Comparison of the Energetics of Avidin ... - ACS Publications

Journal of the American Chemical Society 2014 136 (4), 1409-1417 .... Membrane Tethered Delta Activates Notch and Reveals a Role for Spatio-Mechanical...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Comparison of the Energetics of Avidin, Streptavidin, NeutrAvidin, and Anti-Biotin Antibody Binding to Biotinylated Lipid Bilayer Examined by Second-Harmonic Generation Trang T. Nguyen, Krystal L. Sly, and John C. Conboy* Department of Chemistry, University of Utah, 315 South 1400 East, Rm. 2020, Salt Lake City, Utah 84112, United States S Supporting Information *

ABSTRACT: A comparison of the binding properties of avidin, streptavidin, neutrAvidin, and antibiotin antibody to a biotinylated lipid bilayer was studied using second-harmonic generation. Protein binding assays were performed on a planar supported lipid bilayer of 1,2dioleoyl-sn-glycero-3-phosphocholine (DOPC) containing 4 mol % biotinylated-cap-1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (biotincap-DOPE). The equilibrium binding affinities of these biotin−protein interactions were determined, revealing the relative energetic contributions for each protein to the biotinylated lipid ligand. The results show that the binding affinities of avidin, streptavidin, and neutrAvidin for biotin were all strengthened by protein−protein interactions but that the stronger protein−protein interactions observed for streptavidin and neutrAvidin make their binding more energetically favorable. It was also shown that neutrAvidin has the highest degree of nonspecific adsorption to a pure DOPC bilayer, compared to avidin and streptavidin. In addition, the biotin-binding affinity of the antibiotin antibody was found to be of the same order of magnitude as that of avidin, streptavidin, and neutrAvidin. These findings provide important new insights into these biotin-bound protein complexes commonly used in several bioanalytical applications.

B

Streptavidin, which has a similar functional domain to avidin (∼33% identical residues),17 is a nonglycosylated protein with a slightly acidic pI of ∼5−6.14 NeutrAvidin is a commercially available deglycosylated form of avidin with a pI of 6.3.19 The lower pI values and the absence of the sugar groups in streptavidin and neutrAvidin are intended to lower the nonspecific binding, relative to avidin, without significantly affecting the affinity for biotin. As such, streptavidin and neutrAvidin are commonly used in biosensing applications, as an alternative to avidin, as a way to reduce nonspecific binding to the sensor surfaces. Antibiotin antibody, which is an immunoglobulin protein that is generated by plasma cells as part of the immune response to the antigen biotin, has also been used as an alternative linker in biosensors6,12 and immunoassays,20,21 since it also has a high binding affinity for biotin (∼108 M−1).22 Surprisingly, little work to date has compared the specific and nonspecific binding affinities of avidin, streptavidin, neutrAvidin, and antibiotin antibody to biotin at surfaces, despite the use of these proteins in a broad range of biosensing applications. Here, we report the binding affinities of these biotin-binding proteins to biotinylated lipid bilayers while specifically addressing the energetics of the binding process and the nonspecific adsorption of all four proteins.

iotin-bound protein complexes have been used in a wide variety of bioanalytical applications, including monitoring biomolecule conformational changes,1,2 biochip sensor fabrication,3−8 immunoassays,9,10 and targeted drug delivery and screening.11 In these applications, the protein−biotin complex is commonly used to tether biomolecules to a surface1,2 or used as a linker to capture biomolecules3−12 by taking advantage of the high affinity, specificity, and stability of biotin-bound proteins, such as avidin,7−9 streptavidin,1−3,10,11 neutravidin,4,5 and antibiotin antibody.6 Avidin is a tetramer consisting of four identical subunits, each of which has one binding site for biotin. It has an extremely high binding affinity, Ka ≈ 1015 M−1, to free biotin in solution,13 and it forms a stable complex with biotin over a wide range of temperatures and pH values.14 One major drawback to using avidin is the high degree of nonspecific adsorption caused by its basic isoelectric point (pI ∼10).14 At physiological pH, positively charged avidin can bind nonspecifically to negatively charged surfaces such as cell membranes14 or silica substrates.15 Another feature that contributes to avidin’s high nonspecific adsorption is its carbohydrate groups, which contain four mannose residues and three N-acetylglucosamine residues per subunit.16 If the avidin−biotin complex is being used to capture a carbohydratebinding molecule, it can also bind specifically to the carbohydrate groups on avidin, limiting its use in bioassays. Both streptavidin and neutrAvidin, which are analogues of avidin, which also have a high affinity and specificity toward biotin,17,18 have been used as an alternative to avidin.3−5 © 2011 American Chemical Society

Received: September 7, 2011 Accepted: November 28, 2011 Published: November 28, 2011 201

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

conditions to examine all four proteins allows for a direct comparison of the binding properties of these biotin−protein complexes as well as comparison to previous literature studies of these biotin-bound proteins, providing useful information on these tether/linker complexes commonly used in several bioanalytical applications.

