Force Spectroscopy Shows Dynamic Binding of Influenza Hemagglutinin and Neuraminidase to Sialic Acid

Abstract

The influenza A virus infects target cells through multivalent interactions of its major spike proteins, hemagglutinin (HA) and neuraminidase (NA), with the cellular receptor sialic acid (SA). HA is known to mediate the attachment of the virion to the cell, whereas NA enables the release of newly formed virions by cleaving SA from the cell. Because both proteins target the same receptor but have antagonistic functions, virus infection depends on a properly tuned balance of the kinetics of HA and NA activities for viral entry to and release from the host cell. Here, dynamic single-molecule force spectroscopy, based on scanning force microscopy, was employed to determine these bond-specific kinetics, characterized by the off rate k_off, rupture length x_β and on rate k_on, as well as the related free-energy barrier ΔG and the dissociation constant K_D. Measurements were conducted using surface-immobilized HA and NA of the influenza A virus strain A/California/04/2009 and a novel, to our knowledge, synthetic SA-displaying receptor for functionalization of the force probe. Single-molecule force spectroscopy at force loading rates between 100 and 50,000 pN/s revealed most probable rupture forces of the protein-SA bond in the range of 10–100 pN. Using an extension of the widely applied Bell-Evans formalism by Friddle, De Yoreo, and co-workers, it is shown that HA features a smaller x_β, a larger k_off and a smaller ΔG than NA. Measurements of the binding probability at increasing contact time between the scanning force microscopy force probe and the surface allow an estimation of K_D, which is found to be three times as large for HA than for NA. This suggests a stronger interaction for NA-SA than for HA-SA. The biological implications in regard to virus binding to the host cell and the release of new virions from the host cell are discussed.

Introduction

The influenza A virus (IAV) is a highly contagious pathogen that causes seasonal epidemics and irregular pandemics, remaining a significant burden to mankind (1). To successfully replicate, the virus relies on—among other things—two glycoproteins at its surface, the homotrimer hemagglutinin (HA) and the homotetramer neuraminidase (NA) (2, 3). In the first step of infection, the virion binds to the epithelial cells of the respiratory tract through multivalent binding of a multitude of HA trimers to sialic acid (SA) residues of the host cell’s glycocalyx (4, 5, 6, 7, 8). The monovalent interaction between individual HA monomers and SA residues is mediated essentially by hydrogen bonds. The dissociation constant is in the lower millimolar range (3–5 mM) (5, 9). NA is well known to cleave SA from the host cells’ glycoproteins, enabling the release of newly formed virions (10). Indeed, it has been shown recently that the cleavage activity is absolutely required for IAV dissociation (11). It is also indispensable for cleaving SA from the mucus of the host lung (12, 13), increasing the mobility of the virion and, thereby, chances to overcome the defensive mucus barrier to access target sites for infection. NA also helps to separate aggregates of newly formed virions by cleavage of SA residues on viral glycoproteins (10). The role of NA in the attachment of IAV to the host cell is still under debate (11).

Because HA and NA share the same receptor but have opposing functions, it is obvious that a precisely tuned balance between the binding affinity of HA and the enzymatic activity of NA is required for efficient viral replication, spread, and transmission. It has been shown for several strains that reduced HA activity is compensated by a reduction in NA activity (14, 15). The strategy of tuning the balance between the HA and NA activity is of relevance for host adaptation and for increasing drug resistance (16, 17). We hypothesized that the respective kinetics of formation and dissociation of the bonds are a critical factor in achieving this balance.

