Integrin-mediated cell adhesion is central to cell survival, differentiation and motility. Many cell responses induced by integrins require both receptor occupancy and receptor aggregation, and appear to be regulated by both biochemical and biophysical means. Multidomain extracellular matrix molecules may serve to foster integrin aggregation by presenting local clusters of adhesion ligands, a hypothesis supported by studies with synthetic substrates showing that cell adhesion and migration are enhanced when adhesion ligands are presented in nanoscale clusters. Here, we used a novel synthetic polymer system to present the adhesion ligand GRGDSPK in nanoscale clusters with 1.7, 3.6 or 5.4 peptides per cluster against a non-adhesive background, where the peptide is mobile on a 2 nm polyethylene oxide tether. Average ligand density ranged from 190 to 5270 RGD/μm2. We used these substrates to study the effects of ligand density and clustering on adhesion of wild-type NR6 fibroblasts, which expressα vβ3 andα 5β1, integrins known to bind to linear RGD peptides. The strength of cell-substratum adhesion was quantified using a centrifugal detachment assay to assess the relative number of cells remaining adherent after a 10 minute application of defined distraction force. An unusual relationship between cell detachment and distraction force at relatively low values of applied force was found on substrates presenting the clustered ligand. Although a monotonic decrease in the number of cells remaining attached would be expected with increasing force on all substrates, we instead observed a peak (adhesion reinforcement) in this profile for certain ligand conditions. On substrates presenting clustered ligands, the fraction of cells remaining attached increased as the distraction force was increased to between 70 and 150 pN/cell, then decreased for higher forces. This phenomenon was only observed on substrates presenting higher ligand cluster sizes (n=3.6 or n=5.4) and was more pronounced at higher ligand densities. Adhesion reinforcement was not observed on fibronectin-coated surfaces. These results support previous studies showing that biophysical cues such as ligand spatial arrangement and extracellular matrix rigidity are central to the governance of cell responses to the external environment.
Adhesive interactions between cells and the extracellular matrix (ECM) are governed primarily by integrins, a large family of heterodimeric cell surface adhesion receptors with at least 18 types of α subunits and 8 types ofβ subunits, which in combination generate over 20 integrin heterodimers ( Boudreau and Jones, 1999; Green and Humphries, 1999; Hynes, 1992). Upon binding to extracellular ligands, integrins can associate with a diverse set of cytoplasmic proteins, initiate signaling cascades and form associations with the cytoskeleton to provide physical links between the cytoskeleton and the ECM. The nature of adhesive interactions governs the survival, growth, motility and differentiation of cells ( Bourdoulous et al., 1998; Chen et al., 1997; Davey et al., 1999; Dike et al., 1999; Huttenlocher et al., 1996; Palecek et al., 1997).
Integrin-mediated cellular behaviors are regulated by a variety of biochemical and biophysical means. At a basic biochemical level, integrins differ in their affinities for individual matrix components, allowing cells to regulate the number of bonds formed with the matrix by the number and type of integrins expressed. Integrin cytoplasmic domains show a diversity of functional interactions with intracellular structural and signaling molecules, and thus the type and number of bonds formed send information about the nature of the extracellular environment. Integrins are also subject to inside-out affinity modulation via intracellular signaling pathways set off by growth factor binding, mechanical stress or other stimuli ( Danen et al., 1998; Hughes and Pfaff, 1998; Humphries, 1996; Mould, 1996).
Biophysical regulation of integrin function appears to occur via diverse mechanisms, including the physical arrangement of matrix ligands and matrix rigidity. In forming attachments to the matrix, integrins aggregate in micron-scale complexes. The cytoplasmic domains of aggregated integrins nucleate a complex of structural and signaling molecules that link the integrins to the cytoskeleton, forming a continuous bridge between the cytoskeleton and the matrix ( Critchley, 2000; Giancotti and Ruoslahti, 1999; Schoenwaelder and Burridge, 1999). Several studies have shown that clustering or aggregation of ligated integrins in this manner is essential to elicit the full range of integrin-mediated biochemical signaling and subsequent cell behaviors ( Miyamoto et al., 1995a; Miyamoto et al., 1995b). Integrins restrained from forming clusters via α chain tail mutations show markedly impaired cell adhesion ( Yauch et al., 1997). Conversely, cell adhesion is positively regulated byα IIbβ3 receptor clustering with a concurrent increase in tyrosine phosphorylation of pp72 (Syk) and pp 125 (FAK) ( Hato et al., 1998).
The multimeric structure of ECM molecules such as fibronectin, a dimer with dual adhesion sites, and tenascin-C, a hexabrachion that presents six identical cell adhesion domains within ∼ 100 nm, suggests that receptor clustering may be influenced by the physical layout of the extracellular matrix components. This inference is supported by in vitro studies using synthetic Arg-Gly-Asp (RGD) adhesion peptides ( Danilov and Juliano, 1989; Maheshwari et al., 2000). Peptides grafted to albumin at 1-20 peptides per albumin molecule ( Danilov and Juliano, 1989) or to synthetic star-configured polyethylene oxide tethers at one to nine peptides per polymer molecule ( Maheshwari et al., 2000) are more effective in promoting cell adhesion when the valency is high. Peptides presented singly (1 peptide per molecule) are poor substrates for adhesion, whereas peptides presented at cluster sizes of nine peptides per molecule or higher induce comparable adhesion to matrix proteins ( Danilov and Juliano, 1989; Maheshwari et al., 2000). Further, when RGD adhesion ligands are presented in a non-clustered fashion, fibroblast migration is significantly impaired, even at identical average ligand densities ( Maheshwari et al., 2000).