Biosensing applications have employed the use of protein− biotin complexes to various types of surfaces, including functionalized gold,3,4,10,23 glass,1,24 silver nanoprisms,6 as well as planar supported lipid bilayers (PSLBs) or liposomes.25 PSLBs were chosen as a platform for the protein binding assays in this study because of the ease of preparation and the ability to precisely control the biotin density,26 providing better reproducibility and facilitating comparison of the various protein−biotin interactions. The binding affinity between protein−ligand pairs at surfaces is known to be affected by the surface density of the ligand.27−29 Zhao and co-workers have shown that, once the biotin density is high enough to bind a monolayer of avidin, any additional biotin sterically hinders further avidin binding.28 Their study also showed that doubly bound avidin−biotin complex (two biotin molecules for every one avidin molecule) is more stable than the singly bound complex.28 This suggests that the optimal biotin density would be one that allows a monolayer of avidin to bind bivalently. In our study, the appropriate biotin density to form a doubly bound complex monolayer of the protein can be calculated using the area for avidin, 3025 Å2,14 and for a lipid and biotinylated lipid molecule, 70 Å.30 For a monolayer of avidin (3.31 × 1012 molecule/cm2) to bivalently bind to biotin, the biotin density must be twice as large; therefore, the density of the biotinylated lipid in the lipid bilayer should be ∼4.6 mol % [(2 × 3.31 × 1012 biotinylated lipid molecule/cm2)/(1.43 × 1014 lipid molecule/cm2) × 100%]. Previous studies that investigated biotin binding to neutrAvidin and streptavidin in PSLBs have also shown that protein binding increases with biotin density up to 4 mol % and then saturates.31,32 At this density, two biotin molecules effectively bind to every one protein to form a doubly protein−biotin complex monolayer on the lipid bilayer surface. When the biotin density is lower than 4 mol %, fewer protein−ligand complexes are formed; above 4 mol %, steric hindrance from neighboring proteins decreases binding. In order to provide useful information for bioanalytical applications, which aim to maximize the number of captured biomolecules, the optimal biotin density of 4 mol % (a monolayer coverage of bivalently bound protein) was used here. To investigate these protein−biotin interactions, we utilized second-harmonic generation (SHG), which is a surface-specific nonlinear optical spectroscopic technique capable of directly detecting protein interactions with surfaces. As opposed to fluorescence, SHG does not require an exogenous label to be attached to the protein. Attaching a label can alter the native conformation and/or charge of a protein and therefore alter its binding capabilities.33,34 Thus, nonlinear optical spectroscopic techniques, such as sum frequency generation vibrational spectroscopy (SFVS) and SHG, have recently been shown as promising alternatives for label-free biomolecule recognition at interfaces. For example, SFVS has been used to probe the structural interactions between proteins/peptides and lipid monolayers or bilayers.35−37 SHG has also been used to monitor protein adsorption at a solid/liquid interface38,39 and association of biomolecules to a lipid monolayer or bilayer.40−42 SHG provides the surface selectivity and high sensitivity required to study protein−biotin interactions at an interface without using an external label. Therefore, we employed SHG to study the interactions between biotin and the four biotinbinding proteins, avidin, streptavidin, neutrAvidin, and antibiotin antibody, in PSLBs. Using the same experimental

■

EXPERIMENTAL SECTION The materials used and preparation of PSLBs are described in the Supporting Information. Ligand−Protein Binding Assay. PSLBs of 1,2-dioleoylsn-glycero-3-phosphocholine (DOPC) containing 4 mol % 1,2dioleoyl-sn-glycero-3-phosphoethanolamine-N-(Cap biotinyl) for avidin, neutrAvidin, streptavidin, and antibiotin antibody binding were created on a prism surface by vesicle fusion as described in the Supporting Information. Three independent experiments were conducted for each protein, except for the case of neutrAvidin and the antibiotin control, where a t-test was conducted, resulting in the elimination of one dataset for a total of two samples. To reduce nonspecific adsorption of the proteins, the PSLBs were incubated in 1 mg/mL monoclonal IgG (pI ≈ 6.1−6.5)43 from rabbit serum in PBS pH 7.4 for 30 min to block any defects that might exist on the lipid surfaces, except for the antibiotin antibody experiments. The PSLBs were then rinsed thoroughly with PBS to remove any free IgG remaining in solution. Increasing concentrations of avidin, neutrAvidin, and streptavidin ranging from 9.25 nM to 537.6 nM were injected into the flow cell and allowed to reach equilibrium. The same procedure was followed for antibiotin; however, the concentration range was from 4 nM to 121 nM. At each protein concentration, the SHG intensity was recorded at 30 min intervals until a steady-state response was achieved. Generally, low concentrations of the proteins required up to 4 h to reach equilibrium. During this period, at least 1.5 mL (∼3 times the volume of the flow cell) of fresh protein solution was injected every 30 min to account for the bulk depletion caused by surface adsorption of the proteins. It is important to note that protein dilutions were freshly prepared prior to each injection to further prevent the proteins from nonspecifically adsorbing to the vials and syringes. Thermodynamic measurements were performed using the SHG intensity collected at equilibrium at each protein concentration. In addition, kinetic measurements were determined for the SHG intensity collected over time. SHG Measurements. Counter-propagating SHG was employed here. A detailed description of this technique can be found elsewhere.44 Briefly, a 532-nm laser beam (secondharmonic output of a Nd:YAG laser, Surelite I, 10 Hz) with a mixed polarization state (equal amounts of s and p polarized components) was directed at the prism/water interface under total internal reflection. The laser intensity used was 14 mJ/ pulse. The reflected beam was steered back to overlap spatially and temporally with the incidence, generating SHG at 266 nm. Optical filters were used to remove any scattered visible light before the reflected SHG signal was collected with a solar-blind photomultiplier tube. The SHG intensity is proportional to the second-order (2) (2) susceptibility tensor, χijk , which has a nonresonant χNR and (2) resonant χR contribution: 2 (2) (2) 2 ISHG ∝ |χ(2) ijk | ∝ |χ NR + χ R | 202

(1)

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

imaginary components of χR(2) due to the proteins, respectively. max 1/2 ) is the square root of the maximum SHG intensity at (ISHG surface saturation, K0 is the equilibrium association constant, and [P] is the bulk protein concentration. The derivation of this equation can be found in the Supporting Information. For avidin, a previous study demonstrated that protein− protein interactions can modulate the energetics of protein association to a biotinylated surface.27 When interactions between proteins are involved, a cooperative binding model, previously described by Zhao et al.,27 can be incorporated into the SHG intensity expression:

The resonant χR(2) can be expressed as follows:45 (χ(2) ijk )R ∝ N

∑ a,b,c

⟨a|μi|c⟩⟨a|μj|b⟩⟨b|μk|c⟩ (2hω − Eca − i Γca)(hω − Ebc − i Γbc)

(2)

where N is the surface density of molecules; h is Planck’s constant; a, b, and c denote the initial, intermediate, and final states, respectively; μ is the Cartesian coordinate dipole operator; and Γ represents the transition line width. The (2) represent the input (j,k) and output (i) fields, indices on χijk which can assume any of the three Cartesian coordinates (x,y,z). As shown in eq 2, an enhancement in the SHG signal is observed when the incident (ω) or SHG (2ω) frequency is resonant with an electronic transition of a molecule at the interface. Therefore, SHG can be used to detect the presence of bound protein to a lipid membrane if the protein has electronic transitions at the frequency of the incident fundamental (532nm) or the SHG (266-nm) light. As seen in the extinction coefficient spectra of the proteins in Figure 1, the SHG

background

ISHG − ISHG ∝ 2B

background ISHG

max max ( ISHG / ISHG ) ISHG ω K 0[P] max 1 + ω( ISHG / ISHG )K 0[P]

max ⎛ max ( ISHG / ISHG ⎞2 ) ISHG ω K 0[P] ⎟ ⎜ + (B + C )⎜ ⎟ ( I / I max ) ⎝ 1 + ω SHG SHG K 0[P] ⎠

2

2

(4)

where ω = η , by assuming the distribution of the biotin-bound proteins follows a square lattice, and η is the cooperativity coefficient, which characterizes the protein−protein interactions between neighboring protein molecules on the surface.27 K0 is the intrinsic binding affinity of the protein to the ligand barring any protein−protein interactions.27 When η > 1, the binding of a protein to a ligand exhibits a positive cooperativity, demonstrating that the protein−protein interaction enhances the ligand−protein binding. When η < 1, the protein−protein interaction reduces the ligand−protein binding, resulting in a negative cooperativity. The cooperativity model becomes the Langmuir model when η = 1. The derivation of eq 4 was also described in the Supporting Information. 4

Figure 1. Extinction coefficient spectra of the proteins.

■

wavelength at 266 nm is in resonance with the π → π* transitions of the protein’s tryptophan and tyrosine’s aromatic rings,46 resulting in enhancement of the SHG signal when the protein is present at the lipid surface. Although the electricquadrupole response from the bulk medium can contribute to the overall SHG signal,47,48 its contribution to the measured signal is negligible if the SHG frequency is in resonance with the electric-dipole allowed transition of molecules residing at the interface.47 As the SHG frequency used in this study is resonant with the transitions of the proteins adsorbed to the surface, the measured SHG intensity is predominantly dipolar in nature with little to no detectable contribution from the quadrupolar response expected. The SHG intensity expression assumed the Langmuir isotherm model can be given as

RESULTS AND DISCUSSION Thermodynamics of Avidin, Streptavidin, and NeutrAvidin Binding to Biotinylated DOPC Bilayers. The binding isotherms for avidin, neutrAvidin, and streptavidin are shown in Figure 2. Note that the SHG intensities shown in Figure 2 were normalized as described in the Supporting Information. The SHG intensities increase with increasing protein concentration until saturation coverage is achieved at concentrations of >100 nM. The Langmuir and cooperativity models were both used to fit the data, and the f-test was performed to determine the best fit for the data. The Langmuir model (eq 3) was used to fit the data with the max 1/2 ) , and K0 by performing a fitting parameters B, C, (ISHG nonlinear least-squares regression. The cross-term is the product of the nonresonant component (A) and real component (B) of the resonant susceptibility tensor, which describes the interference between the background and protein adsorption responses. Assuming the nonresonant term A is real and positive, the measured SHG intensity can increase or decrease due to constructive (A and B have the same sign) or destructive (A and B have different signs) interference. In the case of our data, as seen in Figure 2, we did not observe any initial decrease in the SHG intensity at low protein concentrations, where the contribution of the resonant term is presumably not much greater than the nonresonant contribution. This observation suggests there is constructive interference between the nonresonant background and a resonant response arising from the adsorbed protein species,

background

ISHG − ISHG ∝2

⎛ I max K [P] ⎞ SHG 0 ⎟ ⎟ 1 [ ] + K P 0 ⎝ ⎠

background ⎜ ISHG B⎜

⎛ I max K [P] ⎞2 SHG 0 ⎟ + (B + C )⎜⎜ ⎟ 1 ⎝ + K 0[P] ⎠ 2

2

(3)

background 1/2 where (ISHG ) = A is the square root of the SHG intensity from the background (nonresonant response) by assuming (2) = A is real and positive. B and C denote the real and χNR

203

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article max 1/2 (ISHG ) . Using these simplifications, the final Langmuir equation used to fit the adsorption data in this study is as follows:

ISHG

⎛ I max K [P] ⎞2 SHG 0 ⎟ ∝ ⎜⎜ ⎟ + 1 K [ P ] 0 ⎝ ⎠

(5)

The cooperativity model (eq 4) was also used to fit the data with the fitting parameters B, C, Nmax, K0, and ω. The fitting results also showed that B is 7 orders of magnitude lower than C, and thus, the contribution of the cross-term in eq 4 is negligible. Therefore, the final form of the cooperativity model used to fit the data has the following form:

ISHG

max ⎛ max ( ISHG / ISHG ⎞2 ) ISHG ω K 0[P] ⎟ ⎜ ∝⎜ ⎟ ( I / I max ) ⎝ 1 + ω SHG SHG K 0[P] ⎠

(6)

Equations 5 (the Langmuir model) and 6 (the cooperativity model) were then used to fit the adsorption data of avidin, streptavidin, and neutrAvidin shown in Figure 2. It was found that the cooperativity model statistically fit the data best, and the results from a nonlinear least-squares regression of the data are given in Table 1. The intrinsic binding affinities (K0) of avidin, streptavidin, and neutrAvidin to a biotinylated DOPC bilayer retrieved from the nonlinear regression are (8.2 ± 2.4) × 107 M−1, (4.3 ± 0.9) × 107 M−1, and (2.6 ± 0.01) × 107 M−1, respectively. The K0 value of avidin to biotin obtained in this study is in good agreement with the value of (2.14 ± 0.20) × 107 M−1 previously reported by Zhao et al. for fluorescently labeled avidin binding to a 0.63 mol % biotinylated lipid incorporated into an arachidic acid monolayer. 27 For streptavidin, the K0 value obtained here is slightly higher than the value of 7.3 ± 0.2 × 106 M−1 reported by Tang and coworkers for streptavidin binding a biotin monolayer functionalized on a SPR gold chip surface.49 The difference in the K0 values of streptavidin between the two studies can be related to the difference in the surface density of biotin used. It is likely that the high biotin density used in the SPR study creates steric hindrance, which reduces the accessibility of an individual biotin to the protein binding sites, resulting in a lower K0 value. For neutrAvidin binding to biotin, the K0 value measured in this study is ∼4.5 orders of magnitude smaller than an affinity of (5.5 ± 0.2) × 1011 M−1 reported by Wayment and Harris, using single-molecule fluorescence.18 To compare the result of the Wayment and Harris study with our study, we performed a similar analysis of the kinetics of protein binding, which is presented in the following section. It was shown that the slower adsorption and desorption rates of biotin−protein measured in this study lead to the lower observed binding affinity. As previously mentioned, the cooperativity binding model was statistically the best fit to the data obtained here. This

Figure 2. SHG intensity vs bulk protein concentration of the proteins binding to DOPC bilayers containing 4 mol % biotin-cap-DOPE (filled circles) and 0 mol % biotin-cap-DOPE (open circles). The lines are fits to the cooperativity binding model. The error bars represent the standard deviation from three independent experiments, except for the specific binding of neutrAvidin, which represents two independent experiments.

indicating the cross-term (2AB) will be positive since A and B have the same sign. In addition, the results from the nonlinear regression indicate that the magnitude of B is ∼8 orders of magnitude smaller than the magnitude of the imaginary portion of the resonant susceptibility tensor (C). Since the contribution background 1/2 ) ) in eq 3 is much smaller, from the cross-term (2B(ISHG compared to the resonant contribution term (B2 + C2), it can be neglected. (B2 + C2), which is related to the surface density (eq 2), serves as a scaling factor and can then be pooled into

Table 1. Measured Intrinsic Binding Affinity (K0), Cooperativity Coefficient (η), Intrinsic Free Energy (ΔG0), Free Energy Due to Protein−Protein Interactions (ΔGη), Total Free Energy (ΔGtotal), and Apparent Binding Affinity (Kapp) for the Proteinsa protein

K0 (× 107 M−1)

η (a.u.)

ΔG0 (kJ/(mol K))

ΔGη (kJ/(mol K))

ΔGtotal (kJ/(mol K))

Kapp (× 107 M−1)

avidin streptavidin neutrAvidin

8.2 ± 2.4 4.3 ± 0.9 2.6 ± 0.01

1.2 ± 0.2 1.8 ± 0.2 1.9 ± 0.01

−45 ± 0.8 −44 ± 0.5 −42 ± 0.01

−1.7 ± 1.4 −5.7 ± 1.1 −6.3 ± 0.04

−47 ± 1.7 −49 ± 0.6 −49 ± 0.05

18 ± 14 44 ± 10 23 ± 0.7

a

Data were obtained with the cooperativity binding model, using the assumption that the distribution of the biotin-bound proteins follows a square lattice. 204

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

Using eq 8, ΔGη was calculated from the binding isotherms in Figure 2 for avidin, streptavidin, and neutrAvidin, and the results are listed in Table 1. The protein−protein interactions for avidin contribute a ΔGη value (−1.7 ± 1.4 kJ/mol) that is ∼3 times less than that for streptavidin and neutrAvidin (approximately −5.7 ± 1.1 and −6.3 ± 0.04 kJ/(mol K), respectively). The total free energy of the protein−biotin binding including the contribution from the protein−protein interactions was determined using the following equation:

indicates that protein−protein interactions are involved in the biotin−protein binding. The value of the cooperativity coefficient (η) obtained from the fit is shown in Table 1. A value of η greater than unity was obtained for avidin (1.2 ± 0.2), streptavidin (1.8 ± 0.2), and neutrAvidin (1.9 ± 0.01), illustrating that all the proteins exhibit positive cooperative binding behavior. This suggests that the intrinsic binding of the proteins to biotin is increased due to protein−protein interactions. Similar cooperative binding behavior has been previously observed for avidin to a monolayer of arachidic acid containing biotinylated lipid.27 The cooperative binding of avidin only was observed if the biotin density was adequately high to bind to a monolayer of avidin so that the distance between the biotin molecules is close enough for the adjacent biotin-bound avidin to interact with an incoming avidin.27 In addition, biotin must be placed above the lipid surface (i.e., using a spacer between biotin and the lipid headgroup, to provide more accessibility for avidin binding).27 The use of a biotin density high enough to form a monolayer of the proteins (∼ 4 mol % biotin) and the spacer between biotin and the lipid headgroup (biotin-cap-DOPE) in our study allows for protein− protein interactions to occur. Interestingly, here, avidin was found to have a smaller η, relative to that for streptavidin and neutrAvidin, indicating that avidin exhibits less positive cooperative binding. The lower cooperativity observed for avidin as compared to streptavidin has been noted before.27,50,51 Blackenburg et al. noted that streptavidin tends to aggregate and form highly ordered domains, which suggests streptavidin−streptavidin interactions.50 Although Blankenburg et al. did not observe any domains for avidin, Ku et al. demonstrated that domains were formed by avidin; however, these domains were very small.50,51 The smaller domain size observed for avidin suggests that avidin has a lower positive cooperativity, compared to streptavidin, congruent to the observations in this study. The lower cooperativity observed for avidin could be related to the high pI (∼10) of the protein. At the neutral pH used here, avidin is positively charged, and thus, the electrostatic repulsion between the charged avidin could repel the other avidin proteins, resulting in a reduction in avidin−avidin interactions. Conversely, streptavidin and neutrAvidin have a pI of ∼7; therefore, the electrostatic interaction between streptavidin and streptavidin or neutrAvidin and neutrAvidin is less pronounced, compared to the avidin−avidin interactions, leading to a higher η obtained for streptavidin and neutrAvidin. The level of cooperative binding observed in avidin, streptavidin, and neutrAvidin can also be evaluated in terms of the binding free energy. Without any contributions from protein−protein interaction, the free energy of protein−biotin binding (ΔG0) can be calculated from K0:

ΔG0 = − RT ln k 0

ΔGtotal = ΔG0 + ΔGη

the results of which are listed in Table 1. The calculated ΔGtotal value of avidin−biotin binding (−47 ± 1.7 kJ/mol) is lower than that of streptavidin (−49 ± 0.6 kJ/mol) and neutrAvidin (−49 ± 0.05 kJ/mol). Accordingly, the effective binding between biotin and streptavidin or biotin and neutrAvidin is more energetically favorable, because of the stronger protein− protein interaction, relative to the avidin−biotin binding. From the ΔGtotal value calculated above, the apparent binding affinity (Kapp) of the proteins to a biotinylated lipid bilayer was calculated, and the results are given in Table 1. The larger calculated values for Kapp, compared to K0, reveal that the affinity of the proteins toward biotin is enhanced by protein− protein interactions. When these protein−protein interactions are taken into account, avidin has the lowest Kapp value ((18 ± 14) × 107 M−1), while streptavidin and neutrAvidin exhibit stronger Kapp values ((44 ± 10) × 107 M−1 and (33 ± 0.7) × 107 M−1, respectively). Nonspecific Adsorption of Avidin, neutrAvidin, and Streptavidin to a DOPC Bilayer. The nonspecific adsorption of avidin, streptavidin, and neutrAvidin to the IgG passivated DOPC bilayer without biotin is shown as open circles in Figure 2. The amount of nonspecific adsorption of the proteins can be evaluated using (ISHG)1/2, which is directly proportional to the protein surface concentration. To compare the amount of nonspecific adsorption, we calculated the percent surface coverage in the absence of biotin at 537.6 nM (above saturation concentration), relative to the calculated monolayer coverage of adsorbed protein in the presence of biotin. Note that the proteins may have a different orientation and conformation when nonspecifically bound to the IgG passivated DOPC bilayer, compared to when specifically bound to biotinylated DOPC. This may alter the conformation and orientation of the tryptophan and tyrosine residues probed in this experiment. If the orientation of the residues becomes more disordered when nonspecifically bound to the IgG passivated DOPC bilayer, this could produce a lower SHG signal, indicating a lower quantity of protein bound at the surface than what is actually bound. Although the orientation of the tryptophan and tyrosine residues could affect the observed SHG signal and care should be taken when comparing nonspecific binding to specific binding, the trends in specific versus nonspecific binding measured by SHG in the current study are consistent with previously reported results in the literature.15 In this study, neutrAvidin was found to have the highest degree of nonspecific adsorption (∼40%) to an IgG passivated DOPC bilayer relative to no appreciable nonspecific adsorption of avidin and streptavidin. In a study by Wolny and co-workers, similar behavior has been reported for the nonspecific adsorption of neutral neutrAvidin to a negative silica surface, where its nonspecific adsorption is ∼3-fold greater than that of positively charged avidin.15 The higher

(7)

The calculated values of ΔG0 for the three proteins are listed in Table 1. When the protein−protein interaction is not included, the ΔG0 value of avidin−biotin binding (−45 ± 0.8 kJ/mol) is similar to that of streptavidin (−44 ± 0.5 kJ/mol) and higher than the ΔG0 value of neutrAvidin (−42 ± 0.01 kJ/mol). The free energy of protein−protein interactions (ΔGη), previously described by Zhao et al.,27 can be expressed as follows:

ΔGη = − 4RT ln(η)

(9)

