Many cell types respond to forces as acutely as they do to chemical stimuli, but the mechanisms by which cells sense mechanical stimuli and how these factors alter cellular structure and function in vivo are far less explored than those triggered by chemical ligands. Forces arise both from effects outside the cell and from mechanochemical reactions within the cell that generate stresses on the surface to which the cells adhere. Several recent reviews have summarized how externally applied forces may trigger a cellular response (Silver FH and Siperko LM. Crit Rev Biomed Eng 31: 255–331, 2003; Estes BT, Gimble JM, and Guilak F. Curr Top Dev Biol 60: 91–126, 2004; Janmey PA and Weitz DA. Trends Biochem Sci 29: 364–370, 2004). The purpose of this review is to examine the information available in the current literature describing the relationship between a cell and the rigidity of the matrix on which it resides. We will review recent studies and techniques that focus on substrate compliance as a major variable in cell culture studies. We will discuss the specificity of cell response to stiffness and discuss how this may be important in particular tissue systems. We will attempt to link the mechanoresponse to real pathological states and speculate on the possible biological significance of mechanosensing.
- matrix compliance
- extracellular matrix
- polyacrylamide gel
cells are subject to a variety of mechanical forces, such as the shear stress due to blood flow over the endothelial cells lining an artery or the compression on chondrocytes in knee cartilage that are impacted as a person runs. Cells are highly dynamic and can reorganize their cytoskeleton and functions in response to these external mechanical signals. Studies of mechanosensing in numerous systems have established that cellular response to mechanical cues can have as large an influence on structure and function as chemical signals. Mechanical force can affect cellular locomotion, morphology, adhesion, and cytoskeletal protein expression (18). Mechanical influences include, but are not exclusive to, fluid shear stress, pressure, and elongational stresses; such forces can be caused either by external effects, such as gravity and impacts with solid objects, or by internally generated forces due to muscle activation or osmotic imbalance. How these forces are distributed on the surface and within the interior of the cell depends not only on the material properties of the cell itself but also on the properties of the surface or matrix to which it attached (6). The cell has adhesion points called focal adhesions that are anchorage points to the substrate on which they lie. By tugging on the matrix at these focal adhesions, the cell creates a tension within its membrane walls (4). The tension that the cell is able to generate depends on the inherent material properties of the matrix: a relatively stiff matrix will resist cellular force more than a soft one, causing the cell to be more rigid and extended about its periphery (36). Various bioengineering approaches have increased our understanding of how matrix mechanical characteristics can affect organization and protein expression in cells. The scope of this review is the evaluation of research describing a role of extracellular matrix stiffness in cell motility, adhesion, and other cytoskeleton-dependent activities. The response specificity of various cell types will be examined, as will the role this specificity may play in the context of tissue disease and injury states.
TEST SYSTEMS: SOFT MATERIALS FOR CELL CULTURE STUDIES
Several systems have been developed to study the influence of outside mechanical forces on individual cells; the category that is the focus of this review varies matrix mechanical properties and examines their effect on cell migration, growth, and cytoskeletal organization.
Extracellular matrix and other natural hydrogels.
Protein-based extracellular matrix gels, such as fibrin, collagen, or a mixture of collagen, laminin, and other proteins forming Matrigel, are commonly used to create two- or three-dimensional cell culture substrates of controlled stiffness. Cross-linked polysaccharides such as alginate and agarose gels can also be manipulated to have varying elastic moduli by altering the polymer mass and are permissive for cull culture.
Solutions of proteins can be induced to polymerize either by addition of specific enzymes or by raising the temperature of the solution. By increasing the total concentration of protein, the resulting stiffness of a polymerized network is increased, with the elastic shear modulus approximately proportional to the square of the protein concentration (28). Gels that are anchored to a surface provide more resistance than free-floating gels to forces generated by cells that are plated on them (21). Other factors can also be adjusted to fine tune the material properties of a biological gel. For example, the addition of fibronectin can increase the tensile strength of collagen matrices (19), and salts can stiffen fibrin matrices at physiological pH, as can activation of the plasma transglutaminase factor XIII (8, 30, 45). The ability to incorporate or attach peptides with specific recognition sites for cell surface proteins allows an additional control of the types of transmembrane complexes, usually integrins, by which the cell engages the extracellular matrix.