In addition to sensing ECM spatial organization, cells exert forces on the matrix and respond to the mechanical properties of their surroundings by regulating adhesive interactions. Compared with compliant matrices of identical composition, rigid matrices have been shown to enhance cell-surface assembly of fibronectin ( Halliday and Tomasek, 1995), provide a preferential substrate for directional cell migration ( Lo et al., 2000), regulate the rates of apoptosis and DNA synthesis ( Wang et al., 2000) and are associated with increased levels of protein phosphorylation at sites of cell-matrix contact ( Katz et al., 2000; Pelham and Wang, 1997). Compliant matrices, on the other hand, promote cell motility ( Pelham and Wang, 1997). Cell speeds are also greater on substrates where fibronectin is adsorbed, and thus compliant, than on substrates where fibronectin is covalently immobilized and thus inflexible ( Katz et al., 2000). Since cell movement requires formation and dissolution of focal adhesions ( Greenwood and Murphy-Ullrich, 1998), substrate properties that foster focal adhesion turnover may also favor cell migration.
The integrated molecular mechanisms underlying these observed integrin-mediated responses to ECM properties are being illuminated using approaches that probe mechanical as well as chemical stimuli. A direct role for a coupling of mechanical and signaling factors in the regulation of focal adhesion dynamics has been shown using a permeabilized cell system to control the local molecular environment at individual focal adhesions ( Crowley and Horwitz, 1995). Control of the phosphorylation levels of cytoskeleton-associated proteins using ATP and/or phosphatases and control of cell contraction using a peptide inhibitor of the actin-myosin interaction suggest that both tyrosine phosphorylations and tension mediate the release of adhesions ( Crowley and Horwitz, 1995). In addition to tension-regulated adhesion turnover, cells respond to external mechanical stimuli by modulating the strength of adhesion sites. Adhesion-ligand-coated magnetic beads, twisted in a magnetic field after cell binding to apply defined stresses to integrin-cytoskeletal linkages, induce increased cytoskeletal stiffness proportional to increases in applied strain ( Wang and Ingber, 1995). Localized stress on integrin-ligand bonds, applied by an optical trap pulling on fibronectin-coated beads bound to lamellipodia, leads to reinforcement in adhesion to ECM ( Choquet et al., 1997). Subsequent optical trap studies have revealed that reinforcement mediated specifically through the vitronectin receptor is regulated by the tyrosine kinase Src ( Felsenfeld et al., 1999). Mechanotransduction via integrin-cytoskeleton linkages has also been cited in the process of microtubule assembly in smooth muscle cells, thereby directly affecting cellular structure and phenotype ( Putnam et al., 2001).
We have previously shown that cell adhesion and migration on substrates presenting a minimal RGD sequence, YGRGD, require ligand clustering on the∼ 50 nm scale ( Maheshwari et al., 2000). In these previous studies, we used a branched from of polyethylene oxide (PEO) to present the ligand in local clusters of one, five or nine peptides per PEO molecule against a background otherwise inert towards cell and protein adhesion. Here, we demonstrate that cell adhesion to the GRGDSPK peptide is reinforced upon application of a distraction force to cells when the ligand is presented in a clustered arrangement. In this work, we use a comb-shaped copolymer comprising a poly(methyl methacrylate) (PMMA) backbone with short PEO side chains (six to nine ethylene oxide units) to present the ligand ( Irvine et al., 2001). The hydrophobic PMMA portion establishes a stable surface film in an aqueous environment, whereas the hydrophilic PEO side chains are extended at the water-polymer interface to present adhesion peptides as well as to prevent non-specific protein adsorption ( Fig. 1A). The average ligand density and ligand cluster size can be independently tuned by controlling the number of ligands per comb polymer molecule and mixing modified and unmodified combs in defined proportions.
We used this system to examine cell adhesion responses to three RGD peptide cluster sizes (n=5.4, 3.6 and 1.7 RGD/comb) with overall RGD surface densities ranging from 5270 to 190 RGD/μm2. In this RGD density regime, we found that ligand clustering increased cell adhesion strength for a given average ligand density, consistent with previous studies at higher ligand densities ( Maheshwari et al., 2000). Furthermore, we found that cells reinforced their integrin linkages to withstand stronger detachment force in a manner that depended on ligand clustering. Our findings implicate nanoscale ligand distribution as an important additional mechanism in controlling the mechanically induced cell adhesion response.
