Comparison of the Energetics of Avidin, Streptavidin, NeutrAvidin,and Anti-Biotin Antibody Binding to Biotinylated Lipid BilayerExamined by Second-Harmonic GenerationTrang 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 biotinylatedlipid bilayer was studied using second-harmonic generation. Proteinbinding assays were performed on a planar supported lipid bilayer of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) containing 4 mol %biotinylated-cap-1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (biotin-cap-DOPE). The equilibrium binding affinities of these biotin−proteininteractions were determined, revealing the relative energetic contribu-tions for each protein to the biotinylated lipid ligand. The results showthat the binding affinities of avidin, streptavidin, and neutrAvidin forbiotin were all strengthened by protein−protein interactions but that thestronger 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 avidinand streptavidin. In addition, the biotin-binding affinity of the antibiotin antibody was found to be of the same order ofmagnitude as that of avidin, streptavidin, and neutrAvidin. These findings provide important new insights into these biotin-boundprotein complexes commonly used in several bioanalytical applications.

Biotin-bound protein complexes have been used in a widevariety of bioanalytical applications, including monitoring

biomolecule conformational changes,1,2 biochip sensor fab-rication,3−8 immunoassays,9,10 and targeted drug delivery andscreening.11 In these applications, the protein−biotin complexis commonly used to tether biomolecules to a surface1,2 or usedas a linker to capture biomolecules3−12 by taking advantage ofthe high affinity, specificity, and stability of biotin-boundproteins, such as avidin,7−9 streptavidin,1−3,10,11 neutravidin,4,5

and antibiotin antibody.6 Avidin is a tetramer consisting of fouridentical subunits, each of which has one binding site for biotin.It has an extremely high binding affinity, Ka ≈ 1015 M−1, to freebiotin in solution,13 and it forms a stable complex with biotinover a wide range of temperatures and pH values.14 One majordrawback to using avidin is the high degree of nonspecificadsorption caused by its basic isoelectric point (pI ∼10).14 Atphysiological pH, positively charged avidin can bind non-specifically to negatively charged surfaces such as cellmembranes14 or silica substrates.15 Another feature thatcontributes to avidin’s high nonspecific adsorption is itscarbohydrate groups, which contain four mannose residuesand three N-acetylglucosamine residues per subunit.16 If theavidin−biotin complex is being used to capture a carbohydrate-binding molecule, it can also bind specifically to thecarbohydrate groups on avidin, limiting its use in bioassays.Both streptavidin and neutrAvidin, which are analogues ofavidin, which also have a high affinity and specificity towardbiotin,17,18 have been used as an alternative to avidin.3−5

Streptavidin, which has a similar functional domain to avidin(∼33% identical residues),17 is a nonglycosylated protein with aslightly acidic pI of ∼5−6.14 NeutrAvidin is a commerciallyavailable deglycosylated form of avidin with a pI of 6.3.19 Thelower pI values and the absence of the sugar groups instreptavidin and neutrAvidin are intended to lower thenonspecific binding, relative to avidin, without significantlyaffecting the affinity for biotin. As such, streptavidin andneutrAvidin are commonly used in biosensing applications, asan alternative to avidin, as a way to reduce nonspecific bindingto the sensor surfaces. Antibiotin antibody, which is animmunoglobulin protein that is generated by plasma cells aspart of the immune response to the antigen biotin, has alsobeen used as an alternative linker in biosensors6,12 andimmunoassays,20,21 since it also has a high binding affinity forbiotin (∼108 M−1).22

Surprisingly, little work to date has compared the specific andnonspecific binding affinities of avidin, streptavidin, neutrAvi-din, and antibiotin antibody to biotin at surfaces, despite theuse of these proteins in a broad range of biosensingapplications. Here, we report the binding affinities of thesebiotin-binding proteins to biotinylated lipid bilayers whilespecifically addressing the energetics of the binding process andthe nonspecific adsorption of all four proteins.

