■ INTRODUCTION

Functionalized gold nanoparticles (AuNPs) have been integrated into many novel biomedical applications, [1][2][3][4] including in vitro diagnostic testing, 5 cellular imaging, 6,7 photothermal therapy, 8 and drug delivery. 9 Central to the success of these emerging AuNP-enabled technologies is the well-controlled surface chemistry to facilitate targeted binding, 10 cellular uptake, 11,12 low toxicity, 13,14 and evasive properties [15][16][17][18] as the downstream application demands. AuNPs are commonly prefunctionalized with proteins to impart the desired functionality prior to implementation in a biological application. Yet, despite efforts to control the surface chemistry, it is well established that a protein corona forms around the bioconjugate upon exposure to a biological fluid that can alter the composition of the adsorbed protein layer and, consequently, the intended function of the bioconjugate. [19][20][21][22][23][24][25] Thus, it is imperative to understand the fundamental mechanism by which proteins interact with the AuNP to mitigate loss of function due to desorption of the preformed protein layer or to enable the controlled release of the therapeutic proteins. 26 An early study reported that the final composition of the protein corona is thermodynamically determined. 27 These studies were based on the introduction of an unconjugated AuNP to a mixture of proteins in which the most abundant protein adsorbs first but evolves such that the final composition of adsorbed proteins correlates with binding affinity, for example, Vrooman effect. The binding affinity is a combination of several forces, including electrostatic, hydrophobic (entropically driven), and covalent Au-S chemisorption through cysteine residues. Of these interfacial forces, the Au-S interaction is the greatest; 28,29 thus, one could conclude that the greater the number of thiols presented by a protein, the greater the adsorption affinity and greater the representation on the AuNP surface. Fitzkee and Zhang present compelling evidence that cysteines are responsible for the affinity of protein to the AuNP surface and ultimately govern the protein corona on the AuNP surface in several recent works using engineered proteins that vary the number of free cysteine residues. [30][31][32][33] The presumed Vrooman effect is thermodynamically driven, which is predicated on reversible protein-AuNP binding. However, published works demonstrate irreversible binding for some protein systems, 31,[33][34][35] which does not represent thermodynamic behavior and precludes Langmuir-type predictive behavior of the protein corona in a multicomponent system. These observations do not contradict the previously observed correlation between protein thiolation and corona composition but suggest that the order of protein adsorption may play a more significant role than protein affinity, assuming that the preadsorbed protein irreversibly adsorbs to the AuNP. To this point, Siriwardana et al. demonstrated that the composition of the final protein adlayer on the AuNP can vary for multiprotein systems with equivalent protein concentrations when the sequence of protein addition is changed. 30 It should be noted that these studies did not remove the excess preadsorbed protein from the as-formed conjugate prior to subsequent exposure to a competing protein and is appropriate for probing equilibrium behavior. However, these experimental conditions are not a representative of a nanobiotechnology application in which the bioconjugate would be purified to remove excess free protein from the suspension of the prefunctionalized conjugate prior to mixing with a biological medium.

Our lab has previously found that IgG antibody irreversibly adsorbs to AuNPs. Moreover, the purified AuNP-antibody conjugate resists protein exchange when incubated with abundant plasma proteins or whole blood serum for a minimum of 24 h. 35 In that work, we postulated that the antibody irreversibly chemisorbs to the AuNP through the formation of Au-S bonds between the cysteines and AuNPs. Moreover, we speculated that the IgG molecule presents a greater number of free thiols than the other plasma proteins to resist protein exchange. However, we did not consider the protein adsorption order. It is possible that the antibody resisted desorption because it was allowed to adsorb first, whereas preadsorption of a serum protein such as human serum albumin (HSA) to form an HSA-AuNP conjugate could potentially resist desorption by the antibody. Here, we aim to systematically vary the number of free thiols on HSA through chemical modification of lysine residues with 2iminothiolane (Traut's reagent) to elucidate the role of free thiols on the synthesis of protein-AuNP conjugates. Additionally, we investigate the potential effect of protein adsorption order on the composition of the final protein layer.

