Article pubs.acs.org/ac

Quantitative Assessment of the Multivalent Protein−Carbohydrate Interactions on Silicon Jie Yang,*,† Jean-Noel̈ Chazalviel,†,‡ Aloysius Siriwardena,§ Rabah Boukherroub,∥ François Ozanam,† Sabine Szunerits,*,∥ and Anne Chantal Gouget-Laemmel*,† †

Physique de la Matière Condensée, Ecole Polytechnique-CNRS, 91128 Palaiseau, France Laboratoire de Glycochimie des Antimicrobiens et des Agroressources (LG2A), (FRE 3517-CNRS), Université de Picardie Jules Verne, 33 Rue St Leu, 80039 Amiens, France ∥ Institut de Recherche Interdisciplinaire (USR CNRS 3078), Université Lille 1, Parc de la Haute Borne, 50 Avenue de Halley, 59658 Villeneuve d’Ascq, France §

S Supporting Information *

ABSTRACT: A key challenge in the development of glycan arrays is that the sensing interface be fabricated reliably so as to ensure the sensitive and accurate analysis of the protein− carbohydrate interaction of interest, reproducibly. These goals are complicated in the case of glycan arrays as surface sugar density can influence dramatically the strength and mode of interaction of the sugar ligand at any interface with lectin partners. In this Article, we describe the preparation of carboxydecyl-terminated crystalline silicon (111) surfaces onto which are grafted either mannosyl moieties or a mixture of mannose and spacer alcohol molecules to provide “diluted” surfaces. The fabrication of the silicon surfaces was achieved efficiently through a strategy implicating a “click” coupling step. The interactions of these newly fabricated glycan interfaces with the lectin, Lens culinaris, have been characterized using quantitative infrared (IR) spectroscopy in the attenuated total geometry (ATR). The density of mannose probes and lectin targets was precisely determined for the first time by the aid of special IR calibration experiments, thus allowing for the interpretation of the distribution of mannose and its multivalent binding with lectins. These experimental findings were accounted for by numerical simulations of lectin adsorption.

S

protein interactions in a high-throughput manner.10−14 A key challenge in the fabrication of such glycan arrays is the development of reliable and reproducible chemistries for the immobilization of chemically and structurally diverse glycans onto an appropriate interface, while at the same time ensuring a selective and multivalent binding with their specific protein partners.15 However, the evaluation of the glycan presentation in terms of spacing, orientation, and density of glycan epitopes remains difficult to predict and to rationalize.9,16−18 The extent of multivalent binding of a lectin on a surface depends on the effective concentration of the corresponding glycan. In principle, this extent can be tuned by diluting the glycan derivative with nonsugar spacer molecules during the immobilization process in which the presentation of glycan probes at low surface density is anticipated to be more accessible for the binding.19 Most of the diluted glycan surfaces were achieved by the use of thiolated self-assembled monolayers (SAMs). Corn and co-workers studied the binding

pecific protein−carbohydrate interactions are important in the normal development and functioning of all living organisms. Such recognition events are also implicated in numerous disease processes, including for example, bacterial and viral infection as well as autoimmunity and cancer metastasis.1−3 These interactions occur between glycan-specific epitopes such as those typically found in complex glycoproteins, glycolipids, or proteoglycans usually present on cell surfaces and various proteins including lectins, which bind to them selectively.4,5 In most of the cases, multivalent interactions take place between partners so as to achieve high-affinity and selective recognition, as individual glycans usually display only low affinity for their partner protein in the millimolar range.6 Multivalent enhancement is most usually observed when multiple copies of a particular glycan interact with its protein counterpart itself featuring two or more glycan recognition sites.7,8 The development of appropriate tools for studying protein−glycan interactions has been keenly researched as they promise to allow for a better understanding and mimicking of such events.9 In this framework, array platforms, displaying carbohydrates on their sensing surface (glycan arrays), are currently being actively pursued so as to ensure the sensitive and accurate mapping of carbohydrate− © 2014 American Chemical Society

Received: July 16, 2014 Accepted: September 12, 2014 Published: September 12, 2014 10340

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 1. Schematics of the fabrication of the diluted mannose-terminated crystalline Si(111) surface and its specific interaction with LENS (pdb file: 1LES).

acid on SiH, a heterobifunctional amino oligo(ethylene oxide) (OEO) analog carrying a terminal azido function was linked via amide coupling, followed by an efficient “click” conjugation with propargyl mannoside moieties. This mannose-terminated surface (SiMan) was shown to interact only with specific lectin partners (such as Lens culinaris, LENS) and not with unspecific lectins (such as Peanut agglutinin, PNA). By taking advantage of this well-controlled glycan architecture, this work aims at probing quantitatively the binding of LENS on various glycan surface concentrations in order to address the requirements for optimal multivalency. As described from Figure 1, LENS possesses a bivalent structure with two mannose-binding sites located at ∼5.9 nm. The surfaces are fabricated by diluting propargyl mannoside in propargyl alcohol at the concentrations of 10 and 1 mol %. Infrared spectroscopy in the attenuated total reflection geometry (ATR-FTIR) is probably the most readily accessible technique for obtaining an absolute quantification of glycans immobilized on the surface and of proteins bound to glycans. This technique will be used in combination with contact modeAFM imaging (CM-AFM) and numerical simulations to explicitly determine the requirements for multivalent binding between mannose and LENS.