Quantitative characterization of the weak noncovalent interactions governing the single protein-SA binding and an accurate determination of their binding affinities is a challenging task that was fulfilled, e.g., by Sauter et al. using NMR for equilibrium binding (9). Single-molecule force spectroscopy (SMFS) based on scanning force microscopy (SFM) has proven to be a powerful technique to quantify the bond strength between very weak ligand-receptor pairs because of its capability to measure forces down to the pN range under near physiological conditions (18, 19, 20, 21, 22). By measuring the forces required to rupture individual molecular bonds, SMFS provides mechanistic insight into dissociation kinetics, even outside the equilibrium (23). Previously, such rupture forces have been measured at slow pulling speeds for IAV bound to cells (19). The rupture forces, which were attributed to single HA-SA bonds, were in the range of 10–25 pN. Molecular dynamics simulations of HA-SA interactions also indicated that at very high pulling speeds, the dependence of the rupture force on the rate of force loading deviates from the Bell-Evans model (19, 23, 24, 25), which is commonly used to determine kinetic parameters. Thus, to provide convincing evidence for this deviation, the experimental range of pulling speeds has to be extended. Furthermore, previous experimental studies on intact IAV have assumed that the measured rupture forces reflect solely interaction between HA and SA (19). However, although no evidence for a second related binding site, perhaps to NA, was found (19) and the ratio of HA monomers/NA monomers is 5:1 (2), it cannot be excluded that interaction between NA and SA has also contributed, in particular if the rupture forces of NA-SA bonds are on the same order as that of HA-SA. Thus, to precisely characterize the interactions between the viral glycoproteins and SA, separate measurements on HA and NA need to be done. This, however, is difficult to achieve with IAV, if possible at all, because a blocking of SA binding sites is only possible for NA by using inhibitors for its catalytic cleft but not yet possible for HA.

In this work, a monovalent receptor was synthesized to test the tensile bond strength between individual SA units and the two major surface proteins HA and NA of IAV A/California/04/2009 (H1N1), using SMFS. The receptor was created by linking SA to one end of a 5 kDa polyethylene glycol (PEG) chain, and lipoic acid (LA) to the other end. The resulting molecule was termed SAPEGLA. To this end, nitrilotriactetic acid (NTA)-coated substrates in combination with the complexation of Ni²⁺ ions were used to immobilize histidine-tagged recombinant HA or NA in an upright orientation, as it is done for the creation of protein arrays, biosensors, and other applications (26). This immobilization procedure has also been established for SMFS (27).

In this study, SMFS was used to explore the dynamic spectrum of rupture forces between SA and both surface proteins HA and NA. The measurements allowed the extraction of intrinsic bond parameters that characterize the SA-HA and SA-NA bonds. These parameters are k_off, which represents the thermal rate of dissociation in the absence of an external force, x_β, which describes the distance from the minimum of the energy landscape to the energy barrier and sets the thermal force scale f_β = k_BT/x_β (28) and ΔG, which is the height of the energy barrier. k_off can be seen as the inverse thermal stability of the bond, which is also expressed by the average bond lifetime τ_off (k_off = τ_off⁻¹). x_β can be taken as a measure of malleability (21). Previously, it was suggested that the common approach to analyze force spectroscopy data, proposed by Bell and Evans, may not provide a sufficient description of the data (19). Indeed, we demonstrate for both viral proteins that their interaction with SA cannot appropriately be described by the Bell-Evans model. We show that the model recently introduced by Friddle, Noy, and De Yoreo (29, 30) matches the full force spectra well by considering rebinding of the protein-receptor complexes at small loading rates and provides mechanistic insights on individual bond strength of IAV HA and NA under equilibrium and nonequilibrium dissociation conditions. We found that SA prominently binds to NA with unbinding forces, which were similar to HA at small rates of force loading but lower at higher rates. This result provides strong indications that even if the enzymatic activity of NA is taken into account, NA could contribute to binding of IAV to the host cell membrane.