■ EXPERIMENTAL SECTION

Reagents. Citrate-capped, spherical AuNPs with a nominal diameter of 60 nm and concentration of 2.6 × 10 10 AuNP/mL were purchased from Ted Pella Inc. (Redding, CA). Mouse antihorseradish peroxidase monoclonal antibody (anti-HRP; clone 2H11) was obtained from MyBioSource with a stock concentration of 4.4 mg/mL. HSA and ethylenediaminetetraacetic acid (EDTA) were obtained from Sigma-Aldrich (St. Louis, MO). Horseradish peroxidase (HRP), 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) (1-Step ABTS), potassium cyanide, 5,5′-dithio-bis(2-nitrobenzoic acid) (Ellman's reagent), 2-iminothiolane (Traut's reagent), and tris(2-carboxyethyl) phosphine hydrochloride (TCEP) immobilized onto 4% crosslinked beaded agarose were purchased from Thermo Scientific (Rockford, IL). Potassium phosphate monohydrate was acquired from Fisher Scientific (Fair Lawn, NJ), and anhydrous potassium phosphate dibasic was purchased from Mallinckrodt Chemicals, Inc. (Paris, KY). All buffers and aqueous solutions were prepared with deionized water (18 MΩ) from a Barnstead water purification system (Thermo Scientific, Rockford, IL).

Chemical Modification of HSA. Twenty-fold and thirty-fold molar excess of Traut's reagent was added to HSA (9.0 mg/mL) and allowed to react for 60 min at room temperature. Traut's reagent reacts with the primary amine of the lysine residue to introduce a free thiol group. The chemically modified HSA was purified by removing excess, unreacted Traut's reagent using a Zeba spin desalting column (Thermo Scientific), following the manufacturer's protocol. Throughout this work, we refer to these chemically modified HSA proteins collectively as T-HSA, and specifically as 20T-HSA and 30T-HSA for preparation with 20-fold and 30-fold molar excess of Traut's reagent, respectively. The concentration of the recovered protein was determined using a Nanodrop 2000 (Thermo Scientific) and the extinction coefficient for BSA (43,824 M -1 cm -1 ) that is preprogrammed in the instrument software. The extent of chemical modification, that is, quantity of free thiols, was determined using Ellman's reagent.

Quantitation of Free Thiol. Ellman's reagent was used to quantify the free thiol groups present in IgG (2.4 mg/mL), HSA (16.0 mg/mL), 20T-HSA (9.0 mg/mL), and 30T-HSA (8.8 mg/mL). Each protein was dissolved in 2 mM phosphate buffer (pH 7.4), and the protein concentration was determined using a Nanodrop 2000. Cysteine standards ranging from 50-800 μM were prepared in 2 mM phosphate buffer (pH 7.4). Ellman's reagent was prepared at 4 mg/ mL in 0.1 M phosphate buffer (pH 8.0), and 150 μL were added to each reaction well of a 96-well plate. The protein sample (10 μL) or cysteine standard (10 μL) was added to the reaction well and allowed to react with Ellman's reagent for 30 min at room temperature. After incubation, the absorbance was recorded at 415 nm with a BioRad iMark Microplate reader. The concentration of free thiol in the protein samples was determined from the calibration curve generated from the cysteine standard solutions, and the ratio of free thiol concentration to protein concentration was calculated to quantify the absolute number of free thiols per protein. The measure-IT Thiol Assay Kit (Thermo Scientific), a fluorescence-based kit, was also used to quantify the number of free thiols per protein and confirm the result of the Ellman's reagent assay. Each protein was prepared and analyzed in triplicate, and the average number of thiols and standard deviations are presented.

Reduction of Solvent Accessible Disulfide on Proteins. A reducing gel slurry (40 μL) consisting of TCEP immobilized onto 4% crosslinked beaded agarose was centrifuged at 1000g for 1 min, the supernatant was discarded, and the resin was resuspended in 2 mM phosphate buffer containing 2 mM EDTA. The resin was centrifuged and washed with buffer two additional times. After the third centrifugation, the supernatant was discarded, and 20 μL of protein (2.4 mg/mL antibody or 16.0 mg/mL HSA) was added directly to the pelleted resin. The protein/resin slurry was vortexed and incubated for 15 min at room temperature to reduce surface accessible disulfide bonds in the protein. Following the reduction step, the suspension was centrifuged at 1000g for 1 min, and the supernatant containing the reduced protein was collected. The concentration of recovered, reduced protein was measured using a Nanodrop 2000, and the quantity of free thiols was determined using Ellman's reagent.