yield of different lectins on glycan mixtures immobilized on SAMs by SPR imaging.20 The amount of proteins loaded onto the surface was found to nonlinearly decrease as the dilution of surface glycans increased. However, Sato et al. diluted the glycan surface concentration using shorter and smaller spacer chains, and they found that the amount of bound proteins was at a maximum at a proper dilution and not for homogeneous glycan SAMs.21 In order to rationalize these crowding effects and their consequence on the multiple-binding capabilities of the lectins, the benefits from absolute quantitative information about the density of glycan probes and lectin targets on the surface are of prime interest. Such a quantitative access into the glycan density and distribution can be evaluated using hydrogenated crystalline (111) silicon (SiH) substrates. They offer welldefined topographies allowing for absolute quantification of molecule surface concentration.22−25 Such a capability potentially gives deeper insight into the conformational effects as compared to a mere measurement of the affinity between probes and targets,26−28 which does not provide any indication concerning the factor limiting the interactions in a specific experimental configuration. The second advantage of using silicon substrates is that they also offer robust and reproducible surface functionalization chemistries based on the formation of organic monolayers through stable Si−C covalent bonds.29−31 For the covalent attachment of glycan analogs to unoxidized silicon surfaces, there already exists a handful of examples in the literature.32−34 The direct attachment of 1-alkene functionalized glycan precursors has been achieved via a hydrosilylation reaction;32,33 alternately, azide-derivatized glycans were attached to a surface prefunctionalized with alkynyl groups via a Cu(I)-catalyzed azide−alkyne cycloaddition (CuAAC) reaction, namely, the “click” reaction.34 This “click” reaction was successfully applied for the glycan immobilization on various solid substrates.35−37 We have previously reported a “click” strategy to develop a mannose-modified crystalline silicon substrate for the study of mannose−lectin interactions.38 Figure 1 depicts our strategy schematically. Starting from the welldefined carboxydecyl-terminated monolayer (SiCOOH) constructed via the photochemical hydrosilylation of undecylenic

■

EXPERIMENTAL SECTION Materials. All cleaning reagents (H2O2, 30%; H2SO4, 96%) and etching (HF, 50%; NH4F, 40%) reagents were supplied by Carlo Erba. Undecylenic acid (99%) was purchased from Acros Organics. All other chemicals (highest grade), lectins (Lens culinaris, LENS, and Arachis hypogaea, PNA), and oligo(ethylene oxide) (OEO, NH2−C2H4−(OCH2CH2)8−N3) were purchased from Sigma-Aldrich. Ultrapure water (MilliQ, 18 MΩ cm) was used for the preparation of the solutions and for all rinses. α-Propargyl mannoside and per-O-acetyl-αpropargyl-mannoside were synthesized as described in ref 38. N-type silicon (111) wafers (525 μm thickness) were purchased from Siltronix (France). Functionalization of Silicon Surfaces. Safety Considerations. The H2SO4/H2O2 piranha solution is a strong oxidant which reacts violently with organic materials. HF is a 10341

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

IR signal. For BSA concentrations >0.95 mg/mL, the IR signal was stabilized immediately after the first measurement. Surface Characterizations. Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR). Home-made doubleside polished n-type (111) silicon prisms (FZ, 30−40 Ω cm) were used as the ATR element. They were shaped as 15 × 14 × 0.5 mm3 platelets; then, two opposite sides were 51° bevelled, providing 22 reflections. The spectra were recorded on a Bruker Equinox FTIR spectrometer coupled to a homemade, nitrogen-gas purged external ATR compartment. The spectra were collected with 100 scans in s- and p-polarization over the 900−4000 cm−1 spectral range with a 4 cm−1 resolution. The calibration and the in situ measurements were performed in a homemade PTFCE IR cell of ∼1.5 mL volume. The top and the bottom of the cell are connected by a PTFE tube (0.8 mm diameter), allowing for the addition of different solutions without breaking the spectrometer purge. On the side, there is an opening of 9 mm diameter against which the prism is pressed via a nitrile O-ring seal (leading to 8 useful reflections for the in situ measurements). Contact-Mode Atomic Force Microscopy (CM-AFM). AFM images were obtained using a Pico SPM microscope (Molecular Imaging, Phoenix, AZ) in contact mode, with silicon nitride cantilevers (Nanoprobe, spring constant = 0.12 N m−1) under a N2 atmosphere. A force of 0.38 nN was applied for imaging the glycan-modified surfaces whereas a smaller one (0.15 nN) was applied for imaging the lectin−glycan interactions. The silicon sample was cut from one-side polished n-type (111) silicon wafers (CZ, 5−10 Ω cm) with a miscut of 0.2° toward the (112̅) direction for obtaining a staircase structure. Numerical Simulation. The issue of steric hindrance among the adsorbed lectins was investigated in a numerical simulation using gfortran with a pgplot graphics library under Linux on a PC computer with quad core processor. Ligand sites were thrown at random on a square area, of size 60 × 60 nm2, with the constraint that each new site must be spaced by a minimum center-to-center distance 2R from all of the other sites (where R represents the Van de Waals radius of a mannose group taken as spherical). The lectins were simulated by ellipses (long axis 8.0 nm, short axis 4.0 nm). The anchoring sites were taken along the long axis, symmetric with respect to the ellipse center, and at a mutual distance d0 = 5.9 nm. For a given concentration of ligand sites on the surface, adsorption of lectins was investigated, by assuming that a lectin can be adsorbed on two ligand sites whenever three conditions are fulfilled: (i) the two ligand sites are located at a distance d0 ± x %, where x is a parameter of the simulation, (ii) there is a free space of width δ around the two ligands (no other ligand at a center-to-center distance lower than 2R + δ), and (iii) there is no overlap of the newly adsorbed lectin with previously adsorbed ones. In a first step, the pairs of ligand sites were explored at random and a lectin was adsorbed irreversibly as soon as a pair was found for which these three conditions are fulfilled. The process stopped when no suitable pair of ligand sites was left free. In a second step, an “anneal” was carried out by allowing a randomly selected lectin to desorb and exploring the various manners of readsorbing one or two lectins instead. When readsorption of two lectins was possible, this new set was chosen, and the process was repeated many times until no further increase in the number of adsorbed lectins was obtained. This process is thought to be realistic, because the lectins are adsorbed weakly enough and desorption followed by rearrangement, leading to an optimization of the surface

