The role of mechanical stimuli in the vascular differentiation of mesenchymal stem cells

Cardiovascular diseases, particularly clogging of the coronary artery, remain the biggest cause of mortality worldwide. Coronary artery bypass grafts (CABGs) are considered gold standard procedures for these diseases. Although the left internal mammary arteries and saphenous veins are still the preferential sources for CABGs, vascular tissue engineering has proven to be an attractive alternative. The engineering of small-diameter (<6 mm) vascular tissue has greatly advanced over the past two decades, with the majority of research efforts having been aimed at the generation of vascular grafts that mimic blood vessels.

Blood vessels typically comprise three main layers – the intima layer, which comprises endothelial cells; the media layer, which comprises smooth muscle cells (SMCs); and the outermost layer (adventitia), which is made up of fibroblasts and extracellular matrix (ECM) components. Endothelial cells are usually characterized by the expression of VE-cadherin, CD31 (also known as PECAM1) and vascular endothelial growth factor receptor 2 (VEGFR2, encoded by KDR), and by their capacity to take up acetylated low-density lipoproteins (ac-LDL), to form tubes in Matrigel^® and to release nitric oxide. Smooth muscle cells, by contrast, are typified by the expression of contractile markers, such as smooth muscle α-actin (SMA, also known as ACTA2), calponin 1, SM22α (also known as TAGLN) and smooth muscle myosin heavy chain (SMMHC, also known as MYH11), and by the ability to synthesize ECM proteins (Owens, 1995). To be viable, the constructed graft needs to exhibit at least the following four properties – (i) a burst pressure superior to that of the saphenous veins; (ii) a long-term potency that resists the pressure of physiological circulation; (iii) avoidance of platelet adhesion that induces clot formation; and (iv) ease of engineering (L'Heureux et al., 2007). The first tissue-engineered blood vessel with these properties was constructed in 1998 and contained SMCs and endothelial cells that had both been isolated from human umbilical veins (L'Heureux et al., 1998). However, the use of mature vascular cells (i.e. those isolated from native blood vessels) is inherently limited because acquiring them demands an invasive procedure, they are typically available in insufficient quantities and they exhibit a low rate of proliferation (Shin'oka et al., 2001; Koike et al., 2004; L'Heureux et al., 2006). Thus, in order to advance vascular tissue engineering, stem cells have been intensively explored as alternative sources for both mature endothelial cells and SMCs (Ferreira et al., 2007). Among the various stem cell types that are suitable for application in vascular grafts, mesenchymal stem cells (MSCs) have been most extensively studied to date (Seifu et al., 2013). MSCs are found in different types of tissue of adult or perinatal origin, such as bone marrow, adipose tissue or Wharton's jelly from the umbilical cord (El Omar et al., 2014). Their high proliferation rate, multipotency and immunomodulatory properties make them a highly suitable candidate for vascular tissue engineering.

A number of factors are typically required to initiate vascular differentiation of stem cells in vitro. VEGF proteins are commonly used for differentiation into endothelial cells (Oswald et al., 2004), and transforming growth factor β1 (TGF-β1) or platelet-derived growth factor (PDGF) for differentiation into SMCs. In addition to these biochemical factors, the microenvironment of vascular cells exerts two main hemodynamic forces, shear stress and cyclic strain, which contribute to the differentiation process. Shear stress is generated by the blood passing through vessels, whereas cyclic strain is created by the pulsatile nature of blood flow, which produces a tensile stress that acts perpendicular to the vessel wall (Fig. 1). In vitro, both of these two mechanical cues can be replicated by different models, such as using a parallel-plate chamber for shear stress and a longitudinal stretch system for cyclic strain (see Box 1). MSCs are highly responsive to their mechanical environment, and different types of forces applied to the same MSC population result in divergent outcomes (Maul et al., 2011). Therefore, in this Commentary, we will summarize the available literature that describes a role for mechanical forces in the differentiation of MSCs into vascular cells and discuss different means to maintain differentiation.

Endothelial cells regulate vascular tone and mural remodelling in a shear-dependent manner, which is commonly assumed to maintain a constant level of wall shear stress across arteries. Recently published data suggest that the flow velocity experienced by endothelial cells varies in a species-dependent manner (Weinberg and Ross Ethier, 2007). In this section, we will summarize the main results published to date pertaining to the possibility of using mechanical stimulation to induce the differentiation of human and animal cells into endothelial cells.