Synthesis of Protein-AuNP Conjugates. The protein-AuNP bioconjugates were prepared by loading 100 μL of 60 nm AuNPs in a low binding microcentrifuge tube followed by centrifugation at 5000g for 5 min. The supernatant was discarded, and the pelleted AuNPs were resuspended in 2 mM phosphate buffer at pH 7.4. Excess protein to ensure monolayer coverage (200 nM antibody, 60 μM HSA, or 60 μM T-HSA) was then added to the AuNP suspension, thoroughly mixed, and incubated for 60 min at room temperature with gentle agitation. 35 After allowing for the protein to adsorb onto the AuNP surface, the suspension was centrifuged at 5000g for 5 min to remove excess protein. The protein-AuNP conjugate was resuspended in 2 mM phosphate buffer (pH 7.4), and the centrifugation/resuspension cycle was repeated three additional times to thoroughly remove excess protein.

Probing Protein Exchange on the Protein-AuNP Bioconjugate. Displacement of the preadsorbed protein from the AuNP surface by the introduction of competing proteins to the protein-AuNP bioconjugate was quantified using a previously reported procedure. 35,36 Briefly, 60 μM HSA or 60 μM T-HSA was added to the suspension of purified anti-HRP-AuNP bioconjugate, vortexed, and incubated for 3 or 24 h to allow for protein exchange on the AuNP surface. Similarly, 200 nM anti-HRP antibody was added to the suspension of prepared HSA-AuNP and T-HSA-AuNP bioconjugates, thoroughly mixed, and incubated for 3 or 24 h. After allowing for potential protein exchange on the AuNP surface, the bioconjugates were centrifuged at 5000g for 5 min, and the supernatant was discarded to remove any unbound or displaced proteins. The bioconjugates were resuspended in 2 mM phosphate buffer (pH 7.4), and three additional centrifugation/resuspension cycles were performed to thoroughly remove any excess protein from the bioconjugates.

The extent of protein displacement was quantified following an enzymatic assay to determine anti-HRP antibody loading on the conjugates. 29,35,36 The purified bioconjugate suspensions were incubated with excess HRP (3 μL, 1 mg/mL) for 1 h to saturate all antigen binding sites presented by anti-HRP antibody on the AuNP bioconjugate. Next, excess HRP was removed from the conjugates via four centrifugation/resuspension cycles. Using a 96-well plate, 10 μL of the conjugate was mixed with 150 μL of 1-Step ABTS solution. The HRP-catalyzed reaction rate for the oxidation of ABTS was monitored by the absorption of the product at 415 nm for 20 min at 10 s intervals. The reaction rate was normalized to the concentration of bioconjugates and measured with inductively coupled plasma-optical emission spectroscopy (ICP-OES). A linear correlation was found between the reaction rate and HRP concentration, thereby providing an indirect measure of antibodies adsorbed onto the AuNP surface. Each protein exchange study was performed a minimum of three times; the relative reaction rates are presented as the average values, and the uncertainty is defined by the standard deviations.

Protein and Bioconjugate Characterization. Bioconjugate Quantitation with ICP-OES. Bioconjugate concentration was determined by ICP-OES. To achieve this, 50 μL of the bioconjugate was loaded into a low binding tube, and 5 μL of 100 mM potassium cyanide was added to dissolve the AuNPs and thus desorb the proteins for direct quantitation of total Au. After dissolution, the clear solution was diluted to a final volume of 5.00 mL with 2% nitric acid. Calibration standards were prepared in 2% nitric acid with Au concentrations ranging from 0.1 to 10 ppm. The prepared samples and Au calibration standards were analyzed via ICP-OES. The samples were introduced at 1 mL/min into the plasma operating at 1500 W, and the emission intensity at 242.795 nm was integrated for 15 s.

