Sensitivity of Protein Adsorption to Architectural Variations in a Protein

ARTICLE pubs.acs.org/Langmuir

Sensitivity of Protein Adsorption to Architectural Variations in a Protein-Resistant Polymer Brush Containing Engineered Nanoscale Adhesive Sites Saugata Gon and Maria M. Santore* Department of Polymer Science and Engineering, University of Massachusetts, Amherst, Massachusetts 01003, United States

bS Supporting Information ABSTRACT: Patchy polymer brushes contain nanoscale (5 15 nm) adhesive elements, such as polymer coils or nanoparticles, embedded at their base at random positions on the surface. The competition between the brush’s steric (protein resistant) repulsions and the attractions from the discrete adhesive elements provides a precise means to control bioadhesion. This differs from the classical approach, where functionality is placed on the brush’s periphery. The current study demonstrates the impact of poly(etheylene glycol) (PEG) brush architecture and ionic strength on fibrinogen adsorption on brushes containing embedded poly-L-lysine (PLL, 20K MW) coils or “patches”. The consistent appearance of a fibrinogen adsorption threshold, a minimum loading of patches on the surface, below which protein adsorption does not occur, suggests multivalent protein capture: Adsorbing proteins simultaneously engage several patches. The surface composition (patch loading) at the threshold is extremely sensitive to the brush height and ionic strength, varying up to a factor of 5 in the surface loading of the PLL patches (∼50% of the range of possible surfaces). Variations in ionic strength have a similar effect, with the smallest thresholds seen for the largest Debye lengths. While trends with brush height were the clearest and most dominant, consideration of the PEG loading within the brush or its persistence length did not reveal a critical brush parameter for the onset of adsorption. The lack of straightforward correlation on brush physics was likely a result of multivalent binding, (producing an additional dependence on patch loading), and might be resolved for univalent adsorption onto more strongly binding patches. While studies with similar brushes placed uniformly on a surface revealed that the PEG loading within the brush is the best indicator of protein resistance, the current results suggest that brush height is more important for patchy brushes. Likely the interactions producing brush extension normal to the interface act similarly to drive lateral tether extension to obstruct patches.

’ INTRODUCTION A fundamental understanding of protein adsorption can be complicated by processes such as unfolding, binding or releasing ions, and restructuring of water. Equally challenging has been the design of biomaterial surfaces that avoid protein adsorption altogether. Ideally, preventing protein adsorption might be as simple as preventing the first, short-time steps of adsorption. Likewise, manipulation of protein-containing surfaces could benefit from control of the first few steps of protein adsorption. While time scales over which proteins unfold on surfaces can last from minutes to hours, we focus here on engineered surfaces that control initial protein capture. The relevant time scales are the first few seconds of the protein surface encounter. A popular strategy for imparting biocompatibility is the modification of surfaces with polymer brushes that sterically repel approaching proteins. For brushes to be effective, both their chemistry and interfacial architecture must be appropriate. The polymer itself must be hydrophilic and well-solubilized in water, charge-neutral, and a hydrogen-bond acceptor.1 Polyethylene glycol (PEG) is one of a handful of polymers meeting these criteria, making it a popular choice for the creation of protein-resistant brushes. The brush must further have an appropriate architecture to screen the full range of van der Waals and electrostatic interactions r 2011 American Chemical Society

between proteins and the underlying substrate.2 This latter criterion has been usually interpreted to mean that the brush extension must exceed the range of substrate protein attractions; however, it is understood that proteins must not penetrate or compress the brush.2 A tall but insufficiently dense brush may not be fully effective. Indeed, some reports suggest that the mass of the tethered polymer (such as PEG) is a better predictor of brush performance than the calculated brush height.2,3 In the case of PEG chains tethered to surfaces by different anchoring chemistries, a PEG chain mass exceeding 1.1 mg/m2 led to almost complete repellence of serum proteins, independent of the length of the tethers themselves, in the range from 2000 to 10 000 molecular weight.3 The ease by which brush architecture can be altered (both in terms of tether length and density) makes polymer brushes an attractive choice for manipulating the initial encounters of proteins with surfaces. Functionalization of the free brush ends has traditionally allowed tuning of bioadhesion.4 A second materials-based strategy to manipulate protein adsorption is control over the area of protein surface contact. Received: August 22, 2011 Revised: October 28, 2011 Published: October 31, 2011 15083

dx.doi.org/10.1021/la203293k | Langmuir 2011, 27, 15083–15091

Langmuir

Figure 1. Interfacial structure and binding energy per patch with increased patch loading: (A) Full brush containing no patches; (B) full brush contains some buried patches; (C) increased patches reduces backfill, giving shorter brushes; (D) substantial patch loading causes patches to lose identity while brush structure is lost (PEG chains in the “mushroom” regime).

