Chapter 12
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
Protein Surface Interactions and Biocompatibility: A Forty Year Perspective John L. Brash* McMaster University, School of Biomedical Engineering, Dept. of Chemical Engineering Hamilton, Ontario, Canada L8S 4L7 *E-mail:
[email protected]
Research in the author’s laboratory on proteins at interfaces over the past four decades, with biocompatibility as the underlying theme, is reviewed. A principal focus in fundamental studies has been the dynamics of adsorbed protein layers in both simple systems and real biofluids. Investigations on protein resistant surfaces based on surface modification with poly(ethylene oxide) and poly(2-methacryloyloxyethyl phosphorylcholine) (polyMPC) have also been a major theme. Most of our work has been done in the context of biomaterials for use in blood contact and the hypothesis that blood compatibility, and biocompatibility in general, depends on control of protein adsorption has guided much of our more recent efforts. Specifically we have hypothesized that surfaces which prevent non-specific protein adsorption and promote specific adsorption of target proteins, should have improved biocompatibility. Blood contacting surfaces that promote fibrinolysis by specific adsorption of plasminogen and tissue plasminogen activator, or that prevent coagulation by adsorption of antithrombin have been investigated on this basis.
© 2012 American Chemical Society In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
Introduction Over the past four decades our lab has carried out research aimed at understanding how proteins behave at interfaces. Such understanding is crucial for many applications in biomedicine and biotechnology, or indeed for any situation in which a surface is in contact with a protein-containing fluid. The list of such applications is long and continues to grow. It includes: medical devices (blood contacting, ophthalmic, soft tissue, orthopedic, dental), biosensors, protein purification (chromatography, filtration), therapeutic apheresis, hemodialysis, biomicrofluidic systems, drug delivery, solid phase immunoassays, biofouling (biomedical & non-biomedical (marine fouling bio/food processing)), and liposomes. Clearly motivation for studying proteins at interfaces is strong and research continues at a brisk pace. Very little was published on this topic prior to 1970. Early on I wrote a review on the subject; published in 1971 it has the title “Adsorption of Proteins and Lipids to Nonbiological Surfaces” (1). There are 127 references in total and of these, only 67 are papers reporting original research on protein adsorption. A search of Pubmed under “protein adsorption” returns a total of more than 20,000 papers, ranging from 3 to 8 per year in the 1950s to over 1100 in 2011. The field “took off” in the 1960s and has continued to grow over the ensuing 40 plus years. In this chapter our work on proteins at interfaces is reviewed briefly under several headings that represent the broad themes of the research. Most of the work has been done in the context of understanding tissue-material interactions, with particular emphasis on blood-material interactions. Accordingly blood proteins and the solid-solution interface are emphasized.
Fundamental Aspects: Single Proteins and Simple Mixtures The main “fact” or axiom of proteins at interfaces is that proteins are highly surface active: they adsorb. Alternatively it may be stated that “any protein will adsorb to any surface”. This behavior is attributable to the macromolecular and amphipathic character of these molecules. Besides adsorption itself, interactions include desorption (at some interfaces), re-orientation with respect to the interface, exchange of adsorbed and dissolved proteins, diffusion of adsorbed proteins over the interface, aggregation, conformational change, and denaturation. Some of our early work addressed the question of reversibility of adsorption (2). This has been an issue of considerable interest (3). It is not clear that it is entirely settled, but the consensus appears to be that desorption is slow, perhaps infinitely so, to the point where adsorption can be considered irreversible. One must be careful in these discussions to distinguish between protein leaving the surface in response to a decrease in solution concentration (true desorption) and protein being displaced by adding a surfactant such as sodium dodecyl sulfate (SDS). Besides its fundamental importance, reversibility is of interest for two main reasons: (1) It is of interest to know whether protein that adsorbs and then desorbs conserves its conformational and biological properties: knowledge in this regard is 278 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
lacking; (2) The application of equilibrium thermodynamics requires reversibility. Estimates of affinity constants from adsorption “isotherms” have frequently been reported (4, 5), but generally speaking these are “apparent”, not true affinities. The finding by many groups that adsorption versus concentration data fit well to the Langmuir equation is in general fortuitous and does not mean that the Langmuir model with its assumptions of reversibility, no protein-protein interactions etc describes the interactions. The “isotherm” or adsorption-concentration behavior of proteins can be summarized as showing that adsorption increases (usually rapidly) with concentration, and reaches a plateau or quasi plateau indicating a limiting surface coverage. This limit has been interpreted in terms of the number of molecules per unit area and the molecular dimensions of the protein, often suggesting the formation of more or less tight protein monolayers (6–9). Notwithstanding the fact that adsorbed proteins tend not to desorb, they nonetheless can exchange with other proteins in solution either of the same or different species. Using double radiolabeling methods we showed in early work that in the albumin-polyethylene system a dynamic equilibrium existed with equal rates of adsorption and desorption, and that the rate and extent of exchange increased with increasing shear rate and concentration (10). Self exchange was also demonstrated for fibrinogen interactions with polyelectrolyte complexes (11). In this work it was also shown that several populations of adsorbed fibrinogen molecules existed: rapidly exchanging, slowly exchanging and non-exchanging. Extrapolating from this work it is likely that there is in general a range of behavior with exchange occurring rapidly in some systems and in some not at all.