hazardous acid, which can result in serious tissue damage if burns are not appropriately treated. They must be handled with extreme care in a well-ventilated fume hood, while wearing appropriate chemical safety protection. Etching of Oxidized Silicon Substrates SiOx/Si(111). The oxidized Si(111) sample was cleaned in a piranha solution (1/3 H2SO4/H2O2) at 100 °C for 10 min and copiously rinsed with Milli-Q water. Subsequently, it was immersed in a 50% HF solution for 5 s to obtain an atomically rough hydrogenterminated SiHx surface or in an oxygen-free 40% NH4F solution containing 50 mM ammonium sulfite for 15 min to obtain an atomically flat hydrogen-terminated SiH surface. The sample was then quickly rinsed in water and blown dry under nitrogen. Preparation of the SiPEG Surface. The preparation of SiCOOH and Si-NHS ester surfaces was performed as described previously.23,39 The freshly Si-NHS ester surface was immersed in a solution of 20 mM methoxypoly(ethylene glycol) amine (MW = 750) overnight. After reaction, the sample was washed in 1× PBS/0.1% SDS for 5 min, 0.2× PBS for 2 min, 0.1× PBS for 2 min, and finally rinsed thoroughly with Milli-Q water. The surface was then dried under a stream of nitrogen. Preparation of Glycan-Derivatized Surfaces. Si-OEO-N3 Surface. The Si-NHS ester surface was reacted with 10 mM NH2−C2H4−(OCH2CH2)8−N3 in 1× PBS at pH ∼ 8 for more than 3 h at room temperature. The resulting surface was copiously rinsed with 1× PBS, followed by a surfactinated rinse (1× PBS/0.1% SDS for 15 min, 0.2× PBS for 5 min, 0.1× PBS for 5 min) and finally with Milli-Q water. The Si-OEO-N3 surface was dried under a stream of nitrogen. Clicking of Alkynyl-Terminated Mannose. The Si-OEO-N3 surface was immersed in an outgassed solution of 3 mM αpropargyl mannoside in 1× PBS containing 5 mol % CuSO4· 5H2O, 20 mol % sodium ascorbate. After 5 h, the sample was washed twice with EDTA solution (0.1 M), 1× PBS/0.1% SDS for 10 min, 0.2× PBS for 2 min, 0.1× PBS for 2 min, and finally rinsed thoroughly with Milli-Q water. The clicked surface SiMan was then dried under a stream of nitrogen. To click 10 and 1 mol % mixtures of α-propargyl mannoside in propargyl alcohol, the concentration of total alkynyl functionalities was kept at 3 mM in 1× PBS. To click propargyl acetate and per-Oacetyl-α-propargyl-mannoside at different dilutions in propargyl alcohol, the concentration of total alkynyl functionalities was kept at 3 mM in DMSO/1× PBS (v/v = 1/3). Interaction with Lectins. The SiMan surface was placed in direct contact with a solution of PNA (1 mg/mL) or LENS (0.1, 0.3, 0.6, 1, and 1.5 mg/mL) in 1× PBS and left for 1 h in a hybridization chamber. The cover slide was removed in 1× PBS, and the surface was washed in 1× PBS for 2 min and then in 1× PBS/0.1% SDS for 10 min, 0.2× PBS for 2 min, 0.1× PBS for 2 min, and finally with Milli-Q water. The lectinmodified surface was dried under a stream of nitrogen prior to analysis. ATR-FTIR Calibration of per-O-Acetyl-α-propargylmannoside, Methylactamide, and BSA. Increasing concentrations of BSA solution (0.325, 0.5, 0.725, 0.95, 1.3, 1.475, 1.775, 2.025, 2.4, 3.1, 3.6, 3.875, 4.375, 5.2, 5.52, 5.9, 7.075, 8.1, 8.85, and 10.15 mg/mL in water) were injected into the in situ IR cell, using either an oxidized SiOx/Si(111) silicon prism or a PEG750-terminated silicon prism as the IR window. For the BSA concentrations up to 0.95 mg/mL, continuous measurements were performed for 20 min until the stabilization of the 10342

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 2. (A) ATR-FTIR spectra in p-polarization of SiCOOH (a), Si-NHS ester (b), Si-OEO-N3 (c), and SiMan surfaces (d), followed by the interaction with PNA (1 mg/mL, gray spectrum, e) and LENS (1 mg/mL) (f). The reference spectra are those of the SiH surface alone. (B) CMAFM images of the corresponding IR spectra, the profiles being exhibited in the Supporting Information S1.

which appears as an increased band around 1050 cm−1 (Figure 2A-d). The determination of the corresponding surface coverage is described in the section Determination of Surface Glycan Density. In addition, the progress of each modification step was monitored by CM-AFM (Figure 2B). Monohydrogen-terminated Si(111) surfaces exhibit a well-defined staircase structure with atomically flat terraces separated by monoatomic steps when they are chemically prepared by etching in ammonium fluoride.40 After hydrosilylation (Figure 2B-a), activation (Figure 2B-b), and aminolysis reactions (Figure 2B-c), the staircase structure is preserved confirming the formation of a uniform organic layer without physisorbed molecules. After clicking the mannoside groups (Figure 2B-d), the flatness is altered due to the presence of 0.5 nm depth variations appearing as black holes, suggesting the formation of glycan domains. The profiles of the corresponding CM-AFM images are shown in the Supporting Information S1. Capture of specific and nonspecific lectins by the SiMan surface was investigated using both ATR-FTIR and AFM (Figure 2). After interaction with the nonspecific PNA, there is no modification in the ATR-FTIR spectrum (Figure 2A-e) and in the CM-AFM image (Figure 2B-e). Once again, the staircase structure is preserved, indicating the absence of nonspecific adsorption of PNA due to the antifouling property of the OEO segments incorporated in the monolayer. In the case of the mannose−LENS interaction, an increase by a factor of 1.9 of the amide I band absorbance is observed. In parallel, the CMAFM image shows the presence of a substantial quantity of lectin at the surface (Figure 2B-f). This indicates that the SiMan surface is able to selectively recognize mannose-specific lectins. The average height of LENS is estimated to be ∼1 nm, which is much lower than the theoretical dimension of LENS (3 nm). This may suggest the deformation or the spreading of the LENS structure during our drying and imaging conditions (contact mode).41,42 Determination of Surface Glycan Density. The principle of IR quantification is to calculate the number of molecules on

