Biomaterials: Protein Surface Interactions

Transcription

1 iomaterials: Protein Surface Interactions Robert A. Latour, Jr. Department of ioengineering, Clemson University, Clemson, South Carolina, U.S.A. INTRODUCTION The study of protein surface interactions represents one of the most important topics in the field of biomaterials, and as such, it has been a focus of intensive study for several decades. The reason for the great interest in this topic is the realization that protein surface interactions are fundamentally responsible for the biocompatibility of medical devices, or the lack thereof. When a solid material (e.g., a catheter, stent, hip joint replacement, or tissue engineering substrate) comes in contact with a fluid that contains soluble proteins (e.g., blood, interstitial fluid, cell culture media), proteins rapidly adsorb onto the surface of the material, saturating the surface within a time frame of seconds to minutes. Therefore, when living cells (which are much larger than proteins and thus much more slowly moving) approach the biomaterial surface, they do not actually contact the molecular structure of the material surface itself, but rather they contact and interact with the molecular structure of the adsorbed protein layer. Cells, of course, cannot see the adsorbed protein layer, but rather they interrogate their surroundings by way of membrane-bound receptors that can bind to specific bioactive features presented by the adsorbed proteins. Then, through a series of orchestrated molecular mechanisms, these receptor protein binding events are transduced through the cell membrane in a manner that stimulates specific intracellular processes that then determine a cell s response. Accordingly, at the most fundamental level the key to controlling cellular response is to control the type of bioactive sites that are presented by the adsorbed layer of proteins. This, in turn, can be controlled by controlling the amounts and the types of proteins that are adsorbed and their orientation, conformation, and packing arrangement on the biomaterial surface. While this is conceptually simple to understand, the numerous types of soluble proteins contained in physiological fluids combined with their structural complexity has made, and continues to make, this an extremely challenging problem. Many excellent reviews and entire books have been written over the past couple of decades that address the multitude of issues related to the interaction of proteins with surfaces. Several of these reviews and books are referenced in the Further Readings section at the end of this article and the readers interested in this topic are encouraged to seek out these additional sources of information. Given the vast number of studies that have been conducted on this topic, it is of course impossible to provide a complete review of the literature in one article. Instead, the goal here is to provide an overview of the basic understanding that has been achieved over the years regarding how proteins interact with surfaces. This will be accomplished by first providing an overview of protein adsorption processes and then by addressing the special case of the design of surfaces to prevent protein adsorption. Following these topics, attention will be focused on highlighting several of the most interesting relatively recent techniques that have been developed and applied to further our understanding of the submolecular-level mechanisms involved in how surface chemistry influences the orientation, conformation, and organization of adsorbed proteins. The continued development of our understanding of these processes is critical if we are to get beyond the current era of surface design largely by trial and error, and move into an era where surfaces are proactively designed to directly control adsorbed protein bioactivity, and thereby control cellular response. This article will then close with a conclusion section that addresses future directions toward the continuing goal of the development of truly biocompatible materials for the design of medical devices for improved patient care. OVERVIEW OF PROTEIN SURFACE INTERACTIONS As introduced in the prior section, the surfaces of synthetic biomaterials (e.g., polymers, metals, and ceramics) are generally not bioactive themselves. Rather, surface bioactivity is provided by the proteins that adsorb to the biomaterial surface following exposure of the surface to biological fluids. The types and the amounts of proteins adsorbed determine the types and surface density of the bioactive sites that may be available for cell interactions, and the orientation, conformation, and packing density of the adsorbed proteins determine whether the available bioactive sites are presented in a manner such that they can Encyclopedia of iomaterials and iomedical Engineering DOI: /E-EE Copyright # 2005 by Taylor & Francis. All rights reserved. 1

2 2 iomaterials: Protein Surface Interactions be recognized by the membrane-bound receptors of cells as they interrogate the adsorbed protein layer. Accordingly, to control cellular response, it is important to first understand how surface chemistry and surface topology influence the formation of the adsorbed protein layer and the bioactive sites presented by this layer. Although the understanding of protein adsorption to biomaterial surfaces is still far from complete, research over the past several decades has led to a general understanding of the complex and multifaceted processes involved in the interactions between proteins and surfaces. In this section, an overview of protein structure is first presented to establish a framework from which protein adsorption behavior can begin to be understood. Following this, an overview of the current understanding of protein adsorption behavior to biomaterial surfaces is presented, initially for the relatively simple case involving protein adsorption from single-component protein solutions, and then for the more complex situation involving competitive adsorption from multicomponent protein solutions. This section will then be completed by addressing the special case of the development of nonfouling surfaces that are highly resistant to protein adsorption. Protein Structure efore protein adsorption behavior can be understood, it is first important to understand the basic makeup of protein structure. Proteins are complex copolymers that are made up of four levels of structure, designated as the primary, secondary, tertiary, and quaternary structures. [1] The primary structure involves the