Figure 1. Explanation for non-occurrence of desorption and possible mechanism of self exchange. Reproduced with permission from reference (12). Copyright 1985 Plenum. 279 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
It seems paradoxical that proteins can exchange but not desorb, since in order to exchange, the protein has to leave the surface, i.e. desorb. A reasonable explanation is shown in Figure 1 due to Andrade (12). The key idea is that proteins are attached multivalently to the surface, i.e. these very large molecules are bound to the surface not just by one interaction but several. So for desorption to occur all of the bonds must be broken simultaneously or in rapid succession, an event of very low probability. However if protein molecules are present in solution, as in the exchange scenario, one can imagine a single adsorbed protein-surface bond breaking and simultaneously a single bond forming with an adjacent solution molecule. Repetition of this process for the same adsorbed protein would result in whole molecule exchange, i.e. one molecule would leave and another arrive to take its place. This constitutes a concerted, synergistic process involving interactions between the dissolved and adsorbed molecules. Our interests in real, as opposed to model biological systems led us to focus our attention on adsorption from systems of more than one protein and ultimately plasma and blood. We showed that in binary solutions of albumin and fibrinogen where the ratio of the protein concentrations was similar to that in blood, fibrinogen was adsorbed preferentially on both glass and polyethylene surfaces with glass showing the stronger preference (13). Surface enrichment of fibrinogen was also observed by Lee et al (14) for several hydrophobic surfaces in solutions of fibrinogen, gamma globulin and albumin. From these and other early observations it was tempting to conclude that fibrinogen would always be a major component of the protein layers adsorbed from blood. And indeed this is often assumed to the present day. However, this is much too simplified a view as is elaborated below in the discussion on plasma and blood.
Adsorption from Plasma and Blood Since we were interested mainly in unraveling the complexities of blood-material interactions we turned our attention to the adsorption of proteins from plasma and blood. Our experimental technique of choice was (and is) trace labeling of proteins with radioiodine, of which there are two convenient isotopes, I-125 and I-131, allowing measurement of the adsorbed quantities of two proteins simultaneously with excellent precision. This remains, in our view, probably the best available method for the measurement of one protein in a complex system of many. In work on adsorption from plasma (15), we investigated the kinetics of albumin, IgG and fibrinogen interactions with several surfaces, some hydrophilic, some hydrophobic. We discovered, contrary to predictions from single protein studies, that fibrinogen was not always a major component of the adsorbed layer. It was not detected on any of the hydrophilic surfaces, and, of the hydrophobic surfaces, it was adsorbed substantially to polystyrene but only transiently to polyethylene and siliconized glass. Albumin and IgG also showed lower adsorption than expected based on their concentrations in plasma. These results suggested that the plasma itself interacts with initially adsorbed proteins and that 280 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
since the major plasma proteins were adsorbed in relatively small quantities, proteins of lower concentration, and perhaps even trace proteins, might be important. The concept of displacement of one protein by another during plasma-surface interactions was also hinted at by these data. Beginning in the 1960s Vroman had been working on the adsorption of fibrinogen from plasma, noting in particular that fibrinogen, detectable on the surface at short exposure time by immunological methods, could not be detected at longer times (16). This process was referred to as “conversion” of the protein to a form no longer able to bind antibody. Using radiolabeling methods it was shown by us and by Horbett that “conversion” was in fact material loss of fibrinogen from the surface (17, 18). From data on fibrinogen adsorption versus plasma concentration or time (Figure 2) we concluded that fibrinogen was adsorbed substantially at short time or from highly diluted plasma. At longer time or in more concentrated plasma the adsorbed fibrinogen was displaced by other plasma components (presumably proteins). These components would, perforce, be of higher binding affinity and lower concentration than fibrinogen. These phenomena constitute the basis of the Vroman effect which in essence describes competitive adsorption in multi-protein systems.