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

Article

pubs.acs.org/ac

© 2011 American Chemical Society 201 dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208

Biosensing applications have employed the use of protein−biotin complexes to various types of surfaces, includingfunctionalized gold,3,4,10,23 glass,1,24 silver nanoprisms,6 as wellas planar supported lipid bilayers (PSLBs) or liposomes.25

PSLBs were chosen as a platform for the protein binding assaysin this study because of the ease of preparation and the abilityto precisely control the biotin density,26 providing betterreproducibility and facilitating comparison of the variousprotein−biotin interactions.The binding affinity between protein−ligand pairs at surfaces

is known to be affected by the surface density of theligand.27−29 Zhao and co-workers have shown that, once thebiotin density is high enough to bind a monolayer of avidin, anyadditional biotin sterically hinders further avidin binding.28

Their study also showed that doubly bound avidin−biotincomplex (two biotin molecules for every one avidin molecule)is more stable than the singly bound complex.28 This suggeststhat the optimal biotin density would be one that allows amonolayer of avidin to bind bivalently. In our study, theappropriate biotin density to form a doubly bound complexmonolayer of the protein can be calculated using the area foravidin, 3025 Å2,14 and for a lipid and biotinylated lipidmolecule, 70 Å.30 For a monolayer of avidin (3.31 × 1012

molecule/cm2) to bivalently bind to biotin, the biotin densitymust be twice as large; therefore, the density of the biotinylatedlipid in the lipid bilayer should be ∼4.6 mol % [(2 × 3.31 ×1012 biotinylated lipid molecule/cm2)/(1.43 × 1014 lipidmolecule/cm2) × 100%]. Previous studies that investigatedbiotin binding to neutrAvidin and streptavidin in PSLBs havealso shown that protein binding increases with biotin density upto 4 mol % and then saturates.31,32 At this density, two biotinmolecules effectively bind to every one protein to form adoubly protein−biotin complex monolayer on the lipid bilayersurface. When the biotin density is lower than 4 mol %, fewerprotein−ligand complexes are formed; above 4 mol %, sterichindrance from neighboring proteins decreases binding. Inorder to provide useful information for bioanalyticalapplications, which aim to maximize the number of capturedbiomolecules, the optimal biotin density of 4 mol % (amonolayer coverage of bivalently bound protein) was usedhere.To investigate these protein−biotin interactions, we utilized

second-harmonic generation (SHG), which is a surface-specificnonlinear optical spectroscopic technique capable of directlydetecting protein interactions with surfaces. As opposed tofluorescence, SHG does not require an exogenous label to beattached to the protein. Attaching a label can alter the nativeconformation and/or charge of a protein and therefore alter itsbinding capabilities.33,34 Thus, nonlinear optical spectroscopictechniques, such as sum frequency generation vibrationalspectroscopy (SFVS) and SHG, have recently been shown aspromising alternatives for label-free biomolecule recognition atinterfaces. For example, SFVS has been used to probe thestructural interactions between proteins/peptides and lipidmonolayers or bilayers.35−37 SHG has also been used tomonitor protein adsorption at a solid/liquid interface38,39 andassociation of biomolecules to a lipid monolayer or bilayer.40−42

SHG provides the surface selectivity and high sensitivityrequired to study protein−biotin interactions at an interfacewithout using an external label. Therefore, we employed SHGto study the interactions between biotin and the four biotin-binding proteins, avidin, streptavidin, neutrAvidin, andantibiotin antibody, in PSLBs. Using the same experimental

conditions to examine all four proteins allows for a directcomparison of the binding properties of these biotin−proteincomplexes as well as comparison to previous literature studiesof these biotin-bound proteins, providing useful information onthese tether/linker complexes commonly used in severalbioanalytical applications.

■ EXPERIMENTAL SECTION

The materials used and preparation of PSLBs are described inthe Supporting Information.

Ligand−Protein Binding Assay. PSLBs of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) containing 4 mol % 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(Cap biotinyl)for avidin, neutrAvidin, streptavidin, and antibiotin antibodybinding were created on a prism surface by vesicle fusion asdescribed in the Supporting Information. Three independentexperiments were conducted for each protein, except for thecase of neutrAvidin and the antibiotin control, where a t-testwas conducted, resulting in the elimination of one dataset for atotal of two samples. To reduce nonspecific adsorption of theproteins, the PSLBs were incubated in 1 mg/mL monoclonalIgG (pI ≈ 6.1−6.5)43 from rabbit serum in PBS pH 7.4 for 30min to block any defects that might exist on the lipid surfaces,except for the antibiotin antibody experiments. The PSLBswere then rinsed thoroughly with PBS to remove any free IgGremaining in solution.Increasing concentrations of avidin, neutrAvidin, and

streptavidin ranging from 9.25 nM to 537.6 nM were injectedinto the flow cell and allowed to reach equilibrium. The sameprocedure was followed for antibiotin; however, the concen-tration range was from 4 nM to 121 nM. At each proteinconcentration, the SHG intensity was recorded at 30 minintervals until a steady-state response was achieved. Generally,low concentrations of the proteins required up to 4 h to reachequilibrium. During this period, at least 1.5 mL (∼3 times thevolume of the flow cell) of fresh protein solution was injectedevery 30 min to account for the bulk depletion caused bysurface adsorption of the proteins. It is important to note thatprotein dilutions were freshly prepared prior to each injectionto further prevent the proteins from nonspecifically adsorbingto the vials and syringes. Thermodynamic measurements wereperformed using the SHG intensity collected at equilibrium ateach protein concentration. In addition, kinetic measurementswere determined for the SHG intensity collected over time.