Fibroblasts in collagen gels.
It is well documented that fibroblasts embedded in collagen gels show distinct morphologies from those cultured on tissue-culture plastic (20, 23). Cells on free-floating collagen gels lack the pronounced F-actin stress fibers that are seen on tissue-culture plastic. Fibroblasts on constrained collagen gels are able to generate stress fibers (21). The mechanism by which stress fibers are reduced in fibroblasts on gels may involve focal adhesions. Focal adhesion complex proteins are downregulated on both collagen gels and Matrigel, although not on plastic dishes coated with a solution of either of the two (47). On unconstrained gels, fibroblasts are also less susceptible to transforming growth factor-β-stimulated smooth muscle α-actin production, a hallmark of contractile behavior. This finding provides some evidence that matrix stiffness appears to be a key component of contractile behavior and associated protein expression (1).
Synthetic substrates: ligand-coated polyacrylamide gels.
Polyacrylamide (PA) gels have emerged as important tools for testing the compliance dependence of cytoskeletal-regulated activities of cells. Of significance for this work is the ability to separate the chemical signals received by cells from the mechanical signals. Protein and polysaccharide gels can interact directly with the cell surface or bind serum proteins that then act as cellular ligands in a manner that is difficult to control or quantify. In addition, manipulations that alter gel stiffness also alter parameters such as fiber thickness or density that could impact cellular response independent of stiffness changes. PA gels have two important features that allow separation of chemical from mechanical signaling. The gel itself is nearly completely inert as an adhesive surface. The same chemical stability and nonadherence to other macromolecules that enable its use for electrophoretic separation of proteins and nucleic acids also ensure that neither cell surface receptors nor adhesive proteins present in serum can bind directly to the gel. Therefore, only those adhesive molecules covalently grafted to the gel surface can act as ligands for the cell. The second convenient feature of PA gels is that their stiffness, quantified by an elastic modulus, can be varied over a wide range by changing the small fraction of dimeric bisacrylamide that cross-links PA chains, while keeping constant the polymer concentration and so avoiding changes in the surface texture or distribution of surface ligand sites. A disadvantage of PA is that its chemical inertness makes covalent attachment of fragile proteins sometimes difficult, but a variety of chemistries has been developed to allow more facile conjugation of adhesion molecules to the gel surface (34, 48).
A pioneering use of PA gels as culture substrates was reported in 1978 (35) by creating N-acetylglutaminase-coated gels to show that chicken hepatocytes specifically bound the gels, whereas other cell types were unable to engage this glycopeptide. Another early study examined the conjugation of 3T3 fibroblasts to PA gels derivatized at their surfaces with carboxylic acids and hydroxyl groups (7). This study also used the PA gel strictly as a substrate on which specific extracellular molecules could be isolated as adhesive conduits. In these studies, the PA concentration was very high (30%), and the modulus, although not reported, can be estimated as >30,000 Pa (13). It is likely that such a high stiffness was needed to produce cells that most closely resembled those grown on plastic.
The first systematic study using PA gels of different stiffness was undertaken by Pelham and Wang (32). The first of these studies also described PA gels that were chemically inert to cells until coated with a ligand, in this case collagen, via a modified surface-coating technique (32). By altering the cross-linker concentration of the acrylamide, the stiffness was varied by an order of magnitude while keeping the total polymer concentration constant. With the use of NIH3T3 fibroblasts as well as normal rat kidney epithelial cells, compliance-dependent changes in both motility and cytoskeletal adhesion were detected. Most notably, the normal rat kidney cells on soft gels were unable to assemble focal adhesions at their periphery, but their lamellipodia were more motile. Tyrosine kinase substrates are implicated in this difference because cells treated with the tyrosine phosphatase inhibitor phenylarsine oxide overcome the soft-gel phenotype and regain focal adhesions (32).
TRACTION MAPPING ON DEFORMABLE SUBSTRATES
Deformations of the substrate created by tugging from cells had previously been documented by visualizing wrinkling of thin silicone membranes (22), and recent studies have shown that the elasticity of such membranes can be varied by changing the time that the membrane is heated (25). A limitation of this system is that stiffness needs to be measured separately for each sample by use of calibrated glass needles to calculate an effective stiffness quantified as the force (in nN) needed to deform the membrane a given distance (in μm) (25). In addition, although in principle the force on a membrane of known stiffness is related to the wavelength of the wrinkles (9), deformations caused by a cell that exerts variable forces in different directions are not likely to be quantifiable in this system.