Materials and Methods

Synthesis and characterization

The SA-displaying receptor SAPEGLA was synthesized by linking SA to one end of a PEG chain (molar weight: 5 kDa, contour length: ∼40 nm) and LA to the other end. To test the specificity of the interaction, the control molecule HOPEGLA, displaying a hydroxyl end group instead of SA, was prepared. Synthetic details are shown in the Supporting Materials and Methods, Section 1. Chemical structures of SAPEGLA and HOPEGLA are shown in the Results (Fig. 2). All reagents and solvents were purchased from commercial suppliers and used without further purification. Reactions requiring dry or oxygen-free conditions were carried out under argon in Schlenk glassware. ¹H spectra were recorded on Bruker AMX 500 (500 MHz; Bruker, Billerica, MA) and Delta Jeol Eclipse 700 (700 MHz; Jeol, Tokyo, Japan) spectrometers at 25°C and calibrated by using the deuterated solvent peak. Infrared spectra were recorded with a Nicolet AVATAR 320 FT-IR 5 SXC (Thermo Fisher Scientific, Waltham, MA) spectrometer with a deuterated triglycine sulfate detector from 4000 to 650 cm⁻¹. Naturally occurring SAs constitute a family of more than 50 structurally distinct nine-carbon 3-deoxy-ulosonic acids. We used the most widespread derivative, 5-N-acetyl-neuraminic acid, which is also termed SA. Heterobifunctional PEG (MW = 5 kDa) was purchased from Rapp Polymere (Tübingen, Germany).

Figure 2.

Structures used as SA-displaying receptor and control. A schematic presentation of (a) the control and (b) the monovalent SA receptor is shown. To see this figure in color, go online.

Cantilever functionalization

The SA receptor SAPEGLA or the control molecule HOPEGLA was covalently bound to the SFM tip by thiol groups present on the LA of the receptor and the gold coating of the SFM tip. The gold-thiol bond is known to be stable enough for SMFS and frequently used for tip functionalization (31). For functionalization, tips were exposed to ultraviolet (UV) radiation (wavelength is 185 nm) for 60 min. The ozone created by UV radiation does not only clean the tips from organic contaminations (31) but also oxidizes the gold surface, enhancing the bond strength of immobilized ligands (32). Immediately after UV radiation, the cantilevers were immersed in a 1 mg/mL solution of SAPEGLA (or HOPEGLA) and incubated for 12–24 h. Before measurements, the cantilevers were gently dipped into MilliQ water (MilliporeSigma, Darmstadt, Germany) to remove unbound molecules.

Protein immobilization

The following reagent was obtained through BEI Resources, National Institute of Allergy and Infectious Diseases, National Institutes of Health (Bethesda, MD): H1 HA Protein with C-Terminal Histidine Tag (histag) from H1N1 pdm09, Recombinant from Baculovirus, NR-15749 (https://www.beiresources.org/Catalog/BEIProteins/NR-15749.aspx). The HA sequence contains also a trimerization domain, favoring that the HA monomers assemble to trimers and resemble the native structure of the HA ectodomain found on the influenza virus envelope.

Purified and lyophilized NA from H1N1 pdm09 was purchased from Sino Biological (Beijing, China) and, according to the manufacturer’s instructions, diluted in MilliQ water (MilliporeSigma) to a final concentration of 0.25 mg/mL.

Immobilization of proteins was achieved by histag-metal chelation (26). As a chelating adsorbent, NTA, covalently bound to glass slides was purchased from Nanocs (New York, NY). As a complexation ion, Ni²⁺ was chosen.

Even though the histag-Ni²⁺-NTA bond is substantially weaker than a covalent bond, studies using SMFS have proven that it can sustain forces up to ∼500 pN (29, 33, 34), depending on the applied rate of force loading. Because the expected rupture forces of the bond between SA with HA or NA are well below that number (19), the strength of the immobilization is sufficient.

Fig. 1 shows a schematic representation of the SMFS setup. For immobilization of the proteins, the NTA glass slides were cut into squares of 1 cm² using a diamond stylus, rinsed with MilliQ water (MilliporeSigma), and loaded with Ni²⁺ ions through incubation with a 10 mM NiCl₂ solution in MilliQ water for 2 h. After loading, the slides were rinsed with MilliQ water, dried, and incubated for 1 h by drop-casting 2.5 μL protein solution onto the activated surface, resulting in a circular area with diameter around 2 mm covered with solution. To remove loosely and unspecific bound proteins, slides were rinsed with phosphate-buffered saline (PBS; 10 mM phosphate buffer, 137 mM NaCl, 2.7 mM KCl (pH 7.4) at 25°C; Sigma-Aldrich, Burlington, MA). Neutral pH was sustained throughout all further steps to prevent protein denaturation or conformational changes, in particular of HA. To suppress the unspecific binding, the slides were incubated with a 6 mg/mL solution of bovine serum albumin (BSA) in PBS for 30 min after incubation with his-tagged proteins and before any SMFS experiments. Afterwards, slides were rinsed with PBS to remove unbound BSA.