SHG Measurements. Counter-propagating SHG wasemployed here. A detailed description of this technique canbe found elsewhere.44 Briefly, a 532-nm laser beam (second-harmonic output of a Nd:YAG laser, Surelite I, 10 Hz) with amixed polarization state (equal amounts of s and p polarizedcomponents) was directed at the prism/water interface undertotal internal reflection. The laser intensity used was 14 mJ/pulse. The reflected beam was steered back to overlap spatiallyand temporally with the incidence, generating SHG at 266 nm.Optical filters were used to remove any scattered visible lightbefore the reflected SHG signal was collected with a solar-blindphotomultiplier tube.The SHG intensity is proportional to the second-order

susceptibility tensor, χijk(2), which has a nonresonant χNR

(2) andresonant χR

(2) contribution:

∝ |χ | ∝ |χ + χ |I ijkSHG(2) 2

NR(2)

R(2) 2

(1)

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208202

The resonant χR(2) can be expressed as follows:45

∑χ ∝⟨ |μ | ⟩⟨ |μ | ⟩⟨ |μ | ⟩

ω − − Γ ω − − ΓN

a c a b b c

h E i h E i( )

(2 )( )ijka b c

i j k

ca ca bc bc

(2)R

, , (2)

where N is the surface density of molecules; h is Planck’sconstant; a, b, and c denote the initial, intermediate, and finalstates, respectively; μ is the Cartesian coordinate dipoleoperator; and Γ represents the transition line width. Theindices on χijk

(2) represent the input (j,k) and output (i) fields,which can assume any of the three Cartesian coordinates(x,y,z). As shown in eq 2, an enhancement in the SHG signal isobserved when the incident (ω) or SHG (2ω) frequency isresonant with an electronic transition of a molecule at theinterface. Therefore, SHG can be used to detect the presence ofbound protein to a lipid membrane if the protein has electronictransitions at the frequency of the incident fundamental (532-nm) or the SHG (266-nm) light. As seen in the extinctioncoefficient spectra of the proteins in Figure 1, the SHG

wavelength at 266 nm is in resonance with the π → π*transitions of the protein’s tryptophan and tyrosine’s aromaticrings,46 resulting in enhancement of the SHG signal when theprotein is present at the lipid surface. Although the electric-quadrupole response from the bulk medium can contribute tothe overall SHG signal,47,48 its contribution to the measuredsignal is negligible if the SHG frequency is in resonance withthe electric-dipole allowed transition of molecules residing atthe interface.47 As the SHG frequency used in this study isresonant with the transitions of the proteins adsorbed to thesurface, the measured SHG intensity is predominantly dipolarin nature with little to no detectable contribution from thequadrupolar response expected.The SHG intensity expression assumed the Langmuir

isotherm model can be given as

−

∝+

+ ++

⎛⎝⎜⎜

⎞⎠⎟⎟

⎛⎝⎜⎜

⎞⎠⎟⎟

I I

I BI K P

K P

B CI K P

K P

2[ ]

1 [ ]

( )[ ]

1 [ ]

SHG SHGbackground

SHGbackground SHG

max0

0

2 2 SHGmax

0

0

2

(3)

where (ISHGbackground)1/2 = A is the square root of the SHG intensity

from the background (nonresonant response) by assumingχNR(2) = A is real and positive. B and C denote the real and

imaginary components of χR(2) due to the proteins, respectively.

(ISHGmax )1/2 is the square root of the maximum SHG intensity at

surface saturation, K0 is the equilibrium association constant,and [P] is the bulk protein concentration. The derivation of thisequation can be found in the Supporting Information.For avidin, a previous study demonstrated that protein−

protein interactions can modulate the energetics of proteinassociation to a biotinylated surface.27 When interactionsbetween proteins are involved, a cooperative binding model,previously described by Zhao et al.,27 can be incorporated intothe SHG intensity expression:

−

∝ω

+ ω

+ +ω

+ ω

⎛

⎝⎜⎜

⎞

⎠⎟⎟

I I

B II K P

K P

B CI K P

K P

2[ ]

1 [ ]

( )[ ]

1 [ ]