As pointed out earlier, the endothelium is in direct contact with blood and is therefore exposed to shear stress, which is in turn required for the development of new blood vessels both in embryos and in adults (le Noble et al., 2004; Folkman and Haudenschild, 1980). Blood flow in small-diameter arteries generates shear stresses that are within the range of 10–20 dyn/cm² in humans (Giddens et al., 1993). Such hemodynamic shear stress on endothelial cells is in fact important for the maintenance of the phenotype, orientation, metabolic activities and homeostasis of the vascular endothelium (Davies, 2009). Furthermore, blood flow exerts an anti-thrombotic effect in vessels by inducing the release of thrombomodulin by endothelial cells, which inactivates the pro-coagulant factor thrombin and activates the anticoagulant protein C (Yamawaki et al., 2005). As endothelial cells are an essential component of the vessel wall, much of the research has focused on obtaining endothelial cells in vitro after stem cell differentiation. Given the crucial role of shear stress in the differentiation of endothelial cells in vivo, it is believed that replicating shear stress in cell culture (in vitro) could be crucial for differentiating the stem cells towards the endothelial phenotype (Ahsan and Nerem, 2010; Yamamoto et al., 2005; Cheng et al., 2013).

The effect of shear stress on different types of MSCs has been studied previously. For example, bone-marrow-derived MSCs from a number of species are able to differentiate into endothelial-like cells when they are stimulated through physiological shear stress conditions (Engelmayr et al., 2006; Bai et al., 2010; Huang et al., 2010; Maul et al., 2011; Dong et al., 2009; Kim et al., 2011). Likewise, MSCs derived from human adipose tissue (hASCs) (Bassaneze et al., 2010; Zhang et al., 2011; Shojaei et al., 2013) and MSCs from other sources, such as amniotic fluid cells (Zhang et al., 2009) and human placenta (Wu et al., 2008), have also been shown to differentiate into endothelial-like cells under shear stress (see Table 1 for a summary).

In addition to shear stress, the physiological microenvironment of the vessels also plays a role in endothelial cell differentiation, particularly if it contains VEGF proteins. This effect has been shown in a study that compared the consequences of shear stress, treatment with VEGF-family members or a combination of both stimuli on rat MSCs (Bai et al., 2010). In that report, the two stimuli are shown to have a synergistic effect on endothelial cell differentiation. Furthermore, when VEGF was removed from the culture, and shear stress was continued for up to 48 h, loss of endothelial markers was observed, suggesting that de-differentiation had begun. By contrast, another study did not observe any endothelial cell markers on hASCs that had been treated with 10 dyn/cm² of shear stress alone, even after 96 h of stimulation (Bassaneze et al., 2010). The authors suggest that the failure to differentiate is caused by the lack of VEGFR2 expression in hASCs, which is essential to induce their differentiation into endothelial cells (Bassaneze et al., 2010).

A synergy between VEGF proteins and shear stress had already been shown in two studies (Wu et al., 2008; Fischer et al., 2009). When cells were cultured in endothelial cell growth supplement for 3 days under static conditions and subsequently exposed to shear stress of 12 dyn/cm² for at least 24 h, expression of CD31 and increased ac-LDL uptake were observed, as well as tube formation in vitro. Analogous results have been obtained by Zhang et al., who report that MSCs derived from human amniotic fluid and hASCs that have been cultivated in endothelial cell differentiation medium EGM™-2 for two weeks express von Willebrand factor (vWF) and CD31, and gain the ability to release endothelial nitric oxide synthase (eNOS, also known as NOS3) and to form tubes on Matrigel^®. The subsequent application of shear stress further increases the expression of endothelial markers and enhanced functional characteristics (Zhang et al., 2009).

Taken together, these studies suggest that mechanical and biochemical influences synergize in order to increase endothelial cell functionality and the expression of endothelial cell markers.

