In Search of the Pivot Point of Mechanotransduction: Mechanosensing of Stem Cells

Abstract

Stem cells are undifferentiated cells with the ability to self-renew and to differentiate into diverse specialized cell types; hence, they have great potential in tissue engineering and cell therapies. In addition to biochemical regulation, the physical properties of the microenvironments, such as scaffold topography, substrate stiffness, and mechanical forces, including fluid shear stress, compression, and tensile strain, can also regulate the proliferation and differentiation of stem cells. Upon physical stimuli, cytoskeleton rearrangements are expected to counterbalance the extracellular mechanical forces, trigger signaling cascades, and eventually cause epigenetic modifications. This article mainly focuses on the mechanosensing, which is the upstream event of stem cell mechanotransduction and the downstream one of physical stimuli. Putative mechanosensors such as ion channels, integrins, and cell membrane as well as primary cilia are discussed. Because mechanical environment is an important stem cell niche, identification of mechanosensors not only can elucidate the mechanisms of mechanotransduction and fate commitments but also bring new prospects of the mechanical control as well as drug development for clinical application.

Mechanosensing: From Sensory Neurons to Stem Cells

Mechanosensing is an essential physiology function in daily life as it keeps the organism informed as to the environment and protects it from danger. Mechanotrans-duction is involved in two of the five senses: hearing (audition) and touch (tactition). In addition to touch sensing, our bodies experience mechanical forces generated from daily activities, such as walking, heart beating, and breathing, which are for maintaining physiological function. In addition to somatosensory mechanosensing, exploration of mechanosensing at the cellular level is also of interest. Cells do not have eyes or ears but can sense both endogenous forces mainly generated from cytoskeletal contractility and exogenous ones originating from the extracellular environment. The exogenous mechanical cues include tissue-specific elasticity depending on the extracellular matrix (ECM) and constituent cells, matrix stiffness changes during embryonic development or wound-healing process, compression forces from cell– cell contact, shear stress from blood flow, and stretch due to organ inflation. These physical cues are mainly transduced into biochemical signals, and cells respond to the cues in the form of morphology changes, migration, or replicative senescence through activating signaling pathways as well as changes in gene expression (67).

In sensory neuron cells, physical forces are transduced by mechanosensors in the form of electrical signals. The information of stimuli is then processed in the brain via sensory nerves located in the spinal cord and translated as proprioception, touch, and pain. The mechanotransduction and mechanoresponses can be analyzed in real time by electrophysiology and fluorescence microscopy as well as animal behavior experiments (15,56,88). There have been several mechanosensors of somatosensory mechanotransduction proposed under different physical stimuli (8,12,76). However, even though it is known that rhodopsin is the sensor for sight (90), G protein-coupled receptors are for smell, and transient receptor potential (TRP) channels are for taste (9,20), the identification of mechanical sensors and the underlying mechanotransduction mechanisms is still elusive.

Stem cells are undifferentiated cells that have the ability to self-renew and to differentiate toward diverse specialized cell types (pluripotency). The self-renewing and pluripotency provide a great potential in tissue engineering and regenerative medicine. In adults, quiescent stem cells can be activated by tissue injury and function as a repair system to maintain or regenerate damaged organs. In addition to the known soluble molecules such as growth factors and cytokines regulating stem cell differentiation, there are accumulating studies demonstrating that the physical properties of the microenvironment can also regulate the proliferation and differentiation of stem cells (61). Considering the plasticity of stem cells to differentiate into diverse cell types, the ability of stem cells to sense and tolerate the dynamic mechanical environment and to fine-tune the mechanoresponses against multiple and constantly changing physical stimuli implies the existence of underlying cross-talk signaling networks (115). In this review, we focus on the mechanical control of stem cells and discuss the possible mechanoreceptors on the basis of the knowledge derived mainly from the studies of somatosensory cells. Understanding the mechanism of mechanotransduction, especially the initial mechanosensing upon stimulus, could enable us to manipulate the differentiation and proliferation of stem cells both in vivo and in vitro physically. In addition, drug development for cell fate regulation will be more realistic once the specific mechanosensors of stem cells are identified. Moreover, due to their multilineage differentiation potential, stem cells can also be a rich model system for the study of mechanobiology.