I I

I I

I I

I I

SHG SHGbackground

SHGbackground SHG

max ( / )0

( / )0

2 2 SHGmax ( / )

0

( / )0

2

SHG SHGmax

SHG SHGmax

SHG SHGmax

SHG SHGmax

(4)

where ω = η4, by assuming the distribution of the biotin-boundproteins follows a square lattice, and η is the cooperativitycoefficient, which characterizes the protein−protein interac-tions between neighboring protein molecules on the surface.27

K0 is the intrinsic binding affinity of the protein to the ligandbarring any protein−protein interactions.27 When η > 1, thebinding of a protein to a ligand exhibits a positive cooperativity,demonstrating that the protein−protein interaction enhancesthe ligand−protein binding. When η < 1, the protein−proteininteraction reduces the ligand−protein binding, resulting in anegative cooperativity. The cooperativity model becomes theLangmuir model when η = 1. The derivation of eq 4 was alsodescribed in the Supporting Information.

■ RESULTS AND DISCUSSION

Thermodynamics of Avidin, Streptavidin, and Neu-trAvidin Binding to Biotinylated DOPC Bilayers. Thebinding isotherms for avidin, neutrAvidin, and streptavidin areshown in Figure 2. Note that the SHG intensities shown inFigure 2 were normalized as described in the SupportingInformation. The SHG intensities increase with increasingprotein concentration until saturation coverage is achieved atconcentrations of >100 nM. The Langmuir and cooperativitymodels were both used to fit the data, and the f-test wasperformed to determine the best fit for the data.The Langmuir model (eq 3) was used to fit the data with the

fitting parameters B, C, (ISHGmax )1/2, and K0 by performing a

nonlinear least-squares regression. The cross-term is theproduct of the nonresonant component (A) and realcomponent (B) of the resonant susceptibility tensor, whichdescribes the interference between the background and proteinadsorption responses. Assuming the nonresonant term A is realand positive, the measured SHG intensity can increase ordecrease due to constructive (A and B have the same sign) ordestructive (A and B have different signs) interference. In thecase of our data, as seen in Figure 2, we did not observe anyinitial decrease in the SHG intensity at low proteinconcentrations, where the contribution of the resonant termis presumably not much greater than the nonresonantcontribution. This observation suggests there is constructiveinterference between the nonresonant background and aresonant response arising from the adsorbed protein species,

Figure 1. Extinction coefficient spectra of the proteins.

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208203

indicating the cross-term (2AB) will be positive since A and Bhave the same sign. In addition, the results from the nonlinearregression indicate that the magnitude of B is ∼8 orders ofmagnitude smaller than the magnitude of the imaginary portionof the resonant susceptibility tensor (C). Since the contributionfrom the cross-term (2B(ISHG

background)1/2) in eq 3 is much smaller,compared to the resonant contribution term (B2 + C2), it canbe neglected. (B2 + C2), which is related to the surface density(eq 2), serves as a scaling factor and can then be pooled into

(ISHGmax )1/2. Using these simplifications, the final Langmuir

equation used to fit the adsorption data in this study is asfollows:

∝+

⎛⎝⎜⎜

⎞⎠⎟⎟I

I K P

K P

[ ]

1 [ ]SHGSHGmax

0

0

2

(5)

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

∝ω

+ ω

⎛

⎝⎜⎜

⎞

⎠⎟⎟I

I K P

K P

[ ]

1 [ ]

I I

I ISHGSHGmax ( / )

0

( / )0

2SHG SHG

max

SHG SHGmax

(6)

Equations 5 (the Langmuir model) and 6 (the cooperativitymodel) were then used to fit the adsorption data of avidin,streptavidin, and neutrAvidin shown in Figure 2. It was foundthat the cooperativity model statistically fit the data best, andthe results from a nonlinear least-squares regression of the dataare given in Table 1. The intrinsic binding affinities (K0) ofavidin, streptavidin, and neutrAvidin to a biotinylated DOPCbilayer 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 thisstudy is in good agreement with the value of (2.14 ± 0.20) ×107 M−1 previously reported by Zhao et al. for fluorescentlylabeled avidin binding to a 0.63 mol % biotinylated lipidincorporated into an arachidic acid monolayer.27 Forstreptavidin, the K0 value obtained here is slightly higher thanthe value of 7.3 ± 0.2 × 106 M−1 reported by Tang and co-workers for streptavidin binding a biotin monolayer function-alized on a SPR gold chip surface.49 The difference in the K0values of streptavidin between the two studies can be related tothe difference in the surface density of biotin used. It is likelythat the high biotin density used in the SPR study creates sterichindrance, which reduces the accessibility of an individualbiotin to the protein binding sites, resulting in a lower K0 value.For neutrAvidin binding to biotin, the K0 value measured in thisstudy is ∼4.5 orders of magnitude smaller than an affinity of(5.5 ± 0.2) × 1011 M−1 reported by Wayment and Harris, usingsingle-molecule fluorescence.18 To compare the result of theWayment and Harris study with our study, we performed asimilar analysis of the kinetics of protein binding, which ispresented in the following section. It was shown that the sloweradsorption and desorption rates of biotin−protein measured inthis 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 proteinsbinding to DOPC bilayers containing 4 mol % biotin-cap-DOPE(filled circles) and 0 mol % biotin-cap-DOPE (open circles). The linesare fits to the cooperativity binding model. The error bars representthe standard deviation from three independent experiments, except forthe specific binding of neutrAvidin, which represents two independentexperiments.