Although shear stress plays a role in endothelial cell differentiation, the underlying mechanism remains unclear. According to two independent studies, it appears that shear stress promotes endothelial differentiation, while downregulating differentiation towards a SMC phenotype (Wang et al., 2008; Dong et al., 2009). Specifically, the first study showed that murine MSCs subjected to a shear stress of 15 dyn/cm² express VEGF, but significantly decrease their expression of other growth factors – such as TGF-β1, as well as PDGF and its receptor – which are known to guide stem cells towards the SMC lineage (Wang et al., 2008). Similarly, Dong et al. have shown an increase in endothelial markers at both the mRNA and protein levels, whereas SMC markers, such as SMA and calponins, decrease after subjecting canine MSCs to a shear stress of 15 dyn/cm² over 2 days (Dong et al., 2009). Furthermore, the effect of shear stress on differentiation has been shown to vary depending on stress intensity. For example, Kim et al. report that MSCs that have been exposed to low (2.5 dyn/cm²) or high (10 dyn/cm²) shear stress express endothelial markers, including CD31, vWF and VEGFR2. The expression of CD31 under low shear stress is considerably higher than that under high shear stress, whereas the expression of vWF and VEGFR2 is only slightly higher under low shear stress. Furthermore, MSCs that have been exposed to high shear stress show a significantly higher expression of SMC markers (myocardin, SMMHC and SM22α) (Kim et al., 2011). By contrast, other studies have suggested that shear stress increases the amount of TGF-β1 and of SMC markers (Park et al., 2007; Kobayashi et al., 2004). These apparent inconsistencies regarding the expression of TGF-β1 and SMC-specific markers could be attributable to the use of different types of stem cells, culture conditions or shear stress regimes.

The effects of mechanical forces on stem cell differentiation depends on the way the force is applied to cells (i.e. continuous stimulation or intermittent) and on its intensity; however, the fate of the stem cells might also be influenced by the stiffness of the substrate on which they are cultured. For example, two studies have reported that shear stress applied to stem cells leads to osteoblastic differentiation of MSCs (Lim et al., 2011; Yeatts et al., 2012). Furthermore, two studies by Kreke et al. have demonstrated that intermittent or low shear stress (below 3 dyn/cm²) induces osteocalcin expression by bone marrow MSCs that have been cultured on fibronectin and enhances the expression of other bone proteins compared with that of cells cultured with osteogenic factors under static conditions (Kreke et al., 2005, 2008). McBride et al. have followed a similar approach and investigated shear-stress-mediated gene expression changes in the MSC line C3H/10T1/2 in order to determine the role of shear stress in osteogenesis (McBride et al., 2008). They find that exposure of cells to a continuous shear stress of either 0.2 dyn/cm² or 1 dyn/cm² for 30 or 60 min induces the upregulation of a number of ossification-related genes, including those encoding collagen type I, collagen type II, Runx2 and Sox9. In this experiment, the length of exposure to shear stress (30 or 60 min) was a stronger predictor of the genes that are upregulated than the intensity of the applied shear stress. By contrast, in an earlier study that exposed the same cell line to laminar shear stress (15 dyn/cm²) for 6 or 12 h, shear-stress-mediated upregulation of endothelial instead of bone differentiation markers was observed (Wang et al., 2005).

Taken together, these results indicate that the intensity and duration of the shear stress employed affects the balance between osteogenic and endothelial differentiation and could thus lead to different outcomes depending on the regime used.

On the basis that shear stress alone can direct differentiation towards an endothelial cell phenotype, the underlying molecular mechanism has been explored. Mechanosensitive cell surface molecules, such as primary cilia, integrins, ion channels and focal adhesion proteins, are able to mediate cellular responses to various forces. The primary cilium, which is ubiquitous in mammals, is an immotile microtubule-based organelle that protrudes from the apical surface and is able to bend in response to flow (Praetorius and Spring, 2005; Satir et al., 2010). A model for primary-cilia-based mechanosensing has been described in which bending of the cell cilium by shear flow can activate ion channels embedded in the membrane (Janmey and McCulloch, 2007). More recently, a study has provided evidence that primary cilia are involved in the osteogenic differentiation of human bone marrow MSCs (hBMMSCs) (Hoey et al., 2012). Other studies have suggested that transmembrane proteins, including ion channels and integrins, could be activated directly by shear flow as well as cyclic strain because these mechanical forces induce changes to the membranes – including altering membrane tension, which activates ion channel conductance (Liu and Lee, 2014). Opening of these channels increases the cytoplasmic concentration of the associated ion, which then changes the membrane potential, or, alternatively, the ions themselves can act as second messengers (Wu et al., 1999). For example, it has been shown that Ca²⁺-signalling pathways can be activated through the activation of the mitogen-activated protein kinase (MAPK) pathway, which controls osteogenic differentiation of MSCs (Barradas et al., 2013).