specific sequence of the 20 L-amino acids coded for by the DNA of a cell. As part of a protein, a given amino acid is referred to as a peptide residue. Each amino acid, or residue, has the general backbone structure of f NH C a HR CO g, with R designating a specific side-group structure that gives the residue its specific functional characteristics. Accordingly, the amino acids are subcategorized into three primary types: nonpolar (i.e., hydrophobic), -charged, and polar. Examples of each type of amino acid are shown in Fig. 1A. The polypeptide chain formed by the primary sequence is then Fig. 1 Protein structure. (A) Primary structure showing examples of four types of residues with designated three-letter and one-letter codes: nonpolar (alanine, Ala, A), positively charged (lysine, Lys, K), negatively charged (aspartic acid, Asp, D), polar (serine, Ser, S). Color code: gray ¼ C, white ¼ H, blue ¼ N, red ¼ O. () Secondary structures composed of an alanine sequence: a-helix (views from side and end), b-sheet. Same color code as in (A). (C) Tertiary structure (lysozyme, PD ID 7LYZ). Color code: yellow ¼ b-sheet, pink ¼ a-helix, blue and white ¼ connecting loops. (View this art in color at

3 iomaterials: Protein Surface Interactions 3 organized into three basic types of secondary structure: a-helices, b-sheets, and loops that connect helix and sheet elements (Fig. 1). These secondary elements are then organized together to form the tertiary structure (Fig. 1C). Finally, more than one polypeptide chain can be organized together to form quaternary structure, with each individual chain having a separate beginning (N-terminus) and ending (C-terminus), which are typically positively and negatively charged, respectively. The specific functional group character of the peptide residues combined with their spatial organization creates specific bioactive domains in a protein s structure that enable it to perform its specific biological function, whether that is as an enzyme, growth factor, intracellular structure, or extracellular matrix (ECM) assembly, or cell adhesion molecule for cell cell and cell ECM interactions. While the subject of how a polypeptide chain goes about folding into its native structure is still a very active area of research, it is well understood that one of the primary driving forces that causes a protein to fold into its native state, and to be maintained in that state in aqueous solution, is the reduction in free energy due to a decrease in the solvent accessible surface area of the nonpolar residues contained in the protein s structure, i.e., hydrophobic effects. [1] Accordingly, proteins are generally structured with their hydrophobic residues buried within the core of the protein and their hydrophilic residues (charged and polar) lining the protein s solvent accessible surface. While this generally describes a protein s structural makeup, it is common to have some hydrophobic residues on the surface and hydrophilic residues buried within the protein s core. Thus, a protein s surface is highly amphiphilic, meaning that it displays a myriad of different types of functional groups (nonpolar, -charged, polar) on its surface. As a further feature that can greatly influence protein adsorption behavior, each charged residue on a protein s surface has a designated pk a value, which defines the residue s protonation state (i.e., whether it is in its charged or neutral state). The overall charged state of a protein is thus sensitive to the ph of the surrounding solution, with the ph value that results in the protein having a net zero value of charge being designated as the isoelectric point (or pi) of the protein. Further details of protein structure are provided in standard biochemistry texts and readers interested in protein adsorption behavior are encouraged to study this material to provide a foundation from which protein adsorption behavior can then be better understood. [2] Protein Adsorption ehavior from Single-Component Protein Solutions The general model used to describe protein adsorption to a surface is illustrated in Fig. 2. [3] This process P P S k r k f solid material Solid Material surfacesurface can be described by the following chemical reaction equation: P ðm þ nþh 2 O þ S ðr þ vþh 2 O 0 k f k i nh 2 O þðm þ rþh 2 O! k i H 2 O Release to ulk Water k f P S P S þðn þ vþh 2 O P S Fig. 2 Illustration of the protein adsorption process. A protein molecule in solution (P) adsorbs to the material surface ðp SÞ in a reversible manner with forward and reverse reaction rate contants, k f and k r, respectively. The adsorbed protein then may transition to an irreversibly adsorbed state ðp SÞ as a result of orientational or conformational changes that further reduce the free energy of the system, which can be described with a reaction rate constant k i. The release of structured water from the protein and surface back to bulk water represents an important factor that greatly influences these processes. (From Ref. [3].) (View this art in color at ð1þ where P is the protein concentration in solution, S is the concentration of available surface area sites for adsorption, P S is the concentration of reversibly adsorbed protein on the surface with k f and k r designating reaction rate constants for the forward and reverse reactions, respectively, and P S is the concentration of irreversibly adsorbed protein on the surface with k i designating the reaction rate constant for the transition of the adsorbed protein from a reversibly bound to an irreversibly bound state. Water is shown complexed to the protein and surface, which is then released to bulk solution during the various stages of adsorption. This is included to emphasize its importance as a molecular constituent in the adsorption process rather than simply being an inert media that the protein and surface happen to be immersed in. ecause of the presence of the irreversible step shown at the right-hand side of this equation, a surface that adsorbs a protein in a manner that includes this part of the process will inevitably become saturated with the adsorbed protein if sufficient interaction time is provided. The state of the final adsorbed protein layer will be determined by the manner in which the proteins are able