Table 1. Measured Intrinsic Binding Affinity (K0), Cooperativity Coefficient (η), Intrinsic Free Energy (ΔG0), Free Energy Dueto 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 8.2 ± 2.4 1.2 ± 0.2 −45 ± 0.8 −1.7 ± 1.4 −47 ± 1.7 18 ± 14streptavidin 4.3 ± 0.9 1.8 ± 0.2 −44 ± 0.5 −5.7 ± 1.1 −49 ± 0.6 44 ± 10neutrAvidin 2.6 ± 0.01 1.9 ± 0.01 −42 ± 0.01 −6.3 ± 0.04 −49 ± 0.05 23 ± 0.7

aData were obtained with the cooperativity binding model, using the assumption that the distribution of the biotin-bound proteins follows a squarelattice.

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208204

indicates that protein−protein interactions are involved in thebiotin−protein binding. The value of the cooperativitycoefficient (η) obtained from the fit is shown in Table 1. Avalue 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 cooperativebinding behavior. This suggests that the intrinsic binding of theproteins to biotin is increased due to protein−proteininteractions. Similar cooperative binding behavior has beenpreviously observed for avidin to a monolayer of arachidic acidcontaining biotinylated lipid.27 The cooperative binding ofavidin only was observed if the biotin density was adequatelyhigh to bind to a monolayer of avidin so that the distancebetween the biotin molecules is close enough for the adjacentbiotin-bound avidin to interact with an incoming avidin.27 Inaddition, biotin must be placed above the lipid surface (i.e.,using a spacer between biotin and the lipid headgroup, toprovide more accessibility for avidin binding).27 The use of abiotin density high enough to form a monolayer of the proteins(∼ 4 mol % biotin) and the spacer between biotin and the lipidheadgroup (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 thatavidin exhibits less positive cooperative binding. The lowercooperativity observed for avidin as compared to streptavidinhas been noted before.27,50,51 Blackenburg et al. noted thatstreptavidin tends to aggregate and form highly ordereddomains, which suggests streptavidin−streptavidin interac-tions.50 Although Blankenburg et al. did not observe anydomains for avidin, Ku et al. demonstrated that domains wereformed by avidin; however, these domains were very small.50,51

The smaller domain size observed for avidin suggests thatavidin has a lower positive cooperativity, compared tostreptavidin, congruent to the observations in this study. Thelower cooperativity observed for avidin could be related to thehigh pI (∼10) of the protein. At the neutral pH used here,avidin is positively charged, and thus, the electrostatic repulsionbetween the charged avidin could repel the other avidinproteins, resulting in a reduction in avidin−avidin interactions.Conversely, streptavidin and neutrAvidin have a pI of ∼7;therefore, the electrostatic interaction between streptavidin andstreptavidin or neutrAvidin and neutrAvidin is less pronounced,compared to the avidin−avidin interactions, leading to a higherη obtained for streptavidin and neutrAvidin.The level of cooperative binding observed in avidin,

streptavidin, and neutrAvidin can also be evaluated in termsof the binding free energy. Without any contributions fromprotein−protein interaction, the free energy of protein−biotinbinding (ΔG0) can be calculated from K0:

Δ = −G RT kln0 0 (7)

The calculated values of ΔG0 for the three proteins are listed inTable 1. When the protein−protein interaction is not included,the ΔG0 value of avidin−biotin binding (−45 ± 0.8 kJ/mol) issimilar to that of streptavidin (−44 ± 0.5 kJ/mol) and higherthan 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 asfollows:

Δ = − ηηG RT4 ln( ) (8)

Using eq 8, ΔGη was calculated from the binding isotherms inFigure 2 for avidin, streptavidin, and neutrAvidin, and theresults are listed in Table 1. The protein−protein interactionsfor 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 wasdetermined using the following equation:

Δ = Δ + Δ ηG G Gtotal 0 (9)

the results of which are listed in Table 1. The calculated ΔGtotalvalue of avidin−biotin binding (−47 ± 1.7 kJ/mol) is lowerthan that of streptavidin (−49 ± 0.6 kJ/mol) and neutrAvidin(−49 ± 0.05 kJ/mol). Accordingly, the effective bindingbetween biotin and streptavidin or biotin and neutrAvidin ismore 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 wascalculated, and the results are given in Table 1. The largercalculated values for Kapp, compared to K0, reveal that theaffinity of the proteins toward biotin is enhanced by protein−protein interactions. When these protein−protein interactionsare taken into account, avidin has the lowest Kapp value ((18 ±14) × 107 M−1), while streptavidin and neutrAvidin exhibitstronger Kapp values ((44 ± 10) × 107 M−1 and (33 ± 0.7) ×107 M−1, respectively).

Nonspecific Adsorption of Avidin, neutrAvidin, andStreptavidin to a DOPC Bilayer. The nonspecific adsorp-tion of avidin, streptavidin, and neutrAvidin to the IgGpassivated DOPC bilayer without biotin is shown as opencircles in Figure 2. The amount of nonspecific adsorption of theproteins can be evaluated using (ISHG)

1/2, which is directlyproportional to the protein surface concentration. To comparethe amount of nonspecific adsorption, we calculated thepercent surface coverage in the absence of biotin at 537.6nM (above saturation concentration), relative to the calculatedmonolayer coverage of adsorbed protein in the presence ofbiotin. Note that the proteins may have a different orientationand conformation when nonspecifically bound to the IgGpassivated DOPC bilayer, compared to when specifically boundto biotinylated DOPC. This may alter the conformation andorientation of the tryptophan and tyrosine residues probed inthis experiment. If the orientation of the residues becomesmore disordered when nonspecifically bound to the IgGpassivated DOPC bilayer, this could produce a lower SHGsignal, indicating a lower quantity of protein bound at thesurface than what is actually bound. Although the orientation ofthe tryptophan and tyrosine residues could affect the observedSHG signal and care should be taken when comparingnonspecific binding to specific binding, the trends in specificversus nonspecific binding measured by SHG in the currentstudy are consistent with previously reported results in theliterature.15 In this study, neutrAvidin was found to have thehighest degree of nonspecific adsorption (∼40%) to an IgGpassivated DOPC bilayer relative to no appreciable nonspecificadsorption of avidin and streptavidin. In a study by Wolny andco-workers, similar behavior has been reported for thenonspecific adsorption of neutral neutrAvidin to a negativesilica surface, where its nonspecific adsorption is ∼3-foldgreater than that of positively charged avidin.15 The higher

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208205

degree of nonspecific adsorption of neutrAvidin is unexpected,because the deglycosylation and lower pI of neutrAvidin areintended to decrease its nonspecific adsorption, relative toavidin (information obtained from Pierce, Rockford, IL, USA).It is important to mention that, although streptavidin is also anonglycosylated protein and has a similar pI to neutrAvidin, thenonspecific adsorption of streptavidin is much less in both thisstudy and the work by Wolny and co-workers. Given the resultsfrom both this study and that of Wolny and co-workers, theremoval of the carbohydrate groups and the lower pI ofneutrAvidin cannot account for the greater nonspecificadsorption of the protein observed here.Kinetics of Avidin Binding to a 4 mol % Biotinylated

DOPC Bilayer. In addition to obtaining the binding affinity ofthe proteins to biotin from the equilibrium thermodynamics,we obtained the binding affinity from kinetic measurements.The comparison between the binding affinities obtained fromthe thermodynamic and kinetic measurements indicateswhether or not the thermodynamic study was performedunder steady-state equilibrium conditions. The kinetics ofprotein binding to a biotinylated DOPC bilayer can bedescribed by the Langmuir kinetic model,52 where the proteinadsorption rate (kon) is first-order with respect to the bulkprotein concentration (Cbulk) and the fraction of unboundbiotin (1 − θ) and the protein desorption rate (koff) are first-order, with respect to the fraction of protein-bound biotin (θ).The rate of change in the fraction of protein-bound biotin is thedifference between the adsorption and desorption rates, givenby the following:

θ = − θ − θt

k C kdd

(1 )on bulk off (10)

