Engineering the surface chemistry of a material so that it can interface with cells is an extraordinarily demanding task. The surface of a cell is composed of thousands of different lipids, proteins and carbohydrates, all intricately (and dynamically) arranged in three dimensions on multiple length scales. This complexity presents both a challenge and an opportunity to chemists working on bioactive interfaces. Here we discuss how some of these challenges can be met with interdisciplinary material synthesis. We also review the most popular classes of functional molecules grafted on engineered surfaces and explore some alternatives that may offer greater flexibility and specificity. Finally, we discuss the emerging field of dynamic surfaces capable of stimulating and responding to cellular activity in real time.
In the aftermath of the Second World War, a physician named Nicholas Ridley made a surprising observation. Many of the pilots he treated had shards of poly(methyl methacrylate) (PMMA) embedded in their eyes from shattered aircraft canopies, but, unlike most materials, these fragments were not eliciting adverse biochemical responses1. This serendipitous finding led Ridley to conclude that PMMA must be biocompatible, and today it is used in a variety of medical applications, from contact lenses to bone cements. In the half-century since this discovery, we have made great progress in materials synthesis and, it is to be hoped, are approaching a time when we can move past empirical engineering of surfaces for biological applications. However, the formulation of overarching design principles is an unmet challenge and many questions remain unanswered. What does a cell perceive when it lands on a solid support? What cues are cells searching for and how do the components of the cell membrane contribute to this search? These are not questions of purely academic interest. The formation and evolution of the interface between cells and artificial materials is of profound significance for a variety of applications, from routine cell culture to tissue engineering.
From a systems biology perspective, there is nearly endless complexity in the crosstalk between cells and dynamic functional materials. Yet it may be possible to clarify the situation somewhat by stepping back from biology and treating both the cell and the material surface as chemical systems. Extending this conceptual framework may allow the routine design of surfaces with specific cellular activity. In the short term, a more modest goal that is already being realized is the creation of chemically defined surfaces that eliminate the need for animal-derived substrates and soluble factors. In addition to allowing simpler, better-defined experiments, this capability will significantly diminish the risks associated with the medical application of engineered tissue and cell-based therapies2.
A physical model of the cell surface
As most cell–material interactions are mediated by the cell membrane, it is worth considering which molecules comprise the membrane and how they are arranged. There is enormous diversity among cell types, especially when considering bacteria, which possess a cell wall that is exterior to the cell membrane. In this Perspective, we will focus only on cells from multicellular organisms, such as those that may be found in human tissue. One of the most significant features of the cell membrane is the incorporation of membrane proteins, which can account for over a third of a cell's surface area. Figure 1a illustrates the strikingly high protein density3 (more than 30,000 molecules per square micrometre) of a typical cell membrane, although this figure can vary significantly with cell type and disease state (Fig. 1c). It is primarily through these proteins that cells interact with surfaces, typically through the creation of attachment points linking the cytoskeleton (the cell's mechanical framework) to extracellular binding sites.
When a cell begins to adhere to a surface, transient focal complexes are formed through the binding of a family of membrane proteins called integrins. Given proper conditions, the cell will flatten and spread on the surface. Eventually, if presented with the proper ligands, focal complexes mature into focal adhesions, the main attachment points between cells and two-dimensional substrates4. This transition requires integrin clustering on tightly regulated length scales, and stimulating this process in an artificial system requires spatial as well as chemical specificity. Several studies have demonstrated a critical spacing of 50–70 nm (ref. 5), beyond which cells no longer recognize individual ligands as being 'clustered'. However, it is not clear what happens mechanistically with a ligand spacing of less than 10–15 nm because individual integrins bind only one ligand and steric exclusion would prevent closer approach of two integrins. These values impose tight limits on the engineering of surfaces or molecules to encourage cell adhesion.
One simple method of studying ligand spacing is to use self-assembled monolayers (SAMs) with the ligand mixed at varying ratios with an inert filler molecule. Although this is a popular method6, it can be difficult to control local ligand ordering and presentation precisely. Some of these limitations have been addressed through the use of star polymer bundles that allow independent control of bulk properties and ligand clustering. In one study, cells recognized ligands within polyvalent bundles as being clustered even though the bundles were spaced hundreds of nanometres apart. This finding confirms that cells are capable of recognizing local order even on a sparsely functionalized surface. To study this idea, researchers are using orthogonal synthetic strategies to shape and functionalize surfaces. For instance, peptide SAMs have been generated on gold nanoparticles that were then placed in a precise gradient spacing using block copolymer nanolithography7. In copolymer lithography, phase separation of amphiphilic block copolymers drives ordering on the 1–10-nm scale, a range that is difficult to access with conventional chemistry or 'top-down' nanofabrication. One exciting possibility is that such techniques could be integrated with recent advances in molecular-level templating in which bioactive ligands are selectively grafted onto positions in a preformed molecular backbone such as a cyclic peptide8. The resulting systems could provide a generic platform with defined chemical features from the ångström level to the macroscopic level.