4 4 iomaterials: Protein Surface Interactions to organize themselves on the surface, which is influenced by both protein surface and protein protein interactions in the presence of the surrounding aqueous solution. The proteins that adsorb to the surface in the latter stages of this process will tend to be increasingly inhibited from the irreversible step of this process by steric restrictions caused by the previously adsorbed proteins. ecause of this, the final fraction of the adsorbed protein layer is more likely to be maintained in its native and possibly reversibly adsorbed state. [4] This behavior, which is illustrated in Fig. 3, explains the frequently observed phenomenon that a fraction of protein is found to desorb from an adsorbed protein layer following exposure of the surface to pure buffer solution while another fraction of the protein layer remains irreversibly adsorbed to the surface. This type of behavior can also be caused by a nonhomogenous surface (e.g., owing to defects in the surface chemistry or phase separation), thus presenting one surface phase that irreversibly adsorbs the protein and another more weakly interacting phase that reversibly adsorbs the protein. Assuming the conditions of adsorption from a dilute solution and unity activity coefficients of each species, and defining the standard states of the products and reactants of Eq. (1) as being 1.0 M (protein solution) and 1.0 mole fraction (adsorbed protein and surface), the kinetics of protein adsorption for Eq. (1) can be expressed as: [2,5] d½p SŠ dt ¼ k f ½PŠ½SŠ k r ½P SŠ k i ½P SŠ; d½p SŠ dt ¼ k i ½P SŠ ð2þ where [S], ½P SŠ, and ½P SŠ are the mole fraction of surface sites occupied by bare surface and the reversibly and irreversibly adsorbed protein, respectively, and [P] is the molar concentration of the protein in solution. The change in the free energy for the reversible part of the adsorption process can then be expressed as: DG ads ¼ DG o ads ½P SŠ þ RT ln ½PŠ½SŠ and at equilibrium: DG o ½P SŠ q ads ¼ RT ln ¼ RT ln ½PŠ½SŠ eq C e ðq qþ G ¼ RT ln ¼ RT ln ðkþ ð3þ C e ð1 GÞ Protein Solution Exposure uffer Wash Amount of Protein Adsorbed A Time C D A C D Fig. 3 Illustration of the development of an adsorbed protein layer formed when a surface is exposed to a single-component protein solution of constant concentration followed by a rinsing step of pure buffer solution. (A) Proteins initially adsorb to random areas of the surface, possibly in a reversible fashion. () The previously adsorbed proteins begin reorienting and spreading out on the surface to become irreversibly adsorbed (black ellipses) while new protein molecules (gray circles) continue to adsorb to open surface areas. (C) This process continues until most of the available surface area is covered with irreversibly adsorbed proteins. The last proteins to adsorb may be sterically prevented from reorienting or spreading on the surface by the previously adsorbed proteins, thus being maintained in a more weakly held, reversibly adsorbed state by the surface (white circles). (D) When the surface is exposed to pure buffer solution, the reversibly adsorbed fraction is able to desorb from the surface with the irreversibly adsorbed fraction remaining.

5 iomaterials: Protein Surface Interactions 5 where DG ads and DG o ads are the change in adsorption free energy and change in standard state adsorption free energy, R is the ideal gas constant, T is the absolute temperature, q is the amount of protein per unit area adsorbed in equilibrium with the molar concentration of protein in solution (C e ), Q is the amount of protein per unit area adsorbed at monolayer saturation, G is the fraction of surface saturation (or q=q), and K is the equilibrium constant, which also equals ðk f =k r Þ. Eqs. (2) and (3) provide a means to quantitatively describe the reversible part of the protein adsorption process for a given set of conditions. For example, Table 1 provides values for DG o ads and KC e that correspond to various values of G for two different proteins, lysozyme (14.6 kda) or fibrinogen (340 kda), reversibly adsorbing to a surface in equilibrium with their respective protein solution concentrations of 1.0 mg=ml (commonly used solution concentration for experimental studies). Here, KC e, which is derived from Eq. (2) for the reversible part of the adsorption process, expresses the ratio of the initial adsorption rate of protein ðd½p SŠ=dtÞ onto a completely bare surface compared to the initial desorption rate from a saturated surface into a pure buffer solution over the surface. Taking the energy of a single hydrogen bond to be on the order of 3.5 kcal=mol, which is also on the order of the energy of adsorption of a single hydrophobic peptide residue onto a hydrophobic surface, the values shown in Table 1 indicate that if the interaction of either of these proteins with a surface involves a decrease in DG o ads of the system no more than that of the formation of about one additional hydrogen bond, or one hydrophobic contact, then the adsorbed amount can be kept to about 0.1% of a monolayer of adsorbed protein. [6] In contrast to this, a DG o ads value of only 15 kcal=mol (i.e., equivalent to approximately four hydrogen bonds or hydrophobic interactions) will result in essentially 100% monolayer coverage of the surface. Furthermore, considering the ratio of the rates of adsorption to desorption, as indicated by KC e, an adsorption energy of only 15 kcal=mol results in an adsorption rate that is about one million times faster than desorption. In other words, the number of protein macromolecules adsorbed in 1 sec would take over 10 days to desorb, thus effectively representing an irreversible condition. Therefore, considering the size of many proteins (such as fibrinogen) and the resulting large number of functional group interactions involved when such a protein contacts a surface, it is not surprising that the adsorption of proteins to biomaterial surfaces generally occurs in an effectively irreversible manner, even without considering the actual irreversible stage of protein adsorption indicated in Eq. (1). ased on the adsorption processes indicated by Eq. (1), when a solid surface is placed in a solution containing a given protein, the protein will generally tend to rapidly adsorb until it saturates the surface. If the surface is hydrophobic, the protein will tend to adsorb by the various hydrophobic patches of residues present on the protein s amphiphilic surface, with the protein then tending to unfold and spread its hydrophobic core over the surface owing to the thermodynamic driving force to reduce the net hydrophobic surface area of the system exposed to the solvent. [7] Hydrophilic surfaces, on the other hand, tend to interact with the charged and polar functional groups of the protein s surface, thus influencing adsorbed protein orientation, but with a lower tendency to cause the protein to unfold and spread over the surface. [7] ecause proteins generally do not tend to adsorb nonspecifically to themselves (which is why they are soluble in solution Table 1 Corresponding values of the standard state free energy of adsorption ðdg o adsþ and relative adsorption=desorption rates (KC e ) for a given percent surface saturation Percent saturation (C 100) Lysozyme (14,600 Da, 1 mg/ml solution) Fibrinogen (340,000 Da, 1 mg/ml solution) DG o ads (kcal/mol) KC e (unitless) DG o ads (kcal/mol) KC e (unitless) , , ,000, ,000,000