(8) 205

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

degree of nonspecific adsorption of neutrAvidin is unexpected, because the deglycosylation and lower pI of neutrAvidin are intended to decrease its nonspecific adsorption, relative to avidin (information obtained from Pierce, Rockford, IL, USA). It is important to mention that, although streptavidin is also a nonglycosylated protein and has a similar pI to neutrAvidin, the nonspecific adsorption of streptavidin is much less in both this study and the work by Wolny and co-workers. Given the results from both this study and that of Wolny and co-workers, the removal of the carbohydrate groups and the lower pI of neutrAvidin cannot account for the greater nonspecific adsorption of the protein observed here. Kinetics of Avidin Binding to a 4 mol % Biotinylated DOPC Bilayer. In addition to obtaining the binding affinity of the proteins to biotin from the equilibrium thermodynamics, we obtained the binding affinity from kinetic measurements. The comparison between the binding affinities obtained from the thermodynamic and kinetic measurements indicates whether or not the thermodynamic study was performed under steady-state equilibrium conditions. The kinetics of protein binding to a biotinylated DOPC bilayer can be described by the Langmuir kinetic model,52 where the protein adsorption rate (kon) is first-order with respect to the bulk protein concentration (Cbulk) and the fraction of unbound biotin (1 − θ) and the protein desorption rate (koff) are firstorder, with respect to the fraction of protein-bound biotin (θ). The rate of change in the fraction of protein-bound biotin is the difference between the adsorption and desorption rates, given by the following:

Figure 3. Fraction of surface coverage (θ) versus time of avidin binding to a 4 mol % biotinylated DOPC bilayer at the following bulk avidin concentrations: 9.25 nM (red), 18.5 nM (blue), 37.0 nM (pink), 73.7 nM (dark red), 137.7 nM (dark yellow), 273.2 nM (green), and 537.6 nM (black). The solid lines are the global fits to eq 11.

1.8) × 10−5 s−1. From the measured kon and koff, the affinity of biotin−avidin binding can be calculated by Ka = kon/koff = (16 ± 10) × 107 M−1. This value is consistent with the Kapp of avidin ((18 ± 14) × 107 M−1) obtained from steady-state equilibrium thermodynamic measurements presented in the previous section. The agreement between the binding affinity obtained from the kinetics and thermodynamics indicates that there was sufficient time for each protein concentration to reach equilibrium in the thermodynamic measurement. As previously mentioned, there is a large discrepancy between the binding affinity measured in our work and that of Wayment and Harris.18 Our measured kon is ∼4.5 orders of magnitude slower than that reported by Wayment and Harris ((2.1 ± 0.5) × 108 M−1 s−1).18 The much-slower adsorption rate at a high density of biotin to form a monolayer coverage of the protein suggests that the binding affinity of avidin to biotin is much weaker in this work. A speculative explanation for this behavior is that the avidin−biotin binding might involve two regimes: (i) at a very low bulk protein concentration, the protein binds to biotin with a very high affinity, as observed by Wayment and Harris;18 and (ii) at a higher bulk protein concentration, the protein binds to biotin with a lower affinity, as seen here. Since we did not observe a biphasic binding isotherm, it is possible that the protein with the high binding affinity to biotin saturates at very low surface coverage. Using a picomolar range of bulk protein concentration to bind to a very small biotin density to form 0, the surface coverage fraction is θ. These boundary conditions allow the kinetics of protein binding at each Cbulk to be evaluated. The solution for eq 10 with the above stated boundary conditions is as follows:

θ=

konCbulk [1 − exp( − konCbulk konCbulk + koff + koff )t ] + θ0 exp( − konc bulk + koff )t

(11)

where θ0 is zero for the first protein concentration. When the next protein concentration is added, θ0 is the maximum fraction of surface coverage at the previous protein concentration. Using eq 11 to simultaneously fit all adsorption data in Figure 3 gives kon = (9.8 ± 5.3) × 103 M−1 s−1 and koff = (6.0 ± 206

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

by a hinge that allows the Fab segments to move from ∼0° to 180°.57 The flexibility of the Fab segments, including the arm rotation, elbow bend, and arm wagging,58 allows the antibody to adjust the spacing and orientation of its binding sites and thus facilitates the ligand−antibody binding.56,57 Therefore, the biotin-binding of antibiotin antibody is not significantly affected by the accessibility of biotin, whereas the lower biotin accessibility in the biotinylated lipid bilayer causes a reduction in the binding affinity of avidin, streptavidin, and neutrAvidin to a biotinylated lipid bilayer relative to free biotin. This results in the same order of magnitude of the binding affinity for all four proteins to the biotinylated lipid bilayer obtained here. Another interesting feature observed is the nonspecific adsorption of antibiotin antibody to a pure DOPC bilayer is negligible (Figure 4). Note that no IgG from rabbit serum was used to suppress the nonspecific adsorption of antibiotin antibody in this experiment. Therefore, this feature could make the antibiotin antibody−biotin complex a prominent tether/ linker in bioanalytical applications, as the time and cost in minimizing the protein’s nonspecific adsorption can be avoided.

Binding of Antibiotin Antibody to Biotinylated DOPC. The binding isotherm for antibiotin antibody binding to a DOPC bilayer containing 4 mol % biotin-cap-DOPE is shown in Figure 4. The binding of antibiotin was fit using the

Figure 4. Corrected SHG intensity versus bulk protein concentration for antibiotin antibody binding to DOPC bilayers containing 4 mol % biotin-cap-DOPE (filled circles) and 0 mol % biotin-cap-DOPE (open circles). The line is the fit to the Langmuir adsorption isotherm. The error bars represent the standard deviation from three independent positive experiments and two control experiments.