Figure 1.

Experimental setup and technique. (a) Setup for SMFS experiments: the proteins were immobilized through histag onto a Ni²⁺-activated NTA-functionalized glass surface. The SFM tip, functionalized with receptor molecules, is brought in contact with the immobilized proteins so that SA can bind to HA or NA. Upon tip retraction, a tensile force mediated by the PEG spacer is applied to separate the bond and to induce bond rupture. (b) Schematic representation of the essential steps within a force-separation cycle and typical data to be obtained: during approach (step 1), no force is detected by the cantilever. Contact (step 2) between tip and sample is determined by an increase in force. Contact can be maintained at a constant force to evaluate the time dependence of bond formation. Retraction of the SFM tip (step 3) results in relaxation of the cantilever, and in the case of successful receptor-ligand bond formation, a negative increase of the force is measured. Elongation of the flexible PEG spacer provides a characteristic signal in the force-separation curve that shows a nonlinear increase and a final stepwise drop of the force indicating bond rupture.

SFM imaging

SFM images were taken with a Multimode 8 SFM by Bruker, operated in force spectroscopy mode (named PeakForce mode by Bruker) (35, 36). With minimal lateral forces and sample damage caused by the tip, the mode is ideal for protein imaging (36). For low-resolution imaging, silicon tips on silicon nitride cantilevers of the type SNL-10 with nominal spring constant of 0.12 N/m and tip radius of 2 nm were used, whereas high resolution was achieved using cantilevers of the type PEAKFORCE-HIRS with spring constant 0.1 N/m and nominal tip radius of 1 nm. The maximal PeakForce set point during imaging was (100–250) pN, 512 points were recorded per line, and scan rates were 0.7 Hz. Postprocessing of images, such as linewise flattening and extraction of particle heights, was done using NanoScope Analysis software.

SMFS

SMFS was performed on a Nanowizard III SFM and a ForceRobot 300 system with motorized precision stage from JPK Instruments AG (Berlin, Germany), using cantilevers of the type OBL-10 (Bruker Nano, Tucson, AZ) with nominal spring constants of 6 and 30 pN/nm. Before measurements, the cantilevers were calibrated using the contact-free thermal noise method (37) so that sensitivity and spring constant were obtained without risking damage or contamination of the tip surface coating. To calculate kinetic parameters using the Friddle-De Yoreo model (29), the effective spring constant k_c was determined from the dependence of the most probable loading rate r on the pulling speed v through k_c = r/v. The effective spring constant describes the stiffness of the force transducer system more accurately because it takes the elasticity of the PEG spacer into account. Fig. 1 b shows a schematic representation of a typical force-separation cycle. The tips were approached toward the substrate at constant speed of 0.5 μm/s until a set point force of 100 pN was reached. To investigate the dependence of intrinsic bond parameters on the rate of applied force loading, the retraction speed of the SFM tip was varied from 0.1 to 10 μm/s. In most experiments, to minimize the formation and rupture of multiple protein-receptor bonds during one force-separation cycle, the cantilever was retracted immediately after reaching the set point. Additionally, in several experiments, contact between tip and sample was maintained for up to 10 s at constant force to study the probability of bond formation as a function of contact time. Successful bond formation was determined when the retraction curve showed at least one characteristic unfolding signal within the maximal possible PEG-spacer extension of 40 nm. The binding probability p_b was set as the number of curves that showed at least one specific rupture event divided by the total number of force-separation curves in one experiment. The specificity of the interaction was confirmed through several control experiments, as described in the Supporting Materials and Methods, Section 3.