Dynamic Light Scattering and Zeta Potential. A Malvern Zetasizer Nano ZSP was used to analyze the mean hydrodynamic diameter and zeta potential of HSA, T-HSA, anti-HRP antibody, and protein-AuNP bioconjugates. All protein samples were prepared with pre-filtered buffer (0.020 μm Anodisc filter) at ∼1 mg/mL. A folded capillary cell was filled with buffer, and a 20 μL aliquot of sample was carefully added to the bottom of the folded capillary cell which was interrogated by the laser (diffusion barrier technique). Dynamic light scattering (DLS) was performed first to determine the hydrodynamic diameter of the proteins and bioconjugates using the intensity distribution. Subsequently, the zeta potential was measured, followed by an additional DLS measurement to confirm that the proteins did not aggregate in the applied electric field during the zeta potential measurement. Each size and zeta potential measurement was performed in triplicates where each measurement is the average of 15 separate 10 s runs.

■ RESULTS AND DISCUSSION

Chemical Modification and Characterization of HSA. Our group recently established that an antibody exhibits greater adsorption affinity to AuNPs than HSA to form a robust conjugate. 35 We hypothesize that free thiols on the protein are responsible for the protein-AuNP affinity; thus, we postulate that adsorption affinity can be controlled through chemical modification of proteins to increase the number of free thiol functional groups. To test our hypothesis, we first quantified the number of free thiols on native HSA and anti-HRP antibody using Ellman's reagent (Figure S1 and Table 1). 37 Surprisingly, no free thiols were detected on the antibody, while HSA presented 1.3 ± 0.6 free thiols per HSA molecule (Table 1). These results were confirmed with a more sensitive fluorescence-based thiol quantification assay (Figure S2), which resulted in 0.9 ± 0.1 free thiols per HSA and no detectable thiols on the antibody. Figure 1A shows a molecular model of HSA (Protein Data Bank ID 1UOR 38 ) and reveals that 34 of 35 cysteine residues are involved in disulfide bonding and therefore not detectable by Ellman's reagent nor the thiol-reactive fluorescent dye. 39 The anti-HRP antibody used in our experimental studies is not sequenced; however, the structure is approximated by the molecular simulation of a fully characterized mouse IgG protein (Protein Data Bank ID 1IGT). 40 Figure 1B shows that all cysteine residues in the IgG protein are involved in disulfide bonding and not present as free thiols. The results of these protein structural models support the experimentally measured free thiol functional groups on native HSA and anti-HRP antibody.