Figure 2. Fibrinogen adsorption from plasma to glass versus time at varying plasma concentration (plasma diluted with buffer). The higher the plasma concentration, the more rapidly the fibrinogen is displaced. Reproduced from reference (17). Copyright 1984 Schattauer. 281 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
We subsequently developed a more extensive data base (19) and found that most surfaces, except for some hydrophilic ones, showed a Vroman effect, but with quantitative differences from surface to surface, i.e. the position and height of the adsorption maxima with respect to time and plasma concentration were variable. Identification of the displacing protein or proteins became a topic of considerable interest. Vroman et al suggested that proteins of the intrinsic coagulation pathway, particularly high molecular weight kininogen (HMWK), were responsible (20, 21). This was supported by work in our lab using plasma deficient in, or deficient in and then reconstituted with, HMWK (22) which showed that the effect was greatly diminished in HMWK-deficient plasma and then was restored when purified HMWK was added back. This work also showed that fibrinogen was not removed from surfaces by plasmin digestion as might have been suspected. Further support for HMWK in this role came from experiments with binary solutions of fibrinogen and HMWK which showed that HMWK displaced initially adsorbed fibrinogen from a glass surface (23). Indeed, fibrinogen adsorption in this system as a function of concentration was exactly analogous to that in plasma. Bantjes et al proposed that, in addition to HMWK, high density lipoprotein (HDL) might be an important fibrinogen displacing species (24). Recent work in our lab has shown that HDL is highly surface active and that apolipoprotein AI (and therefore HDL itself) is extensively adsorbed from plasma to a wide range of materials (25, 26). Despite this and other work, including on the Vroman effect (27), it cannot be said that competitive adsorption is understood in any rigorous sense. Noh and Vogler (27) have argued that, due to the finite capacity of the surface, smaller proteins are adsorbed in preference (number basis) to larger ones and smaller proteins are exchanged for bigger ones. Vroman has suggested that in plasma, adsorption is sequential and that over time more abundant proteins are replaced by less abundant ones (28). This is supported by observations indicating the sequence albumin > IgG > fibrinogen > fibronectin > HMWK. That there is a sequence seems beyond dispute, but it seems likely that binding affinity as well as concentration should be a factor. Indeed there may well be several other factors that combine to determine the composition of the protein layer on any surface at any time. There is also little doubt that the layer is complex and contains many proteins as opposed to one or even a few regardless of the time of contact. Related to competitive adsorption our lab has expended much effort in attempts to determine the composition of protein layers adsorbed from plasma. It is of course important to understand how surface properties affect the competition so that ultimately adsorption can be controlled. In these experiments surfaces are exposed to plasma under defined conditions and the adsorbed proteins are eluted with sodium dodecyl sulfate (SDS). The eluates are then subjected to SDS polyacrylamide gel electrophoresis (SDS-PAGE) for separation and to immunoblotting for identification. We have used panels selected from among 25 antibodies directed against some of the more abundant proteins, coagulation factors, complement proteins, cell adhesive proteins and others. Of course this approach is limited in the sense that only proteins probed for can be identified, leaving many others (on the order of several hundred) as “not tested”. A second 282 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
limitation is that proteins must be SDS-elutable to be identified. Given the surfactant power of SDS this is less of a concern. In general we have shown that any protein probed for on any surface is found to be present, although quantitatively there is variation from protein to protein and surface to surface (29–40). Figure 3 shows SDS-PAGE data for several surfaces including hydrophobic and hydrophilic polymers, liposomes, a heparinized surface (CBAS), and a polysulfone hemodialyzer membrane (25). The gel patterns appear similar with a concentration of bands in the 50-70 kDa region and a prominent band at ~27 kDa which has been identified as apolipoprotein AI. However there are important differences as revealed by immunoblotting.