Cytoskeleton-Related Mechanotransduction in Stem Cells

Signaling Molecules

Mechanical stimuli are first physically sensed by mechanical sensors around the cell membrane interface, where the stimuli pass from the extracellular environment to the intracellular cytoskeleton systems. Unlike rigid objects, biological cells are considered as prestressed materials with dynamic organization of the cytoskeleton. Therefore, the rearrangement of cytoskeleton upon physical stimulation also provides internal stress as feedback against the external forces. The dynamic balance between internal and external strains provides spatial and temporal cues to switch on and off the biochemical signal cascades, which may be through secondary messengers or force-induced conformational changes of the associated enzymatic proteins. With transmission by cytoskeleton, the external signal could therefore quickly pass to distant parts of the cells without the limitation of diffusion rate. Studies using fluorescence resonance energy transfer (FRET)-based c-Src homology domain 2 (SH2) reporter in a living cell have shown that stress-induced Src activation (< 0.3 s) at remote cytoplasmic sites from where the force applied is at least 40 times faster than epidermal growth factor (EGF)-induced Src activation (84). The rate-limiting step of the stress-induced activation could be conformational changes of the enzyme, and the activation depends on the prestressed cytoskeleton, whereas the relay of growth factor-induced activation is limited by diffusion rate.

In response to the mechanical stimulation, cells start to produce biochemical signals in order to tune their physiological function by either biochemical reactions or epigenetic modification via sending the signals across the nucleus (17,107,120,128). A number of signaling molecules have been reported to be involved in cytoskeleton-related mechanotransduction. Src, a nonreceptor tyrosine kinase located at focal adhesions and an integrin–cytoskeleton interaction regulator, has been shown to play important roles in cell adhesion and intracellular stress redistribution (6,21,49,121). Focal adhesion kinase (FAK), another nonreceptor tyrosine kinase, is involved in mechanosensing of shear stress and matrix rigidity (106,130,131). Members of the mitogen-activated protein kinase (MAPK) family, such as extracellular signal-regulated kinase (ERK) and c-Jun N-terminal kinase (JNK), are related to cell stretch and compression (30,44,52,68). Rho family guanosine-5′-triphosphatases (GTPases), such as ras homolog gene family, member A (RhoA)/Rho-associated, coiled-coil containing protein kinase (ROCK) complex, and ras-related C3 botulinum toxin substrate 1 (Rho family, small GTP binding protein Rac1; Rac), are also crucial in mechanosensing of matrix stiffness, cell shapes, and cytoskeleton tension (42,54,80,96,106,126). These signaling molecules have been shown to play important roles in stem cell proliferation and differentiation. Moreover, it has been proposed that these signaling molecules may work together as a complex network in various forms, depending on the subcellular location or the status of cells (29,41,104). The complex network could provide cells the flexibility to finetune cellular responses to stabilize themselves under various physiological conditions.

Calcium Ions

Calcium ions play an important role in cellular physiological function such as muscle contraction and cell movement (13). Upon ion channel activation, calcium is mobilized into the cytoplasm either from outside the cell membrane or from intracellular reservoirs including the endoplasmic reticulum, sarcoplasmic reticulum, and mitochondria. In addition, calcium is an indirect byproduct from enzymatic activities like phospholipase C. As a consequence of the intracellular calcium increase, the calcium acts as a second messenger by allosterically binding to proteins including calmodulin, troponin-C, annexin, and calpain, and then the protein complex triggers downstream cellular processes (33,123). Changes in intracellular free calcium concentration are one of the first responses against environmental stress and are an important biological signal in mechanotransduction (109,118). Interestingly, it has been reported that the magnitude and frequency of calcium oscillations is regulated by substrate rigidity via the RhoA pathway in human mesenchymal stem cells (MSCs, also referred to as marrow stromal cells) and sometimes are not necessarily correlated with cytoskeleton activities (55,57). Calcium signaling is also related to mechanical forces including compression, stretch, and shear stress in different cell types (36,72,88,98,100). The response time of calcium influx upon mechanical stimuli as well as the frequency and amplitude of the calcium oscillation could be rational parameters to analyze the underlying mechanotransduction mechanisms (72,118).