Ordering on the cell surface is also important in the vertical axis, that is, away from the membrane plane. Many membrane lipids and proteins are conjugated to polysaccharides, which comprise the glycocalyx, or cell coat, of all cells9. The glycome, or complement of sugars in the cell, varies among cell types but also within cell types owing to differences such as stage of differentiation and malignant transformation10. This variation can in turn affect the dimensions of the glycocalyx, which can be hundreds of nanometres thick in some cells11 (Fig. 1b). Even given the more typical cell dimension of a few tens of nanometres, this layer can regulate the approach of the cell to a solid support. Interference microscopy studies have shown that for many cell types the membrane is initially held around 50 nm away from a solid surface12, 13. Over several minutes, this spacing decreases12, allowing proteins in some regions of the cell membrane to contact the support. Understanding how and why this process occurs is of critical importance in determining which surface-grafted ligands will be accessible to the cell, and when.
One way to study the effects of glycocalyx-mediated structuring at the cell interface is by modifying ligand tethers. In one such investigation14, glycine spacers were systematically added to the end of an adhesion-promoting peptide and it was found that, as expected, adding a small spacer increased cell binding. Notably, spacers more than nine amino acids in length actually decreased cell binding. It was posited that this effect may be due to the peptide folding over, underlining the need to consider the mechanical as well as dimensional properties of even a supposedly inert spacer. One study of this effect looked at polymer brushes with a bimodal length distribution (22- or 10-mer) where only one of the two chain populations bore bioactive ligands15. Changing the length of the 'background' chain affected ligand presentation on the 'active' chain (Fig. 2b). Similar studies using advances in 'grafting-from' polymerization will be an active research area in coming years. For instance, by applying recent techniques to generate gradient polymer brushes16, 17, researchers could create variable tether lengths across a substrate, probing the vertical as well as planar effects of patterning.
It is tempting to view the glycocalyx in simple, structural terms but, as usual in biology, it is now becoming clear that it is a multifunctional system. In addition to regulating bulk mechanical contact, the glycocalyx also has specific biochemical functions. The extended conformation of the glycoproteins is enhanced by the negative charge density borne by many chains owing to extensive sulfation. Studies have shown that the distribution of sulfation on chains is not random but is organized into domains that seem to control specific binding with proteins18. The nature and degree of this specificity are still a subject of investigation, but the maintenance of proper sulfation is clearly important for cell behaviour. In one striking demonstration of this significance, a recent study19 on embryonic stem cells showed that chemically desulfated cells lost the ability to differentiate (Fig. 1d).
There is an important lesson for modern chemists in the multifunctionality of cellular structures: creating useful biologically relevant systems will require continued crossing of disciplines. Polymer chemistry alone may enable an elegantly structured tissue-engineering scaffold, but the complementary contribution of peptide chemistry to functionalize that scaffold affords a much more powerful material. Often the most efficient solutions will pack multiple functions into a single system, as in the above example of ligand tethers15. Through clever application of physical chemistry principles, simple tethers can become active components of the cell–material interface. Figure 2 illustrates a few of the successful interdisciplinary investigations made so far; future work will undoubtedly build on these concepts. Evolution has produced remarkably elegant systems through the mixing, modification and re-application of a comparatively limited biochemical toolbox. We must now strive to do the same with the much greater array of chemical methods at our disposal.
Applying a range of biochemical interactions
Many cell types will not grow if floating freely, and must be attached to a surface to function normally. However, some surfaces support cell growth much better than others. One important parameter is the material's wettability, which is determined mostly by the charge, polarizability and polarity of the surface functional groups. These factors do not, however, directly mediate cell response. The effect of surface wettability is largely to alter the type (and state) of proteins adsorbed from solution20, which in turn affects cell adhesion. Hydrophobic surfaces irreversibly adsorb large quantities of albumin, an abundant serum protein that does not support cell attachment. By contrast, moderately hydrophilic surfaces such as glow-discharge-treated polystyrene (that is, tissue culture plastic) tend to adsorb adhesion-promoting proteins such as fibronectin. It is unsurprising that fibronectin-rich surfaces would allow adhesion because fibronectin is a component of the extracellular matrix (ECM), the fibrillar material that surrounds and connects cells in most tissues. Surfaces coated with other ECM proteins, including laminin and collagen, also promote adhesion and growth of many cell types.