With an eye toward constraining protein denaturing5,6 or tuning the adsorbing sites,7 a few groups have created interfaces where the protein-attracting regions are small and the remaining surface is neutral or repulsive. Restricted contact areas have been potentially achieved by employing phase-separated polymer8 and silane ligands,9,10 surface-immobilized or suspended nanoparticles,11 15 and other strategies.16 (Antibody-based and receptor capture of targeted proteins represents a special case outside the current scope.) Varied observations regarding adsorption (or lack of it) on surfaces whose adhesive elements are similar to the protein size have not produced a consensus on the importance of the relative sizes of proteins and adhesive islands. Some argue that the adsorption site must be at least as large as the protein;17 however, counter examples exist.18 The philosophy in the Santore lab, generally, is that the energy rather than the size of the initial contact is critical.19 While increased contact area will generally provide an opportunity for strengthened interactions, sufficiently strong attractions might result from contact regions smaller than the size of the target. For example, we demonstrated that micrometer-scale silica spheres could be captured and held in flow by electrostatic attractions to single surface-immobilized 10 nm nanoparticles.20,21 More recently we demonstrated bacterial capture by single nanoparticles on a weakly repulsive surface.22 The current study focuses on surfaces containing protein-adhesive cationic elements or “patches”, about 10 nm in size (roughly the same as the protein), randomly positioned at the base of a PEG brush whose steric repulsion limits patch protein interactions. This approach in Figure 1 opposes the conventional wisdom of placing attractive functionality on a brush periphery.4 While tethered ligands can selectively bind targets via biomolecular specificity, the adhesive elements buried within patchy brushes bind targets with sharp selectivity, as a result of competing steric repulsion from the brush and attractions to the adhesive elements.23 Without prerequisite biofunctional specificity, patchy brushes can be fabricated at smaller expense and tend to be more robust. Previous studies of patchy brushes revealed a threshold in the patch surface loading necessary for protein adsorption.24 That is,

ARTICLE

if the PEG brush contained less than the critical surface concentration of patches, no protein adsorption occurred. This was preliminary evidence for multivalent binding: It was speculated that a protein needed to adhere to multiple patches simultaneously in order to be retained. In a second work,23 the utility of patchy brushes with protein adsorption thresholds was demonstrated: The thresholds for a series of proteins correlated with their size, further supporting the hypothesis of multivalent protein capture. Additionally, surfaces engineered between the two thresholds of the competing species in a protein mixture could selectively adhere one protein. Sharp selectivities exceeding 100 (the ratio of the surface to bulk composition) were demonstrated for proteins with similar charge. To date we have examined the influence of cationic patches on protein capture in only one type of brush and at a single ionic strength.23,24 The current study, employing fibrinogen as a model protein (chosen because it is well studied on a variety of surfaces and because its dimensions are similar in magnitude to the patch size), varies the relative ranges of electrostatic attractions and steric repulsions, via systematic variation in ionic strength and brush architecture. Given the complexity of the results, further variations in the patch composition and size are addressed separately. The current study reveals how variations in electrostatic and steric forces shift the adhesion thresholds in ways that might be engineered to fine-tune selectivity. Additionally, the work provides fundamental perspective into the properties of polymer brushes: the ability of tethered chains to extend laterally on a surface and the accessibility of bare protein-sized spots, spatial fluctuations, or flaws as the base of a brush to proteins in solution. Technical Background and Strategy. The patchy brushes in this study, represented schematically in Figure 1, were created as documented previously:24 Tightly controlled amounts of a cationic polymer, poly-L-lysine (PLL), were adsorbed on a silica surface at well-characterized mass transport conditions so the surface loading is limited well below saturation. With PLL coverage sufficiently low that adsorbed coils are isolated, each coil acts as a randomly situated cationic patch, about 10 nm in size, an estimate based on its free solution size from light scattering. (Though ionic strength was varied during the protein adsorption portions of the study, all patches and brushes were deposited at uniform conditions, in pH 7.4 phosphate buffer of ionic strength 0.026 M, having a Debye length of 1.96 nm. These electrostatically screened conditions impart flexibility to the PLL backbone so that it is in a random coil conformation, rather than a rigid rod. Indeed, the observed 10 nm radius is consistent with a well-solvated random coil rather than a rod.) The remaining surface was backfilled via adsorption from solution, with a PLL PEG graft copolymer to prevent protein adsorption on the bare silica. Key in this approach, the PLL component of the patches was the same as the anchoring part of the PLL PEG backfill brush. This led to exclusion of PLL PEG from the regions where PLL is already adsorbed. Notable features of these interfaces24 are (1) the net positive charge of a saturated layer of PLL on silica (0.4 mg/m2), suggesting a local positive charge in the vicinity of the isolated patches of the current study; (2) retention of patches on the surface during backfilling and subsequent use of the surfaces at the conditions of these studies, with the backfill brushes also being robust at the conditions of interest;25 and 3) substantial fibrinogen adsorption onto saturated PLL layers on silica, suggesting protein patch attractions, at least in the absence of the brushy backfill. 15084

dx.doi.org/10.1021/la203293k |Langmuir 2011, 27, 15083–15091

Langmuir

ARTICLE

In this study, three different backfill brushes were studied. The copolymers from which these brushes were created are described in Table 1. Here the molecular weight of the PLL anchor group was fixed and the amount and molecular weight of PEG side chains was varied. Table 2 describes the brushes formed from the adsorption of these copolymers. With the molecular compositions established via NMR in Table 1, the PEG and PLL content of each brush follows from measurements of the adsorbed amount, for instance via reflectometry. Other brush properties, such as brush height and persistence length, were calculated according to the Alexander DeGennes model26,27 as described previously24 and summarized here in the Supporting Information. Worth remembering concerning the use of terms like “brush height” is that the picture of the brush itself is approximate. The Alexander DeGennes brush model27 we implement, while unrealistic in its treatment of the interface as a sharply defined (stepfunction) region having a constant polymer solution concentration, is powerful in its nearly accurate predictions of key brush properties, as reviewed by Milner.28 The parabolic brush treatment, thought to give a more accurate description of brush features including the interfacial concentration profile of segments (and estimating a nominal brush height to be about 25% greater than the Alexander DeGennes treatment), has been shown by Kent29 to be achieved only in rare experimental instances where the loading of chains at the interface is much greater than we achieve in the current work. The significance of this for the current work is that the description of our brushes is imprecise because of the lack of an appropriate model. On the other hand, the three brushes presented in this work are distinct in their architectures, protein interactions, and, separately, their dynamics.25 The brushes in Table 2, without the incorporation of PLL patches, completely resist the adsorption of serum proteins such Table 1. Copolymer Samples and Schematics