coverage being possible. The number of adsorbed lectins, divided by the surface area of the square, readily gave the surface concentration.

■

RESULTS AND DISCUSSION Glycan-Terminated Surfaces and Their Interactions with Lectins. The chemical functionalization of the SiH surface was first studied by ATR-IR as shown in Figure 2A. Direct photochemical hydrosilylation of undecylenic acid on a SiH surface results in the incorporation of carboxydecyl chains for which the carbonyl stretching mode νCO of the acid and the symmetric and antisymmetric methylene stretching modes νCH2 are detected at 1715, 2853, and 2921 cm−1, respectively (Figure 2A-a). Activation of the acid end groups with NHS/EDC leads to the formation of succinimidyl esters with a characteristic triplet band in the ATR-FTIR spectrum at 1744, 1788, and 1816 cm−1 assigned to the νCO of the ester-NHS functions (Figure 2A-b). The next step is the aminolysis of the ester-NHS functions with an azide precursor containing eight ethylene oxide units (OEO). The success of the reaction is confirmed by the appearance of the band at 2105 cm−1 due to the stretching mode of the azide function together with the bands at 1545 and 1642 cm−1 related to the amide II and I bands (Figure 2A-c). The IR contributions of OEO units appear as broad bands at ∼1080−1150 cm−1 assigned to the νC−O−C of the OEO chains and around 2820−2960 cm−1 assigned to the νCH2 from both the alkyl chains and OEO segments. From a quantitative ATR-FTIR analysis of the carbonyl bands in p- and s-polarization, the surface concentration of grafted chains is found to be N = 2.4 × 1014 acid chains per cm2, corresponding to a coverage of 30% (referred to the number of surface Si atoms) and an activation and aminolysis yield of 93% and 71%, respectively, corresponding to N = 1.7 × 1014 cm−2 for azide termination.21,30 The “click” immobilization of propargyl mannoside onto the SiOEO-N3 surface was confirmed by the complete disappearance of the νN3 band in the ATR-FTIR spectrum and the presence of mannosyl units related to the νC−C and νC−O of the ring 10343

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 3. (A) ATR-FTIR spectra in p-polarization of diluted SiManOAc (black) and SiOAc surfaces (gray). The reference spectra are the Si-OEON3 surface. (B) The molar fraction of “clicked” ManOAc on the surface as a function of that in solution. The data are fitted according to the equation obtained in ref 23, where the α value characterizes the kinetic ratio of ManOAc over propargyl alcohol.

Figure 4. (A) CM-AFM images and (B) ATR-FTIR spectra in p-polarization of SiMan surfaces (black) at 100 mol % (a), 10 mol % (b), and 1 mol % (c) and after their interactions with LENS (1 mg/mL, gray). The profiles of the AFM images are shown in Supporting Information S2. The fitting of the amide bands is shown in red for SiMan surfaces and in blue for SiMan-LENS surfaces. The reference spectra are the SiHx surfaces. The AFM images were created by several scans of the AFM tip on the locked center square. (C) Histogram of integrated absorbance of amide I band from LENS for the three surfaces.

the surface from the absorbance of their characteristic vibrational bands. The cross-section of these bands has to be extracted first from a calibration by measuring the same molecule in liquid phase with known concentration.23 For the “clicked” SiMan surface, the ATR-FTIR spectrum does not contain a band suitable for the quantitative determination of

probe density. Therefore, we proposed an indirect quantification method by “clicking” a structurally similar molecule, the per-acetylated propargyl mannose (ManOAc), offering a characteristic peak at ∼1750 cm−1 for the acetyl νCO band.38 We earlier proved that the surface concentration on a “clicked” 100 mol % ManOAc surface is equal to that on a 100 10344

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 5. (A) Isotherm of the interaction of 100 mol %, 10 mol %, and 1 mol % SiMan surfaces with LENS. The data were fitted by using the Langmuir model (solid lines) or FFG model (dashed lines). (B) Scatchard plots for the three surfaces where 100 and 1 mol % SiMan surfaces are linearly fitted (inset).