■

SUMMARY We have investigated the binding of avidin, streptavidin, neutrAvidin, and antibiotin antibody to a biotinylated lipid bilayer using SHG. The binding affinities of avidin, strepatvidin, and neutrAvidin to 4 mol % biotin-cap-DOPE incorporated into a DOPC lipid were determined. A positive cooperative binding behavior was obtained for these three proteins, demonstrating that protein−protein interactions enhance the biotin−protein binding. Moreover, the binding of streptavidin and neutrAvidin to biotin is more energetically favorable than that of avidin, because of the stronger protein−protein interactions. Although neutrAvidin is designed to lower the nonspecific adsorption, we found that it exhibits the greatest degree of nonspecific adsorption to a DOPC bilayer relative to avidin and streptavidin. A similar study of antibiotin antibody revealed that it binds to biotin with a similar affinity as avidin, streptavidin, and neutrAvidin. Furthermore, antibiotin antibody showed negligible nonspecific adsorption to a DOPC bilayer without the use of an additional agent to reduce the protein’s nonspecific adsorption. This study presented a deeper understanding about the binding properties of avidin, streptavidin, neutrAvidin, and antibiotin antibody to biotin at lipid bilayer surfaces, which can further assist in selecting an appropriate tether/linker for biosensing and other bioanalytical applications.

Langmuir model from eq 5. The K0 of antibiotin antibody to the biotinylated lipid bilayer measured here of (1.0 ± 0.4) × 108 M−1 is close to the value of (2.8 ± 0.8) × 108 M−1 obtained by Jung et al. for fluorescently labeled antibiotin antibody binding to 5 mol % biotin-cap-PE incorporated into a lipid bilayer.22 A direct comparison between the binding affinities measured in this study for antibiotin antibody, avidin, streptavidin, and neutrAvidin can be made, as they were all determined under similar experimental conditions. The Kapp values of avidin, streptavidin, and neutrAvidin were used in the comparison, as it takes into account protein−protein interactions involved in the binding of these proteins. Since no protein−protein interactions were observed for antibiotin antibody, Kapp is equal to K0. Interestingly, we found that the binding affinity of antibiotin antibody to a biotinylated lipid bilayer has the same order of magnitude (∼10 8 M−1 ) as the K app values of avidin, streptavidin, and neutrAvidin. As a result, the free energy of antibiotin antibody binding to biotin (46 ± 1.0 kJ/mol) is also very close to those of avidin, streptavidin, and neutrAvidin (see Table 1). Streptavidin and antibiotin antibody have also been found to have the same affinity to biotin when biotin is attached to bovine serum albumin via a linker53 even though it is wellknown that antibiotin antibody binds to free biotin in solution with a much lower affinity than avidin.53,54 This could be explained by the higher dependence of avidin’s binding affinity on the biotin accessibility, i.e., lower accessibility of biotin when it is attached to a macromolecule and higher biotin accessibility when it is free in solution. Avidin’s binding sites are located in a depression near the end of the β-barrels,17 which makes it more difficult to bind with biotin when biotin is bound to a macromolecule, therefore reducing its binding affinity, relative to that of free biotin.53,55 Unlike avidin, antibiotin antibody’s binding sites are located on the end of the Fab segments not in a depression,56 which makes the binding affinity of antibiotin antibody less dependent on the biotin accessibility.53 Furthermore, the Fab segments of the antibody are connected

■

ASSOCIATED CONTENT S Supporting Information * Materials used and preparation of PSLBs, derivations of eqs 3 and 4, and normalization of the SHG intensities shown in Figure 2. This material is available free of charge via the Internet at http://pubs.acs.org.

■ ■

AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]. ACKNOWLEDGMENTS The authors acknowledge the financial support from the National Institutes of Health (NIH) (No. R01-GM068120). Any opinions, findings, conclusions, or recommendations 207

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Analytical Chemistry

Article

(39) Hicks, J. M.; Petralli-Mallow, T. Appl. Phys. B: Lasers Opt. 1999, 68, 589. (40) Fujii, S.; Morita, T.; Kimura, S. J. Pept. Sci. 2008, 14, 1295. (41) Kriech, M. A.; Conboy, J. C. J. Am. Chem. Soc. 2003, 125, 1148. (42) Mitchell, S. A. J. Phys. Chem. B 2009, 113, 10693. (43) Wu, X.-Z.; Huang, T.; Mullett, W. M.; Yeung, J. M.; Pawliszyn, J. J. Microcolumn Sep. 2001, 13, 322. (44) Kriech, M. A.; Conboy, J. C. J. Opt. Soc. Am. B: Opt. Phys. 2004, 21, 1013. (45) Shen, Y. R. The Principles of Nonlinear Optics; Wiley: New York, 1984. (46) Havel, H. A.; Chao, R. S.; Haskell, R. J.; Thamann, T. J. Anal. Chem. 1989, 61, 642. (47) Guyot-Sionnest, P.; Shen, Y. R. Phys. Rev. B: Condensed Matter 1987, 35, 4420. (48) Shen, Y. R. Appl. Phys. B: Lasers Opt. 1999, 68, 295. (49) Tang, Y.; Mernaugh, R.; Zeng, X. Anal. Chem. 2006, 78, 1841. (50) Blankenburg, R.; Meller, P.; Ringsdorf, H.; Salesse, C. Biochemistry 1989, 28, 8214. (51) Ku, A. C.; Darst, S. A.; Kornberg, R. D.; Robertson, C. R.; Gast, A. P. Langmuir 1992, 8, 2357. (52) Langmuir, I. J. Am. Chem. Soc. 1918, 40, 1361. (53) Vincent, P.; Samuel, D. J. Immunol. Methods 1993, 165, 177. (54) Bagci, H.; Kohen, F.; Kuscuoglu, U.; Bayer, E. A.; Wilchek, M. FEBS Lett. 1993, 322, 47. (55) Brigati, D. J.; Myerson, D.; Leary, J. J.; Spalholz, B.; Travis, S. Z.; Fong, C. K.; Hsiung, G. D.; Ward, D. C. Virology 1983, 126, 32. (56) Tsai, C. S. Biomacromolecules: Introduction to Structure, Function and Informatics; John Wiley & Sons: Hoboken, NJ, 2007. (57) Hanson, D. C.; Yguerabide, J.; Schumaker, V. N. Biochemistry 1981, 20, 6842. (58) Burton, D. R. Mol. Immunol. 1985, 22, 161.

expressed in this material are those of the authors and do not necessarily reflect the views of the NIH.