Integrins not only bind to many components of the ECM, such as collagen and laminin, but also to other proteins, such as VEGF-A or insulin-like growth factor binding proteins (IGFBP1 and IGFBP2) (Vlahakis et al., 2007). Integrins link the ECM to the cytoskeleton and so facilitate the formation of focal adhesions between the cell and the ECM (Hoffman et al., 2011). At focal adhesions, integrins can also translate physical forces into biological signals (Schwartz and Assoian, 2001). Specific interactions formed by the α and β subunits of integrins that are activated by shear stress trigger the focal adhesion kinase (FAK)–Src and the caveolin 1 (Cav1)–Fyn kinase pathways, thereby eliciting phosphorylation cascades of various downstream effectors and their interaction with each other through SH2- and SH3-domain-mediated interactions. The two pathways converge at the Raf–MEK–ERK axis in endothelial cells (Shyy and Chien, 2002). Cheng et al. have shown that the integrin–Ras–ERK pathway also has a role in endothelial cell differentiation. They report that exposure of endothelial progenitor cells (EPCs) to shear stress results in an increase in the endothelial maturation markers CD31 and vWF (Cheng et al., 2013). In this work, shear stress was also found to activate several mechanosensitive molecules, including integrin β1, Ras, ERK1/2, paxillin and FAK, which are all involved in cytoskeletal rearrangements, as well as in the differentiation into late EPCs (Cheng et al., 2013).

Other cell surface molecules, such as G-protein-coupled receptors (GPCRs) and the glycocalyx, also serve as sensors of shear stress (Ando and Yamamoto, 2009; Stolberg and McCloskey, 2009). Mechanical stress and/or ligand binding induces conformational changes in the GPCRs, which transduce a signal to the cytoplasmic tail. Guanosine triphosphate (GTP) can provoke further conformational changes of the receptor, leading to its association with Ras, which in turn activates phosphoinositide 3-kinase (PI3K)–Akt pathway. This pathway has been shown to regulate MSC survival, proliferation, migration and other cellular fates through crosstalk with the Wnt-signaling pathway (Huang et al., 2013). Furthermore, it has been shown that during endothelial cell differentiation of hASCs in response to shear stress, the acquisition of endothelial cell characteristics also depends on the PI3K–Akt pathway (Zhang et al., 2011). In this context, it is also worth mentioning that other protein kinases that are involved in cell growth and differentiation are recruited following mechanical stimulation of GPCRs, including p38-family MAPKs, ERK1/2, JNK-family kinases and YAP/TAZ (Kaneko et al., 2014). Other pathways may be involved downstream of these surface molecules as recently reviewed in (Shah et al., 2014).

SMCs are located in the middle layer of small-diameter arteries and play an important role in maintaining the elasticity and homeostasis of blood vessels. Their limited proliferative capacity and reduced collagen production thus prevent the formation of neointima (scar tissue) and luminal narrowing (Paszkowiak and Dardik, 2003).

Biochemical factors such as TGF-β1 and PDGF have a well-established role in the differentiation of MSCs into SMCs. Successful SMC differentiation that is mediated by these growth factors is usually assessed by the expression of major SMC-specific proteins – including SMA, SM22α, calponin-family proteins, heavy caldesmon (h-caldesmon) and SMMHC – contraction in response to carbachol and KCl, and the production of ECM (Owens, 1995; Huang and Niklason, 2014). SMCs experience constant mechanical stimulation in blood vessels and undergo cyclic strain that results from pulsatile stresses from the systolic ejection of the heart. Under physiological conditions, the strain levels in blood vessels are in the range 5–30% at frequencies of 30–90 cycles per minute (0.5–1.5 Hz) (Huang and Li, 2008). This phenomenon has generated interest in recent years, and a number of studies have investigated the effect of cyclic strain on MSC differentiation into SMCs (summarized in Table 2).