Figure 3. SDS-PAGE (12%, gold stained, reduced) of proteins eluted from surfaces after plasma contact. Reproduced from reference (25). Copyright 2002 Elsevier.
Figure 4 shows immunoblots of proteins eluted from plasma-exposed PVC. It is seen that virtually all of the proteins tested for were detected. Strong responses were seen in particular for fibrinogen, plasminogen, C3, transferrin, albumin, IgG, vitronectin, and apolipoprotein A-I. Other surfaces showed different patterns but responses were almost always strong for fibrinogen, albumin, vitronectin and apolipoprotein AI.
283 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
Figure 4. Immunoblots of proteins associated with PVC following incubation in human plasma. Reproduced from reference (33). Copyright 2002 Wiley.
A more recent approach to the analysis of protein layer composition is the use of proteomics based methods. These often involve elution of adsorbed proteins, separation by 2-D gel electrophoresis, excision of individual spots (proteins) from the gel, proteolytic digestion to give smaller peptide fragments and identification by mass spectrometry. The latter requires the use of large data bases and informatics analysis (41). These methods have been widely applied to biofluids such as plasma, serum, tear fluid etc. Application to adsorbed proteins is less developed although a few reports have appeared (42–45). These studies emphasize the huge complexity of the adsorbed layers. Zhang et al identified 88 different proteins adsorbed from plasma to polystyrene nanoparticles (45), and Capriotti et al identified about 130 different plasma proteins adsorbed to cationic liposomes (43). Interestingly it was found in the latter work that differences among different cationic liposome compositions were minimal. It is clear that much work remains to be done to unravel competitive protein adsorption and to determine the “proteome” of adsorbed proteins. Given the huge range of surfaces of interest, the variables such as contact time and concentration, the different bio-environments such as blood, tissue, tear fluid etc, this is a formidable and forbidding task. However, given the widely held belief that biocompatibility is strongly influenced by adsorbed proteins it remains true that detailed knowledge of the protein layers and their dependence on surface properties may be the key to solving this problem. 284 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
Resistance to Adsorption Beginning in the 1980s a new impetus for research on proteins at interfaces came from interest in so-called “protein resistant” surfaces, and the discovery that certain hydrophilic polymers (notably polyethylene oxide) incorporated at the interface could inhibit adsorption. This has since become a major focus of research. Non-fouling surfaces are of interest for a number of applications including marine structures, bioprocessing equipment, biosensors, and biocompatible materials. Prevention of nonspecific or indiscriminate protein adsorption is the primary requirement for such surfaces. Indeed nonspecific protein adsorption may be regarded as the “enemy” of biocompatibility since in general, tissue-material interactions begin with the rapid formation of an adsorbed protein layer, followed by cell interactions which are strongly dependent on the layer composition. For example platelets in blood adhere specifically to adsorbed fibrinogen. A key objective in biocompatibility research is to gain control over protein adsorption (46) and one element of such control is to minimize nonspecific adsorption. The creation of a protein resistant surface is problematic since proteins, being amphiphilic and macromolecular, are highly surface active. Thus any interface in a protein containing system will tend to accumulate protein, and as of now there is no known surface which can prevent protein adsorption completely. As mentioned, the most successful strategy is to modify the material by the incorporation of various polymers (generally hydrophilic polymers) at the surface. Examples are polyethylene oxide (PEO), also referred to as polyethylene glycol (PEG), and its derivatives, polysaccharides (47), polyacrylamide (48), polyvinylpyrrolidone (PVP) (49), polyzwitterionic materials such as polybetaines (50–52), and mixed charge SAMs with exactly balanced positive and negative charges (53). A number of methods have been used to incorporate these modifiers including simple coating, blending and surface grafting by photochemical, radiation, plasma and chemical techniques. Work in our lab has used several of these approaches and has focused on PEO/PEG and poly(2-methacryloyloxyethyl phosphorylcholine) (polyMPC) as modifiers. In early work PEO was grafted to a polyurethane-urea by introducing isocyanate groups into the surface followed by reaction with hydroxy- or amino-terminated PEO (35, 54). On these surfaces adsorption from single protein solutions was reduced by up to 95%; adsorption of fibrinogen from plasma was also greatly reduced. Making use of gold-thiol chemistry, we prepared PEO-modified surfaces by grafting thiol-terminated PEO to gold. An interesting result from this work was that for 750 and 2000 MW PEO layers, resistance to fibrinogen increased with chain density, but at densities greater than ~0.5/nm2 adsorption increased again (55). It was suggested that at higher chain density, beyond the maximum in protein resistance, the PEO layer lost water giving a surface having reduced protein resistance. Using the same gold-PEO system, we also showed that the “distal” chain end group of PEO had an effect on protein resistance. For hydrophobic, methoxy, chain ends the maximum in resistance at a “critical” PEO density was again observed, whereas for hydrophilic, hydroxyl, chain ends, 285 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