Materials and Methods
Cell line and culture
The fibroblastoid wild-type (WT) NR6 cell line was generated by stable transfection of NR6 cells, a murine 3T3-derived line that lacks endogenous epidermal growth factor receptors (EGFR), with wild-type human EGFR ( Chen et al., 1994; Pruss and Herschman, 1977). As previously described, the WT NR6 line expressesα 5β1 andα vβ3 integrin surface receptors and migrates on the adhesion peptide YGRGD ( Maheshwari et al., 2000). WT NR6 cells were cultured in minimum essential medium-α (MEMα) supplemented with 7.5% fetal bovine serum (FBS), 350 μg/ml G418, 1 mM sodium pyruvate, 2 mM L-glutamine, 1 mM non-essential amino acids, 100 i.u./ml penicillin, and 200 μg/ml streptomycin. In serum-containing assay medium, FBS was replaced by 1% dialyzed serum and 1 mg/ml bovine serum albumin (BSA) in MEMα with 25 mM HEPES.
Preparation of clustered RGD adhesion substrates
Polymers were synthesized and modified with RGD-containing adhesion peptides as previously described ( Irvine et al., 2001). Briefly, comb terpolymer was synthesized with free radical polymerization using methyl methacrylate, polyethylene glycol methacrylate (HPOEM, Mn ∼360 g/mol), polyethylene glycol methyl ether methacrylate (POEM, Mn ∼475 g/mol); Azo(bis)isobutyronitrile (AIBN) was used as an initiator. The molecular weight and polydispersity of the resulting polymer were determined by gel permeation chromatography to be Mn=93,900 g/mol and PDI=2.0, respectively, on the basis of the polystyrene standards ( Irvine et al., 2001). The weight ratios of the three monomers in the resulting polymer were determined by NMR to be 66:16:18 (MMA:HPOEM:POEM). Comb polymer was carboxylated via hydroxyl groups of HPOEM using succinic anhydride and was activated using dicyclohexylcarbodiimide (DCC) and N-hydroxysuccinimide (NHS). NHS-activated polymer was stored at -20°C until used. GRGDSPK peptide (American Peptide Company) was coupled to the NHS-activated HPOEM side chains of the polymer in solution via the N-terminus. An average number of peptides per comb molecule (n) was determined on the basis of the elemental analysis performed by Quantitative Technologies, Inc. Variations in peptide cluster size were achieved by varying the peptide to polymer concentration ratio during the solution coupling reaction.
To vary overall RGD density independently of cluster size, ligand-modified combs were mixed with unmodified combs in defined proportions to achieve a total combined final comb polymer concentration of 2 mg/ml in 50:50 water:ethanol. Polymer films were prepared in 96-well plates by solvent casting from water/ethanol and drying at room temperature for 5 hours followed by drying in vacuo for 24 hours. Films cast at this concentration had a thickness ∼150 nm. The diameter of an expanded quasi-2D comb `island' at the interface ( Fig. 1B) was approximated to be ∼32 nm ( Irvine et al., 2001), which can accommodate roughly 10 closely packed integrin receptors. For all cluster sizes examined in our study, the estimated center-to-center distance between two RGD peptides within an RGD-bearing comb disk ranged from 14 to 25 nm ( Table 1); therefore most tethered RGD peptides should be available for binding to integrin receptors (∼10 nm head diameter) without steric hindrance. The average densities of RGD peptides on the substrates used in adhesion studies were determined using quantitative reaction with a fluorescent reporter and are summarized in Table 1.
Fibronectin surface preparation
Fibronectin diluted in phosphate buffered saline (PBS) (0.1, 0.3 μg/ml) was allowed to adsorb to Nunc tissue-culture-treated polystyrene 96-well plates for 18 hours at 4°C. Wells were washed twice with cold PBS and blocked for 1 hour at 37°C with 1 mg/ml heat-inactivated bovine serum albumin (BSA) in PBS. Wells were washed twice with warm PBS prior to cell plating.
Cell adhesion strength was determined using a centrifugation assay ( Chu et al., 1994). Cells were incubated in serum-free medium containing 1 mg/ml BSA instead of FBS for 12 hours post-seeding (5,000 cells per well on RGD-comb surfaces and 2,000 cells per well on fibronectin-coated surfaces). Serum-free medium was then changed to HEPES assay medium containing 1% dialyzed serum (dserum) and cells were incubated for 8 hours. This incubation protocol with its particular time points was designed to correspond to other assays we plan to perform in the future to study cell migration and growth factor signaling on the RGD-comb substrate ( Maheshwari et al., 2000). At the end of the incubation, the wells were filled with 1% d-serum medium and covered with sealing tape to avoid medium loss and air bubbles during spinning. Inverted plates were spun in a Sorvall centrifuge for 10 minutes at 25°C. For each adhesion experiment performed on RGD-comb films, control wells containing surfaces with 5.4 RGD/comb and 1050 RGD/μm2 were subjected to gravity (1 g) by simple inversion to account for inter-experiment variations. After centrifugation, the number of cells left adherent in the wells was quantified using the CyQuant nucleic assay according to the manufacturer's instructions. A cell adhesion index (CAI) was used to measure surface adhesiveness. To obtain a CAI, the post-assay number of adherent cells left in sample wells was divided by the number of adherent cells left in the control wells. Each surface condition was examined at least in triplicate in every experiment, and each error bar indicates the standard deviation of the mean. The assay was repeated for selected points (labeled with an asterisk `*' in data plots). Post-centrifugation, the RGD-substrates usually retained 25% or less of the original cell seeding number. The normal detachment force was calculated using the equation f=RCF × V × (ρc -ρ m), where f is the force exerted on a cell, RCF is the relative centrifugal force, V is the cell volume (∼500 μm3),ρ c is the density of the cell (∼1.07 g/ml) andρ m is the density of the medium (∼1.00 g/ml).