Prior to molecular modeling, we had expected to measure greater number of free thiols on the antibody than the HSA molecule given the relative adsorption affinity of these two proteins to AuNPs; 35 however, it is important to note that the thiol quantification assays are not sensitive to disulfides. 37 It is well-established that disulfide bonds cleave in the presence of a gold surface to form a more stable S-Au bond. 41 To more accurately measure the number of thiol points of contact, including those present as disulfides in the native protein, the proteins were treated with TCEP to reduce disulfides and form thiols. In these experiments, TCEP was immobilized on a resin rather than free in solution. This immobilization of the reducing agent prevented penetration of TCEP into the protein and only allowed reduction of surface accessible disulfides, simulating the effect of the AuNP contact. HSA and antibodies were both treated with resin-supported TCEP to reduce accessible disulfides, and Ellman's reagent was used to quantify the free thiols of the TCEP-reduced proteins. Table 1 shows that the TCEP-treated HSA did not increase the number of free thiols (1.000 ± 0.003 thiol/HSA); however, TCEP-treated antibody produced 10.0 ± 0.06 free thiols per antibody. This result suggests that the disulfide bridges are more surface accessible on IgG than the HSA molecules. To corroborate this experimental result, the solvent accessibility of each amino acid in HSA and IgG was calculated and given a value ranging from 0 to 9 to semiquantitatively describe solvent accessibility for the native protein, where 0 indicates that the amino acid is completely inaccessible, and 9 indicates the amino acid is fully exposed. 42,43 With this model, only three cysteine residues in HSA yielded a value of 3 or greater, indicating low to moderate surface accessibility for interacting with AuNPs, while all other cysteine residues were buried within the folded HSA molecule and were not accessible to bind to AuNPs (surface accessibility value < 3) (Figure S3A). The IgG contains six cysteine residues with a value of 3 or greater, suggesting that these amino acids may interact with the AuNP surface to trigger disulfide reduction and formation of a Au-S bond, and potentially lead to structural changes that further expose additional disulfides (Figure S3B). These computational results are consistent with the experimental observations regarding disulfide reduction of HSA and IgG to generate free thiols. Collectively, these data suggest that native HSA binds to AuNPs through a single thiol moiety. Moreover, each antibody is immobilized to the AuNP through multiple S-Au interactions, although we note that it is not probable that all thiols are located in a position to simultaneously interact with the AuNP. Next, we aimed to systematically vary the number of thiols through chemical modification of HSA and identify any correlation between the number of thiols per protein and protein-AuNP binding affinity. To this end, HSA was treated with Traut's reagent, which reacts with primary amines present on lysine residues to produce a terminal thiol (Figure S4). Traut's reagent was selected as a chemical modifier because its chemistry does not alter the protein charge; the positive charge of the lysine is retained in the modified amino acid. This is an important consideration given that previous reports have implicated protein charge in the kinetics, orientation, and affinity of protein-AuNP binding. 29,31,44 HSA was first modified with 20-fold molar excess of Traut's reagent to result in a total of 5.50 ± 0.01 thiols per HSA molecule (Table 1). To increase thiolation, HSA was also treated with 30-fold molar excess of Traut's reagent at pH 9.0, which has been reported to react with phenols in addition to amines. 45 Under these reaction conditions HSA was thiolated to introduce a total of 19.5 ± 2.8 thiols groups per HSA molecule (Table 1). Generating this highly thiolated HSA species was of particular interest because it contained more thiols than the antibody and enabled us to further test the hypothesis that the number of thiols govern protein-AuNP adsorption affinity.

After establishing that thiolation can be controlled through chemical modification, it was necessary to confirm that the modification does not induce protein unfolding or aggregation. The hydrodynamic diameters of anti-HRP antibody, HSA, 20T-HSA, and 30T-HSA were analyzed by DLS and measured 12.3 ± 0.2, 8.1 ± 0.2, 7.6 ± 0.3, and 7.7 ± 0.2 nm, respectively (Figure 2A). These results are consistent with the known size of HSA (8 nm × 3 nm) 38 and the previously measured hydrodynamic diameters of IgG antibody and HSA. 29,44 The similar sizes of HSA and both T-HSAs confirmed that chemical modification with Traut's reagent did not cause the protein to unfold or aggregate. Our results are consistent with previously reported chemical modification to succinylate and aminate HSA which did not result in protein unfolding or aggregation. 44 While unfolding of HSA and T-HSA is not paramount to elucidate the role of thiolation on adsorption affinity, it is significant to demonstrate that thiolation is possible without protein denaturation. This provides a potential pathway to enhance protein adsorption to AuNP while maintaining the biofunctionality of the bioconjugate for downstream applications.

The protein charge was also measured to evaluate the impact of chemical modification. Previous works have established that protein charge significantly impacts adsorption affinity and protein orientation 29,44 and likely contributes to the adsorption kinetics for protein-AuNP interactions. 31 Guided by these previous findings, we aimed to maintain a constant protein charge among all HSA species to eliminate electrostatic interactions as a variable during protein adsorption studies, and modification with Traut's reagent was not expected to alter the protein charge. Zeta potential was measured to assess any changes in the protein surface charge, and Figure 2B shows a measured zeta potential of -4.6 ± 0.4, -5.9 ± 0.5, and -5.9 ± 0.3 mV for HSA, 20T-HSA, and 30T-HSA, respectively. The protein charge did not statistically differ after chemical modification, and this result alleviates any concern regarding charge-dependent adsorption effects. The anti-HRP antibody was also analyzed, and the zeta potential for this protein measured -8.0 ± 0.2 mV (Figure 2B).