Types of Mechanical Forces in Mechanotransduction

Substrate Stiffness

Most tissue cells have to attach to a solid surface or communicate with proximal cells to maintain their viability, although exceptions include cells derived from the hematopoietic lineage or certain tumor tissues. Tissues like brain, muscle, and bone represent different substrate stiffnesses, and their elasticity is contributed by the composed differentiated cells as well as the surrounding ECM. The Discher's and Wang's laboratories have carried out a series of experiments to demonstrate that cells feel and respond to substrate stiffness, and therefore, stiffer substrates induce stiffer cells (7,22,74). For example, MSCs differentiate into neuron-like cells when cultured on soft substrates mimicking brain tissue. On the contrary, rigid substrates with stiffness similar to that of collagenous bone matrix direct MSCs toward the osteogenic fate (22,26). This suggests that elasticity regulation toward the osteogenic phenotype is likely to be through ROCK and FAK signaling. α₂ integrin has been shown to contribute to elasticity sensing as well (106). Differentiation regulation by substrate stiffness is also observed in embryonic stem cells (ESCs) and neural stem cells (NSCs) (28,63). Scaffolds with various stiffness thresholds can direct ESCs to differentiate into different germ layers (132); NSCs prefer to differentiate into oligodendrocytes instead of astrocytes when cultured on stiffer substrates.

Recent work performed by the Engler lab demonstrated that, similar to shear flow, stiffness variation, not just overall stiffness, is also an important factor in regulating MSC fates (116). Instead of being stationary, cells tend to migrate to stiffer substrates but retain the memory of the substrate elasticity in which it used to reside. The durotaxis of stem cells indicates a potential sensing mechanism. Nevertheless, whether the destination of differentiation is directed by the mechanotransduction of stiffness starting from a mechanosensor, perhaps integrin, or is directly associated with cytoskeleton arrangement due to the balance of intracellular tension, needs to be further elucidated.

Cell Morphology

Other than substrate elasticity, mechanical forces capable of changing cell morphology may also be sufficient to direct stem cell differentiation under certain culturing conditions. Mouse embryonic MSCs are inclined toward smooth muscle myogenesis upon elongation when confined within an area of 20 μm in diameter (area of 314 μm²) in contrast to that of 10 μm in diameter (129). Chen and colleagues also showed that human MSCs prefer adipogenic commitment when confined within an ECM patch of 1024 μm² and are biased toward osteogenesis when cultured within an ECM patch that is 10 times bigger (80). The switch between commitments of chondrocytes and smooth muscle cells is also observed following a similar setup of ECM size. It is proposed that cell shape regulates MSC fates swinging between osteoblasts and adipocytes through RhoA signaling. Similarly, the commitment of either chondrocytes or smooth muscle cells is mediated by cell shape, Rac1, and N-cadherin (31).

One plausible mechanism of matrix stiffness sensing is that cells have to generate intracellular tension in order to counterbalance the substrate stiffness in case of mechanical perturbation. A cell generates intracellular tension via cytoskeleton rearrangement and strengthened focal adhesion in response to stiffer substrates (10,54). As a consequence of cytoskeleton arrangement, a cell changes its morphology and spreading area. Changes in cell morphology also modulate cytoskeleton and focal adhesion reorganization. However, in contrast to previous studies observed in conventional two-dimensional (2D) cell culturing, the Mooney lab used 3D alginate matrices covalently coupled with integrin-binding peptide to encapsulate stem cells and demonstrated that cell fate was not correlated with morphology in 3D microenvironments but with matrix elasticity. The reorganization of adhesion clusters on the cell–matrix interface was traction dependent and correlated with MSC osteogenic commitment. The work indicates that integrin-mediated adhesion bonds are morphology-independent sensors of elasticity and dimensionality (48,58).