For adhesion studies performed on fibronectin-coated surfaces, each experiment was repeated independently three times, and each surface condition was examined in triplicate in every experiment. Control surfaces coated with 1μ g/ml of fibronectin in PBS were subjected to gravity (1 g) by simple inversion in each experiment to account for inter-experiment variations. The number of cells left adherent post-assay was determined as described previously. The number of cells in sample wells was normalized by the number of cells in control wells. The resultant percentage of cells adherent was used to measure surface adhesiveness.
To ascertain the statistical significance of adhesion reinforcement and strengthening owing to mechanical stimuli and ligand clustering, a student t-test was performed with a 95% confidence level on data points adjacent to either side of the reinforcement peak for a given substrate and on adhesion data derived from different substrates at a given detachment force.
Cell adhesion is reinforced in response to mechanical stimulus
Comb substrates bearing no ligand are completely cell resistant even in the presence of 7.5% serum over the course of 24 hours ( Irvine et al., 2001). These same substrates, when modified with covalently attached GRGDSPK peptides, induce cell attachment and spreading, and these behaviors are inhibited by soluble RGD peptide. Comparable comb substrates presenting clustered RGE peptide are non-adhesive ( Irvine et al., 2001), indicating that cell attachment to these engineered substrates is RGD specific.
Here, we quantified cell attachment to these highly specific RGD-modified substrates using a centrifugal detachment assay over a wide range of forces. Cell adhesion was quantified by enumerating the cells left adherent after a given detaching force had been applied for 10 minutes, and the results were plotted as a cell adhesion index (as described in Materials and Methods) versus normal detachment force to generate an adhesion profile for each set of conditions. The expected profile was a monotonically decreasing curve of adhesion index with increasing force. Surprisingly, a definitive peak in the WT NR6 adhesion strength profile was found on most of the RGD-comb substrates investigated ( Fig. 2A,B). Peaks occurred within a detachment force range of 70 to 150 pN/cell, depending on the average cluster size and overall RGD density.
The presence of a peak in adhesion strength implies a reinforcement of cell connections to the substrate, namely the RGD-integrin and/or integrin-cytoskeletal linkages, in response to externally applied mechanical forces. Reinforcement was most prominent on surfaces with the highest peptide density and the highest cluster size ( Fig. 2A; n=5.4 RGD/comb, 5270 RGD/μm2). However, mechanically induced adhesion reinforcement occurred at all densities varying from 260 to 5270 RGD/μm2 when five RGD peptides were clustered in 32 nm islands. Following reinforcement, cell adhesion decreased monotonically and eventually leveled off with increasing detachment force. A similar reinforcement phenomenon was found on 3.6 RGD/comb surfaces, with densities ranging from 190 to 2100 RGD/μm2. The position of the reinforcement peak remained relatively insensitive to RGD density and cluster size. However, the amplitude of the peak increased with increasing ligand density. For each clustered surface (n=5.4, 3.6 RGD/comb), the peak presence was verified with a student t-test at a 95% confidence level as described in the Materials and Methods.
Adhesion reinforcement was not detectable on surfaces with the lowest RGD cluster size (n=1.7), even when overall RGD densities were comparable to those of the clustered substrates ( Fig. 2C). Thus adhesion reinforcement appears to be a phenomenon promoted by ligand clustering.
Cell adhesion depends on ligand density and ligand clustering
For higher cluster sizes, cell adhesion to RGD-comb surfaces increased with greater values of overall RGD density ( Fig. 2A,B). The baseline at which the cell adhesion profile leveled off at high detachment forces also exhibited a density dependence. Usually, a baseline in the cell adhesion profile is expected to occur when all cells are removed from the substrate. However, cell detachment on comb surfaces reached a plateau at a non-zero value. Furthermore, this baseline value increased with increasing peptide density. This observation further suggested that cells were actively responding to the applied forces in addition to the simple process of physical detachment in a substrate-dependent manner.
To discriminate the effect of ligand clustering from that of ligand density, we examined adhesion strength measured on RGD substrates with comparable average ligand density but different average cluster size ( Fig. 3). To most effectively demonstrate clustering-promoted cell adhesion, we compared surfaces with highest (n=5.4) and lowest (n=1.7) cluster size. At all three densities examined (260-1660 RGD/μm2), nanoscale ligand clustering enhanced cell attachment to the substrate when the ligand density remained the same or was slightly lower.