For all pulling speeds and contact times, force-separation curves were measured on several positions of the slide given by several grids of 10 × 10 points over an area of (10 × 10) μm² and on different slides with different cantilevers to avoid influence of sample preparation. Data were processed with the software JPK Data Processing and OriginPro 2017 from OriginLab Corporation (Northampton, MA). The force-separation curves were baseline corrected, and the point of contact between tip and surface was determined. Absolute height of the z-piezo was converted to actual tip-sample separation between tip and surface. The nonlinear retraction traces showing sawtooth-shaped peaks were fitted with the freely jointed chain model for small forces (<50 pN) and with the extended freely jointed chain model for larger forces (38). To obtain more accurate and specific fitting, the parameter Kuhn length was fixed at l_K = 680 PM, in agreement with results as reported by Liese et al. (39). The fit determined not only the rupture force but also the rate of critical force loading from the slope of the retraction curve at bond rupture.

Results

Synthesis of the monovalent receptor SAPEGLA and control molecule HOPEGLA for SMFS

In the following, the synthesis of the monovalent SA receptor SAPEGLA and the control receptor HOPEGLA is described. For SMFS studies, heterobifunctional N₃PEGNH₂ (5 kDa) and HOPEGNH₂ (5 kDa) were used as a spacer for the synthesis of monovalent SA receptor SAPEGLA and the control molecule HOPEGLA, respectively. LA-N-hydroxysuccinimide ester was conjugated to the terminal of PEG spacers for linking the molecule with the gold coated cantilever. Heterobifunctional N₃PEGNH₂ (5 kDa) was coupled with SA via Cu²⁺-assisted click chemistry on one end, and the LA via N-hydroxysuccinimide ester coupling on the other end for the synthesis of monovalent SA receptor SAPEGLA (Fig. 2, a and b). Synthetic details are available in the Supporting Materials and Methods, Section 1.

SFM imaging

NTA glass loaded with Ni²⁺ was imaged in PBS before and after incubation with HA or NA. SFM images of the Ni-NTA surface (Fig. S7) revealed a smooth and homogeneous surface with a root mean-square roughness of 0.77 nm and a maximal peak-to-valley roughness of 8.6 nm. Large particles found on the NTA glass surface may be attributed to contaminations and inhomogeneity of the surface coating before incubation. Most of these contaminations could be removed with MilliQ water (MilliporeSigma). More aggressive cleaning reagents were not used to avoid damage to the surface coating. Cross-sectional profiles (Fig. S7) together with additional surface parameters, such as root mean-square roughness (R_q), average roughness (R_a), and peak-to-valley roughness (R_max), allowed us to compare the Ni-NTA surface before and after attachment of HA or NA. After attachment of influenza proteins, the surface topography usually showed a drastic change in surface roughness and a dense coverage with proteins.

SFM imaging in PeakForce mode, taken after incubation of a Ni²⁺-NTA surfaces with HA, was conducted to characterize the protein distribution on the surface (Fig. 3). Individual influenza proteins can be clearly identified with a regular distribution across the surface with a surface density of ∼10³ proteins/μm². Protein heights, taken from 50 proteins, exhibit a narrow distribution (see Fig. 3 b) with a mean height of (12.8 ± 1.9) nm. Here and for all following fits to histograms, the error accounts for SD. These height measurements are consistent with an expected two-dimensional protein array of HA proteins with a reported height of ∼13.5 nm (6, 40, 41). The mean width of the proteins exhibited a broad distribution with a value of (30.9 ± 4.0) nm. Although height measurements with SFM are accurate within the subnanometer range, measured widths are directly affected by the dimensions of the tip apex. Despite the fact that the sharpest tips commercially available were used and a simple geometrical approximation for the “true” width was made, calculated values are still larger than expected. Using the tip radius r, tip opening angle α, height of the protein h_p, and the measured width of the protein w_{p_measured}, the true width w can be calculated.

$$w=w_{p\_measured}-2\left\{\left(h_p-r\right)\tan \alpha +r\right\}$$ (1)

Figure 3.

SFM imaging and identification of HA on Ni²⁺-NTA. Ni-NTA surface after incubation with HA imaged using PeakForce mode in PBS is shown. (a) SFM image of a glass substrate coated with HA is shown. Exemplarily, some HA are highlighted with colored circles. Scale bar represents 200 nm. (b) Histograms for measured height and width from cross-sectional profiles of individual HAs are shown. The errors refer to standard deviation. (c) Typical cross-sectional profiles of three HA enclosed in circles in Fig. 3a are shown.