Here, θ is the ratio of (ISHG)1/2 and (ISHG

max )1/2. The Langmuirkinetic model used here assumes Cbulk is unchanged with time.This assumption is valid for this study, as the proteinconcentration was replenished during the adsorption process.Since the SHG intensity (and, subsequently, θ) was

measured with the increasing bulk protein concentrationthroughout the course of a single experiment, the change inθ as function of time at each Cbulk can be plotted separately bysetting the time t at which a new Cbulk was introduced into thesystem to zero. The corresponding θ0 at t = 0 at each Cbulk isthe maximum θ obtained at the previous bulk proteinconcentration. θ for avidin binding to a 4 mol % biotinylatedDOPC bilayer versus time in the concentration range from 9.25nM to 537.6 nM is plotted in Figure 3.Equation 10 was solved with the boundary conditions such

that, at time t = 0, the initial surface coverage fraction is θ0 and,at time t > 0, the surface coverage fraction is θ. These boundaryconditions allow the kinetics of protein binding at each Cbulk tobe evaluated. The solution for eq 10 with the above statedboundary conditions is as follows:

θ =+

− −

+ + θ − +

k Ck C k

k C

k t k c k t

[1 exp(

) ] exp( )

on bulk

on bulk offon bulk

off 0 on bulk off (11)

where θ0 is zero for the first protein concentration. When thenext protein concentration is added, θ0 is the maximum fractionof 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 ±

1.8) × 10−5 s−1. From the measured kon and koff, the affinity ofbiotin−avidin binding can be calculated by Ka = kon/koff = (16± 10) × 107 M−1. This value is consistent with the Kapp ofavidin ((18 ± 14) × 107 M−1) obtained from steady-stateequilibrium thermodynamic measurements presented in theprevious section. The agreement between the binding affinityobtained from the kinetics and thermodynamics indicates thatthere was sufficient time for each protein concentration toreach equilibrium in the thermodynamic measurement.As previously mentioned, there is a large discrepancy

between the binding affinity measured in our work and thatof Wayment and Harris.18 Our measured kon is ∼4.5 orders ofmagnitude slower than that reported by Wayment and Harris((2.1 ± 0.5) × 108 M−1 s−1).18 The much-slower adsorptionrate at a high density of biotin to form a monolayer coverage ofthe protein suggests that the binding affinity of avidin to biotinis much weaker in this work. A speculative explanation for thisbehavior is that the avidin−biotin binding might involve tworegimes: (i) at a very low bulk protein concentration, theprotein binds to biotin with a very high affinity, as observed byWayment and Harris;18 and (ii) at a higher bulk proteinconcentration, the protein binds to biotin with a lower affinity,as seen here. Since we did not observe a biphasic bindingisotherm, it is possible that the protein with the high bindingaffinity to biotin saturates at very low surface coverage. Using apicomolar range of bulk protein concentration to bind to a verysmall biotin density to form <10−6 of a protein monolayer, thehigh affinity binding regime of the protein could be detected, asseen by Wayment and Harris.18 In contrast, with a higher bulkprotein concentration (at approximately the nanomolar level)and a large surface coverage of bound protein, the high affinitysites cannot be observed, because they do not account for muchof the surface density of bound protein. This explains themeasured lower affinity of the protein binding to biotin in thisstudy, because the more moderate affinity sites are moreprevalent. In addition, the koff value in this study is ∼1.5 ordersof magnitude slower than the value measured by Wayment andHarris ((3.8 ± 0.5) × 10−4 s−1),18 illustrating the formation ofthe longer-lived doubly biotin−neutrAvidin complex in thiswork. The slower adsorption and desorption rates of avidinbinding obtained here result in a smaller binding affinity relativeto the high binding affinity seen in Wayment and Harris’work.18

Figure 3. Fraction of surface coverage (θ) versus time of avidinbinding to a 4 mol % biotinylated DOPC bilayer at the following bulkavidin 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 eq11.

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208206

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

Langmuir model from eq 5. The K0 of antibiotin antibody tothe 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 obtainedby Jung et al. for fluorescently labeled antibiotin antibodybinding to 5 mol % biotin-cap-PE incorporated into a lipidbilayer.22