6 6 iomaterials: Protein Surface Interactions in the first place), protein adsorption tends to stop once complete monolayer coverage is achieved. The final organization of the adsorbed protein layer (i.e., the orientation, conformation, and packing arrangement of the adsorbed proteins) depends on the chemical and physical structure of the protein, the surface, and the aqueous solution, and the thermodynamics of the interactions between these system components. It is also strongly influenced by the kinetics of mass transport of the protein to the surface from solution relative to the kinetics of the protein s movement on the surface (i.e., orientational and translational motions), the kinetics of protein spreading on the surface (i.e., conformational changes), and the physical constraints imposed by protein protein interactions on the surface. [8] ecause protein adsorption is influenced by so many factors, many of which result in irreversible processes, the final state of an adsorbed protein layer is history dependent, meaning that very different results can be obtained for the same protein surface system depending on how the adsorption process is carried out. For example, it is generally true that a hydrophobic surface will more strongly adsorb proteins than a neutrally charged hydrophilic surface. ecause of this, it is usually believed that a hydrophobic surface will then adsorb a greater amount of protein. [9] This will tend to be true, however, only if the protein is adsorbed from a solution with high protein concentration such that mass transport to the surface is much faster than the rate of protein spreading=reorientation on the surface. Under this condition, the hydrophobic surface will tend to adsorb more protein than the hydrophilic surface because its stronger attraction for the protein will result in a more densely packed layer of protein at surface saturation. This situation is illustrated in Fig. 4 for adsorption to a hydrophobic surface a compared to a hydrophilic surface b from a relatively concentrated protein solution. The opposite result can occur, however, if the same protein is adsorbed from a very dilute solution, which will greatly slow down the rate of mass transport of the protein to the surface. If the rate of mass transfer is slow compared to the rate of protein spreading=reorientation on the surface once it adsorbs, then the hydrophilic surface may actually adsorb more protein than the hydrophobic surface. This occurs because spreading tends to occur to a much greater extent on the more hydrophobic surface, thus causing an adsorbed protein molecule to occupy a greater area of surface, leading to a lower amount of protein that can be adsorbed before the surface is saturated (illustrated in Fig. 4; conditions c and d ). As a further consequence of this concentration effect, a plot of the amount of adsorbed protein vs. solution concentration can erroneously give the appearance of a Langmuir isotherm adsorption plot (see Fig. 4III), which has often mistakenly been taken to imply a reversibly adsorbing system. [10] In fact, many investigators have wrongly applied Eq. (3) to such isotherms and calculated DG o ads for the adsorption process. [11,12] As illustrated in Fig. 4, however, a Langmuir-looking isotherm will often occur from an irreversible protein adsorption process simply owing to protein spreading effects. In such cases, the calculation of DG o ads is meaningless; complete reversibility must be established as a necessary (but still not sufficient) requirement for the application of the Langmuir adsorption model for the meaningful calculation of DG o ads. The types of processes described in the preceding paragraph have been clearly demonstrated in experimental studies by Wertz and Santore. [7] These investigators compared the kinetics and adsorption behavior of albumin and fibrinogen on hydrophobic and hydrophilic surfaces (CH 3, CF 3, OH silane-functionalized glass) using total internal reflectance fluorescence spectroscopy. In these studies, the rate of mass transfer of the protein to the surface was controlled by varying both the solution concentration and flow rate over each surface, with the adsorption process being continued until the condition of surface saturation was achieved. The total amount of protein adsorbed to each surface was found to directly correlate with the rate of mass transport provided, with the surface area per adsorbed protein (referred to as the protein s footprint) thus increasing as the mass transport rate was decreased. Their results also indicated that the increase in the size of the adsorbed protein s footprint as the adsorption process was slowed down was associated with the reorientation of the adsorbed protein on the hydrophilic surface and conformational spreading on the hydrophobic surface. Similar events have been demonstrated by Agnihotri and Siedlecki, who used tapping-mode atomic force microscopy (AFM) to observe timedependent structural differences in fibrinogen adsorbed to a hydrophilic (mica) vs. a hydrophobic (graphite) surface. [13] In this study, four different orientational states were observed on each surface, with the height of the adsorbed protein on the hydrophobic surface then undergoing a significant reduction over a time period of 2 hr following initial adsorption, while a height increase was actually observed over time on the hydrophilic surface. It was also observed that the AFM tip tended to move the protein around on the hydrophilic surface, but not on the hydrophobic surface, thus suggesting a more tightly bound interaction of the protein with the hydrophobic surface. These results are again consistent with the understanding that hydrophobic surfaces tend to more strongly adsorb proteins and induce conformational spreading, while hydrophilic surfaces tend to more weakly adsorb protein and induce orientational changes in the adsorbed protein as a function of time.