mol % “clicked” SiMan surface obtained by a colorimetric experiment, which was found at 1.2 × 1014 cm−2, corresponding to a click yield of 75%. When the mannosyl probes were diluted with shorter propargyl alcohol at different ratios (10 and 1 mol %), the colorimetric quantification is not sensitive enough and becomes impracticable. Therefore, we chose to quantify the diluted SiManOAc surfaces by the IR method (Figure 3A). We obtain a density of 5.2 × 1013 and 7.2 × 1012 cm−2 for 10 and 1 mol % SiManOAc surfaces, respectively. To determine the “click” yield of propargyl alcohol, in parallel, we clicked propargyl acetate (cf. Figure 3A) and we found a density of 1.5 × 1014 cm−2. Starting from a number of 1.7 × 1014 azide per cm2, the yield of the “clicked” acetylated SiOAc surface is almost quantitative (∼90%) due to the lowered steric hindrance as compared to 100% mannose (yield of 75%). For the “clicked” 10 and 1 mol % ManOAc-terminated surfaces, Figure 3B represents the percentage of “clicked” ManOAc as a function of molar fraction of ManOAc in solution. The nonlinear variation indicates that the diluted ManOAc monolayers are richer in glycan than the mixture solution, which is attributed to faster kinetics for glycan coupling than for propargyl alcohol. These enhanced kinetics could be due, e.g., to favored surface adsorption, as seen previously in other mixed monolayers on silicon.23 Influence of the Glycan Density on the Binding Efficiency. The binding efficiency of the diluted SiMan surfaces with LENS was then evaluated by AFM as represented in Figure 4A. The AFM images were captured after wiping the tip for several times at a smaller locked region, so as to reveal the underlying silicon surface. With regard to the newly formed layer of adsorbed protein, the 10 mol % SiMan surface is much more efficient to load the lectins than the 1 and 100 mol % SiMan surfaces. For the 10 mol % SiMan surface, the LENS covers completely and uniformly (at least at the scale of the tip resolution) the silicon surface with a molecular “carpet”. In all cases, the average height of LENS is again around 1 nm. The difference was quantitatively assessed by ATR-FTIR (Figure 4B). By fitting the amide peaks of the three SiMan surfaces and of those exposed to LENS, the sole contribution from LENS is obtained by subtraction. The loading of LENS is found to be

enhanced by a factor of 3.5 on the 10 mol % SiMan surface as compared to 100 and 1 mol %. The observation that the optimal binding activity appears at a properly diluted glycan surface is in accordance with the earlier reports by Sato et al.21 and Svedhem et al.43 and different from that of Corn and coworkers20 who observed the maximal binding on a 100% mannose surface. Furthermore, the isotherm of the mannose-LENS binding was studied by ATR-FTIR for the three surfaces. Figure 5A represents the integrated absorbance of the amide I band as a function of LENS concentration. The curves were fitted with the classic Langmuir adsorption model (solid lines).44 The Ka values obtained for the three surfaces are around 105 M−1 as listed in Table 1. Alternatively, Scatchard plots were used to Table 1. Association Constants of Mannose-LENS Binding Measured by Different Isotherm Models KA 100 mol % SiMan 10 mol % SiMan 1 mol % SiMan

Langmuir (105 M−1)

Scatchard plot (105 M−1)

FFG (105 M−1)

3.04

3.16

2.6

5.93 1.23

nonlinear 1.29

4.42 1.66

test the plausibility of the Langmuir isotherm, as shown in Figure 5B. The 100 and 1 mol % SiMan surfaces lead indeed to linear plots with Ka values close to those directly obtained from the fitting in Figure 5A, whereas the 10 mol % SiMan surface curve exhibits a nonlinear shape. This effect may be a consequence of the interaction between proteins since the 10 mol % SiMan surface is fully covered with LENS as shown in Figure 4A-b. The fitting of the 10 mol % surface curve can be largely improved by using the Frumkin-Fowler-Guggenheim (FFG) isotherm model, which takes into account lateral interactions among adsorbate molecules, indicated by the β value (a negative β implies attractive forces between adsorbate molecules whereas a positive β implies repulsive forces).20,45 The best results are obtained for β = −0.1 (dashed lines in Figure 5A). From the FFG model, the Ka values of the three 10345

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 6. (A) IR calibration spectra in s-polarization of BSA at various concentrations on the SiOx/Si (a) and SiPEG (b) surfaces, the reference spectrum being water. Each spectrum was taken after the absorption reached stabilization. The inset in (a) is the fitting result of amide I and II bands at C = 10.15 mM. (B) The integrated absorbance of the amide I band from the two surfaces as a function of concentration is fitted for the range after surface saturation.

Table 2. Summary of the Measured Concentrations of Mannose and LENS on the Surface from the Analysis of the Spectra in Figure 4Ba 100% SiMan 10% SiMan 1% SiMan a

Cmannose [cm−2]

CLENS [cm−2]

mannose−LENS ratio

occupied area per mannose [nm2]

1.2 × 10 5.2 × 1013 7.2 × 1012

7.7 × 10 2.7 × 1012 7.8 × 1011

156:1 19:1 9:1

0.83 (0.91 × 0.91 nm) 1.96 (1.4 × 1.4 nm) 14 (3.74 × 3.74 nm)

14

11

Each mannose site is considered as a square grid.

signal within 20 min. Figure 6B shows the amide absorbance due to BSA as a function of its concentration on the two different surfaces. In both cases, two stages are observed. A first steep increase of absorbance at low concentrations is due to the adsorption of BSA (presaturation). After saturation, the amount of BSA adsorbed at the surface is constant and the increase in the IR amide band is solely related to the increased BSA concentrations in solution. The saturation on the SiPEG surface was observed at a smaller BSA concentration than on the SiOx surface and the amount of adsorbed BSA was lower, revealing the capability of the PEG monolayer to partly resist the BSA adsorption. The slope of the linear increase after saturation is the calibration value which is used to deduce the quantification equation (see Supporting Information S3). This increase in amide I band absorbance amounts to 0.0051 (cm mg/mL)−1 on the SiOx/Si surface and 0.0055 (cm mg/mL)−1 on the SiPEG surface. The close values of the calibration on the two surfaces confirm the reliability of the experiments since the absorbance of BSA in solution is independent of the surface type. The calibration result for BSA can be applied for other proteins like LENS. Considering that the protein is the assembly of a certain number of amino acids (AA) through the peptide bond, the average MW of AA of BSA molecule (MW = 133 kDa, 1166 AA) is 114 g/mol, whereas that of LENS (MW = 52 kDa, 466 AA) is 112 g/mol. Thus, the conversion from BSA to LENS is simply a factor of 114/112 = 1.02, a negligible influence for the quantification of LENS. On the basis of these estimates, we use the value 0.0051 (cm mg/mL)−1 as the calibrated amide IR absorption in solution, with an estimated uncertainty of ±10%. The deduced quantification equation is