Cyclic strain is usually generated in vitro by the repetitive extension and constriction of the elastic substrates onto which cells have been seeded. Cyclic strain can be applied and investigated in both two- and three-dimensional (2D and 3D, respectively) systems, with elastic silicone membranes most commonly used as a scaffold for 2D models and tubular matrices for 3D models. In 2D stretch models, cyclic strain is divided into uniaxial and equiaxial strains. Uniaxial strain refers to the fluid force along only one axis, whereas equiaxial strain describes uniform strain in all directions. Park et al. have compared the efficacies of equiaxial and uniaxial strain in inducing differentiation of hBMMSCs into SMCs in a 2D model. Equiaxial strain decreases the expression of the SMC markers SMA and SM22α, whereas uniaxial strain transiently increases their expression after 1 day of treatment. Furthermore, hBMMSCs align perpendicularly to the uniaxial strain axis but not that of the equiaxial strain (Park et al., 2004). In fact, the majority of studies using 2D models have been performed with uniaxial cyclic strain and result in SMC-like cells (Hamilton et al., 2004; Kurpinski et al., 2009; Jang et al., 2011; Ghazanfari et al., 2009). 3D models that are designed to replicate the status of SMCs in vivo because the radial distension of the tubular grafts that is induced by blood flow pressure allow cells to experience an effective uniaxial cyclic stain, similar to that in their natural environment (Du et al., 2011; Zheng et al., 2012). However, unlike the perpendicular orientation that is observed in most 2D uniaxial cyclic strain experiments, SMCs in native vessels align themselves parallel to the principal direction of the strain. Furthermore, comparison between the effects of 2D uniaxial strain and 3D radial distension on SMC differentiation has shown that the latter results in a significantly lower level of expression of SMC-associated markers (O'Cearbhaill et al., 2010). Although 2D uniaxial strain induces a different response from that of the in vivo situation in terms of cell alignment, it is nevertheless more effective in mediating differentiation than cyclic strain applied in a 3D model. It is thus clear that additional 3D models need to be developed which are able to better replicate in vivo conditions and thus might be more effective in directing differentiation towards SMCs (Smith and Gerecht, 2014).

In the case of rat and human MSCs, cyclic strain alone has been shown to induce the expression of some but not all SMC markers, in particular markers of mature SMCs such as SMMHC (Nieponice et al., 2007; Hamilton et al., 2004; Ghazanfari et al., 2009; O'Cearbhaill et al., 2010). In order to increase differentiation efficiency, cyclic strain can be applied in combination with other factors, such as growth factors. For instance, it has been shown that in order to maintain the SMC phenotype induced by growth factors (e.g. TGF-β1, BMP-4), that is to avoid a loss of SMC-related markers, this stimulation needs to be followed by pulsatile strain (Wang et al., 2010). However, a different study has been successful in obtaining differentiation only by combining TGF-β1 and uniaxial strain, as these factors showed a positive synergistic effect (Kurpinski et al., 2009).

Based on the above studies, we can conclude that two methods can be used to ensure long-term (or stable) SMC differentiation – sequential stimulation (first by chemical then mechanical means) or synergistic stimulation (simultaneous chemical and mechanical stimulation).

As a general rule, the duration of mechanical stimulation that is required to achieve differentiation into SMCs is longer than that for differentiation into endothelial cells. Cyclic strain should be applied for at least 3 days in order to mediate the expression of some SMC markers, such as SMA, but even this is insufficient for the expression of mature markers, such as SMMHC (Hamilton et al., 2004). Cyclic strain has also been used to induce other types of differentiation, such as differentiation into skeletal muscle cells (Haghighipour et al., 2012), neuron-like cells (Leong et al., 2012) and osteogenic lineages. It is worth noting that cyclic strain alone can evoke considerable expression of SMC markers, as well as upregulation of the osteoinductor bone morphogenic protein 2 (BMP-2) and the early matrix protein osteopontin (OPN) (Maul et al., 2011). In recent work, the role of cyclic strain intensity on MSC differentiation was investigated. It was shown that cells that are stretched at 0.1 Hz express higher levels of osteoblast-specific genetic and protein markers compared with cells that are stretched at 1 Hz, which express significantly higher levels of vascular-SMC-specific genetic and protein markers (Yao and Wong, 2015). These results also support the findings of an earlier study, in which cyclic strain of a low magnitude (0.26 Hz, 3%) mediated the differentiation of MSCs into osteogenic cells, whereas greater cyclic strain (0.26 Hz, 10%) favoured differentiation into SMCs (Jang et al., 2011).