7 iomaterials: Protein Surface Interactions 7 I High Solution Concentration Low Solution Concentration T1 a b c d T2 II Amount of Protein Adsorbed a b d c III Amount of Protein Adsorbed Time Solution Concentration T1 T2 Kinetics of Protein Adsorption c a Protein Adsorption Isotherm Fig. 4 Illustration of the history dependence of the final state of a protein layer adsorbed from a single-component protein solution. (I) Comparison of the resulting adsorbed protein layers when the same protein is adsorbed from a high vs. low protein concentration solution on hydrophobic (a, c) vs. hydrophilic (b, d) surfaces at time points T1 and T2 (T1 < T2), with T2 representing the final condition at surface saturation. (II) The amount of adsorbed protein vs. time for conditions (a) (d) showing that mass transport controls the kinetics of the initial adsorption process as opposed to surface chemistry, with the final amount adsorbed at surface saturation dependent on both mass transport and surface chemistry. (III) Resulting irreversible adsorption isotherm with reversible Langmuir isotherm-like appearance. The concentration dependence is due to a change in the molecular footprint of the adsorbed protein as opposed to an equilibrium exchange with the protein in solution. Each point on the isotherm plot represents a fully saturated irreversibly adsorbed protein layer. (From Refs. [7,10].) Protein Adsorption ehavior from Multicomponent Protein Solutions When more than one type of protein is present in solution, a competitive process occurs between the different proteins for adsorption to the surface. The first important factor that influences the adsorption of proteins from a multicomponent system is that the mass transfer rate of a given solute molecule to a surface is directly related to its solution concentration and inversely related to its molecular weight. [14] Accordingly, when a material is exposed to a solution containing several different soluble proteins, such as blood plasma, the more concentrated and smaller proteins tend to adsorb to the surface first, and then be displaced by larger, more strongly interacting proteins that may arrive at the surface at a later point in time. [15] This exchange process, known as the Vroman effect, was first recognized by Vroman and Adams in the late 1960s for the case of fibrinogen exchange from various types of surfaces. [16,17] Although initially believed to be unique for fibrinogen exchange, numerous subsequent studies have indicated that this is actually a general phenomenon that occurs for many other types of proteins, including albumin, IgG, and fibronectin. [16] Interestingly, this process is observed to occur even for proteins that are otherwise irreversibly adsorbed to a surface. Thus, while an adsorbed protein may not tend to desorb from the surface under the flow of pure buffer solution, it may be displaced from the surface by another more strongly interacting protein that can more successfully compete for the same surface functional groups, thus exchanging itself for the previously adsorbed protein. This process is illustrated in Fig. 5. Many interesting studies have been conducted over the past several decades to investigate the competitive nature of proteins; only a few of which will be highlighted here. Readers are therefore referred to the Further Readings section for further references related to this subject. As an example of the intriguing types of behavior described by Vroman effects, fibrinogen preadsorbed

8 8 iomaterials: Protein Surface Interactions A A surface Surface surface Surface A surface Surface A surface Surface Fig. 5 Illustration of the Vroman effect. Protein, which is initially adsorbed to the surface, is displaced from the surface in an exchange reaction by protein A owing to the ability of the residue functional groups presented by A to form more stable bonds with the available binding sites on the surface than the residue functional groups of. (View this art in color at on a negatively charged, hydrophilic glass surface has been observed to be readily displaced following exposure of the surface to various dilutions of blood plasma. [16] Immunochemistry studies have revealed that fibrinogen is primarily displaced by high molecular weight kininogen (HMWK) under these conditions. In other studies involving the preadsorption of fibrinogen to polyethylene, the behavior is quite different. [16,18] While fibrinogen is also displaced from a polyethylene surface following exposure to blood plasma solutions, it has been shown that HMWK is not involved in the exchange process. Instead, other proteins, especially high-density lipoproteins, have been shown to be responsible for fibrinogen exchange. Warkentin et al. have conducted studies with multicomponent protein solutions that shed some light on why this difference occurs. [19] Using hydrophobic-gradient surfaces, they found that HMWK exhibited distinctly higher affinity for the hydrophilic end of the gradient-surface compared to other proteins, thus demonstrating that HMWK s ability to displace fibrinogen from hydrophilic but not hydrophobic surfaces is simply due to HMWK s inherent ability to bind most strongly to hydrophilic surfaces rather than due to a specific interaction between HMWK and fibrinogen. More recent studies by Cremer and coworkers have provided insights into some of the actual molecular-level mechanisms that influence Vroman effects with regard to fibrinogen displacement from surfaces. [20] Their studies showed that at a ph of 8.0, fibrinogen adsorbs to a negatively charged silica surface via the relatively small positively charged ac domains of the protein, while the larger D and E domains of fibrinogen, which are negatively charged, are not adsorbed to the surface because of electrostatic repulsion. This results in the protein being only weakly bound and enables it to be readily displaced from the surface by other, more strongly adsorbing proteins like HMWK. However, at a ph of 3.2, the negative charges of both the protein and this surface are largely neutralized, thus eliminating the electrostatic repulsive effects that previously prevented the D and E domains from adsorbing to the surface. Under these conditions, the D and E domains of fibrinogen are able to participate in the adsorption process via hydrogen bonding and van der Waals interactions with functional groups on the silica, which enables fibrinogen to bind much more strongly to the silica surface. This was observed to reduce the rate of fibrinogen displacement from the silica surface by other plasma proteins by over two orders of magnitude. Other studies have shown that Vroman effects are not only influenced by surface chemistry and solution ph, but also, not surprisingly, by many other variables including solution composition, temperature, and wall shear strain rate. [16] Other interesting protein adsorption effects due to protein protein interactions have been shown by Wertz and Santore. As part of the same study referred to in the previous section, Wertz and Santore studied the competitive adsorption behavior between fibrinogen and albumin. [7] In these studies, the surfaces were first saturated with albumin and then aged under an albumin containing solution, following which the surfaces were exposed at various time points to a fibrinogen solution. The amount of fibrinogen adsorption and albumin desorption was then monitored as a function of time by separate radiolabeling studies. The results from these studies indicated that albumin saturated each surface very quickly and then maintained a constant surface density irrespective of aging time. Following exposure to the fibrinogen solution, they found that