Cell adhesion reinforcement is not observed on fibronectin substrates
Clustering-dependent cell response to a mechanical stimulus suggested a modulatory function of ligand distribution in the integrin-ligand interaction and the subsequent signaling processes. To test this phenomenon in a physiological context, we measured cell adhesion on substrates adsorbed with fibronectin, a natural ECM adhesion ligand, using a protocol similar to that described for measurement of adhesion on RGD-polymer substrates. Cell adhesion was quantified after a 12 hour serum starvation period followed by an 8 hour incubation in the presence of 1% d-serum. The adhesion profiles for the various surface conditions were nearly indistinguishable from one another, and the number of cells attached remained essentially unchanged within the range of detachment forces applied ( Fig. 4). Cell adhesion dependence on fibronectin density was greatly diminished. Furthermore, while 70% to 80% of the cells remained adherent on fibronectin after centrifugation, heat-inactivated BSA, a protein that does not support cell adhesion, retained 60% of the cells. Thus under the given conditions, we were not able to determine whether fibronectin could promote reinforcement in the presence of serum. Unlike the synthetic RGD polymer substrates, proteins could adsorb to the substrates used in these experiments. Therefore we conclude that adhesion proteins present in the 1% d-serum have contributed to the increased adhesion on these substrates.
Since this adhesion assay entailed two incubation periods, a 12-hour serum-starvation phase to quiesce cells and an additional 8 hour incubation in 1% d-serum, we examined adhesion to fibronectin at the end of the first phase, when protein adsorption from serum would not be present to influence the results. In the absence of serum, the adhesion strength profile for WT NR6 cells exhibited a monotonic decrease without reaching a plateau as the detachment force was increased ( Fig. 5). As expected, cell adhesion also depended on surface fibronectin concentration. When the applied force was under 300 pN/cell, substrates adsorbed with 0.3 μg/ml fibronectin solution retained four to five times more cells than those coated with 0.1 μg/ml solution. Thus, in the absence of serum, a physiological ECM protein such as fibronectin failed to induce adhesion-strength reinforcement. Additionally, this result confirmed that the peak observed in adhesion to RGD was not an artefact of the centrifugation assay or due to changes in the viscoelastic properties of the cells.
In this study, we have confirmed previous findings that cell attachment and adhesion strength can be enhanced through nanoscale clustering of adhesion ligands ( Danilov and Juliano, 1989; Maheshwari et al., 2000). More significantly, we have measured a dramatic increase, or reinforcement, in cell adhesion strength in response to mechanical forces applied to pull cells away from their substrate. This phenomenon depends on ligand clustering as well as ligand surface density.
The phenomenon we observe here appears to be similar to the adhesion reinforcement observed in response to forces exerted on integrins by pulling on ligand-coated beads using optical tweezers after bead attachment to lamellipodia ( Choquet et al., 1997; Felsenfeld et al., 1999). In these studies, Sheetz and colleagues simulated extracellular attachment sites with different rigidities by using an optical trap to restrain the movement of beads that presented covalently linked fibronectin type III repeat 7-10 fragments and had formed linkages to the retrograde-moving cytoskeleton through integrins ( Choquet et al., 1997). In response to increased restraining force, integrin-cytoskeleton linkages stiffened and strengthened proportionally. This adhesion reinforcement is an acute, sustained and localized phenomenon that requires both receptor occupancy and aggregation. Only a 10-second induction time is required after a bead initially contacts the cell surface in order to elicit such a strengthening response to restraining force. The reinforcement phenomenon is inhibited by phenylarsine oxide (PAO), an inhibitor of tyrosine phosphatases, and is inhibited by the tyrosine kinase Src when adhesion is mediated by the vitronectin receptor ( Felsenfeld et al., 1999).
From the perspective of this work, an interesting facet of the results presented by Sheetz and colleagues arises from the method used to systematically vary ligand density on the beads in optical trap force experiments ( Choquet et al., 1997). For one set of the ligand-modified beads, fibronectin fragments were linked onto albumin molecules to achieve several peptides per albumin molecule. Then, multivalent peptide-modified albumin was mixed with unmodified albumin in defined proportions and coupled onto the bead surface to vary overall average ligand density. This is similar in principle to the methodology employed in making the clustered RGD-polymer surface for this study, where peptide-modified synthetic polymers were mixed with unmodified polymers to create clusters on a flat adhesion substrate. It also bears resemblance to the use of RGD-peptide modified albumin in the study by Danilov and Juliano, where the number of peptides per albumin was systematically varied over a wide range and coated onto substrates at defined average densities ( Danilov and Juliano, 1989). Thus, ligand clustering could conceivably be a key parameter in the reinforcement phenomenon observed by Sheetz and colleagues, although they do not explicitly discuss this possibility. Results with beads prepared in this fashion were pooled with results obtained using a more homogeneous method of ligand attachment, and thus it is difficult to tease out clustering effects from the data presented ( Choquet et al., 1997; Felsenfeld et al., 1999).
To our knowledge, the phenomenon of adhesion strength reinforcement under detachment stress has not been reported previously in published shear flow or radial flow detachment assays using surfaces adsorbed with natural ECM adhesion proteins. The phenomenon we observe depends on both ligand density and nanoscale ligand clustering. Reinforcement occurs at very low centrifugal forces; hence, this phenomenon could easily be missed in such an assay unless a comprehensive range of forces is used.