Further research efforts are thus required to develop protocols that promote reliable cyclic-strain-induced differentiation of cells into SMCs that exhibit a contractile phenotype for use in vascular tissue engineering.

For the purpose of vascular tissue engineering, an ideal multilayer vessel would comprise an innermost layer of endothelial cells, a middle layer of SMCs and an outer layer comprising ECM and fibroblasts. The studies summarized above suggest that shear and cyclic stress can be used to differentiate MSCs into endothelial cells and SMCs, respectively. Questions remain, however, as to whether the differentiation that has been initiated in vitro can be maintained long term and how to avoid cellular dedifferentiation once the stimulation is terminated. One of the answers could be to maintain the tissue-engineered blood vessels (TEVs) under continuous stimulation with shear stress or cyclic strain. This can be achieved in vitro with the use of bioreactors or by grafting the vessel into a host organism in vivo, thereby exploiting the host's physiological environment to generate either shear stress by blood flow or cyclic strain by pulsatile stimulation.

In the area of bioreactors, Huang et al. have recently designed a bioreactor that can better simulate the physiological stresses acting on native arteries and that allows biaxial stretching during culture (Huang and Niklason, 2014; Huang et al., 2015). Biaxial stretching induces a greater wall thickness compared with stimulation with static or uniaxial stretch. Furthermore, this stimulation leads to the development of undulated collagen fibres and mature elastin in the ECM, which might contribute to the observed vascular compliance in the biaxial TEVs obtained with this bioreactor. These results suggest that this type of bioreactor could be used to optimize biomechanical conditioning of TEVs and, thus, could help to maintain cells in a differentiated state.

Regarding the in vivo implantation of vascular grafts to improve their patency after being engineered in vitro, interesting results have been obtained. For instance, Udelsman et al. have recently engineered a vessel by using a polymer that had been seeded with bone marrow mononuclear cells, which they then implanted into mice. After 7 months, they observed the development of a luminal endothelial cell layer, as well as a SMC-populated ‘medial’ layer that associated with components of ECM, including elastin and collagen type I and type III, suggesting that cells that have been biomechanically stimulated in vivo could differentiate into endothelial cells or SMCs (Udelsman et al., 2014). Zhao et al. have also shown that it is possible to maintain cell differentiation after implantation in vivo. These authors constructed a TEV by using endothelial cells and SMCs that had been differentiated from MSCs and seeded onto a decellularized ovine artery, which was subsequently interposed in the carotid arteries in an ovine host model. After 5 months, the analysis of the graft showed the existence of endothelium, smooth muscle, collagen and elastin, indicating that differentiation had been maintained over this period of time (Zhao et al., 2010).

Shear stresses at physiological levels can differentiate MSCs into endothelial cells by inducing the expression of endothelial-cell-specific markers and endothelial-cell -specific functional properties. Cyclic strain, both in 2D and 3D models, can guide MSCs to differentiate into SMCs by inducing the expression of several SMC-associated markers. However, compared with shear stress-mediated endothelial cell differentiation, cyclic strain alone has only a relatively limited ability to induce SMC differentiation. It is clear that mechanical force can increase vascular differentiation in combination with biochemical factors, such as growth factors. Taken together, the studies described here suggest that the most effective device for generating high-functioning TEVs would be one that is capable of applying shear stress and cyclic strain on seeded MSC-derived endothelial cells and SMCs, in order to maintain their activity and differentiation state before use in tissue engineering. However, much remains to be done to progress from 2D to 3D models and from in vitro to in vivo applications in order to be able to produce a mechanical-force-mediated tissue-engineered vessel for clinical use.

We thank Professor Kira Weissman for careful reading of the review.