Stretch and Compression

Compression and tension (stretch) forces provide dynamic cues for proliferation–differentiation balance and directing fate commitment of stem cells, compared with the static mechanical control cues of substrate stiffness and size of the ECM patch. Compression forces may be from cell–cell contact as well as from growth resistance against the boundary contributed by neighboring tissues, different types of cells, or healing wounds; tension forces occur when tendon, muscle, and epithelium respond to mechanical loading and movement. Stem cells have been shown to have the ability to distinguish between compressive and tensile strains and respond accordingly in terms of morphology and gene expression changes (38). Moreover, variance in orientation, magnitude, frequency, and duration of the compressive or tensile forces could have cells responding differently. The elasticity of substrates where cells are cultured also affects the degrees of morphology changes when forces are applied on a cell population (114). Dynamic compression has been shown to upregulate genes related to chondrogenesis of MSCs cultured in gel scaffolds (39,83,93). On the contrary, applying a 5% increase in cyclic tensile strain has been shown to promote the osteogenic differentiation of MSCs (46,122). Cells even favor a tendon/ligament-like phenotype when experiencing higher tension forces (23). Moreover, recent developments in techniques of micromanipulating mechanical loading on a single cell, such as optical tweezers, magnetic tweezers, and atomic force microscopy, enable further understanding of force-related signaling processes in living cells at the single-cell level (61).

Fluid Shear Flow

Physiologically, the fluid flow comes from blood flow and interstitial fluid. For example, during early development, the intake of amniotic fluid into the fetal lung generates fluid flow; blood flow in blood vessels generates shear force; and the mechanical loading of bones perturbs interstitial fluid flow around cartilage as well as in bone marrow. To date, in addition to embryogenesis, the effects of shear stress to the pathophysiology of vascular endothelial cells and to the differentiation of stem cells have been studied extensively (11,110). Shear stress has been shown to regulate osteogenic, chondrogenic, and endothelial differentiation of MSCs as well as promote endothelial and cardiovascular commitment of ESCs (1,2,4,16,59). It is also suggested that the shear-flow promotion of cardiovascular commitment is via histone modifications (50).

The application of shear stress in vitro is to generate a fluid flow on attached cells cultured as a monolayer. Therefore, the force is first sensed at the cell–fluid interface, not at the cell–cell or cell–matrix interfaces where focal adhesion or cell junction is involved, even though the latter two interfaces cannot be excluded from the mechanotransduction of fluid shear stress in vivo. In other words, shear stress could be considered as one type of tension strain that is mildly acted on by the cell– fluid interface. It has been reported that a cell changes its morphology by elongating and aligning along the shear flow (87). Moreover, a cell senses differences in fluid flow such as flow rate, flow viscosity, and flow patterns as well as types of fluid flow (45). For example, laminar flow promotes wound healing and cell migration of endothelial cells. On the other hand, disturbed flow increases cell turnover rate and enhances atherogenesis (18,34). A recent review by Chiu and Chien has extensively discussed the effects of shear flow on endothelial cells (11). In addition, MSCs have been shown to respond more to the changes of shear flow than to steady flow, and thus the intermittent fluid flow can better promote the osteogenic differentiation compared to the continuous one (59,69,73). It is worth noting that the Voldman lab used low continuous fluid flow (0.007 dyn/cm²), which is two orders of magnitude lower than what is considered low fluid shear stress for cells, to globally diminish diffusible cell-secreted molecules and demonstrated that autocrine signaling is necessary to continuously remodel the ECM and thus prevent ESC differentiation (5,99). Even though mass transfers of nutrition and oxygen as well as secreted signaling factors are affected by shear flow, and it is challenging to distinguish these biochemical factors from the biophysical ones, accumulating evidence has suggested an overlooked mechanosensing mechanism of flow shear stress.