9 iomaterials: Protein Surface Interactions 9 about the same amount of albumin was displaced by the fibrinogen irrespective of prior albumin exposure time. However, the amount of fibrinogen adsorbed was found to be strongly dependent on the aging time of the adsorbed albumin. This was interpreted to indicate that although the amount of albumin adsorbed did not change with time during the aging phase of the experiment, the preadsorbed albumin was able to undergo substantial reorganization during this time period, more effectively covering the surface area, and perhaps strengthen its interaction by interprotein associations, such that less surface area was then available for subsequent fibrinogen adsorption. This effect was observed to occur most strongly on the hydrophobic surfaces, with the implication that surface reorganization on the hydrophilic surface mainly occurred as a result of reorientational effects, while the hydrophobic surfaces induced additional effects due to conformational spreading. A final study that will be addressed here, reported by Lee et al., compared the competitive binding capabilities on glass for a wild-type T4 lysozyme and two single-residue mutants, one designed to be more conformationally stable and one less conformationally stable than the wild-type lysozyme. [21] Using binary mixtures of each combination of the proteins, the authors were able to show that in each case, the more stable protein was more easily displaced from the surface by the less stable protein, thus nicely demonstrating that conformation stability plays an important role in the competitive adsorption behavior of proteins. Surfaces that Are Resistant to Protein Adsorption Surfaces that are resistant to protein adsorption, also referred to as nonfouling surfaces, represent a special case in the study of protein surface interactions. The development of nonfouling surfaces is of substantial interest for marine, food processing, biomems, and biosensor applications, in addition to their potential application for the development of medical devices with improved biocompatibility. Of the various types of surface chemistries that have been studied, poly(ethylene glycol) (PEG; [ (O CH 2 CH 2 ) n OH]) has received the greatest amount of attention because it has been widely demonstrated to be very resistant to the adsorption of proteins from aqueous solution. Many of the earlier attempts to understand and explain the resistance to protein adsorption imparted by PEG-coated surfaces centered on the Alexander-De Gennes theory of polymer interfaces. [22] This theory essentially presents a physical view that supposes that protein resistance is provided by steric repulsion of the protein due to the conformational entropy of the very flexible PEG chains on the surface, which acts as a physical barrier that prevents a protein from approaching the underlying substrate. The development of this theory, however, does not address why the protein does not favorably interact with the PEG molecule itself. In fact, its development is based on the a priori assumption of equal attraction between the PEG chains, water, and protein, thus inherently neglecting any enthalpic contribution to this process. In a relatively recent paper by Morra (22), this deficiency is highlighted and the viewpoint expressed that a more sophisticated theory is needed to explain the nonadsorptive behavior of PEG. [22] Morra suggests that the physical view be combined with a chemical view that includes the strongly favorable enthalpic interactions between the PEG chains and water. This position is supported by a report by Kjellander that addresses the thermodynamics of PEG water interactions and indicates that PEG apparently interacts with water structure in a very energetically favorable manner that is rather unique among the class of polyether polymers, which are otherwise largely insoluble in water. [23] While providing an improvement over the relatively simple steric repulsion theory of PEG, this approach still neglects the other side of the issue, which is equally important, namely, why are PEG protein interactions thermodynamically less favorable than the interactions of PEG with water? Although a definitive theory regarding how to design molecular structures to provide resistance against protein adsorption has not yet been agreed upon, several attempts have been made to characterize the general qualities necessary for the design of such surfaces. Several studies using oligo(ethylene glycol) (OEG) functionalized alkanethiol self-assembled monolayer (SAM) surfaces have demonstrated that protein adsorption resistance of OEG with either hydroxyl or methoxy end groups ([ (O CH 2 CH2) n X; X: OH, O CH 3 ) is dependent on the degree of polymerization and chain packing density. [24 27] Herrwerth et al. have shown that protein resistance requires that the degree of OEG polymerization be 2 or more (i.e., n 2) and suggested that the chains must be sufficiently spaced to enable water to penetrate the OEG layer, especially for the methoxy-terminated surfaces. These findings were summarized by stating that protein resistance requires both internal and external hydrophilicity. Similar findings for hydroxyl- and methoxyterminated OEG, and other functional groups, have been reported by Whitesides and coworkers. [27,28] Of particular interest, Ostuni et al. have reported values for the protein resistance provided by 48 different surface functional types covalently linked to SAM surfaces, many of which were very different from OEG, although still providing a high level of resistance to protein adsorption. [27] In particular, several compounds with dimethyl amine ( N(CH 3 ) 2 ) and methoxy

10 10 iomaterials: Protein Surface Interactions ( O CH 3 ) end groups were shown to adsorb very low levels of protein compared to the other functionalities tested. ased on these findings, these authors stated that four common elements were present in each of the protein-resistance functionalities found in their studies: the functional groups 1) were hydrophilic; 2) included hydrogen bond acceptors; 3) did not include hydrogen bond donors; and 4) were neutrally charged. One immediate problem with this set of four necessary characteristics, however, is that condition (3) is violated by the most protein-resistant molecular structure tested, namely hydroxyl-terminated OEG. Thus, while these studies provide very interesting results, the requirement that protein-resistant functional groups not be hydrogen bond donors obviously cannot be universally applied. Another limitation to the abovedescribed studies is that they again focus on functional group water interactions and do not address the issue concerning why these interactions are thermodynamically preferred over interactions between the functional groups and the