We have shown that agrin presented as a nanopatterned substrate mediates adhesion for a variety of cell types, including primary neurons and neuroblastoma cell lines. In addition, we have shown that the dependence of adhesion on the spacing of the agrin nanopattern has a sharp spacing-dependent threshold, similar to that of the canonical integrin ligand RGD, a component of many ECM proteins [14
]. In contrast, the cell-surface homophilic adhesion molecule N-Cadherin has a more linear dependence on spacing, with no significant spacing-dependent threshold. Nanopatterned spacing also influences cell motility and morphology, but not simply in parallel with adhesion. Consistent with cells responding to agrin as they would to integrin-dependent extracellular matrix ligands, adhesion was inhibited by competing IKVAV peptides. Finally, nanopatterned substrates are capable of highlighting differing cell responses that are masked when the substrate is uniformly presented.
In addition to the effects on cell adhesion that we show, cell proliferation had a similar dependence on the spacing of agrin substrates (not shown). Both proliferation and adhesion are often integrin-mediated processes, and the inhibition of adhesion by the IKVAV peptide is consistent with this. The IKVAV sequence was identified as an active motif in the laminin alpha1 chain that promoted cell adhesion, proliferation, and neurite extension, and these effects depend, at least in part, on Beta1 integrins [15
]. As such, the RGD and IKVAV peptides have been previously reported to be similarly effective [15
]. Agrin contains neither an RGD nor an IKVAV sequence; however, previous studies have shown that adhesion of chick primary ciliary ganglion neurons to agrin was largely dependent on Beta1 integrins, and was sensitive to RGD peptide competition at a site that was also sensitive to function blocking αV integrin antibodies. A second, non-RGD-sensitive site of integrin-dependent cell-adhesion mapped to a more C-terminal portion of agrin, although both sites are contained within the C50 fragment [5
]. The insensitivity of rat B35 neuroblastoma cells to RGD peptide competition in this study is interesting and may reflect a greater dependence in this cell type for adhesion to the more C-terminal domain of agrin. Together, these results strongly suggest that adhesion to agrin is at least in part integrin dependent, and the absence of an IKVAV or RGD peptide may indicate that the interaction is either indirect or mediated by a less-well defined domain. Despite the lack of primary sequence homology, agrin and laminins do share many structural similarities, including EGF-like repeats and laminin-type globular domains, found in the C-terminus of agrin.
The N-terminus of agrin and its heparan sulfate additions have been extensively studied for their role in stopping neurite outgrowth and promoting nerve terminal differentiation [18
]. Agrin has also been used in other biophysical studies in which its presentation was spatially restricted either by using microcontact printing to pattern substrates [20
], or microfluidics to restrict the point of agrin application [21
]. These studies were designed to investigate the spatial properties of agrin signaling in synaptogenesis on muscle cells, an activity that maps to the C-terminal portion of the protein, which is not subject to glycosaminoglycan addition.
Since agrin exists as both a matrix protein and a transmembrane protein in vivo, we felt it was particularly important to determine in which context cells are responding to it [5
]. The extracellular matrix bound isoform is the active form for neuromuscular junction synaptogenesis [4
]. However, agrin's role in the central nervous system, where the transmembrane isoform predominates, remains less clear. The loss of agrin effects synapse density in the brain, and through interactions with the K-Na-ATPase, is proposed to mediate activity dependent plasticity [23
]. Agrin also activates both C-Fos and CREB signaling in neurons [25
]. Given these disparate effects, it is difficult to know what in vitro
assays may most accurately measure agrin's true function. We chose cell adhesion as a first attempt to examine how neurons respond to agrin, in part because this assay also allowed the comparison to well-known components of the extracellular matrix and to well-established transmembrane adhesion molecules, such as L1 and cadherins. It is interesting that even neurons respond to agrin more as cells respond to the ECM ligand RGD than as they do to the transmembrane proteins N-cadherin.
