor declared a past co-authorship with one of the authors LS.