Mechanosensor Candidates

Cytoskeleton rearrangements for counterbalancing the extracellular mechanical stimuli may be the key step of mechanotransduction (51). However, accumulating evidence has indicated that there are more molecules involved in mechanosensing than just cytoskeletonrelated proteins. Because the mechanical stimulation is transmitted from extracellular regions into the cytoplasm, one may first assume that the sensors are located in the proximity of cell membrane. There are various transmembrane receptors, ion channels, and glycocalyx components buried in or connected to the cell membrane, and most of them are also connected to cytoskeletons, ECM, and other membrane proteins. Furthermore, the composition of phospholipids and membrane-associated molecules changes during the differentiation process of stem cells (91). Therefore, the intrinsic complexity and dynamic properties of the cell membrane make it challenging to pinpoint the mechanosensitive molecules that play roles in regulating stem cell differentiation.

Lipid Membrane

The cell membrane is the outer boundary of the cell and is the first subcompartment encountering the extracellular mechanical forces. The complexity of the cell membrane is no less than that of the cytoplasm and its importance is often overlooked (37). Unlike most plants and prokaryotic cells, which have rigid cell walls as a protection against sudden mechanical stress, eukaryotic cells are more sensitive to mechanical forces due to their membrane flexibility. The properties of the cell membrane, such as membrane fluidity, membrane thinning, curvature of a membrane leaflet, lipid microdomains, and hydrophobic mismatch, may be locally changed upon physical stimuli (35,102,103). In a lipid monolayer model, the binary mixers of phospholipids have lower shear viscosity (higher fluidity) compared to the single-component monolayer. In addition, cholesterol, a major component of lipid rafts, may act as a fluidity regulator (27). In osteoblastic cells, lipid rafts are speculated to play an important role in transducing shear stress into cellular responses, and the depletion of membrane cholesterol inhibits oscillatory fluid flow-induced intracellular calcium mobilization as well as ERK phosphorylation (127). Furthermore, there are several studies suggesting the links between membrane cholesterol and activities of shear stress-sensitive ion channels as well as volume-regulated ion channels (47).

Although the membrane fluidity may not be sufficient to elicit the conformational changes of mechanosensitive proteins upon physical stimuli, it is thought to nonspecifically regulate the response time from the stimulation onset to the protein activation taking place. In addition to fluidity, the enthusiastic studies of bacterial small- and large-conductance mechanosensitive (MscS and MscL, respectively) channels reveal a detailed gating mechanism resulting from lipid bilayer deformation upon mechanical stimuli (94,95,119). MscS and MscL are the two channels identified in Escherichia coli by Kung and colleagues that are responsible for the mechanoresponse upon osmotic stress to prevent cell lysis (79,111). The two types of channels have different conductance and share no significant similarity in sequence and conformation, but both have been suggested to transduce the lateral tension forces via lipid–protein interaction by tilting the transmembrane helices to generate a permeable pathway through cell membrane. For example, hydrophobic mismatch of lipid– protein interface destabilizes the closed state of MscL, and then asymmetry in transbilayer pressure caused by the membrane curvature further activates the channel. Similarly, asymmetric incorporation of lysophospholipid keeps MscS in the open state. Moreover, the tension thresholds of channel opening are around 12 dyn/cm² for MscL and 5 dyn/cm² for MscS (94), which also fall within the physiological range of eukaryotic cells. Yet, there are no homologous mechanosensitive channels linked to osmolarity shock identified in eukaryotic systems; the activities of several putative eukaryotic mechanosensitive channels, such as tandem pore domains in a weak inward rectifying K⁺ channel (TWIK-1)-related K⁺ channel (TREK-1) and TWIK-related arachidonic acid-stimulated K⁺ channel (TRAAK), are found to be directly regulated by membrane curvature (77). Additionally, two recent studies provide evidence that mechanosensitive channels act as bridges to link membrane tension and biochemical signaling: MscS C terminus homolog PII-associated membrane protein A (PamA) is related to nitrogen and sugar metabolism (89); MscS homologs MscS-like 2 (MSL2) and MSL3 were recently found to play roles in regulating the formation of filamenting temperature-sensitive mutant Z (FtsZ) ring, which is a prokaryotic homolog to eukaryotic protein tubulin (125).