Despite the differences in methodology used in the optical trap experiments conducted by Sheetz and colleagues and the detachment studies reported here, we can compare the forces associated with adhesion reinforcement by estimating the approximate force per bond in each case. In the bead assay, restraining forces ranging from 5 pN to 60 pN were applied to 1 μm diameter beads coated with peptides at densities of up to 5000 ligands/bead. Assuming uniform ligand distribution, hemispherical contact with the cell membrane and full engagement of available ligands with integrins, a simple calculation of force per integrin-ligand bond for the `high density' ligand regime (1000-5000 ligands/bead) where adhesion reinforcement was reliably observed in the optical trap assays, leads to an estimate of 0.002 pN to 0.12 pN per bond. In a centrifugation assay, detachment force encountered by a single cell is calculated using the equation f=RCF × V × (ρc-ρm), where f is the force exerted on a cell, RCF is the relative centrifugal force, V is the cell volume, ρc is the density of the cell, and ρm is the density of the medium. Typical integrin expression is of the order of 100,000 receptors per cell ( Akiyama and Yamada, 1985a). If we presume that ∼50,000 integrins/cell (∼50 bonds/μm2 for a spreading cell) are engaged, forces that induce adhesion reinforcement during centrifugation range from 0.0028 pN to 0.0042 pN per bond. Here we assume that most available integrins are engaged, hence the force range calculated is likely to be a substantial underestimate. Nevertheless, these forces are within the estimated range of those observed in the optical trap assays.
Several factors may also contribute to the more modest force range observed in our study. (1) The centrifugation assay measures the average response from a cell population, whereas the optical trap method determines a single-cell response. For example, focal adhesion development and dynamics are drastically different for mobile and stationary cells ( Balaban et al., 2001; Smilenov et al., 1999), and such a distinction is not feasible to identify when averaged over a population of cells. (2) Trapped beads only measure a localized cell response from the upper surface of the lamella, but the strength of fibronectin-integrin-cytoskeleton linkages are regionally specific, especially in motile cells where preferential binding at the leading edge and release at the tail are usually observed ( Galbraith and Sheetz, 1997; Nishizaka et al., 2000). (3) The optical trap exerts a combination of normal and shear forces as opposed to the pure normal force applied in the centrifugation method. (4) Short linear RGD peptides such as the one used in this work are known to have lower affinity for integrin receptors than their parental proteins or larger fragments of the parental proteins such as the fibronectin 7-10 fragment used in the optical trap experiment. Further, the isolated adhesion domain contained in an RGD peptide is likely to elicit incomplete signaling responses compared to the fibronectin 7-10 fragment, which contains both RGD and adhesion synergy motifs ( Leahy et al., 1996). Finally, we note that the time scale of adhesion strengthening in the optical trap studies is ∼10 seconds. This is consistent with the time scale observed in our centrifugation study, where reinforcement must develop before equilibrium conditions for detachment are achieved (possibly on the order of seconds to minutes) and must be stable over the course of force application (10 minutes).
Reinforcement is most prominent on the substrate with the largest cluster size and the highest surface peptide density, and it diminishes when either parameter is reduced. One mechanism for mechanotransduction in adhesion modulation is the ability of the cell to dynamically form, mature, sustain and disassemble adhesion structures in response to chemical and mechanical signals. Visualization of GFP-tagged vinculin or paxillin dynamics has provided direct evidence of focal adhesion growth in response to local centripetal force ( Riveline et al., 2001). To further understand whether focal adhesion size correlates to force transmission, Geiger and colleagues have measured forces exerted by stationery cells on a deformable substrate ( Balaban et al., 2001). A striking finding is that local forces are related linearly to focal adhesion area, and a constant force of the order of 1 pN is estimated for each integrin bond. Also consistent with the time scales observed in adhesion reinforcement studies, local adhesion assembly was extremely rapid, on a time scale below seconds ( Balaban et al., 2001).
In light of the striking effect that clustered RGD peptides had on cell adhesion in response to mechanical stimulus, we wanted to investigate possible adhesion reinforcement induced by multidomain adhesion proteins such as fibronectin. Fibronectins are dimers of two similar polypeptide chains, each containing an RGD recognition moiety with a nearby synergy site. Thus a cluster of adhesion motifs is effectively present in each fibronectin molecule over a molecular dimension of hundreds of nanometers. In addition to the molecule's multimeric structure, Monte Carlo simulation has predicted that random adsorption at low fibronectin densities may provide stochastic clustering of fibronectin molecules when they are randomly deposited in close proximity ( Irvine et al., 2002a). Using fibronectin coating concentrations of 0.1 and 0.3μ g/ml under the coating conditions described in this paper, approximately 50 and 100 fibronectin molecules/μm2 were adsorbed onto the substrate ( Asthagiri et al., 1999). These densities are lower than the average RGD densities used (500-5000 molecules/μm2) in order to compensate for the much higher integrin-binding affinity of fibronectin compared with its peptidyl counterpart. According to competitive inhibition studies, there is a 10- to 100-fold difference in affinity between fibronectin and linear RGD peptides such as GRGDSP ( Akiyama and Yamada, 1985b; Hautanen et al., 1989; Pierschbacher and Ruoslahti, 1984), suggesting that significantly fewer fibronectin molecules are required on the surface in order to form the same number of bonds with integrin receptors. In light of these findings, we selected a fibronectin density range that would generate comparable surface adhesive properties to the RGD-comb substrates by using commensurably lower fibronectin densities. Furthermore, the fibronectin densities chosen provided a wide range of cell adhesion responses to the range of forces we applied; ∼50% of the cells seeded adhered to the 0.3 μg/ml substrates at low distraction forces, whereas only ∼15% adhered at 0.1 μg/ml. Higher fibronectin densities would not provide as large a dynamic range in the response and were not investigated. Although in our results we observed an absence of adhesion reinforcement on fibronectin after serum-free treatment, we were not able to determine the effect of serum, as serum-derived adhesion proteins apparently confounded substrate chemistry through non-specific adsorption. Consistent with our finding, detachment of erythroleukemia cells from fibronectin-adsorbed surfaces by shear stress under serum-free conditions exhibits sigmoidal detachment profiles without reinforcement ( Garcia et al., 1998a; Garcia et al., 1998b). Serum-derived non-specific adhesion also greatly increases substrate adhesiveness such that no significant changes are observed in cell detachment within the force range employed in this study. Thus, it remains inconclusive whether or not fibronectin promotes adhesion reinforcement, although in the absence of serum such a phenomenon was not established. It is also possible that spatial arrangement of the ligand clustering is indeed required, as in the case of RGD-comb substrates.