as fibrinogen and albumin (within detectible limits of 0.01 mg/ m2), and related brushes have been shown to eliminate protein adsorption from serum below the detectible limits.30,31 The nearly perfect protein resistance of the brushes themselves is a key element in this work. The current study focuses on the ability of the patches to interact with and adsorb proteins in the presence of protein-repelling brushes. The only exception to the nearly perfect protein resistance is brush 1, which, when k 1 = 4 nm, adsorbs 0.08 mg/m2 of fibrinogen. This adsorption is, however, reversible, with fibrinogen washing off the surface (giving coverages below 0.01 mg/m2) when the ionic strength is increased. This suggests that only with shorter brush 1, when the ionic strength allows electrostatic interactions with the base of the brush to be felt by proteins at the periphery, does slight and weak protein adsorption occur on the outside of the brush. For thicker brushes 2 and 3 and for higher ionic strengths, all brushes screen electrostatic and other protein substrate interactions. As a general rule, adsorbed PLL patches occupy the negative silica and reduce the amount of PLL PEG backfill necessary to saturate the interface. The details of the progressive reduction in backfill with increasing PLL patches are presented in Figure 2, reproduced from a recent paper on brush protein interaction physics.25 An interesting recent finding is that low levels of PLL patches do not alter the brush; however, at above some level of PLL coverage, the amount of PLL PEG needed to saturate the surface is reduced substantially. The amount of PLL that can be incorporated at fixed PLL PEG loading and the degree to which PLL PEG backfill is reduced depends on the PLL PEG architecture and the brush itself, varying substantially among the three samples in our system. Since brushes without patches contain limited amounts of PLL anchoring groups on the silica beneath the brush relative to a saturated PLL layer (0.4 mg/m2), small amounts of homopolymer PLL can be accommodated before the backfill is affected.

’ MATERIALS AND METHODS Poly L-lysine hydrobromide (PLL) with a molecular weight of 20 000 was purchased from Sigma and used for both the adhesive cationic patches and the anchoring component of the three copolymer brushes in this study. Three PEG brushes were created using the three different PLL PEG copolymers listed in Table 1. The 2000 MW PEG component was the N-hydroxysuccinimidyl ester of methoxypoly(ethylene glycol) acetic acid (Laysan BioInc.). This reactive compound was not available with a 5000 molecular weight PEG polymer, so PEG sodium valeic acid (PEG SVA) from Laysan was employed. The copolymers were synthesized following the original procedure by Huang et al.30 and modified slightly by us.24 Notably, Huang et al. and Kenausis et al.,30,31 in

Table 2. Brush Architecture

designation saturated adsorption (mg/m2)

PLL

brush 1

brush 2

brush 3

homopoly 20K 0.4

PLL (2.7)PEG(2K) 1.1

PLL (2.2)PEG(5K) 0.9

PLL (4.7)PEG(5K) 1.3

adsorbed PEG (mg/m2)

0

0.94

0.85

1.16

adsorbed PLL (mg/m2)

0.4

0.16

0.05

0.14

area/copolymer (nm2) area/PEG tether (nm2)

83

206 3.6

680 9.6

247 7.2

“blob” diameter, or tether spacing (nm)

1.9

3.1

2.7

number of blobs

4.7

5.1

6.4

brush height, nm

9

15.5

17.2

15085


Langmuir

Figure 2. The amount of PLL PEG copolymer adsorbed after PLL patches are deposited on the surface. Adapted from ref 29. Lines guide the eye. the their development of protein-resistant coatings formed by PLL PEG copolymers, determined the proportions of PEG and PLL that ultimately lead to complete protein resistance, guiding our choice of molecular architectures: Of the full library of compounds studied, they report protein resistance for certain copolymers having 2000 and 5000 molecular weight PEG tethers, as long as the grafting ratio (of total lysine units to the number of PEG chains) is appropriate for each tether length. Copolymer synthesis was carried out by dissolving the desired amounts of PLL and end-reactive PEG in pH 9.1 sodium borate buffer, stirring for 6 h, dialyzing against pH 7.4 phosphate buffer, and then dialyzing against deionized (DI) water. The product was freeze-dried and stored at 20 °C. Copolymers were characterized using 1H NMR, with D2O solvent and a Bruker 400 MHz instrument. The lysine side chain peak areas ( CH2 N ) at 2.909 ppm and the PEG peak areas ( CH2 CH2) at 3.615 ppm were compared to determine the grafting ratio. Bovine serum fibrinogen from Sigma (F8630-1G) was used as received. Acid-etched (overnight in concentrated sulfuric acid, and well-rinsed in DI water) microscope slides (Fisher Finest) were used as the adsorption substrate. XPS has revealed these to have a silica surface.32 Polymer and protein adsorption was carried out in a laminar slit shear flow cell at a wall shear rate of 5 s 1 using polymers dissolved in 0.01 M pH 7.4 phosphate buffer (0.002 M KH2PO4 and 0.008 M Na2HPO4). Notably, all patchy brushes were created in this buffer, and for protein adsorption studies at other ionic strengths, the buffer was switched subsequently, a process which did not alter the originally formed patchy brushes. This buffer has a Debye length of k 1 = 2 nm. Buffers with k 1 = 4 or 1 nm were created by diluting to an overall concentration of 0.005 M or operating at a greater concentration. Patchy brushes were created by flowing a 5 ppm PLL solution over the surface for a specific amount of time, allowing only the desired amount of PLL to adsorb and be retained before the flow was switched back to buffer. Then a 100 ppm buffered solution of the particular PLL PEG of interest was flowed to backfill the remaining surface before the flow was again switched to buffer. This procedure has been documented and studied in detail,23,24 including a study of the brush stability.25 The buffer was then switched to that of the protein adsorption study and fibrinogen introduced at 100 ppm in the buffer of interest for 20 30 min, prior to switching back to buffer. Adsorption of polymer and protein was observed using a custom-built near-Brewster optical reflectometry instrument in which 633 nm parallelpolarized laser light impinges on the interface from the substrate side.33 The reflected intensity is proportional to the square of the adsorbed interfacial mass, with the calibration constant determined from the optical properties of the solution and the adsorbed layer or