Ion Channels

In addition to the gating of mechanosensitive ion channels by the cell membrane, there are two other possible mechanisms through which the mechanosensitive channels are activated by mechanical stimuli: activation by direct physical interaction and gating by extracellular membrane components or by cytoskeleton. These three mechanisms are not mutually exclusive and may cooperate with each other. The extremely fast response (40 ms) of inner-ear hair cells indicates direct ion channel activation upon mechanical force (14). However, the microsecond-to-millisecond response time resulting from protein conformational rearrangements makes it challenging to observe channel activities in situ and in real time. Patch clamping is a conventional technique for studying ion channel activities, but the gigaohm seal for recording could shield the channel membrane from certain external mechanical perturbations such as shear stress and could alter membrane tension due to changes in membrane geometry (35). Alternatively, mechanical forces regulate the channel-associated cytoskeleton or ECM components; therefore, movements of these components couple the gates of ion channels via propagation of conformational changes, resulting in switching between channel open and closed states. By this mechanism, ion channels play more passive roles in mechanotransmission, and the response time is expected to be longer compared to that by direct physical activation.

Several ion channels have been reported as mechanosensitive ion channels in eukaryotic systems. Mechanosensory abnormality protein 4 (MEC4) and MEC10 of degenerin/epithelial sodium channel (DEG/ENaC) superfamily form a functional complex together with MEC2 and MEC6 and are responsible for Caenorhabditis elegans light touch sensation (8). No mechanoreceptor potential C (NOMPC) channels, a member of TRP channel subfamily N (TRPN) expressed in Drosophila melanogaster, C. elegans, and zebrafish, share conserved but undefined roles in mechanosensation among different species. The ankyrin repeat motifs of NOMPC have been hypothesized as the gating spring for mechanosensing, and the same motif is found in TRP channel subfamily A (TRPA) as well (43). There are quite a few types of channels in mammals that had been proposed to be sensitive to mechanical forces. However, either their roles in mechanotransduction are undefined or the inconsistent experimental results in heterologous cell systems as well as in the animal models left the channel functions in need of further confirmation. The discrepancy may be due to the following: (1) Protein expression is too low in heterologous cell systems to study. (2) The channels of interest require other proteins or components to function, which are not present in heterologous cell systems. (3) Functional redundancy exists; therefore, gene disruptions have only mild effects in animal models. (4) The channels are mechanical modulators but are not primary mechanosensors. These channels include but are not limited to acid-sensing ion channels (ASICs) and TRP channels such as TRPA1, TRPC1, and TRPV4. For example, ASIC2, a member of the DEG/ENaC superfamily, is a putative mechanosensor and has been shown to be responsible for the arterial baroreceptor reflex, but loss of ASIC2 protein does not cause touch defects in mice (75). TRPV4, a member of the TRP channel subfamily V (TRPV), is an osmotransducer, but disruption of TRPV4 gene only shows modest effect on mechanosensing in mice (70). Additionally, two-pore-domain potassium channels, TREK1, TREK2, and TRAAK [also known as potassium channel, subfamily K, member 2 (KCNK2), KCNK10, and KCNK4, respectively], are mechanically gated ion channels in vitro that are related to membrane stretch and crenation, but their roles in mechanosensory as well as mechanotransduction are still unknown (3,19).