A direct comparison between the binding affinities measuredin this study for antibiotin antibody, avidin, streptavidin, andneutrAvidin can be made, as they were all determined undersimilar experimental conditions. The Kapp values of avidin,streptavidin, and neutrAvidin were used in the comparison, as ittakes into account protein−protein interactions involved in thebinding of these proteins. Since no protein−protein inter-actions were observed for antibiotin antibody, Kapp is equal toK0. Interestingly, we found that the binding affinity of antibiotinantibody to a biotinylated lipid bilayer has the same order ofmagnitude (∼108 M−1) as the Kapp values of avidin,streptavidin, and neutrAvidin. As a result, the free energy ofantibiotin antibody binding to biotin (46 ± 1.0 kJ/mol) is alsovery close to those of avidin, streptavidin, and neutrAvidin (seeTable 1).Streptavidin and antibiotin antibody have also been found to

have the same affinity to biotin when biotin is attached tobovine serum albumin via a linker53 even though it is well-known that antibiotin antibody binds to free biotin in solutionwith a much lower affinity than avidin.53,54 This could beexplained by the higher dependence of avidin’s binding affinityon the biotin accessibility, i.e., lower accessibility of biotin whenit is attached to a macromolecule and higher biotin accessibilitywhen it is free in solution. Avidin’s binding sites are located in adepression near the end of the β-barrels,17 which makes it moredifficult to bind with biotin when biotin is bound to amacromolecule, therefore reducing its binding affinity, relativeto that of free biotin.53,55 Unlike avidin, antibiotin antibody’sbinding sites are located on the end of the Fab segments not ina depression,56 which makes the binding affinity of antibiotinantibody less dependent on the biotin accessibility.53

Furthermore, the Fab segments of the antibody are connected

by a hinge that allows the Fab segments to move from ∼0° to180°.57 The flexibility of the Fab segments, including the armrotation, elbow bend, and arm wagging,58 allows the antibodyto adjust the spacing and orientation of its binding sites andthus facilitates the ligand−antibody binding.56,57 Therefore, thebiotin-binding of antibiotin antibody is not significantly affectedby the accessibility of biotin, whereas the lower biotinaccessibility in the biotinylated lipid bilayer causes a reductionin the binding affinity of avidin, streptavidin, and neutrAvidin toa biotinylated lipid bilayer relative to free biotin. This results inthe same order of magnitude of the binding affinity for all fourproteins to the biotinylated lipid bilayer obtained here.Another interesting feature observed is the nonspecific

adsorption of antibiotin antibody to a pure DOPC bilayer isnegligible (Figure 4). Note that no IgG from rabbit serum wasused to suppress the nonspecific adsorption of antibiotinantibody in this experiment. Therefore, this feature could makethe antibiotin antibody−biotin complex a prominent tether/linker in bioanalytical applications, as the time and cost inminimizing the protein’s nonspecific adsorption can be avoided.

■ SUMMARYWe have investigated the binding of avidin, streptavidin,neutrAvidin, and antibiotin antibody to a biotinylated lipidbilayer using SHG. The binding affinities of avidin, strepatvidin,and neutrAvidin to 4 mol % biotin-cap-DOPE incorporatedinto a DOPC lipid were determined. A positive cooperativebinding behavior was obtained for these three proteins,demonstrating that protein−protein interactions enhance thebiotin−protein binding. Moreover, the binding of streptavidinand neutrAvidin to biotin is more energetically favorable thanthat of avidin, because of the stronger protein−proteininteractions. Although neutrAvidin is designed to lower thenonspecific adsorption, we found that it exhibits the greatestdegree of nonspecific adsorption to a DOPC bilayer relative toavidin and streptavidin. A similar study of antibiotin antibodyrevealed that it binds to biotin with a similar affinity as avidin,streptavidin, and neutrAvidin. Furthermore, antibiotin antibodyshowed negligible nonspecific adsorption to a DOPC bilayerwithout the use of an additional agent to reduce the protein’snonspecific adsorption. This study presented a deeper under-standing about the binding properties of avidin, streptavidin,neutrAvidin, and antibiotin antibody to biotin at lipid bilayersurfaces, which can further assist in selecting an appropriatetether/linker for biosensing and other bioanalytical applica-tions.

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

■ AUTHOR INFORMATIONCorresponding Author*E-mail: conboy@chem.utah.edu.

■ ACKNOWLEDGMENTSThe authors acknowledge the financial support from theNational Institutes of Health (NIH) (No. R01-GM068120).Any opinions, findings, conclusions, or recommendations

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

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208207

expressed in this material are those of the authors and do notnecessarily 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. C2011, 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.; Traeger-Synodinos, 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: BocaRaton, 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.

(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 Matter1987, 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, Functionand Informatics; John Wiley & Sons: Hoboken, NJ, 2007.(57) Hanson, D. C.; Yguerabide, J.; Schumaker, V. N. Biochemistry1981, 20, 6842.(58) Burton, D. R. Mol. Immunol. 1985, 22, 161.

Analytical Chemistry Article

dx.doi.org/10.1021/ac202375n | Anal. Chem. 2012, 84, 201−208208