ARTICLE

Figure 3. Reflectometry traces for fibrinogen adsorption on brush 2, with different amounts of PLL patches listed. The ionic strength of 0.026 M corresponds to k 1 = 2 nm. from an appropriate kinetic-based calibration. Notably, slightly different calibration constants apply to polymer- and protein-adsorption portions of the runs.

’ RESULTS Figure 3 presents a series of raw reflectometry data for fibrinogen adsorption onto different surfaces with varying amounts of PLL patches in a PLL PEG brush. This example, with brush 2, is typical of all the data obtained with the other brushes. Most importantly, without any patches at the base of the brush, there is no protein adsorption. With increasing amounts of PLL patches, fibrinogen adsorbs more readily, both in terms of the initial rate and the coverage after approximately 20 min (at which time buffer was reinjected). Indeed with 10 700 patches/ μm2, the fibrinogen coverage approaches that on a saturated PLL layer without PLL PEG.24 The 0.4 mg/m2 of PLL at saturation on silica translates to 12 000 patches/μm2. The nearly linear initial data in Figure 3 allow determination of the initial protein adsorption rate, providing insight into single protein surface encounters at short times. The data from Figure 3, along with other data for brushes 1 and 3 and data at different ionic strengths, are summarized in Figure 4, which plots the initial fibrinogen adsorption rate as a function of PLL patch density within each brush. Each part of Figure 4 summarizes data for a different brush, highlighting the influence of the Debye length. (The corresponding data summaries based on the adsorbed fibrinogen amounts at 20 min, rather than the initial rates, are included in the Supporting Information and mirror the features we discuss here.) The main feature of each data set is its extrapolation to a finite x-intercept. We term this x-intercept the “adhesion threshold”, the minimum surface loading of adhesive patches needed to produce protein adhesion. Previous studies of the fibrinogen adhesion threshold in a brush similar to brush 1at k 1 = 2 nm supported the interpretation that the threshold is indicative of multivalent protein capture. Because they are buried within the brush, individual patches are too weakly binding to each adhered fibrinogen. For several patches to simultaneously engage a protein (providing the requisite binding energy), surfaces must be loaded with sufficient patch density, so that the 47 nm length 15086


Langmuir

Figure 4. Summary of initial fibrinogen adsorption rates, comparing adsorption at different Debye lengths, for the three brushes in parts A, B, and C. One example of threshold quantification is illustrated in red, explaining the range in threshold values obtained either by extrapolating data to the x-axis or by using the conditions where fibrinogen adsorption was the smallest measurable.

of the fibrinogen molecule exceeds the average patch spacing. This interpretation, which was reasonable for brushes like brush 1, might also hold for brushes 2 and 3, discussed below. Figure 4 examines the impact of Debye length on the fibrinogen adsorption rates for the three different brush types. As a general trend, a particular brush is most adhesive at low salt conditions, where the Debye length is 4 nm, and least adhesive at higher ionic strengths, where the Debye length is 1 nm. This is a result of the range and strength of electrostatic attractions from the PLL patches toward fibrinogen. While there may be some interpretation as to the exact value of each threshold (for instance illustrated in Figure 4C), it is remarkable that a 3 nm change in Debye length shifts the position of the threshold by several thousand patches/μm2 for any particular brush. For instance, in the case of brush 1 in Figure 4A, the threshold shifts from near

ARTICLE

Figure 5. Summary of initial fibrinogen adsorption rates, focusing on brush architecture for different Debye lengths in the different parts of the figure.

1000 patches/μm2 at k 1 = 4 nm to around 5000 patches/μm2 or greater when k 1 = 1 nm. The effect is even greater for the thicker brushes 2 and 3. This range of threshold shifts corresponds to about 50% of the possible range of surface composition or capacity for patches. Adhesion onto patchy brushes is therefore highly sensitive to ionic strength, likely a result of the comparable ranges of Debye lengths and brush thicknesses in this study. Fibrinogen adsorption is slight when the Debye length is 1 nm: with hardly any adsorption up to 8000 patches/um2. This patch density was generally the maximum tested because 8000 patches/μm2 comprise two-thirds of a fully saturated PLL layer. Adsorbed PLL coils are not positioned as isolated adhesive islands at high PLL coverages, and further, the polymer brush is not well-established because the backfill amount has become small. In Figure 4C, nonetheless, we extended the range of study slightly, because even with 8000 PLL/μm2 there was no evidence for fibrinogen adsorption on brush 3. The observations of negligible fibrinogen adsorption for k 1 = 1 nm with as many 15087