Excitingly, the recent identification of piezo-type mechanosensitive ion channel component 1 (Piezo1) and Piezo2 in a mouse neuroblastoma cell line as the components of mechanically activated cation channels provides direct evidence that the ion channel is one of the mechanosensors in mammals (15). Piezos are multipass transmembrane proteins, and Piezo1 has been shown to be involved in integrin activation. Piezos can be activated by compression forces introduced by piezoelectrically driven probes and are responsible for the mechanically activated cation currents. Even though the channel function of Piezos still needs to be confirmed due to the lack of a putative pore-forming domain, it is expected to act as an accessory subunit of the mechanosensitive channel—if it is not the channel itself—which is similar to the β subunit of the voltage-activated channel.

In human MSCs from bone marrow, ion channel genes that are highly expressed include but are not limited to potassium voltage-gated channel, shaker-related subfamily, member 4 (Kv1.4); potassium voltage-gated channel, shal-related subfamily, member 2 (Kv4.2); potassium voltage-gated channel, subfamily H [ether-a-go-go (eag)-related], member 1 (heag1); potassium large-conductance calcium-activated channel, subfamily M, alpha member 1 (MaxiK); hyperpolarization-activated cyclic nucleotidegated potassium channel 2 (HCN2); and the L-type calcium channel (40,66). Moreover, the expression patterns of ion channels in pluripotent cells are species and origin dependent. The expression pattern varies in mice, rats, and rabbits and is also distinct in adipose tissue-derived, bone marrow-derived, and umbilical cord-derived MSCs. Species and origin variations are reported in ESCs as well (65,92). Even though the variation may result from discrepancies in culturing conditions and the cells' undifferentiated states, it also suggests stem cell plasticity by which ion channel gene expression can respond to slight differences in physical and chemical environmental cues. More specifically, a recent study has suggested that ion channels play roles in stem cell fate determination. Inhibition of voltage-gated potassium channels decreases ESCs in S phase and increases cells in G₀/G₁ phase. In addition, Kv channel blockage depolarizes the membrane potential, which favors ESCs to differentiate (85,113). Even though there are several putative mechanosensitive ion channels expressing in stem cells, such as ASICs, TREKs, TRAAK, TRPMs, TRPVs, TRPCs, and several Cl^– channels [(3) and unpublished data], the link between channel-related mechanotransduction and fate determination in stem cells remains elusive.

Primary Cilia

Primary cilia are thin, hair-like organelles that have been viewed as cellular antenna that sense biochemical and biomechanical signals and integrate these stimuli into determination of left–right asymmetry, proliferation, or differentiation (124). In vertebrates, nearly every cell is ciliated at some point during its life cycle. When cells are confluent and reach G₁ phase, primary cilia are formed from the cells' mother centrioles, anchored to the basal body, and then maintained in G₀ phase for differentiation (64). The cilia have been shown to play roles in paracrine, wingless-type mouse mammary tumor virus (MMTV) integration site family (Wnt), as well as sonic hedgehog (Shh) signaling, and its malfunctions are now linked to ciliopathies, such as polycystic kidney disease (PKD) and Bardet–Biedl syndrome (BBS) (32). In mouse embryos, the left–right asymmetric development is established by extraembryonic fluid flow, which is generated and sensed by the cilia (81). In endothelial cells (ECs), the primary cilia are present in the areas of low or disturbed flow. They are able to sense mild fluid shear as low as 0.0007 Pa and regulate calcium signaling as well as nitric oxide production (97). In the embryonic heart where high flow occurs, cells do not present primary cilia and coincidentally undergo endothelial-to-mesenchymal (EndoMT) transition. Correspondingly, shear-induced EndoMT transition has also been observed in the mutated nonciliated mouse embryonic ECs in response to shear stress as low as 0.5 Pa in vitro (25).