surfaces are still of the same order of magnitude with minor modifications (Table 1). Ka values on the order of 105 M−1 suggest that the binding is relatively strong due to multivalency for all of the surfaces instead of weak monovalency. Absolute Quantification of LENS on the Surface. The absolute quantification of lectin bound on the surface is obtained from its characteristic vibrational amide I mode, arising mainly from the νCO with minor contributions from other nitrogen-related vibrations. The amide I band is most commonly used for secondary structure analysis since its position is essentially determined by the backbone conformation.46,47 The major difficulty of this calibration is the concomitant uncontrolled adsorption of proteins on the surface.48,49 Two different surface terminations of the IR substrate were chosen for the calibration: an oxidized silicon surface SiOx/Si and a PEG750-terminated silicon surface (SiPEG, see Supporting Information S3) known to limit the nonspecific adsorption of proteins.22 In both cases, however, we observed the uncontrolled adsorption process of the lectin by the continuous increase of the amide I and II bands as a function of time on a few hours time scale (see Supporting Information S4). Therefore, the absence of steady-state IR response prevents one from using these experiments to obtain the IR cross-section of LENS. To circumvent this problem, bovine serum albumin (BSA) has been used for the calibration since it is often used as blocking agent to limit the nonspecific adsorption of proteins and turns out to readily reach a saturation concentration at the surface.50−52 In contrast with the LENS case, BSA exhibits a faster adsorption which yields a steady-state value of the IR 10346

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

Figure 7. (A) Schematic model for the binding of LENS onto the “clicked” SiMan surfaces. (B) Numerical simulation results: (a−c) typical images corresponding to the three ligand concentrations used in our experiments. The ligand sites are represented by the small red dots and the adsorbed lectins by the green ellipses; distance of suitable mannose pairs d0 = 5.9 nm ± x%. Spacing exclusion is defined as 2R + δ. (d) Resulting lectin concentration Cads as a function of ligand concentration Clig, obtained by averaging the results of 100 simulations. Two varying parameters, x% (2.5% and 5%) and δ (0.14 and 0.19 nm), are envisaged.

therefore, N∥= 3.53 × 1013 As, N⊥= 7.69 × 1013 Ap − 6.99 × 1013 As, and Ntotal = 7.69 × 1013 Ap − 3.46 × 1013 As, where As and Ap are the integrated absorbance of the amide I band of LENS in s- and p-polarization and Ntotal is the surface density of LENS (see Supporting Information S3). Multivalent Binding of Mannose-LENS. The determined densities of mannose and LENS on the surface are listed in Table 2 which allows one to obtain the probe−target ratio and the distribution of mannose. The three surfaces exhibit a surface−concentration ratio of mannose to LENS much larger than that needed for the bivalency binding (2:1). Given the lateral dimensions of LENS (8 × 4 nm), the maximal surface concentration of LENS is typically 3.1 × 1012 cm−2. This suggests that the concentration of 2.7 × 1012 cm−2 found at the 10 mol % SiMan surface is close to saturation, as suggested indeed by AFM (Figure 4A-b). These absolute quantitative data now allow for one to discuss how the bivalent binding proceeds at various surfaces. Two factors are addressed: the distance between two mannose binding sites (5.9 nm) and the size of the carbohydrate recognition domain (1.4 nm) (Figure 1). By considering the size of mannose (0.7 nm) and its occupied area on the surface, such a spacing is available around the mannose molecules at the 10 and 1 mol % SiMan surfaces but not at the nondiluted SiMan surface where the mannose is distributed densely and the click yield of 75% corresponds to the limit of steric hindrance (Figure 7A). Moreover, by considering the mannose pair distance, the largest distance between adjacent mannoses is on the 1 mol % SiMan surface (3.7 nm). This means that, in spite of the statistical fluctuations of positions and the flexibility of the molecular system, it is not possible to always find a

couple of mannose sites away from each other by a distance of around 5.9 nm. The constraints of spacing and pair distance are evidenced by the low coverage and nonuniformity of lectin captured at the 100 and 1 mol % SiMan surfaces (Figure 4A-a, b). Such constraints are certainly much less stringent on the 10 mol % SiMan surface where the average distance between two adjacent mannoses (1.4 nm) seems to be optimum to fulfill the two opposite requirements: large enough for providing sufficient binding space and small enough for enabling bivalent adsorption. These assumptions have been confirmed quantitatively by numerical simulations (Figure 7B). The minimum center-tocenter distance between two mannose ligands 2R was determined by choosing the value that yields the maximum surface concentration found experimentally (1.2 × 1014 cm−2). This value 2R = 0.76 nm (effective diameter of the ligand group) is in typical agreement with the known size of a mannose (0.7 nm) molecule. Bivalent adsorption was allowed whenever two mannoses were spaced by a distance of 5.9 nm ± x%, and the minimum distance between these two sites and all of the other mannoses was larger than 2R + δ. Various values of x% and δ were explored. The obtained concentration of adsorbed lectins Cads was generally found to exhibit a flat maximum when plotted as a function of the concentration of ligand sites Clig (Figure 7B-d). The curves actually exhibit three regions. At low Clig (region I), Cads is weak and increases with increasing Clig. On the other hand, when Clig approaches its maximum value (region III), Cads decreases with increasing Clig, an obvious consequence of the impossibility for a ligand site to be active when a ligand neighbor is located in its vicinity closer than 2R + δ. In the mid range (region II), Cads is at a maximum 10347

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

and hardly sensitive to Clig, an indication that there is steric hindrance between the adsorbed lectins. As it was to be expected from these considerations, regions I and III are found to be sensitive to the choice of x% (governing the distance of a suitable mannose pair), while region III is the only part sensitive to the choice of δ (governing the available binding space). On the other hand, region II is a little sensitive to x% and δ. Values x% ∼ 2.5% and δ = 0.14 nm are found to yield fair agreement with the experimental findings. These values sound reasonable in view of the geometry of the lectin and its adsorption sites. Note that the effective diameter of the coordination site of the lectin, 2(R + δ), appears somewhat smaller from the simulations (1.04 nm) than expected from the structure of LENS (1.4 nm). This may be due to the fact that the steric hindrance effect is less severe than the estimation from the measured diameter of the coordination shell, due to the tilting of its axis with respect to the surface normal.