Langmuir

ARTICLE

Figure 6. Impact of Debye length on the adhesion thresholds for 20K PLL patches in the three brushes. There are two data sets for each brush, showing the range of interpretation when identifying thresholds. Each upper data set represents a linear extrapolation of the rate data to the xaxis. The lower data set indicates when fibrinogen adsorption becomes nonzero, within reasonable limits of detection. A large difference between the two indicates curvature in the rate data of Figures 4 and 5.

as 8000 or 9000 PLL patches/μm2 (and as little as 0.15 mg/m2 of PLL PEG brush) are unexpected, given the lack of an established PEG brush in this regime. Even though the electrostatic attractions between fibrinogen and PLL are short-range at k 1 = 1 nm, the attractions are still expected to be substantial. Figure 5 recapitulates the data from Figure 4, providing a perspective on the impact of the brush choice on the fibrinogen capture rate. For a fixed Debye length in a single panel of the figure, the trends are generally clear. For instance at a Debye length of k 1 = 2 nm, the threshold increases with increases in the brush thickness, with the thresholds and brush numbers ranking in order 1, 2, and 3. The ranking of the thresholds is different, however, at 4 nm: the threshold order is brush 1, brush 3, and brush 2. While there are complexities in the ranking of the thresholds at different ionic strengths, the basis for discussion in the next section, there is a zero-order ranking of the thresholds with the brush thickness: brush 1 always has the smallest thresholds. Brushes 2 and 3, which exhibit slight differences in height, have greater thresholds than brush 1. Figure 6 summarizes the impact of Debye length on the thresholds for the three brushes. The two data sets for each brush (with the curves through the data set drawn only to guide the eye) demonstrate the range of PLL loadings at the threshold, depending on the particular criterion for the threshold. The upper data sets represent extrapolations of the linear part of the data to the x-axis. The lower data sets represent the PLL coverage when fibrinogen adsorption first becomes noticeable, as shown in Figure 4C. The difference between the two varies with conditions. Important to note, the crossing of the data for brushes 2 and 3 is modest if one consistently employs a single criterion for the threshold (linear extrapolation or rigorous onset of protein adsorption). Figure 3 suggests a decay-type functionality for the impact of Debye length. However, with a practical limited range (1 4 nm) in the Debye length, speculating on the functional form is premature.

’ DISCUSSION Complex Influence of Brush Features. To attempt a quantitative interpretation of the protein capture mechanism and the

Figure 7. (A) The adhesion thresholds as a function of (A) the PEG content of the patchy brushes, (B) the corrected brush height, as calculated in the Supporting Information, based on the backfill amounts in Figure 2, and (C) the persistence length, as calculated in the Supporting Information and based on the backfill in Figure 2. The PEG content, brush height, and persistence length without patches is shown on the x-axis in red. In parts A and B points are connected in order of brush number. In part C they are connected in order of persistence length.

impact of steric forces, this study considered systematic variations in brush architecture. Because the brushes are created by adsorption, governed by the competition between entropic stretching of the tethers and the enthalpic binding of the anchors, a nonlinear relationship links the adsorbed amount of the brush and the molecular parameters of the graft copolymers. This translates to simultaneous variation in some brush properties. 15088


Langmuir For instance, Table 2 lists the brushes in order of their height and persistence lengths. The PEG content of the brushes is ordered differently. Prior studies suggest that either brush height or the amount of PEG in a brush correlates well with protein resistance.2,3 That this classical perspective would apply to patchy brushes is unclear: All three brushes in this study were chosen because of their protein repellence before the incorporation of adhesive patches.25,30,31 Adhering proteins are directed onto patchy “flaws”. Figure 5 suggests, at first glance, that the protein adsorption thresholds rank with brush height, rather than PEG content. Indeed, Table 2 shows that brush 1, with its 2K molecular weight tethers, is by far the thinnest brush, though it ranks between brushes 2 and 3 in terms of the amount of PEG it contains. Complicating the correlation is the fact, in Figure 2, that, as the amount of patches increases, the amount of PLL PEG backfill is reduced, reducing the brush thickness. The proper way to consider the impact of brush parameters is to document the condition of the brush at each threshold, rather than the brush height without the addition of patches. Figure 7 therefore summarizes the adhesion thresholds (both the high and low limits as presented in Figure 6) in terms of the calculated brush thickness, PEG content, and brush persistence length (average tether spacing), based on the amount of backfill, from Figure 2, at each threshold. In Figure 7A, with the thresholds for each Debye length presented as a function of the PEG content of the brush at each threshold; in Figure 7B, where thresholds are presented as a function of brush height at each threshold; and in Figure 7C, which tests dependence on persistence length, there appear different behaviors at different Debye lengths. At Debye lengths of 2 and 4 nm, the data exhibit the most nearly clear functionality on brush height in Figure 7B. For the data set with k 1 = 1 nm in any part of Figure 7, the data turn back on themselves (if the data are connected in the same sequence as the ordering for the longer Debye lengths). Alternately, if the data are considered in the order of the brush feature (i.e., rank 1, 3, 2), then the data are nonmonotonic in brush height or PEG content and sometimes reverse their ordering at different Debye lengths. An explanation starts with the observation that, for k 1 = 4 nm, the electrostatic attractions between the patches and the fibrinogen are the strongest. As a result of the thresholds occurring at relatively low PLL loadings, with k 1 = 4 nm, the amounts of adsorbed PLL PEG at these thresholds are similar to the adsorbed amounts in the “full” brushes of Table 2 and also on the x-axes of Figure 7. This suggests that while individual patches attract fibrinogen (at k 1 = 4 nm) too weakly to hold fibrinogen onto a single patch, they are still able, from their positions at the base of the brush, to exert weak attractions toward fibrinogen. While one patch cannot capture fibrinogen, several of them can, as evidenced by the existence of a threshold. Thus, at k 1 = 4 nm, fibrinogen adsorption near the threshold follows the intended patchy brush concept and can be modeled simply by plain brush parameters (Table 2) and a PLL-loadingindependent fibrinogen patch attraction. At the other extreme, with k 1 = 1 nm, the fibrinogen patch interactions are weak, and in the limit of low patch loadings (backfilled by the full brushes), the attractions may be completely screened. Protein adsorption occurs only when there is so much PLL that the brushes are severely compromised compared with their “full properties” in Table 2. In most cases, there is so little interfacial PEG that tethers are likely configured as nonstretched