Characterization of Protein-AuNP Conjugates. Native HSA, thiolated HSA, or antibody was mixed with AuNPs to allow spontaneous adsorption of the protein onto the AuNP surface, thereby forming bioconjugates. The bioconjugate was formed with a large excess of protein to ensure that the protein saturates the AuNP at monolayer coverage. [46][47][48][49] The hydrodynamic diameters of the HSA-AuNP, 20T-HSA-AuNP, and 20T-HSA-AuNP bioconjugates measured 68.5 ± 0.2, 68.9 ± 0.2, and 68.9 ± 0.2 nm, respectively (Figure 3A). This is an increase of 6 nm in the diameter relative to the unconjugated AuNP which measured 62.3 ± 0.2 nm and is representative of a fully formed monolayer of protein. These data also confirm that the HSA-AuNP and both T-HSA-AuNP conjugates were stable, and the protein did not induce aggregation. Given the inherent sensitivity of DLS to aggregates, the presence of only a few aggregates would lead to greater than expected measured value of the population mean hydrodynamic diameter, for example, ∼68 nm. 50 The antibody-AuNP conjugate was stable and formed a bioconjugate with a mean hydrodynamic diameter of 87.6 ± 0.2 nm which reflects the larger size of IgG compared to HSA. 48 We measured the zeta potential of the unconjugated and protein-AuNP bioconjugates as an orthogonal metric to confirm the formation of protein-AuNP conjugates. Unconjugated AuNPs exhibited a zeta potential of -43 ± 3 mV that resulted from the adsorbed citrate capping agent that formed during synthesis (Figure 3B). The adsorption of HSA, 20T-HSA, 30T-HSA, and antibody onto AuNPs shifted the zeta potential to -27 ± 2, -26 ± 2, -26 ± 2, and -19 ± 2 mV, respectively (Figure 3B). These increases in zeta potential are consistent with other reported zeta potential values for protein-AuNP conjugates, 27,51 and the similar values for HSA and both T-HSA formed bioconjugates suggest similar protein loading for both proteins given that each protein has a similar charge.

The binding of HSA, 20T-HSA, 30T-HSA, and antibody to AuNPs was also assessed by extinction spectrophotometry. The unconjugated AuNP displayed an extinction maximum at 537 nm (Figure 3C). The formed HSA-AuNP and both T-HSA-AuNP conjugates resulted in a 2 nm red shift relative to the unconjugated AuNP to exhibit extinction maxima at 539 nm (Figure 3C). This shift in extinction is characteristic of a change in the local refractive index at the AuNP surface and is expected for the adsorption of a protein layer with a modest thickness such as HSA. 52,53 Adsorption of a larger protein, such as an antibody, results in a thicker proteinaceous layer to produce a greater red-shift in the extinction, for example, λ max = 542 nm (Figure 3C).

The saturation of the AuNP surface with proteins was further assessed by evaluating the stability of the protein-AuNP conjugates in a high ionic strength environment. After allowing adsorption of each protein to the AuNP and subsequent centrifugation to remove the unbound protein, NaCl was added to the conjugates to a final concentration of 150 mM. Extinction spectra were collected after the addition of NaCl and show that each conjugate resisted salt-induced aggregation (Figure S5). The stability of the conjugate in the saline environment further confirms that the loading of each protein saturated the AuNP surface.

Protein Displacement. In our previous work, we established that the antibody irreversibly adsorbs to AuNPs. 35 Upon removal of excess antibody to purify the antibody-AuNP conjugate and resuspension in buffer, the antibody did not desorb over a period of seven days. Moreover, introduction of plasma proteins into a suspension of the purified antibody-AuNP conjugate did not disrupt the antibody-AuNP interaction, and no protein exchange was detected. In that work, we attributed the resilient antibody-AuNP interaction to greater number of cysteine residues on the antibody compared to the other plasma proteins. However, we did not rule out the possibility that any pre-adsorbed protein, bound to the AuNP through a thiol, could resist protein exchange and that protein displacement from the AuNP is independent of the number of thiols contained in the protein. Here, a series of protein exchange studies were performed with antibody, native HSA, and thiolated HSA proteins to elucidate the role of accessible free sulfhydryl and to identify the key factors that lead to irreversible adsorption, for example, order of protein adsorption.