Using a star polymer system, we have previously demonstrated that adhesion enhancement induced by ligand clustering was concurrently accompanied by a higher degree of focal adhesion and stress fiber formation ( Maheshwari et al., 2000). Numerous studies have shown that such cytoskeletal changes are mediated by Rho family GTPases ( Hall and Nobes, 2000; Tapon and Hall, 1997). In particular, integrin-promoted Rho activity drives actin stress fiber assembly and focal adhesion maturation ( Clark et al., 1998; Hotchin and Hall, 1995). Rho may also play a key role in adhesion phenomena induced by ligand clustering. Reinforcement observed in the optical trap studies also exhibits dependence on ligand concentration and can be inhibited by low doses of PAO, suggesting the involvement of a ligand-mediated enzymatic signaling cascade ( Choquet et al., 1997). Biochemical studies to gain mechanistic understanding of clustering-mediated cell response in our system are in progress. In our study, the monotonically decreasing adhesion profile of long-term cell adhesion to fibronectin under serum-free conditions supports the notion that reinforcement on RGD-comb is not an artefact of the system or the assay, but rather is a biochemically based phenomenon.
We propose a mechanism by which a mechanical stimulus might trigger intracellular signaling leading to adhesion reinforcement ( Fig. 6). The adhesion profile for an attached object undergoing physical detachment by an external force is expected to decrease monotonically as a function of detachment force until all the objects are removed (blue solid line). However, for living cells there may exist signaling cascades that lead to the strengthening of cell attachment to a substratum, for example, through strengthening of ligand-receptor and/or receptor-cytoskeleton linkages. This signaling is induced by mechanical stimuli such as an external force or substrate rigidity. In our model, signaling reinforces cell adhesion in response to force in a sigmoidal fashion (pink solid line), similar to an `on/off' switch. When this signaling effect is compounded to the purely physical process of detachment (blue solid line), an adhesion profile characterized by an initial drop followed by a distinct reinforcement peak is developed (red solid line, region A). Further, if signaling for reinforcement itself can be reinforced at a higher mechanical threshold, then we will observe a non-zero plateau in the resultant adhesion profile (red solid line, region B). We are planning future studies to test these ideas.
Using a multi-arm YGRGD-modified star PEO polymer with ligand tethers of∼ 60 nm (fully extended length), we have previously demonstrated enhanced cell adhesion to nanoscale clustered adhesion ligands compared with singly presented ligands, but adhesion reinforcement was not studied in this system ( Maheshwari et al., 2000). Although both star and comb systems focus on the regime of small clusters (<10 ligand/polymer molecule) that are approximately 30-50 nm in diameter, there are marked distinctions in their structures, properties and display of ligand that may influence cell responses. A quasi-2D comb polymer at the interface with short ∼2 nm ligand tethers provides a more rigid substrate than a 3D star polymer with flexible long tethers linked to a 300 nm thick hydrogel base, and substrate flexibility modulates cell behaviors such as focal adhesion formation and motility ( Lo et al., 2000; Pelham and Wang, 1997). In addition to substrate mechanical properties, tether length also affects ligand mobility and spacing. For systems with long PEO tethers such as the star polymer, the ligand is in constant, rapid motion over a volume defined by the ∼60 nm extended length of the highly flexible tether, leading to a high probability that multiple ligands on the same star can `adjust' their positioning in response to integrin clustering. In comparison, the short (∼2 nm) PEO tether of the combs constrains ligand mobility; therefore integrin spacing in integrin-ligand clusters formed on the comb systems closely matches that of the side-chain spacing in the comb itself. We speculate that there may be specific minimum peptide densities within a cluster to achieve appropriately close ligand spacing, which in turn allows proper integrin clustering. The average spacing between adjacent ligands within a cluster ranged from approximately 14 nm (n=5.4) to 17 nm (n=3.6) to 25 nm (n=1.7) ( Table 1). For comparison, the head dimension of an integrin receptor is of the order of 10 nm ( Erb et al., 1997). We are currently using the comb system to vary ligand spacing and the number of peptides per cluster independently.