ARTICLE

Figure 8. Comparing the capture probability for two, three, or five patches or (B) three, four, or five patches in the contact area. The black lines show the result for a contact area of 200 nm2, while gray dashes are for a contact area of 150 nm2.

chains or “mushrooms”. This loss of brushy character suggests data at k 1 = 1 nm be omitted from any attempt at a correlation on physical brush features. The data are retained in the graphs because of their potential practical utility in protein separations. One sees in Figure 7 for k 1 = 2 or 4 nm that, regardless of whether one considers PEG loading, brush height, or persistence length at the fibrinogen capture threshold, there is not an obvious correlation between the threshold and brush architecture. The trend with brush height is, however, better than those with PEG surface loading or brush persistence length. Brush height screens protein substrate interactions in the absence of patches, and it also determines the extent to which small adhesive islands can be accessed. The forces driving chain stretching perpendicular to the substrate also determine lateral chain stretching to obstruct patches. The illustrations in Figure 1 schematically interpret how PLL PEG loading, via brush height, could influence the ability of fibrinogen to access patches. The schematics consider a series of surfaces in which the PLL patch density is increased and the remaining surface is backfilled with one choice of PLL PEG. At low PLL amounts, the effective binding energy per isolated PLL patch is constant. (Both the brush height and its ability to obstruct patches are unchanged.) As more PLL is added to the interface, however, the amount of PLL PEG which can subsequently adsorb is reduced, causing the effective binding energy per patch to increase as the patches become more exposed. A quantitative analysis of the fibrinogen binding energy per patch requires an understanding of the brush structure near the patch, 15089


Langmuir for instance, the extent to which chains extend laterally over the patch. This level of detail is beyond the current scope. Capture Statistics: The Numbers of PLL Patches Needed for Fibrinogen Capture. In general, the binding energy per patch can increase with patch density (due to reduction in backfill), making it challenging to estimate the statistical aspects of fibrinogen adsorption, for instance, the number of patches involved in fibrinogen capture. Simple estimates are possible, especially at k 1= 4 nm, where the per-patch binding energy is nearly constant up to and near the threshold, as a result of the relatively small thresholds and nearly constant amount of backfill over this range, in Figures 2 and 7. Figure 8 explores the statistics of fibrinogen capture, for comparison to the rate data of Figures 4 and 5. Figure 8 is calculated according to a previous statistical model,34 modified slightly as detailed in the Supporting Information. The model calculates the probability of finding a “hot spot” on the surface suitable for the capture of fibrinogen. This probability is expected to be proportional to the surface-controlled capture rate, shown to be the case for multivalent capture of microparticles on heterogeneous surfaces. A hot spot is defined as an area, the size of the fibrinogensurface contact, containing the necessary number of patches for fibrinogen capture. The required number of patches, divided by the contact area, will typically be greater than the average surface concentration of patches, making the adhesive regions “hot spots” relative to the average. As one example, the model can predict the probability of finding a contact region containing at least three patches, as a function of the overall (average) patch concentration (on the x-axis). The model’s assumptions are (1) a random placement of adhesive patches described by a Poisson distribution and (2) a protein substrate contact area of 200 nm2, based on side-on fibrinogen adsorption, with fibrinogen’s measurements being 47 4.5 4.5 nm3. The model does not account specifically for fibrinogen’s elongated shape (it does not require the three sticky patches to be arranged on a line), nor does it make any explicit assumptions about the binding energy, ionic strength, or the effect of the brush. The perspective is purely statistical, but powerful. The resemblance of Figure 8A to Figure 5C, for k 1 = 4 nm, makes the argument that, on brush 1, fibrinogen capture (to the extent that it involves a 200 nm2 protein surface contact area) involves only two patches. This prediction is supported by the observation that, on average, at the fibrinogen capture threshold of brush 1 at 4 nm, the average patch spacing is somewhat less than fibrinogen’s 47 nm length. On brushes 2 and 3, the comparison suggests that a third patch is needed. Brushes 2 and 3 have similar thresholds, and the main difference in the protein capture occurs at higher patch loadings above the thresholds. At these elevated PLL loadings, the structures of brushes 2 and 3 change differently with the incorporation of additional PLL patches, producing the different curvatures of the two data sets. Notably, the assumption of a 200 nm2 contact area for capture does introduce some uncertainty in making these interpretations. However, calculations using a 150 nm2 contact area (gray dashes) give a similar result to that using a 200 nm2 contact area. This supports the interpretation that two or three patches are likely involved in protein capture for brush 1 or brush 2/3, respectively. Further, since at k 1 = 4 nm brush structure is negligibly altered by the presence of patches, it follows that the brushes with 5K tethers (2 and 3) exhibit a reduced per-patching capture energy by about 50% relative to that for brush 1 (with 2K tethers). Figure 8B, which compares the fibrinogen capture probabilities for three, four, and five PLL patches in a 200 nm2 contact