There are more than 200 ciliary proteins and many of them are receptors or components of signaling complexes (32,62). For example, both polycystin-1 and polycystin-2 are expressed in primary cilia and are members of transient receptor potential polycystic (TRPP), a subfamily of TRP ion channels (86,105). Polycystin-1 is a glycoprotein containing numerous N-terminal PKD domains, which are involved in cell adhesion and cell–cell interactions. The C terminus of polycystin-1 is proposed to interact with polycystin-2 and form a flow-sensitive receptor–ion channel complex in epithelial and endothelial cells. In addition, the polycystins are also proposed to be regulators of unidentified stretch-activated cation channels (SACs) in arterial myocytes (105). Quantitatively, if the channels of interest are confined in the primary cilia, which have subfemtoliter volume, even activating one single channel can introduce a dramatic change in local ion concentration and therefore trigger the downstream signaling cascades with high sensitivity.

Integrin and Integrin-Associated Networks

Integrin has been considered as a plausible candidate of mechanosensors for a long time because it physically connects the ECM and cytoskeleton as well as acts as a signal transducer bidirectionally across cell membrane (78). Heterodimeric integrin changes its conformation to the active state under mechanical stimuli, such as mechanical stretch, and interacts with the ECM proteins (53). The formation of integrin–ECM complex therefore promotes the integrin cytoplasmic domain to interact with cytoskeleton and other focal adhesion proteins, such as paxillin, talin, viculin, and α-actin, as well as the formation of stress fibers (71,82). The dynamic formation of the adhesion complexes upon mechanical stimuli triggers the intracellular signaling cascades, which are initiated by FAK and Src family kinases, and therefore regulates cellular migration, proliferation, and differentiation (112).

β₁ and α₆ integrins have been considered as stem cell biomarkers that can support the tissue-specific stem cell niches (101). A recent study showed that stereotopographic cues are able to regulate osteogenic differentiation of hMSCs, which was correlated with the expression of α₂/β₁ integrin heterodimers (60). In addition, α₂ integrin was upregulated on stiffer matrices during osteogenic induction, and gene knockdown downregulated the osteogenic phenotype of hMSCs (106). It is well accepted that the integrin is a mechanotransmitter, but the mechanosensing as well as the mechanism of activation are still elusive (108). A mechanosensory complex platelet endothelial cell adhesion molecule 1 (PECAM-1)/vascular endothelial (VE)-cadherin/vascular endothelial growth factor receptor 2 (VEGFR2) upstream of integrin activation was reported in endothelial cells as a downstream event of shear stress (117). Moreover, a recent study showed that a soft substrate enhances caveolae/raft-mediated integrin internalization and the inhibition of the internalization blocks MSC neurogenic differentiation (24). These studies indicate the possibility that there are other mechanosensing units expressed in stem cells in addition to integrins.

Concluding Remarks

The mechanosensors are directly activated by mechanical forces within milliseconds, not by secondary messengers or other proteins. The sensors should have mechanically sensitive domains and participate in mechanical transduction (35). In undifferentiated stem cells, mechanotransduction includes the signaling pathways related to fate commitments and is associated with cytoskeleton and nucleoskeleton rearrangements. Herein, mechanosensors act as pivot points connecting input extracellular forces and the output forces created by the corresponding cytoskeleton rearrangements. The sensors are expected to be mechanical force type specific, and perhaps magnitude as well as location specific. Integrins, ion channels, and perhaps other membrane-associated proteins like G protein-coupled receptors are candidates of mechanosensors, sensory complexes, or part of ECM–integrin–cytoskeleton linkages. With the accumulating knowledge of signaling networks in cytoplasm, mechanosensing seems to be one of the missing pieces in the puzzle toward the understanding of stem cell fate determination upon mechanical stimuli. In addition, identification of the mechanosensors and understanding their activation mechanisms could shed the light on whether the stem cell fate choice is a deterministic or a stochastic process. Moreover, because mechanical environment is an important niche for tissue engineering and regeneration, identification of mechanosensors of stem cells can not only elucidate the mechanisms of mechanotransduction, but also bring new prospects of the mechanical control of stem cells as well as drug development for clinical application.