PA gels can provide higher precision force mapping because the displacement of fluorescent microbeads embedded in them can be accurately visualized in two dimensions. The use of PA gels was enhanced by development of a cellular traction force mapping system (11). Traction force mapping was used, for example, to determine which cytoskeletal element was predominantly responsible for mechanical forces exerted by fibroblasts (33). By administering micromolar quantities of cytochalasin D, a drug that disrupts the actin cytoskeleton, nearly all measurable traction forces disappeared, suggesting the forces were mostly F-actin dependent. Focal adhesion immunofluorescence showed elongated structures at sites lining strong traction force areas, further implicating focal adhesions as the “roots” of cell tension by stabilizing microfilament structures. Furthermore, focal adhesion kinase-null fibroblasts are unable to mechanosense; they do not move in response to a pulling or pushing, and they migrate between soft and stiff substrates with no apparent preference. Nocodazole, a substance known to destabilize microtubules, had no effect on cell traction, suggesting that microtubule-generated forces do not contribute significantly to forces on the substrate.
Whether microtubules contribute to traction forces on soft substrates may depend on the cell type and the nature of the assay. A tensegrity-based model describes cell contraction as a balance between the external traction forces at the substrate-cell interface with a microtubule-dependent intracellular prestress (41). Evidence to support this model was also derived from use of traction force mapping on PA gels plated this time with human airway smooth muscles cells. Substrate traction mapping after cells were treated with colchicine, another microtubule-destabilizing drug, resulted in increased traction force on the substrate and an increase in strain energy stored in the substrate. The authors use this evidence to suggest that microtubules could act as compression-bearing elements in cell contraction.
Traction force mapping of fibroblasts on PA gels proves that they are able to pull harder on relatively stiffer substrates (27). If a cell can generate higher traction, it has some internal structure that can sense the stiffness of the matrix on which it resides. Fibroblasts also preferentially move toward stiffer areas of a gel as well as areas of a gel that are compressed, a phenomenon analogous to chemotaxis in which a cell can move directionally toward a mechanical signal. The movement has been dubbed “durotaxis.” This finding may have relevance to areas from development to disease where cells migrate to various regions of tissue systems that may differ in material characteristics.
Micropatterned elastomer substrates and theoretical modeling.
Another effective method to track the forces a cell exerts on a substrate uses elastomer substrates that are micropatterned with a design visible in phase contrast (2). A phase-visible pattern allows simultaneous observation of multiple fluorescent cytoskeletal elements like focal adhesions and actin, which can then be correlated to substrate movement. Forces from fibroblasts align in the direction of elongated focal adhesions, and displacement of the substrate is never seen without strong focal adhesion assembly. This focal adhesion assembly apparently follows force on the substrate, which may be driven by actomyosin contraction (2).
Experimental data of traction forces facilitate predictive modeling of forces. Modeling the cellular reaction to substrate stiffness is achieved by making simplifying assumptions about the extracellular matrix and then treating the material as a spring that is stretched by cellular forces (5). First, the substrate material is presumed to be isotropic, meaning its properties are the same in all directions. This is a fair assumption for elastomer, PA, and even elastin substrates, but probably is inaccurate for other biological gels like collagen and fibrin formed by stiffer filaments, which are anisotropic when deformed or when polymerized in the presence of external forces (12). An isotropic extracellular matrix can be mathematically modeled with Hooke’s law as a spring with a spring constant k that exerts a force F proportional to the distance, x, by which it is extended, following the expression F = −kx. The spring constant, k, is related to the material’s shear modulus, a constant for the material that describes its resistance to shear, expressed as stress over strain. The forces on the spring are actomyosin driven, and the actin filaments are anchored at focal adhesion complexes. The model suggests that a cell may be targeting an optimal force by exerting energy (W), which is expressed as W = F2/2K. A stiffer ECM with a higher K value will assist the cell in reaching the desired force with less invested energy. Conversely, the cell may be targeting a certain displacement of the substrate to communicate with neighboring cells. In this case, a lower K, or softer gel, would be energetically favorable because matrix displacement would require less energy (Fig. 1).