Our lab recently developed an enzyme-mediated assay to quantify protein exchange on the surface of AuNPs. 29,35,36 Using this previously developed method, we first investigated the ability of HSA and T-HSA to displace antibody preadsorbed on the AuNP. To this end, the AuNP was fully saturated with an adlayer of anti-HRP antibody via spontaneous adsorption, and excess antibody was removed to purify the antibody-AuNP conjugate. HSA, 20T-HSA, or 30T-HSA was then added to the fully formed antibody-AuNP conjugate at a concentration of 60 μM and incubated for 3 or 24 h to allow for protein exchange on the AuNP surface (Figure 4A). This concentration of HSA, 20T-HSA, and 30T-HSA corresponds to the normal HSA concentration in 10% serum and is ∼30-fold molar excess of the HSA concentration required to form a monolayer on the unconjugated citratecapped AuNP. 35,51 A large excess of antigen (HRP) was added to the suspension to saturate all antigen binding sites that were presented by the antibodies remaining on the AuNP surface after allowing for protein exchange. The captured HRP was quantified by the enzymatic oxidation rate of ABTS (Figure S6) and provided an indirect measure of immobilized antibody on the AuNP. As a positive control and to define the maximum amount of antigen captured by a conjugate, that is, maximum amount of adsorbed antibody, the antibody-AuNP was incubated in buffer without a competing protein. The enzymatic reaction rates for all conjugates involved in protein exchange studies were normalized with respect to the rate measured for this fully coated antibody-AuNP conjugate (Figure 4B). As a negative control, HSA-AuNP conjugates were synthesized and incubated with HRP to allow for nonspecific interactions. Minimal enzymatic activity was detected by the HSA-coated AuNP and confirmed that HRP was only captured by the AuNP conjugate through specific interactions with the immobilized anti-HRP antibody (Figure 4B). Thus, we can conclude any normalized rate of less than 1.0 indicates fewer captured HRP molecules due to the displacement of antibody from the conjugate surface in exchange for HSA. Figure 4B shows that HSA did not displace antibody after allowing 3 or 24 h for protein exchange and is in agreement with our previous work. 35 Similar results were observed for 20T-HSA, in which the antibody resisted displacement by this chemically modified HSA containing ∼6 thiols. However, 30T-HSA displaced ∼85% of antibody from the antibody-AuNP conjugate after allowing only 3 h for protein exchange. Increased incubation to allow 24 h of protein exchange did not lead to further displacement of antibody. These data establish that preadsorbed proteins are susceptible to protein exchange and repudiate the idea that the order of adsorption determines bioconjugate protein composition. These results suggest that the final composition of the protein layer on the AuNP is primarily governed by the total number of free thiols that are solvent accessible (Table 1).