protein. As indicated from their adsorption=desorption plots, the protein-resistant surfaces reported by Ostuni et al. represent systems exhibiting reversible adsorption behavior with low surface coverage, and thus the results from this study can be considered from a thermodynamic perspective with the approximate determination of the standard state free energy of adsorption ðdg o ads Þ calculated from Eq. (3) above. [27] The data provided by Ostuni et al. are reported as the percent monolayer adsorption of protein (lysozyme and fibrinogen) from 1.0 mg=ml solutions, which, fortuitously, are the same example conditions provided in Table 1 of this entry. Accordingly, the DG o ads values in Table 1 indicate that under equilibrium conditions, to achieve 1% monolayer coverage or less requires that DG o ads 3:0 kcal=mol and DG o ads 5:0 kcal=mol for the adsorption of lysozyme and fibrinogen, respectively. Considering fibrinogen, which has a side-on molecular footprint of 235 nm 2 (5.0 nm 47 nm) compared to 0.21 nm 2 per OEG surface group on a homogenously functionalized SAM surface, an adsorbed fibrinogen molecule would potentially interact with over 1000 OEG chains. [7,29] This thus averages out to DG o ads 0:005 kcal=mol per OEG chain, which is approximately stating that DG o ads 0 kcal=mol for this interaction. Considering the thermodynamic relationship of DG o ads ¼ DHads o TDSo ads,itis clear that to provide protein resistance, the interaction of a surface with a protein should thus ideally result in DHads o 0 and DSads o 0; or in other words, the unbound hydrated state of the surface functional group should be lower in enthalpy and higher in entropy than when bound to a protein. Considering enthalpy, this will be provided if the surface functional group is able to bond with water more favorably than with functional groups of the protein. Considering entropy, the functional group must be able to maintain a higher state of system entropy when bonded to water than when bonded to the protein. Of course, the water structure itself must also be considered, with the surface being able to maintain the surrounding water structure in a free-energy state that is equal to or lower than that of bulk water. With this thermodynamic perspective, it is possible to interpret the results shown by Hernwerth et al. and Ostuni et al. somewhat differently, with a key feature being that water is much smaller and more mobile than the functional groups presented by a protein. [26,27] Accordingly, if hydrogen bonding groups are presented by the surface molecules in a manner such that water can access the groups much more readily than can the functional groups of a protein, then a lower state of enthalpy will be provided for the hydrated, nonadsorbed state. Similarly, because of the mobility of water, water is able to maintain hydrogen bonding with a flexible surface chain without substantially restricting either the chain s or the water s configurational space, whereas the bonding of the protein to a flexible surface molecule will substantially restrict the motion of the surface chain, thus decreasing the entropy of the system relative to the nonbonded state. Accordingly, the longer the surface chain, the greater the entropic penalty will be for restricting its motion. These arguments are consistent with experimental evidence regarding the types of molecular structure that provide resistance to protein adsorption. Longer OEG chains that are not restricted by too high a packing density are more resistant to protein adsorption (entropic contribution), and chains that have wellhydrated hydrogen bonding groups that are relatively inaccessible to the protein, such as ( O CH 2 CH 2 X) and ( N (CH 3 ) 2 ), are resistant to protein adsorption (enthalpic contribution). It is interesting to note that similar molecular features are found in phosphatydlcholine molecules presented by cell membranes and 2-methacryloyloxyethyl phosphorylcholine molecules for surface functionalization that have been designed as a biomimetic of the cell surface, both of which provide a very high level of resistance to protein adsorption. [2,30,31] PROING SUMOLECULAR MECHANISMS OF PROTEIN ADSORPTION EHAVIOR While a great deal has been learned from studies focusing on the amount and type of protein that adsorbs to various types of surfaces, if surfaces are to be designed to actually control protein adsorption behavior, it will be necessary to develop a much more detailed understanding of the submolecular mechanisms involved in protein surface interactions, and how these interactions

11 iomaterials: Protein Surface Interactions 11 influence the orientation, conformation, packing, and subsequent bioactivity of adsorbed proteins. Fortunately, several very exciting relatively new technologies are being developed and applied that should greatly advance the understanding of protein surface interactions. Several of these technologies are highlighted in this section. Drobny, Stayton, and coworkers are developing solid-state nuclear magnetic resonance (ssnmr) techniques for the study of peptide surface interactions to enable the distance between isotopically labeled atom types to be measured with subangstrom resolution. [32,33] y the careful selection of which atoms are labeled in their systems, they have been able to quantitatively determine the secondary structure of adsorbed peptides and to identify the types of peptide residues that specifically interact with functional groups of the surface for several different types of surfaces. This method thus enables specific interactions governing the structure of adsorbed peptides and their associated surface functional groups to be targeted and directly probed. Circular dichroism spectropolarimetry (CD) has also been demonstrated by several groups to be able to quantitatively determine the secondary structure of adsorbed peptides and proteins, as well as provide information regarding the adsorbed tertiary structure for proteins containing tryptophan residues. [34 37] In particular, the high sensitivity of this method has enabled its use to be extended to proteins adsorbed onto flat surfaces in addition to particles in solution, thus expanding its versatility to a much broader range of materials while also avoiding issues related to signal noise problems due to light scattering by dispersed solid particles. Recent studies by Hylton et al. have demonstrated that surface chemistry strongly influences the amount of adsorption-induced loss in the a-helix structure of both fibrinogen and albumin and, more importantly, platelet adhesion following protein adsorption was found to directly correlate to the degree of structural change in the adsorbed layer of each protein. [36] In another very interesting study, Engel et al. combined CD with stop-flow fluorescence spectroscopy and demonstrated that up to 70% of adsorption-induced unfolding of bovine a-lactalbumin (LA) on polystyrene nanospheres in aqueous solution occurs within 15 msec of nanoparticle exposure, thus providing direct experimental measurement of the kinetics of the adsorption-induced unfolding process. [35] In a very interesting follow-up experimental study by Engel and coworkers, submolecular events related to adsorption-induced LA