With this equation, a reduced width of ∼18–20 nm was obtained using 1 nm as the tip radius and 20° as the average tip angle. This discrepancy from an expected value of ∼4 nm for a trimeric HA head (41) can be explained by underestimation of the tip angle, due to an additional tilt between the SFM tip and the sample plane. Another source of overestimation of the protein width is the protein waggling during scanning because of protein flexibility. Cross-sectional profiles for three (enclosed in circles) individual proteins are shown in Fig. 3 c.

In the same manner, NA bound to the Ni²⁺-NTA surface was identified (Fig. 4 a). The NA showed a point-like appearance with a much smaller but monodispersed lateral size as the HA. In Fig. 4 b, height and width distributions determined from 51 particles are provided. A height of (7.4 ± 1.5) nm and a width of (21.4 ± 6.7) nm are found for NA. The height correlates well with a height of 6 nm, as has been reported by Harris et al. (2). Using the calculation of the tip influence, described earlier for HA, a reduced width of ∼15 nm is obtained. This reduced width is still too large, even if tetramers with a width of 8 nm are considered (42). This additional broadening may be caused by the same effects as discussed for HA. A zoomed region for imaged NA is shown in Fig. 4 c, with arrows indicating four individual NAs with their profiles shown in Fig. 4 d. SFM results confirmed the efficacy of the Ni²⁺-NTA method for protein immobilization with full control of their orientation on the substrate because single HA and NA units standing upright could be identified.

Figure 4.

SFM imaging and identification of NA on Ni²⁺-NTA. Ni-NTA surface after incubation with NA imaged with PeakForce mode in PBS buffer is shown. (a) NAs are distributed as small monodisperse particles on the background. Scale bar represents 200 nm. (b) Histograms for height and width for individual NAs are shown. The errors refer to standard deviation. (c) Zoomed region from (a) is shown. Scale bar represents 100 nm. (d) Cross-sectional profile of four NAs indicated by arrows in (c) are shown.

Force spectroscopy

First, binding of SAPEGLA-decorated SFM tips to an NTA glass surface loaded with Ni²⁺ ions was measured. This initial control on the Ni²⁺ loaded surface showed a strong interaction between the SA-coated SFM tip and the Ni²⁺-NTA substrate, which was attributed to the interaction of the positively charged NTA glass with the slightly negative SA. The force of the interaction was (46 ± 13) pN at a pulling speed of 500 nm/s. These undesired ionic binding interactions, which could interfere with measuring interaction of SA with viral proteins, could be almost entirely blocked by incubation with a BSA solution (6 mg/mL), causing a drop in the probability of recording a binding event from ∼3 to 1%. Therefore, to minimize these nonspecific interactions, after incubation with the HA or NA solutions, functionalized surfaces were additionally incubated with a BSA solution to block remaining protein-free surface areas. On such surfaces, covered with either HA or NA, SAPEGLA showed a binding probability of ∼4%, regardless of whether the surface was coated with HA or NA. The small binding probability ensured that the detected binding effects were almost entirely single-bond rupture events (43). Rupture events were collected from up to several hundred force-separation curves for every pulling speed and contact time.

Most rupture events occurred at a tip-sample separation smaller or equal to the length of the PEG tether (40 nm). The cumulative probability of the tip-sample separation at bond rupture is provided in the Supporting Materials and Methods for HA-SAPEGLA and NA-SAPEGLA (Fig. S13). Occasionally, rupture events were observed at larger tip-sample separation. These events were removed from further analysis.

The rupture forces and critical loading rates of the remaining events were binned into histograms, and the most probable rupture forces and related critical loading rates were obtained by fitting the resulting histograms with bimodal Gaussians (44).

As the pulling speed was increased, the rupture forces for the protein-SAPEGLA complex increased on average as expected from the Bell-Evans model (25). This behavior is illustrated in Fig. 5.

Figure 5.