We conclude that protein exchange proceeded through an associative exchange mechanism, rather than a dissociative mechanism. We previously demonstrated that the antibody does not spontaneously desorb from the AuNP surface; 35 thus, a dissociative ligand exchange pathway is ruled out. While the antibody is fully saturating the AuNP surface, the random adsorption process does not result in a close-packed adlayer, and it was previously demonstrated that as much as 70% of the AuNP surface is still available for adsorption of small molecules that penetrate the protein layer of a fully formed bovine serum albumin-AuNP conjugate. 32 HSA is smaller than IgG, and therefore, HSA can penetrate between the adsorbed antibody to interact with the AuNP surface. However, displacement of the initially adsorbed antibody is only possible by 30T-HSA which exhibits greater binding affinity than the IgG molecule. Here, we observe ∼85% protein exchange efficiency for 30T-HSA, which is similar to ligand exchange efficiencies of 75-90% measured for small thiol ligands on the AuNP. 54 To further test our suppositions proposed above, HSA or T-HSA was first preadsorbed onto AuNPs to form bioconjugates, followed by the addition of anti-HRP antibody to allow for protein exchange (Figure 5A). Antibody readily displaced the preadsorbed HSA from the AuNP surface within 3 h to gain full antigen binding capacity equivalent to the positive control anti-HRP-AuNP conjugate (Figure 5B). Interestingly, the 20T-HSA displayed greater resistance to displacement by the antibody than the native HSA. After 3 h of incubation, the 20T-HSA was partially displaced by antibody such that the conjugate recovered 31% of its antigen binding capacity compared to the anti-HRP-AuNP conjugate. However, after 24 h of incubation, the antibody displaced 20T-HSA from the F AuNP surface to fully recover antigen binding capacity equivalent to the anti-HRP positive control conjugate. In contrast to HSA and 20T-HSA, Figure 5B shows that the 30T-HSA-AuNP conjugate benefited from a more robust adsorptive interaction, and this conjugate yielded 20-25% of the antigen binding activity of the fully formed antibody-AuNP conjugate after allowing for protein exchange. Clearly, the data show that 30T-HSA resisted displacement by the antibody; however, minimal, yet significantly, antigen binding activity was observed for this conjugate. The time-dependent results suggest that the 30T-HSA and antibody composition of the AuNP protein corona reached a steady-state after 3 h of incubation because the antigen-binding activity did not increase with antibody exchange time. It is possible that this recovered antigen-binding capacity is due to the displacement of a few weakly adsorbed 30T-HSA molecules that were oriented such that only a few thiols interacted with the AuNP surface. Alternatively, the antibody may backfill any available space on the AuNP surface, given that the HSA adlayer is disordered and only passivates ∼70% of the Au surface to result in a mixed protein layer with minimal antigen-binding activity. However, the antibody, with 10 available thiols, does not displace the adsorbed 30T-HSA, which contains 20 free thiols. In contrast, the antibody backfills "pinholes" in the HSA and 20T-HSA conjugates and then displaces these weaker bound HSA variants that display fewer free thiols than the antibody. Notably, these results demonstrate that as the number of thiols on the modified HSA exceeds the number of thiols on antibody, the HSA protein can resist displacement by antibody. Moreover, these protein exchange studies with preadsorbed HSA and chemically modified HSA provide further evidence that protein adsorption onto AuNPs is governed by the number of free and accessible thiols. Most significantly, this model protein system provides evidence for controlled release of immobilized proteins through thiolation, as specifically demonstrated by the 20T-HSA-AuNP conjugate.

■ CONCLUSIONS

Here, we performed ligand exchange studies to probe the role of free thiols on protein adsorption and the impact of protein adsorption sequence using an IgG monoclonal antibody and chemically modified variants of HSA using Traut's reagent. A key finding is that thiols can be added to proteins to control the total number of free thiols on each protein molecule, and the chemically modified protein remains stable, for example, folded and nonaggregated. This chemical modification protocol allows for controlling adsorption affinity through facile chemical treatment of the protein, which is much more time-and cost-effective than protein engineering strategies to synthesize mutant proteins with a defined number of cysteine residues. 31 Ligand exchange studies using the chemically modified HSA led to four key outcomes. First, the preadsorbed protein is only displaced from the AuNP surface by proteins in solution that contain more thiols. This finding is consistent with those reported for the relative stability of engineered GB3 (0-2) proteins adsorbed onto AuNPs. Second, the composition of the final protein corona on the AuNP surface is independent of the order in which the proteins are added to the AuNP. Third, we observe that small proteins with many thiols outcompete larger proteins that have more overall interactions, for example, electrostatic, hydrophobic, and so forth, yet have fewer thiol moieties. Previous research has reported that protein binding affinity correlates with protein molecular weight; however, this is likely due to larger proteins having a greater probability of containing more cysteine residues, unless smaller proteins are engineered or chemically modified to increase the number of free thiols, as was done in this work. Last, we observe that the protein exchange rate is dependent on the difference between the number of thiols on the preadsorbed protein and the competing protein, with slower exchange rates for a similar number of thiols. Collectively, we conclude from our data that the number of Au-S bonds is the dominant attractive force between proteins and AuNPs and are ultimately responsible for the final protein corona. These findings present an exciting opportunity to carefully control protein adsorption affinity through chemical modification to protect against desorption of the preadsorbed protein and loss of function or enable timecontrolled release for therapeutic applications.