France (IUF). A.S. wishes to thank Dr. David Lesur and Dr. Dominique Calieu of the UPJV for expert technical assistance.

■

(1) Lis, H.; Sharon, N. Chem. Rev. 1998, 98, 637−674. (2) Bertozzi, C. R.; Kiessling, L. L. Science 2001, 291, 2357−2364. (3) Lee, Y. C.; Lee, R. T. Acc. Chem. Res. 1995, 28, 321−327. (4) Ambrosi, M.; Cameron, N. R.; Davis, B. G. Org. Biomol. Chem. 2005, 3, 1593−1608. (5) Rudd, P. M.; Wormald, M. R.; Dwek, R. A. Trends. Biotechnol. 2004, 22, 524−530. (6) Jayaraman, N. Chem. Soc. Rev. 2009, 38, 3463−3483. (7) Reynolds, M.; Pérez, S. C. R. Chim. 2011, 14, 74−95. (8) Jimenez Blanco, J. L.; Ortiz Mellet, C.; Garcia Fernandez, J. M. Chem. Soc. Rev. 2013, 42, 4518−4531. (9) Kiessling, L. L.; Gestwicki, J. E.; Strong, L. E. Angew. Chem., Int. Ed. 2006, 45, 2348−2368. (10) Tong, Q.; Wang, X.; Wang, H.; Kubo, T.; Yan, M. Anal. Chem. 2012, 84, 3049−3052. (11) Zhi, Z.-l.; Powell, A. K.; Turnbull, J. E. Anal. Chem. 2006, 78, 4786−4793. (12) Wang, H.; Zhang, Y.; Yuan, X.; Chen, Y.; Yan, M. Bioconjugate Chem. 2010, 22, 26−32. (13) Mercey, E.; Sadir, R.; Maillart, E.; Roget, A.; Baleux, F.; LortatJacob, H.; Livache, T. Anal. Chem. 2008, 80, 3476−3482. (14) Chevolot, Y.; Bouillon, C.; Vidal, S.; Morvan, F.; Meyer, A.; Cloarec, J.-P.; Jochum, A.; Praly, J.-P.; Vasseur, J.-J.; Souteyrand, E. Angew. Chem., Int. Ed. 2007, 46, 2398−2402. (15) Oyelaran, O.; Gildersleeve, J. C. Curr. Opin. Chem. Biol. 2009, 13, 406−413. (16) Almant, M.; Mastouri, A.; Gallego-Yerga, L.; García Fernandez, J. M.; Ortiz Mellet, C.; Kovensky, J.; Morandat, S.; El Kirat, K.; Gouin, S. G. Chem.Eur. J. 2013, 19, 729−738. (17) Deniaud, D.; Julienne, K.; Gouin, S. G. Org. Biomol. Chem. 2011, 9, 966−979. (18) Zheng, H.; Du, X. Biochim. Biophys. Acta, Biomembr. 2011, 1808, 2128−2135. (19) Houseman, B. T.; Mrksich, M. Angew. Chem., Int. Ed. 1999, 38, 782−785. (20) Smith, E. A.; Thomas, W. D.; Kiessling, L. L.; Corn, R. M. J. Am. Chem. Soc. 2003, 125, 6140−6148. (21) Sato, Y.; Yoshioka, K.; Murakami, T.; Yoshimoto, S.; Niwa, O. Langmuir 2011, 28, 1846−1851. (22) Perez, E.; Lahlil, K.; Rougeau, C.; Moraillon, A.; Chazalviel, J.N.; Ozanam, F.; Gouget-Laemmel, A. C. Langmuir 2012, 28, 14654− 14664. (23) Faucheux, A.; Gouget-Laemmel, A. C.; Henry de Villeneuve, C.; Boukherroub, R.; Ozanam, F.; Allongue, P.; Chazalviel, J.-N. Langmuir 2006, 22, 153−162. (24) Yang, M.; Teeuwen, R. L. M.; Giesbers, M.; Baggerman, J.; Arafat, A.; de Wolf, F. A.; van Hest, J. C. M.; Zuilhof, H. Langmuir 2008, 24, 7931−7938. (25) Wallart, X.; de Villeneuve, C. H.; Allongue, P. J. Am. Chem. Soc. 2005, 127, 7871−7878. (26) Linman, M. J.; Taylor, J. D.; Yu, H.; Chen, X.; Cheng, Q. Anal. Chem. 2008, 80, 4007−4013. (27) Liang, P. H.; Wang, S. K.; Wong, C. H. J. Am. Chem. Soc. 2007, 129, 11177−11184. (28) Dhayal, M.; Ratner, D. A. Langmuir 2009, 25, 2181−2187. (29) Gooding, J. J.; Ciampi, S. Chem. Soc. Rev. 2011, 40, 2704−2718. (30) Chazalviel, J.-N.; Allongue, P.; Gouget-Laemmel, A. C.; de Villeneuve, C. H.; Moraillon, A.; Ozanam, F. Sci. Adv. Mater. 2011, 3, 332−353. (31) Thissen, P.; Seitz, O.; Chabal, Y. J. Prog. Surf. Sci. 2012, 87, 272−290. (32) de Smet, L. C. P. M.; Stork, G. A.; Hurenkamp, G. H. F.; Sun, Q.-Y.; Topal, H.; Vronen, P. J. E.; Sieval, A. B.; Wright, A.; Visser, G. M.; Zuilhof, H.; Sudhölter, E. J. R. J. Am. Chem. Soc. 2003, 125, 13916−13917.