Why is there a sharp threshold for adhesion to agrin and RGD, but not N-cadherin? First, we have to acknowledge that the comparison of agrin to one integrin-dependent matrix ligand and one homophilic transmembrane adhesion molecule is not exhaustive, but the consistencies and inconsistencies are noteworthy. Our preferred explanation is that adhesion to agrin requires an intracellular signaling cascade, and the formation of an adhesion complex analogous (if not identical) to that of integrin-mediated adhesion to RGD. The size of these complexes inside the cell may create a lower limit for spacing; presumably their effectiveness is reduced if the sites of adhesion are spaced more closely than the diameter of such a complex. This could explain the saturating effect seen at 30 and 60 nm spacing, which were similar to uniform coating. The threshold for decreasing adhesion would arise when the avidity of these complexes is no longer sufficient to cause adhesion, not through the lack of signal at an individual complex, but from the lack of sufficient cumulative signal from the array of adhesive sites. In contrast, the N-cadherin mediated adhesion may be more analogous to mechanical "Stickiness" at the cell surface, where the force required to disrupt adhesion varies more linearly with the protein's density on the membrane. The absence of an intracellular cascade would eliminate the threshold, and as long as some of the protein is present there will be some amount of adhesion, with the upper limit being defined at the molecular scale and by the affinity of the interaction.
The affinity of the cellular receptors for the ligand may also contribute to the agrin threshold effect. If this is the case, it is interesting to note that integrins have a much lower affinity for RGD peptides than for full-length fibronectin [27
], yet large agrin fragments have an apparently similar threshold to RGD peptide. However, in considering affinity as an explanation, it is also important to consider the constraints of our system. Each gold particle is bound with only one or sometimes two proteins, and this number should not change as the spacing changes [13
]. Therefore, at the molecular level, the number of ligands presented to a cell-surface receptor is not changed; it is the spacing between these sites of interaction this changes. This effectively decreases the density of potential attachment sites seen by a single cell by the square of the change in spacing, but does not change the molecular composition of an individual attachment site. Therefore, avidity, and not classic receptor-ligand binding affinity, is the variable. A biophysical test of this would be to measure the force of cell adhesion as the spacing of agrin and N-cadherin is increased to confirm that our measures of adhesion co-vary with mechanical force.
The subtleties of cellular responses to agrin, such as the orientation of the protein, may have further implications for function. For instance, the anti-parallel presentation created by the N-terminal 6X His tag of C100 may more closely resemble membrane-to-membrane interactions or suggest an orientation of agrin within the meshwork of matrix that is optimal for cellular recognition and binding.
In considering molecular spacing, laminin polymers in a self-assembling meshwork form a polygon pattern with vertices separated by 30–40 nanometers, comparable to the 30–60 nanometer spacing of agrin-coated nanoparticles that conferred optimal cell adhesion. Laminin polymerization also organizes other cell-surface proteins, including dystroglycan, another receptor interacting with the C-terminus of agrin [28
]. Thus our results appear to be in a physiologically relevant range of distances and the patterning of ECM-associated ligands with such spacing may actually provide a more physiological presentation. However, there are also limitations to the use of nanopatterned substrates. For instance, if cell adhesion depends on clustered adhesion molecules, the size of the gold-nanoparticles as well as their spacing would be expected to have a strong influence. Also, if adhesion depends on a mix of proteins, either as homo-dimers or in multiprotein complexes, the stoichiometric presentation of the substrates would be more difficult to control.
An additional distinction between uniformly coated substrates and the nanopatterned substrates such as we have described is the thickness and rigidity of the substrate itself. Homogeneous coating of glass with recombinant proteins applied at 10 μg/ml results in a film of protein that can be 100–200 nm thick (T. W. unpublished observations). However, molecules directly anchored to the nanoparticles are presented in comparatively low copy number and the thickness of the surrounding PEG-passivation can be controlled by polymer length. Therefore, the nanopatterned substrates are presumably more rigid and more reproducible in their thickness and stiffness than homogeneous substrates. Substrate rigidity has been shown to have a variety of effects on cell morphology, proliferation, and differentiation [32
]. These effects may also be influencing the cellular response to homogeneous versus patterned substrates, particularly for parameters like cell motility.