unfolding were probed using hydrogen=deuterium (H=D) amide proton exchange and 2D NMR. [38] This methods provides an excellent technique to probe the effect of adsorption on protein structure because H=D exchange is directly related to the degree of solvent accessibility of a given peptide residue. When the H=D exchange behavior for the native LA in solution was compared to adsorbed LA, the investigators found distinct differences in H=D exchange before and after adsorption, thus providing direct insights into how LA unfolds during adsorption by identifying the specific residues that transitioned from a buried to a solvent-exposed state following adsorption. The results of this study support the presence of distinctly different populations of adsorbed protein one that exhibited substantial unfolding and another, smaller population with nearnative-state H=D exchange behavior. This finding was believed to occur because the last fraction of adsorbed proteins was prevented from unfolding because of the physical restrictions provided by the previously adsorbed proteins, which is in agreement with the process illustrated in Fig. 3. Castner and coworkers are developing a different experimental technique to probe adsorbed protein orientation and conformation using static time-offlight secondary ion mass spectrometry (ToF-SIMS), which uses an ion beam to release molecular fragments from the top nm of an adsorbed protein layer. [39,40] Using multivariate analysis techniques, the released fragments are able to be associated with their parent peptide residue types. [41] Thus, if the adsorbed protein layer is thicker than 1.5 nm (which is smaller than the smallest dimension of most clinically relevant proteins), and if the protein structure has an anisotropic distribution of peptide residues, the specific distribution of peptide residues present in the ToF-SIMS spectrum can be used to provide information relating to the orientation and conformation of the adsorbed protein layer. Castner and coworkers have already demonstrated that differences in both the type and conformational state of adsorbed proteins can be determined with this method, with continued research being conducted to further develop and evaluate its potential as a tool for the characterization of adsorbed protein structure. Advances in biochemistry techniques have also greatly assisted in the study of protein adsorption behavior. As previously noted, extensive peptide residue mutation studies have been conducted with T4 lysozyme. [42] McGuire and coworkers have used this versatile model protein system to elucidate relationships between protein stability, adsorption behavior, and adsorbed-state bioactivity. [21,43] García and coworkers have conducted extensive studies with recombinant fragments of fibronectin, which is a clinically relevant ECM cell-adhesion protein with cellbinding domains recognized by various types of integrins displayed by fibroblasts. [44,45] These studies have clearly demonstrated that surface chemistry substantially influences the adsorption behavior of the cell-binding domains of this protein, with subsequent changes in the

12 12 iomaterials: Protein Surface Interactions integrin expression and adhesion strength of fibroblast cells. In another set of recombinant protein studies, Tang and coworkers have demonstrated that a specific bioactive site located in the C-terminus segment of the D-fragment of fibrinogen, which is not accessible in the soluble state of this protein, becomes exposed when fibrinogen adsorbs to a biomaterial surface, with this bioactive site serving as an important mediator of the inflammatory response. [46] In addition to experimental methods, substantial advances in computational hardware, software, and analysis methods have been made over the past few decades. These advances, coupled with the structural determination of thousands of proteins that are now available via the Protein Data ank (PD), have now provided the capability to conduct all-atom molecular simulations to investigate the atomic level behavior of proteins in solution. [47] While these capabilities have been widely developed and exploited for application to the study of protein folding and ligand receptor interactions for rational drug design, they represent a largely untapped resource for assisting in the understanding of protein adsorption behavior on biomaterial surfaces. [48] Latour and colleagues have been working over the past several years to develop the necessary methods to enable molecular modeling to be accurately applied to simulate protein surface interactions. [49,50] As an example from this work, Fig. 6 shows the interactions of a 30 kda segment of fibrinogen with two different SAM surface chemistries. [49] As illustrated, distinctly different types of peptide residue surface interactions were predicted for each type of SAM surface, with subsequent hydrophobic, electrostatic, and hydrogen-bonding interactions being observed. In other studies, Jiang and coworkers and Kreuzer and coworkers have applied molecular simulation methods to theoretically investigate protein surface and OEG water interactions to gain further insight into the possible submolecular interactions responsible for these types of adsorption systems. [51 54] CONCLUSIONS With the advent of tissue engineering and regenerative medicine, the goals of biomaterial surface design have shifted from a focus on attempting to minimally disturb the surrounding biological processes to the design of surfaces that are able to direct them to attain a desired response. Although the desired type of response varies widely depending on the specific application, be it the incorporation of an implant without fibrous capsule formation or thrombosis, or the regeneration of a targeted tissue or organ, at the fundamental level the biological response will still largely be governed by the bioactive signals that are provided by the biomaterial surface. It has now been widely accepted that protein adsorption behavior must be controlled to control the bioactivity of a biomaterial surface. [55,56] For many applications, the complete prevention of protein adsorption may represent the optimal condition, thus eliminating the presentation of bioactive signals from the surface altogether. Proteins, however, are still the body s natural means of directing biological processes and thus, if properly harnessed, still possess the greatest potential Fig. 6 (A) Molecular model of a 30 kda segment from the C-terminus end of the g-chain of fibrinogen (gfg, PD ID 1FID) over a SAM surface in 150 mm saline. Water molecules and salt ions surrounding the protein and surface, which were present during the simulation, are not shown for clarity. (, C) Close-up views of the resulting peptide surface interactions observed during the 5.0 nsec molecular dynamics simulations for adsorption of gfg to a negatively charged COOH-SAM surface () and a hydrophobic CH 3 -SAM surface (C). Note that positively charged peptide residues (e.g., lysine, arginine) preferentially interacted with the COOH-SAM surface (indicated by in ) while hydrophobic peptide residues (e.g., alanine, leucine) preferentially interacted with the CH 3 -SAM surface (indicated by # in C). (View this art in color at