Characteristic force-separation curves and force distributions. (a) Characteristic force-separation curves for the rupture of the bond of SAPEGLA with HA or NA at three pulling speeds are shown. At low and medium pulling speeds, bonds rupture at similar forces. At high speeds, the HA-SAPEGLA bond sustains higher forces than the NA-SAPEGLA bond. (b) Rupture force histograms for several pulling speeds are shown. The probability measures the relative frequency of rupture forces. Only at high speeds did the most probable rupture force differ between HA and NA.

Fig. 5 a shows three typical force-separation curves obtained for HA-SAPEGLA and NA-SAPEGLA bonds. Only retraction traces are presented for the sake of clarity. Fig. 5 b shows the distribution of the rupture forces collected for each corresponding retraction speed. The histograms of rupture forces typically showed a broad but still distinct peak with a shoulder at higher forces. The most probable peak force was attributed to the rupture of a single protein-SAPEGLA bond. The shoulder was ascribed to the background caused by the remaining undesired but inevitable interactions of SAPEGLA with the surface and simultaneous unbinding of multiple receptors. Indeed, in some cases, it was possible to observe a second peak of the rupture force distribution at about twice the value of the first distribution, suggesting the simultaneous rupture of two bonds (45). The histograms not only showed that the most probable rupture force increases with increasing retraction speed, but also that at low and medium retraction speed, SAPEGLA binds to NA with similar strength as it binds to HA.

This raises the question of whether the forces originate from actual protein-receptor interaction or whether the instrumental noise is hiding the true bond strength (46). By calculating the velocity-dependent noise limit, we can assure that the instrumental noise does not have significant effect on our rupture force histograms. Details are shown in Supporting Materials and Methods, Section 6.

When the forces were applied at a high pulling speed, the NA-SAPEGLA bond ruptured at ∼2/3 the force of the HA-SAPEGLA bond. Additional force histograms at lower and higher loading rates are shown in Fig. S9.

The full spectra of rupture forces are depicted in Fig. 6, in which the most probable rupture force was plotted against the most probable loading rate at bond rupture.

Figure 6.

Dynamic force spectra of HA and NA interacting with SAPEGLA. Dependence of most probable rupture forces on the most probable rate of force loading is shown. (a) Force spectra are fitted with the Bell-Evans model. The solid blue line represents a single barrier of the NA-SAPEGLA bond. Solid red lines represent two individual barriers of the HA-SAPEGLA bond. The dotted red line represents a single barrier of HA-SAPEGLA bond that fails to describe the data. (b) Force spectra are fitted with the Friddle-De Yoreo model that matches both spectra. Error bars account for SD of the Gauss fits to the corresponding histograms.

A first inspection revealed similar force spectra for both proteins at low loading rates, but a clear deviation occurred at higher loading rates above 10,000 pN/s, at which the HA-SAPEGLA bond sustained higher forces. In a first attempt to extract the characteristic kinetic parameters k_off, x_β, and ΔG, the spectra of HA-SAPEGLA and NA-SAPEGLA were both fitted with the Bell-Evans model, assuming a single barrier dominating the protein-receptor interaction. Following this approach, k_off and x_β were obtained using the following equation:

$$\text{F}\left(\text{r}\right)=\frac{{\text{k}}_{\text{B}}\text{T}}{{\text{x}}_{\mathit{\beta }}}\text{ln}\left(\frac{{\text{rx}}_{\mathit{\beta }}}{{\text{k}}_{\text{off}}{\text{k}}_{\text{B}}\text{T}}\right),$$ (2)

and ΔG was obtained using

$$\mathit{\Delta }\text{G}=-{\text{k}}_{\text{B}}\text{T}\text{ln}\frac{{\text{k}}_{\text{off}}\text{h}}{{\text{k}}_{\text{B}}\text{T}}$$ (3)

with the Planck’s constant h (25).