The research discussed to this point has used mostly fibroblasts as the cell type of choice for testing the influence of physical properties on cell behavior. In all systems studied, there has been convincing evidence that focal adhesion complexes and their related proteins are involved in the traction force generation and cytoskeletal organization, especially F-actin structure, on soft materials. Fibroblasts spread more on stiffer substrates whether the substrate is a biological gel or a protein-laminated PA gel. On softer materials, they adopt a more spherical morphology, and their F-actin structure is markedly more diffuse (Fig. 2). Other cell types, however, respond differently to changes in substrate stiffness.
SPECIFICITY OF CELLULAR RESPONSE TO MATRIX COMPLIANCE
Endothelial cells in vivo are attached to compliant surfaces and generally subjected to mechanical force. Shear stress caused by hemodynamic forces modify endothelial cell function and cytoskeletal structure (24). Although shear stresses from fluid flow are imposed on the apical side of the cell layer, they elicit changes throughout the cell interior. Of more recent consideration is determining what interactions are on the basal surface of the endothelial cell layer between cells and extracellular matrix and how important the ECM is in capillary morphogenesis. Endothelial cells plated on collagen gels (44) or fibrin gels (43) of differing flexibility show a decrease in network-like structures on stiffer gels. Softer substrates allow cells to form long capillary-like tube structures. On stiffer gels, endothelial cells from human umbilical vein are more spread, have larger lumens, and exhibit less branching compared with the same cells on soft gels (39). The authors also reported that human blood outgrowth endothelial cells are able to create greater contractile forces on collagen gels than endothelial cells from human umbilical vein. This finding is consistent with the capacity of human blood outgrowth endothelial cells for capillary morphogenesis on higher concentration collagen gels; these cells are able to withstand greater external resistance from the matrix because they can generate more traction on the ECM. The authors speculate that this specificity of feedback between ECM and the two different cell types may point to local adaptation within a blood vessel to changes in basal ECM properties.
Muscle cells also exhibit substrate compliance-dependent behavior, not surprisingly because their main function involves contraction of tissue. Myoblasts plated on PA gels initially spread more on stiffer substrates, although after a day in culture they are able to spread on both soft and stiff substrates (14). Myoblasts fuse to form myotubes on a range from soft to stiff PA gels patterned with strips of collagen. Only on intermediate stiffness, however, do these myotubes exhibit evidence of striation, as indicated by immunocytochemistry against myosin. Although myotubes on soft and stiff gels are multi-nucleated up through 4 wk in culture, they show poor striations not evident of physiological myotubes (Fig. 3). Remarkably, the modulus of healthy muscle tissue falls within this intermediate range of stiffness, whereas the modulus of diseased fibrotic tissue associated with muscular dystrophy is in the stiffer range on which striations form poorly. This study suggests a link between changes in myoblast behavior and differences in stiffness of healthy and diseased muscle tissue.
Liver-derived cell types.
The liver is another organ in which compliance changes are evident in diseases like fibrosis and cirrhosis. As liver disease progresses, extracellular matrix molecules are deposited and overall stiffness of the tissue increases (26). Hepatic stellate cells located within the space of Disse in the liver are mainly responsible for this ECM deposition after differentiation from a quiescent state to a myofibroblast-like phenotype (29). The mechanism for this differentiation is unknown and has been difficult to study because in vitro cells automatically differentiate on tissue culture plastic. On Matrigel, however, it has been shown that stellate cells can remain quiescent and even revert to quiescent phenotype from a reactive one (17). Although it has been speculated that there is a chemical interaction with the basement membrane that keeps these cells from differentiating, it is possible that this response has a mechanical component as well. Uncross-linked Matrigel has a very low shear modulus, corresponding to, and even lower than, what has been described in this review as a soft gel (<35 Pa), and it is possible that hepatic stellate cells need external resistance to differentiate.