In light of the importance of ECM proteins, many studies have focused on engineering protein adsorption at interfaces. Owing to its simplicity and clinical applicability, this remains a popular approach and there have been many studies using SAMs to generate defined surfaces with known functional group identity and density. Ideally, such surfaces will allow the development of quantitative models to predict protein (and, thus, cell) response to a given material. At present, however, this challenge has not been met owing to the complexity and dynamic nature of the processes involved. Thus, there are still few general design principles21, 22. An alternative approach is to synthesize and graft peptides and carbohydrates onto surfaces to impart specific biofunctionality. Rational design of surfaces using this concept may be a more immediately tractable problem, as illustrated by some of successes so far (Fig. 3).
The discovery that short peptide sequences can substitute for whole ECM proteins to induce cell adhesion ushered in a new era of biointerfaces. The use of short peptides is advantageous because it provides a less variable system than do naturally derived proteins. Another advantage of synthetic peptides is that it is simple to introduce a range of functional groups to aid in grafting. Adding a cysteine, for instance, gives a thiol group that can be used to create peptide SAMs on gold or, if added at both peptide termini, to crosslink acrylates by means of a Michael addition. The most widely used cell adhesion sequence is arginine/glycine/aspartic acid6 (RGD). The RGD motif is found in many ECM proteins and is recognized by around one-half of the known integrins. This binding is extremely sensitive, being completely abolished by the substitution of glutamic acid for aspartic acid, a difference of only one methylene23. Recognition seems to be conformationally dependent as the flanking residues, which do not form direct bonds but can alter the peptide conformation, affect binding affinity24. Further evidence for conformation dependence is the fact that cyclic versions are more effective than linear sequences25 (Fig. 2f). RGD-grafted surfaces are now widely used for cell adhesion, but there remain some limitations. First, precisely because RGD is so widely recognized, it offers little possibility for cell-specific adhesion. To address this issue, researchers have created cell-specific surfaces by grafting peptides from less generic adhesion-promoting proteins24. Another limitation of RGD grafting is that it may be insufficient to promote full spreading and focal adhesion formation for some cell types26. In vivo, complementary receptors must be bound and activated in addition to integrins27, implying that it may be necessary to consider some of these other binding motifs when designing growth-promoting surfaces.
The syndecans represent an interesting class of complementary adhesion molecule because they act synergistically with integrins as co-receptors for ECM proteins and because they demonstrate the importance of membrane-bound carbohydrates. Each syndecan contains three to five glycosaminoglycan chains composed of heparan sulfate, and it is through these carbohydrates that they participate in adhesion. ECM proteins such as fibronectin contain peptide regions, known as heparan sulfate binding domains (HSBDs), that interact with these chains. Initial studies indicated a generic electrostatic interaction between the negatively charged heparan sulfate and positively charged protein domains, but further work has shown that there is actually a more complicated peptide binding sequence in HSBDs involving both basic and hydrophobic amino acids28. By grafting HSBD peptide sequences onto a surface together with integrin-binding domains such as RGD, researchers have generated systems that bind cells through multiple pathways, more fully replicating the natural interaction between cells and the ECM29.
In addition to membrane-bound carbohydrates that bind specific peptide sequences, there are also membrane proteins known as lectins that bind specific carbohydrates. This binding is sensitive enough to distinguish between epimers such as glucose and galactose30 and may present an attractive target for rational design as the recognition site seems to be preformed and not to undergo conformational changes during binding31 (Fig. 2e). Exploiting lectin–sugar interactions increases design flexibility by offering the possibility of grafting active sugars as well as peptides. One successful exploitation of lectin binding is the creation of galactose-grafted scaffolds to support hepatocyte (liver cell) growth32, 33. It is notable that the galactose receptor normally functions to bind soluble sugars, not to mediate adhesion. Thus, it is clear that some molecules not naturally involved with adhesion may be pressed into that service through creative engineering. This is not a universal principle, however, and care must be taken in interpreting results. Most molecules that bind to a cell surface component (either specifically or nonspecifically) will allow some degree of adhesion. However, many of these will not induce subsequent spreading32, and even some of those that do result in spreading will not give rise to focal adhesions23, 29. At present, there are few reliable rules predicting which ligand–receptor pairs will yield which behaviour, and each new binding agent must be experimentally validated for every cell type.