ARTICLE

region, somewhat resembles protein capture at k 1 = 1 and 2 nm in Figure 5A,B. The comparison is now, however, qualitative. In the case of k 1 = 2 nm, where brush structure changes only slightly with increased PLL loading up to the thresholds, it is likely that three or four PLL patches are involved in the capture of fibrinogen, with a few more needed on the thicker brushes, compared with brush 1. Above these thresholds at k 1 = 2 nm, however, the brush structure does change with additional PLL loading, producing differences in the curvatures of the predictions versus the experimental observations. Figures 2 and 7 show, for instance, that brush 3 especially is reduced with increased PLL loading above the thresholds at k 1 = 2 nm: This is likely to be accompanied by an increase in the binding energy per patch, producing the sharp upturn in the data of Figure 5B, compared with the calculations of Figure 8B. In the case of k 1 = 1 nm, fibrinogen capture occurs at sufficiently high PLL loadings that the brushes are substantially altered. Figure 8B suggests that three to five patches are required for fibrinogen capture: Though electrostatic interactions are weakened by the added salt (which produces a large threshold), the thresholds themselves occur when PLL PEG is reduced: only three to five patches are actually involved in capture. The reduced binding energy (from the ionic strength) has been compensated by an increase in binding energy from the reduced brush. Significance and Utility of Tunable Thresholds. Capture thresholds are technologically relevant because of their potential to affect sharp separations of proteins and biological objects up to micrometers in size.21,35 We previously demonstrated, for instance, that fibrinogen and albumin could be almost perfectly discriminated by selective fibrinogen adhesion on a surface that exhibited distinct thresholds for the two proteins.23 Surfaces between the two thresholds can give selectivities on the order of 100 or more (defined as the ratio of surface to bulk solution composition during adsorption from a mixture.) The current work demonstrates that ionic strength and brush structure profoundly influence the position of the protein capture thresholds, allowing thresholds to be placed precisely to accommodate separation targets. Remarkably, however, large variations in the threshold position (density of patches) correspond to only small differences in the numbers of patches involved with capture of fibrinogen. This is a result of the reduced brush height at greater PLL loadings. The Biomaterial Utility and Physics of Flawed Brushes. While it is a general technological goal to produce brushes sufficiently dense and tall to avoid protein adsorption, it is worth considering the consequences when this is not achieved. If a brush is, overall, too short or sparse, proteins might adsorb anywhere on the interface. Separately, it is possible that surface contamination could prevent brush placement in small regions, or with use, some tethers could cleave from the surface at isolated locations. This study shows brushes with localized nanoscale flaws to be extremely effective at obstructing those flaws, especially at physiological conditions. Single flaws alone cannot adhere proteins such as fibrinogen, and small proteins generally are more difficult to capture via small brush flaws.23 Only when the brush is substantially compromised by a large number of such flaws is protein adsorption observed.

’ SUMMARY This study examined the impact of ionic strength and brush architecture on the ability of a model protein, fibrinogen, to adsorb onto 10-nm cationic patches at the base of a protein-repellant 15090


Langmuir PEG brush. Beyond confirming the general expectation that taller/denser brushes more effectively hide the buried “stickers,” interesting and technologically useful behaviors were revealed: (1) In all cases fibrinogen capture was multivalent, involving from two to five patches or patch protein interactions; (2) at k 1 = 4 nm, where the electrostatic patch protein interactions were strongest, protein adsorption started at conditions where the brush structure was uninfluenced by the presence of patches, leading to a fixed binding energy/patch; (3) at k 1 = 1 nm, where the protein patch interaction energy was weakest, the protein adsorption thresholds occurred at high PLL surface loadings, where the amount of backfill brush was reduced, strengthening the weak binding energy; and (4) while ionic strength and brush structure profoundly affected the protein adsorption thresholds, producing a range of thresholds that covered half the possible surface loadings of PLL patches, the impact of these parameters on the numbers of patches actually engaged in protein capture was smaller: two to five patches were needed for capture. This was a result of competing effects of ionic strength and steric repulsions due to altered brush structure and the adhesion thresholds. While a statistical treatment of the numbers of patches engaged in protein capture was qualitatively and, in some cases, quantitatively successful, a quantitative criterion for the physical brush features allowing fibrinogen access to buried 10 nm patches was not forthcoming: A clear correlation on brush height, PEG surface loading at the threshold, or brush persistence length was not discovered. For instance, there was no single brush height or PEG surface loading below which the patches were sufficiently accessible to produce adsorption. Likewise, analysis such as that in Figure 7 did not motivate collapse of data using the ratio of the brush height to the Debye length. This complexity is likely a result of the multivalent mechanism of protein capture, which depends on the binding energy of single patches along with the probability of finding several of them in close proximity. One might speculate that, in the case of univalent capture of proteins onto single patches, critical brush parameters might be revealed. Notably, the current investigation revealed that brush height is a more important factor than the amount of PEG at the interface in controlling multivalent protein binding onto buried patches. This finding opposes the importance of PEG surface loading for the adsorption of serum proteins onto uniform (nonpatchy) brushes. This study, revealing a strong sensitivity of threshold position to ionic strength and brush parameters, bodes well for the use of patchy brushes in separations and biomanipulation applications: A broad range of surface conditions can be accessed for precise control of the amount and rate of fibrinogen capture.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