■

REFERENCES

(1) Grunwell, J. R.; Glass, J. L.; Lacoste, T. D.; Deniz, A. A.; Chemla, D. S.; Schultz, P. G. J. Am. Chem. Soc. 2001, 123, 4295. (2) Wennmalm, S.; Edman, L.; Rigler, R. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 10641. (3) Ladd, J.; Boozer, C.; Yu, Q.; Chen, S.; Homola, J.; Jiang, S. Langmuir 2004, 20, 8090. (4) Esseghaier, C.; Helali, S.; Ben Fredj, H.; Tlili, A.; Abdelghani, A. Sens. Actuators, B 2008, B131, 584. (5) Sun, H.; Choy, T. S.; Zhu, D. R.; Yam, W. C.; Fung, Y. S. Biosens. Bioelectron. 2009, 24, 1405. (6) Hall, W. P.; Ngatia, S. N.; Van Duyne, R. P. J. Phys. Chem. C 2011, 115, 1410. (7) Bashir, R.; Gomez, R.; Sarikaya, A.; Ladisch, M. R.; Sturgis, J.; Robinson, J. P. Biotechnol. Bioeng. 2001, 73, 324. (8) Lazcka, O.; Del Campo, F. J.; Munoz, F. X. Biosens. Bioelectron. 2007, 22, 1205. (9) Barton, A. C.; Davis, F.; Higson, S. P. J. Anal. Chem. (Washington, DC, U.S.A.) 2008, 80, 9411. (10) Zhavnerko, G. K.; Yi, S. J.; Chung, S. H.; Yuk, J. S.; Ha, K. S. NATO Sci. Ser., II 2004, 152, 95. (11) Cooper, M. A. Nat. Rev. Drug Discovery 2002, 1, 515. (12) Litos, I. K.; Ioannou, P. C.; Christopoulos, T. K.; TraegerSynodinos, J.; Kanavakis, E. Biosens. Bioelectron. 2009, 24, 3135. (13) Green, N. M. Biochem. J. 1963, 89, 585. (14) Green, N. M. Adv. Protein Chem. 1975, 29, 85. (15) Wolny, P. M.; Spatz, J. P.; Richter, R. P. Langmuir 2010, 26, 1029. (16) DeLange, R. J. J. Biol. Chem. 1970, 245, 907. (17) Green, N. M. Methods Enzymol. 1990, 184, 51. (18) Wayment, J. R.; Harris, J. M. Anal. Chem. (Washington, DC, U.S.A.) 2009, 81, 336. (19) Pierce Biotechnology. Instructions for NeutrAvidin, Product No. 31000. (20) Danilowicz, C.; Manrique, J. M. Electrochem. Commun. 1999, 1, 22. (21) Mushahwar, I. K.; Spiezia, K. S. J. Virol. Methods 1987, 16, 45. (22) Jung, H.; Yang, T.; Lasagna, M. D.; Shi, J.; Reinhart, G. D.; Cremer, P. S. Biophys. J. 2008, 94, 3094. (23) Xu, F.; Zhen, G.; Yu, F.; Kuennemann, E.; Textor, M.; Knoll, W. J. Am. Chem. Soc. 2005, 127, 13084. (24) Osborne, M. A. J. Phys. Chem. B 2005, 109, 18153. (25) Patel, A. R.; Frank, C. W. Langmuir 2006, 22, 7587. (26) Sandrin, L.; Coche-Guerente, L.; Bernstein, A.; Basit, H.; Labbe, P.; Dumy, P.; Boturyn, D. Org. Biomol. Chem. 2010, 8, 1531. (27) Zhao, S.; Walker, D. S.; Reichert, W. M. Langmuir 1993, 9, 3166. (28) Zhao, S.; Reichert, W. M. Langmuir 1992, 8, 2785. (29) Shi, J.; Yang, T.; Kataoka, S.; Zhang, Y.; Diaz, A. J.; Cremer, P. S. J. Am. Chem. Soc. 2007, 129, 5954. (30) Marsh, D. CRC Handbook of Lipid Bilayers; CRC Press: Boca Raton, FL, 1990. (31) Smith, K. A.; Gale, B. K.; Conboy, J. C. Anal. Chem. (Washington, DC, U.S.A.) 2008, 80, 7980. (32) Mao, H.; Yang, T.; Cremer, P. S. Anal. Chem. 2002, 74, 379. (33) Sun, Y. S.; Landry, J. P.; Fei, Y. Y.; Zhu, X. D.; Luo, J. T.; Wang, X. B.; Lam, K. S. Langmuir 2008, 24, 13399. (34) Song, X.; Swanson, B. I. Anal. Chem. 1999, 71, 2097. (35) Ye, S.; Nguyen, K. T.; Chen, Z. J. Phys. Chem. B 2010, 114, 3334. (36) Wang, J.; Paszti, Z.; Clarke, M. L.; Chen, X.; Chen, Z. J. Phys. Chem. B 2007, 111, 6088. (37) Kim, G.; Gurau, M. C.; Lim, S.-M.; Cremer, P. S. J. Phys. Chem. B 2003, 107, 1403. (38) Salafsky, J. S.; Eisenthal, K. B. J. Phys. Chem. B 2000, 104, 7752. 208

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208