■

CONCLUSION We have shown that crystalline Si(111) surfaces are well-suited to the linking of diluted glycans in a “click” chemistry approach where the multivalent interactions with specific lectins can be followed by quantitative ATR-FTIR and AFM. Both technics demonstrated that the binding yield with a specific lectin can be enhanced by a proper dilution of glycans on the surface. In any case, the strong binding of lectins to the surface seems to result from a bivalent interaction, as shown by the measurement of affinity constants. The surface concentration in loaded lectin has been measured by ATR-FTIR using a newly developed calibration method that provides for the first time the ratio of glycan to bound lectin. This quantitative result suggests that optimum conditions for bivalent adsorption are a compromise on the surface concentration of glycan ligands: enough free space around each glycan together with large enough density to provide suitably spaced glycans. Such a compromise is best achieved for the 10 mol % Si-Man surface. These ideas have been assessed by numerical simulations, which account for the experimental results quantitatively.

■

ASSOCIATED CONTENT

S Supporting Information *

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

■

REFERENCES

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. *E-mail: [email protected]. Present Address ‡

25, rue Eugène Combes, 19800 Corrèze, France.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS J.Y. thanks the Ecole Polytechnique for Ph.D. financial support (EDX grant). Catherine Henry de Villeneuve is thanked for fruitful help in the AFM imaging experiments. We gratefully acknowledge financial support from the Centre National de la Recherche Scientifique (CNRS), the Université Lille 1, the Région Nord Pas de Calais, and the Institut Universitaire de 10348

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349

Analytical Chemistry

Article

(33) Funato, K.; Shirahata, N.; Miura, Y. Thin Solid Films 2009, 518, 699−702. (34) Qin, G.; Santos, C.; Zhang, W.; Li, Y.; Kumar, A.; Erasquin, U. J.; Liu, K.; Muradov, P.; Trautner, B. W.; Cai, C. J. Am. Chem. Soc. 2010, 132, 16432−16441. (35) Santoyo-Gonzalez, F.; Hernandez-Mateo, F. Chem. Soc. Rev. 2009, 38, 3449−3462. (36) Debrassi, A.; Ribbera, A.; de Vos, W. M.; Wennekes, T.; Zuilhof, H. Langmuir 2014, 30, 1311−1320. (37) Barras, A.; Martin, F. A.; Bande, O.; Baumann, J. S.; Ghigo, J. M.; Boukherroub, R.; Beloin, C.; Siriwardena, A.; Szunerits, S. Nanoscale 2013, 5, 2307−2316. (38) Gouget-Laemmel, A. C.; Yang, J.; Lodhi, M. A.; Siriwardena, A.; Aureau, D.; Boukherroub, R.; Chazalviel, J.-N.; Ozanam, F.; Szunerits, S. J. Phys. Chem. C 2013, 117, 368−375. (39) Sam, S.; Touahir, L.; Salvador Andresa, J.; Allongue, P.; Chazalviel, J.-N.; Gouget-Laemmel, A. C.; Henry de Villeneuve, C.; Moraillon, A.; Ozanam, F.; Gabouze, N.; Djebbar, S. Langmuir 2009, 26, 809−814. (40) Munford, M. L.; Cortès, R.; Allongue, P. Sens. Mater. 2001, 13, 259−269. (41) Ivanov, Y.; Frantsuzov, P. A.; Zollner, A.; Medvedeva, N. V.; Archakov, A.; Reinle, W.; Bernhardt, R. Nanoscale Res. Lett. 2011, 6, 54. (42) Ngunjiri, J.; Stark, D.; Tian, T.; Briggman, K.; Garno, J. Anal. Bioanal. Chem. 2013, 405, 1985−1993. (43) Svedhem, S.; Ö hberg, L.; Borrelli, S.; Valiokas, R.; Andersson, M.; Oscarson, S.; Svensson, S. C. T.; Liedberg, B.; Konradsson, P. Langmuir 2002, 18, 2848−2858. (44) Liang, P.-H.; Wang, S.-K.; Wong, C.-H. J. Am. Chem. Soc. 2007, 129, 11177−11184. (45) Kurth, T.; Biresaw, G.; Adhvaryu, A. J. Am. Oil Chem. Soc. 2005, 82, 293−299. (46) Barth, A.; Zscherp, C. Q. Rev. Biophys. 2002, 35, 369−430. (47) Ye, M.; Zhang, Q.-L.; Li, H.; Weng, Y.-X.; Wang, W.-C.; Qiu, X.G. Biophys. J. 2007, 93, 2756−2766. (48) Ballet, T.; Boulange, L.; Brechet, Y.; Bruckert, F.; Weidenhaupt, M. Bull. Pol. Acad. Sci. 2010, 58, 303−315. (49) Buijs, J.; Norde, W.; Lichtenbelt, J. W. T. Langmuir 1996, 12, 1605−1613. (50) Park, S.; Lee, M.-R.; Pyo, S.-J.; Shin, I. J. Am. Chem. Soc. 2004, 126, 4812−4819. (51) Zhang, G.-J.; Huang, M. J.; Ang, J. A. J.; Yao, Q.; Ning, Y. Anal. Chem. 2013, 85, 4392−4397. (52) Jeyachandran, Y. L.; Mielezarski, E.; Rai, B.; Mielczarski, J. A. Langmuir 2009, 25, 11614−11620.

10349

dx.doi.org/10.1021/ac502624m | Anal. Chem. 2014, 86, 10340−10349