Given the dearth of guiding principles, it can seem a daunting challenge to create novel cell-adhesive surfaces. However, methods have been developed that do not rely on a-priori functional knowledge. One of the earliest examples of such a high-throughput approach was the screening for cell growth capacity of more than 500 acrylate surfaces comprising 24 different monomers mixed in varying ratios34. This idea was later extended to the screening of peptide monolayer surfaces and, more recently, to the use of phage display to screen millions of peptide sequences2. One exciting new direction is the development of similar approaches using carbohydrate arrays. Printed glycan arrays have been created to identify lectin profiles on cell surfaces35, but this technology has mostly been used for cell identification, not to screen for carbohydrates capable of supporting cell growth. Recent advances in the solid-phase synthesis of carbohydrates36 should serve to facilitate this vein of research by simplifying the creation of polysaccharide libraries analogous to those that exist for peptides.
Another significant development in this area has been the trend towards adaptable systems for highly parallel quantitative immobilization, because molecular spacing and arrangement can drastically affect the way cells interact with a surface. Recently, a novel system giving greater precision than traditional self-assembly schemes was demonstrated37. The method involves generating an alkanethiol monolayer with hydroxyquinone termini that can subsequently react with oxyamine-functionalized ligands to yield an oxime linkage. This approach affords active control because the hydroxyquinone groups must be electrochemically oxidized to quinones before they will react. Furthermore, there is a built-in ability to quantify initial and reacted groups because the quinone and oxime groups can be separately detected with cyclic voltammetry. Using such generic strategies with an increasingly broad range of adhesion molecules will serve to accelerate the discovery of underlying principles in cell adhesion.
Adapting to interfacial dynamics
The surface of a living cell is not merely convoluted and heterogeneous; it is also a site of constant, hectic motion and reordering. Consider, for example, pinocytosis, or 'cell drinking', the process through which cells sample their surroundings by pinching off small folds in their membrane. This is not a slow and subtle process: some cells internalize a membrane area equivalent to their entire surface every thirty minutes38. Cell adhesion and growth can similarly be a rapid, dynamic process, resulting in physical and chemical changes to both the cell and the substrate. Here we discuss some examples of how a cell actively modifies a surface and how materials can be designed to accommodate these changes.
One of the first modifications that a cell performs after adhering is to pull on surface-bound biomolecules, often reorganizing them. Such rearrangement is a critical part of natural growth but is typically not possible with molecules covalently bound to a surface. This immobility has limitations regarding the transition to active clustering but can be overcome by incorporating flexible tethers linking the ligand to the material surface. Even if RGD groups are grafted farther apart than the critical cluster radius, when they are attached to a long, flexible tether the cell can pull them together into an active cluster33. The critical parameter affecting cell adhesion in such cases is not the absolute spacing of grafting points. Rather, adhesion depends on the potential spacing accessible to the ligand given a sampling volume determined by a random coil conformation of the tether (Fig. 4a). Thus, we again see the importance of molecular-scale mechanical properties in addition to the known dependence of cell behaviour on bulk material stiffness39. Future studies will undoubtedly reveal further cooperative effects of ligand presentation and substrate mechanics40.
Mechanical forces can sometimes change the chemical nature of the molecules involved as well as their placement. In vivo, the tension exerted by cells on the ECM can induce conformational changes that expose cryptic binding sites41. Recently, scientists have attempted to use the concept of mechanochemical transduction in the design of actively switchable surfaces. One study42 demonstrated the use of electric fields to modify molecular conformation in a monolayer. Applying a positive bias to the substrate attracted the terminal carboxylate groups, burying them within the monolayer and exposing the methylene backbone of the molecule (Fig. 2c). The resulting transition from a hydrophilic surface to a hydrophobic surface did not involve the participation of specific bioactive moieties, but the techniques developed may be broadly applicable to a wide range of chemistries. Specifically, a precise molecular packing was achieved by incorporating a triphenyl spacer group with a cleavable ester linkage, and the effects on spacing and conformation were verified throughout with sum frequency generation spectroscopy.
Apart from rearrangement and reordering, cells are also capable of covalently modifying their surroundings through enzymatic cleavage. This is an important process in vivo, as it allows the cell to remodel the ECM to accommodate growth and migration (Fig. 4b). Originally, such remodelling was seen as an undesirable side effect in artificial systems because carefully created protein coatings can be degraded over a few days. This concern was in fact one of the motivations for grafting peptides instead; short polypeptides are less susceptible to proteolysis. However, it is also possible to design intentionally degradable materials that take advantage of cell-mediated proteolysis. Incorporating oligopeptides that are recognized by cell-secreted proteases encourages migration and growth into artificial substrates such as poly(ethylene glycol) hydrogels for tissue engineering applications43. Another advantage of using such gels is that it is simple to incorporate integrin adhesion sites in the same material