ARTICLE

’ REFERENCES (1) Ostuni, E.; Chapman, R. G.; Holmlin, R. E.; Takayama, S.; Whitesides, G. M. Langmuir 2001, 17, 5605. (2) Szleifer, I. Curr. Opin. Solid State Mater. Sci. 1997, 2, 337. (3) Pasche, S.; De Paul, S. M.; Voros, J.; Spencer, N. D.; Textor, M. Langmuir 2003, 19, 9216. (4) Lin, J. J.; Bates, F. S.; Hammer, D. A.; Silas, J. A. Phys. Rev. Lett. 2005, 95. (5) Teichroeb, J. H.; Forrest, J. A.; Jones, L. W. Eur. Phys. J. E 2008, 26, 411. (6) Teichroeb, J. H.; Forrest, J. A.; Ngai, V.; Jones, L. W. Eur. Phys. J. E 2006, 21, 19. (7) Dutta, D.; Sundaram, S. K.; Teeguarden, J. G.; Riley, B. J.; Fifield, L. S.; Jacobs, J. M.; Addleman, S. R.; Kaysen, G. A.; Moudgil, B. M.; Weber, T. J. Toxicol. Sci. 2007, 100, 303. (8) Lau, K. H. A.; Bang, J.; Kim, D. H.; Knoll, W. Adv. Funct. Mater. 2008, 18, 3148. (9) Hodgkinson, G.; Hlady, V. J. Adhes. Sci. Technol. 2005, 19, 235. (10) Miller, R.; Guo, Z.; Vogler, E. A.; Siedlecki, C. A. Biomaterials 2006, 27, 208. (11) Linse, S.; Cabaleiro-Lago, C.; Xue, W. F.; Lynch, I.; Lindman, S.; Thulin, E.; Radford, S. E.; Dawson, K. A. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 8691. (12) Lundqvist, M.; Stigler, J.; Elia, G.; Lynch, I.; Cedervall, T.; Dawson, K. A. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 14265. (13) Asuri, P.; Karajanagi, S. S.; Vertegel, A. A.; Dordick, J. S.; Kane, R. S. J. Nanosci. Nanotechnol. 2007, 7, 1675. (14) Fischer, N. O.; McIntosh, C. M.; Simard, J. M.; Rotello, V. M. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 5018. (15) Hong, R.; Fischer, N. O.; Verma, A.; Goodman, C. M.; Emrick, T.; Rotello, V. M. J. Am. Chem. Soc. 2004, 126, 739. (16) Muller, B.; Riedel, M.; Michel, R.; De Paul, S. M.; Hofer, R.; Heger, D.; Grutzmacher, D. J. Vac. Sci. Technol. B 2001, 19, 1715. (17) Wolfram, T.; Belz, F.; Schoen, T.; Spatz, J. P. Biointerphases 2007, 2, 44. (18) Slater, J. H.; Frey, W. J. Biomed. Mater. Res. Part A 2008, 87A, 176. (19) Santore, M. M.; Wertz, C. F. Langmuir 2005, 21, 10172. (20) Zhang, J.; Srivastava, S.; Duffadar, R.; Davis, J. M.; Rotello, V. M.; Santore, M. M. Langmuir 2008, 24, 6404. (21) Santore, M. M.; Zhang, J.; Srivastava, S.; Rotello, V. M. Langmuir 2009, 25, 84. (22) Fang, B.; Gon, S.; Park, M.; Kumar, K.-N.; Rotello, V. M.; Nusslein, K.; Santore, M. M. Colloids Surf., B 2011, 87, 109. (23) Gon, S.; Santore, M. M. Langmuir 2011, 27, 1487. (24) Gon, S.; Bendersky, M.; Ross, J. L.; Santore, M. M. Langmuir 2010, 26, 12147. (25) Gon, S.; Fang, B.; Santore, M. M. Macromolecules 2011, 44, 8161. (26) de Gennes, P. G. J. Phys. (Paris) 1976, 38, 1443. (27) Alexander, S. J. Phys. 1977, 38, 983. (28) Milner, S. T. Science 1991, 251, 905. (29) Kent, M. S. Macromol. Rapid Commun. 2000, 21, 243. (30) Huang, N. P.; Michel, R.; Voros, J.; Textor, M.; Hofer, R.; Rossi, A.; Elbert, D. L.; Hubbell, J. A.; Spencer, N. D. Langmuir 2001, 17, 489. (31) Kenausis, G. L.; Voros, J.; Elbert, D. L.; Huang, N. P.; Hofer, R.; Ruiz-Taylor, L.; Textor, M.; Hubbell, J. A.; Spencer, N. D. J. Phys. Chem. B 2000, 104, 3298. (32) Toscano, A.; Santore, M. M. Langmuir 2006, 22, 2588. (33) Fu, Z. G.; Santore, M. M. Colloid Surf. A 1998, 135, 63. (34) Duffadar, R.; Kalasin, S.; Davis, J. M.; Santore, M. M. J. Colloid Interface Sci. 2009, 337, 396. (35) Santore, M. M.; Kozlova, N. Langmuir 2007, 23, 4782.

’ ACKNOWLEDGMENT This work was made possible by NSF 0932719 and 0805061 and the UMass MRSEC on Polymers. 15091


Sensitivity of Protein Adsorption to Architectural Variations in a Protein

Recommend Documents