resistance increased up to the same “critical” density and then remained constant (56). Thus the distal chain end chemistry, most likely reflecting hydration state in this case, affected protein resistance but only at chain density greater than the critical value. This behavior is reminiscent of the leveling off of resistance as PEO chain length increases (57), which may also reflect a maximum or limiting hydration state. In work on grafting by atom transfer radical polymerization (ATRP) we used MPC and oligoethylene glycol methacrylate (OEGMA) as monomers and silicon as substrate (58–63). ATRP gives grafts of uniform and controlled (“dial in”) chain length. Graft density can be controlled by the initiator density. For poly(MPC) surfaces in the brush regime (~0.4 chains/nm2), we found that the adsorption of fibrinogen and lysozyme decreased with increasing chain length of the grafts (59). Grafts of chain length 200 (MW 59,000 Da) gave very low adsorption levels corresponding to reductions of greater than 98% compared to unmodified silicon. Experiments using mixtures of the two proteins showed that suppression of adsorption on the poly(MPC)-grafted surfaces was similar for both, i.e. it was not dependent on protein size or charge. We also prepared surfaces with a range of chain density and molecular weight of poly(MPC) (61), and showed that fibrinogen adsorption was affected by both graft density and chain length, but more strongly by graft density. OEGMA has PEO side chains and grafting gives a surface with a high density of these short PEOs. Surfaces prepared by graft ATRP of OEGMA on silicon (60) were found to be strongly protein resistant. In experiments to compare PEO and MPC moieties, OEGMA (MW 300, PEO side chains of average length 4.5) and MPC (MW 295, phosphoryl choline side chains) were grafted to silicon with varying density and chain length. Adsorption of fibrinogen and lysozyme to both surface types were found to decrease with increasing graft density and chain length (62). Moreover adsorption on the MPC and OEGMA surfaces for a given chain length and density was essentially the same; surfaces were equally resistant to fibrinogen and lysozyme whether of low or high graft density. The data from experiments with these surfaces suggested that the main determinant of the protein resistance of these surfaces is the “water barrier layer” resulting from their hydrophilic character (63). Analogous poly(OEGMA) surfaces were prepared on a polyurethane-urea rather than the model silicon substrate discussed above (64, 65). OEGMAs of varying MW (i.e. varying side chain length) were investigated. Trends in protein resistance similar to those for the silicon based surfaces were observed. The adsorption of both fibrinogen and lysozyme decreased with increasing poly(OEGMA) main chain length for a given side chain length (number of EO units). For a given main chain length, the fibrinogen adsorption level did not change significantly with increasing side chain length. Surprisingly, lysozyme adsorption increased with increasing side chain length, possibly due to decreasing graft density as monomer size and footprint on the surface increased. Perhaps the most promising materials we have worked with in terms of protein resistance are blends of PEO with polyurethane (7, 66–68). A great advantage of such materials is the simplicity of the preparation method. Solutions of the two components are mixed, a film is cast or a part is coated and the solvent evaporated. 286 In Proteins at Interfaces III State of the Art; Horbett, T., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2012.
Downloaded by UNIV OF CALIFORNIA SANTA CRUZ on October 14, 2014 | http://pubs.acs.org Publication Date (Web): December 12, 2012 | doi: 10.1021/bk-2012-1120.ch012
In an aqueous environment the PEO migrates to the interface producing a surface that is rich in PEO as indicated by water contact angles and surface elemental composition. The PEO component in these materials is a triblock copolymer with a polyurethane-like middle block which interacts with the PU matrix to prevent leaching in aqueous contact. Materials with a range of copolymer loading (PEO content) in the matrix and varying PEO block molecular weight were prepared. On blends containing 20 wt % copolymer, fibrinogen adsorption was found to be reduced by greater than 95% for all PEO block molecular weights from 550 to 50007. At lower copolymer content (