The comb system used here and the star system used in the previous study also differed in their biochemical constitutions. In this study, we used a higher affinity peptide (GRGDSPK) and examined a different ligand-density regime (200 to 5,000 RGD/μm2) than in our previous study (YGRGD; 1,000-200,000 RGD/μm2). Thus, ligand-clustering-promoted cell adhesion was consistently observed in two markedly different substrate systems. Our present study additionally demonstrated adhesion strength reinforcement under applied force - another unexpected cell adhesion response that is dependent on nanoscale ligand organization.
It is conceivable that ligand clustering enhances local ligand avidity, which in turn promotes integrin receptor aggregation to form focal adhesions. Focal adhesions are specialized cell attachment sites where ligated integrin receptors aggregate to link the ECM to a dynamic cytoskeletal architecture, transmitting both force and signals ( Critchley, 2000; Sastry and Burridge, 2000; Yamada and Geiger, 1997). To account for cell-substrate adhesion enhancement owing to focal adhesions where receptor clustering is promoted, Ward and Hammer developed a biophysical model incorporating the effects of ligand density, receptor density and intracellular talin polymerization at focal adhesion sites ( Ward and Hammer, 1993). It predicts that receptor aggregation within focal adhesions would be greatest at high ligand densities and becomes negligible below a certain minimum density. In agreement with their model, Massia and Hubbell have shown that human skin fibroblasts failed to form focal adhesions and stress fibers when the density of RGD peptide on the substrate fell below 60 peptide/μm2, which corresponds to an RGD spacing of 140 nm ( Massia and Hubbell, 1991). Although in our study average ligand densities ranged from 200 to 5,000 peptides/μm2, a more meaningful parameter characterizing overall ligand distribution is the RGD cluster spacing ( Table 1). Presenting adhesion ligands in discrete nanoscale islands (5 RGD/comb) enhances cell adhesion even when RGD cluster spacing is increased to 290 nm ( Fig. 3C; Table 1). The substrates prepared for this study have inappropriate optical characteristics for high-resolution fluorescence microscopy, and thus we did not examine focal adhesion formation. We are currently modifying the preparation protocols to facilitate such studies.
Cell adhesion studies are often carried out using protein-coated surfaces, and adhesion strength is measured after short-term cell attachment (within several hours of cell plating). Serum-free conditions and protein synthesis inhibitors are usually used to ensure negligible non-specific adhesion mediated through deposition of serum-derived or cell-secreted matrix molecules. These requirements place grave restrictions on long-term studies of cell adhesion, multi-molecular synergy and other adhesion-based cell behaviors such as motility. Adhesion profiles presented in Fig. 4 and Fig. 5 demonstrate the `masking' of a bio-specific surface by nonspecific molecules. Applying a cell detachment assay after a 12 hour incubation in serum-free media shows that cell adhesion exhibits unambiguous dependence on fibronectin density and detachment force. However, an 8 hour incubation in 1% dialyzed serum following serum deprivation nearly abolishes fibronectin density dependence. Even substrates coated with heat-denatured BSA, which does not promote cell adhesion specifically, became non-specifically adhesive after an 8 hour incubation in 1% d-serum. Thus protein-coated substrates fail to provide a robust system for long-term studies in which ligand density and specificity are of interest. The PEO/PMMA comb polymer surface is rationally designed to provide a cell inert background on which covalently tethered biological signals remain specific over time ( Banerjee et al., 2000; Irvine et al., 2001; Irvine et al., 2002b). Therefore, over the course of our experiment (>20 hours), surface biochemistry has remained constant and specific in the presence of serum and cell-secreted proteins.
Finally, we noted that the RGD-comb substrates used in the present work were of relatively moderate adhesiveness since the number of cells attached to them was low. However, the development of an adhesion profile is inherently due to a distribution of cell adhesive properties within a cell population. Therefore, even when the percentages of retained cells are moderate, the observed phenomena are still reflective of the behavior of a heterogeneous cell population.
Results from this study emphasize the integrated governance of cell behavior through biochemical and biophysical cues that are present in the ECM. Delineation of the individual biochemical and biophysical contributions remains challenging, and we present here a new tool to allow independent control over ligand clustering and ligand density. Our findings suggest that these two substrate properties control cell adhesion synergistically. Nanoscale ligand clustering enhances cell adhesion even when the ligand density remains constant. In addition, cell adhesion is modulated by mechanical forces transduced through cell-ECM contacts in a manner that requires both ligand distribution and intracellular signaling. These results have implications for rational design of biomaterials that modulate cell adhesion, allowing the efficiency of peptide use to be improved. In this paper, we presented the initial work of an integrated cell behavioral investigation using the comb substrates. We are currently investigating the role of ligand clustering in cell motility and and in the crosstalk between integrin and growth-factor-mediated signaling pathways.
This work was supported by NIH/NIGMS grant 5R01GM59870 to L.G.G. and A.M.M., a Whitaker Foundation Young Investigator grant to A.M.M., and NIH grant GM53905 to D.A.L. L.Y.K. was supported by a Whitaker Foundation graduate fellowship in biomedical engineering and D.J.I. by a NSF graduate fellowship. We are grateful for helpful discussions with A. F. Horwitz (University of Virginia).
- Accepted January 2, 2002.
- © The Company of Biologists Limited 2002