Hepatocytes, the main functional cells in the liver, also exhibit some mechanical-dependent behavior. Hepatocytes plated on Matrigel minimally spread, quickly form spheroidal aggregates, and reorganize the matrix (10). The compliance of the Matrigel can be lowered by cross-linking the gel with glutaraldehyde. On stiffer matrices, hepatocytes become polygonal and do not aggregate as effectively. It is likely that hepatocytes on soft substrates can sufficiently apply forces on the matrix that are felt by neighboring cells, and increasing the substrate stiffness blocks this mechanical communication between cells. Because hepatocytes must aggregate to maintain differentiated function, it is desirable that these aggregates form, especially in the context of functional tissue-replacement design. Additionally, as in the case of myotube formation, there are possibly optimal compliance levels for aggregation cues. Hepatocytes on slightly cross-linked Matrigel are more responsive to growth factor-induced aggregation (37) than on basal Matrigel, which has a lower modulus. Cells on the softest, basal Matrigel substrate are round in morphology and show fewer signs of being differentiated. The basal Matrigel shear modulus was reported as being 34 Pa compared with the slightly cross-linked modulus of 180 Pa, both below the physiological modulus of liver tissue and both relatively soft. The 34-Pa gel, however, may be too soft for hepatocytes to anchor to the substrate and send signals via force transduction to neighboring cells. In summary, current evidence suggests that hepatocytes maintain a differentiated phenotype on soft materials, whereas hepatic stellate cells remain quiescent (Fig. 4). Perhaps mechanical properties that change in liver disease causing a stiffer extracellular matrix for cells to sense are partially responsible for the hypertrophy of hepatic stellate cells and the deterioration of hepatocyte networks.
Neurons and glial cells.
ECM mechanical properties have also been implicated in nerve regeneration in the central and peripheral nervous systems. There has been success in vivo using biomaterial implanted into injury cavities to bridge the gap between severed axons (42, 49). One hypothesis for why hydrogels may be beneficial is that they reduce the level of scar tissue that is considered to be a mechanical barrier at the wound site. Neurons are incapable of crossing dense scar tissue, and this suggests that this cell type may be more able to grow on malleable substrates than on rigid ones, and recent work in the field of neuroscience has supported this hypothesis. The neurite extension of dorsal root ganglion neurons can be correlated to a reduction in the elastic modulus of agarose gels, and a mathematical model can predict the length a neuron will extend based on the modulus of the matrix (3). Primary cultures from mouse spinal cords also show that neurons preferentially branch on soft substrates, in this case Matrigel-coated PA gels (16). In contrast to neurons, glia, the primary cell type in the scar that forms after central nervous system injury, are unable to survive on soft matrices. Perhaps a local mechanical change in vivo in the ECM surrounding the neurons and glia contributes to the response of these cell types to injury that limits neuron regeneration because deposition of ECM molecules stiffens the tissue, making the environment permissive for hypertrophic glia and nonpermissive for neurons, the cell type that needs to be repaired. This differential stiffness preference may also explain why in vivo studies suggest that hydrogel implants are able to inhibit scar formation, i.e., glia are not inclined to migrate into the gels because the mechanical properties are not ideal for their growth, allowing only neurons to extend inward.
Material properties of tissue-engineering constructs have many design criteria, and with the recent understanding of matrix effects on cell function it is clear that matrix compliance is an important variable to control. Optimizing polymer substrates to mimic the complex mechanical attributes of the tissue being replaced is a significant and difficult problem. To achieve this goal, extensive mechanical testing of healthy tissue will need to be performed to properly characterize the compliance, strength, and viscoelasticity of different tissue systems. This may simultaneously advance and complicate the field of tissue engineering as the specificity of response for various cell types becomes more apparent.
Not all cell types appear to be sensitive to substrate stiffness, and not all mechanosensitive cell types respond similarly to changes in substrate stiffness. Most cell types studied thus far spread more, adhere better, and appear to survive better on stiffer matrices, and some cannot grow on very soft (<50 Pa) surfaces. Other cell types, such as neutrophils, appear not to respond to substrate stiffness, at least over the range of 3 to 50,000 Pa that can be accessed experimentally (50). Still other cell types, such as neurons, extend processes more avidly and appear to survive better on soft materials. The contrast between cell-type preferences for certain extracellular matrix flexibilities is significant and could contribute to differentiation of cell-type functionality within a particular tissue. Studying the relationship between physical environment and unique cell types could be important, not only in the case of designing implantable polymers, stents, and liver tissue but also in understanding the pathological sequlae that occur after an insult or in disease progression.
This work was supported by National Heart, Lung, and Blood Institute Grants HL-64388 and HL-007954 and National Science Foundation/Materials Research Science and Engineering Centers Grant DMR-00-79909
- Copyright © 2005 the American Physiological Society