As shown in Fig. 6 a, the single-barrier Bell-Evans fit matched the spectrum of NA-SAPEGLA reasonably well, giving k_off = (1.3 ± 0.7) s⁻¹, x_β = (657 ± 68) pm, and ΔG = (29 ± 1) k_BT. Errors account for standard errors. However, this simple approach failed to describe the force spectrum of the HA-SAPEGLA bond because of the increasingly high forces required to rupture the bond at high loading rates. Such inconsistencies in the Bell-Evans framework have been discussed in literature frequently. An excessive overview with selected examples can be found from Friddle et al. (29). Most commonly, this sudden increase of the rupture forces is explained by the coexistence of multiple barriers in the free-energy landscape of the bond. Kinetic parameters for each individual barrier are then obtained by separating the dynamic force spectrum into regimes of increasing loading rates and fitting each regime separately (28, 47). Assuming a two-barrier scenario, the force spectrum of HA-SAPEGLA is split into a slow regime with loading rates <3000 pN/s and a fast regime with loading rates >3000 pN/s. Fitting each regime separately yielded k_off = (0.7 ± 1.4) s⁻¹, x_β = (1.1 ± 0.5) × 10³ pm, and ΔG = (30 ± 2) k_BT for the slow and k_off = (31 ± 5) s⁻¹, x_β = (206 ± 25) pm, and ΔG = (26.1 ± 0.2) k_BT for the fast regime, respectively. The possibility of negative k_off and the large x_β in the regime of small loading rates are indications that even the two-barrier Bell-Evans approach is not suited well to describe the full force spectra. For NA, this two-barrier approach is shown in Fig. S12 and Supporting Materials and Methods, Section 4.4.

Recently, Friddle, Noy, and De Yoreo have developed a model that accounts for the nonlinear behavior frequently obtained in force spectroscopy experiments (29). They also successfully tested the applicability of their model on a variety of previously reported experimental data. In their model, the nonlinearity is caused by rebinding. In contrast, the Bell-Evans model considers only irreversible unbinding. Even though this is a fair assumption for high loading rates, rebinding becomes increasingly more important at small loading rates. Friddle and colleagues outlined that the full range of individual linear regimes can be described by considering such rebinding (29, 30). Hereby, the force spectrum is divided into an equilibrium regime, in which rebinding is permitted, and a kinetic regime, in which rebinding is negligible because of the high pulling speed. k_off and x_β obtained tend to match k_off and x_β determined with the Bell-Evans model in the fast pulling regime but deviate in the slow pulling regime (48). In this sense, the Bell-Evans model describes a limited section of the full force spectrum. A comparative study (49) showed the universal applicability of the Friddle-De Yoreo model. To prove its applicability to the force spectra presented in Fig. 6, the force-dependent transition rates for binding, k_on(f), and unbinding, k_off(f), were calculated. Details are shown in Supporting Materials and Methods, Section 4.2. It can be seen in Fig. S10: $k_{off}\left(f\right)\ll k_{on}\left(f\right)$ for f < 30pN and ${\text{k}}_{\text{off}}\left(\text{f}\right)\gg {\text{k}}_{\text{on}}\left(\text{f}\right)$ for f > 40 pN. Therefore, the transition from the equilibrium to the kinetic regime is between 30 and 40 pN. As shown in Fig. 6, the rupture forces of HA-SAPEGLA range from ∼17 to 88 pN, and the rupture forces for NA-SAPEGLA range from 22 to 55 pN. The forces required to transit from the equilibrium to the kinetic regime are well within these ranges. Therefore, it can be concluded that the forced dissociation of the HA-SAPEGLA and NA-SAPEGLA bonds transits from the equilibrium to the kinetic regime and that they can be adequately described with the Friddle-De Yoreo model to extract the relevant kinetic parameters.

The force spectra are fitted with the following equation:

$$\text{F}\left(\text{r}\right)\cong {\text{f}}_{\text{eq}}+{\text{f}}_{\mathit{\beta }}\text{ln}\left\{1+\frac{{\text{re}}^{\mathit{\gamma }}}{{\text{k}}_{\text{off}}\left({\text{f}}_{\text{eq}}\right){\text{f}}_{\mathit{\beta }}}\right\}.$$ (4)

Hereby, f_eq represents the equilibrium force, i.e., the force at which k_off(f) exceeds k_on(f), and the dissociation of the protein-receptor complex transits from the equilibrium to the kinetic regime. f_eq depends on the height of the energy barrier, ${\text{f}}_{\text{eq}}\sim \sqrt{\mathit{\Delta }\text{G}}.$ f_β represents the thermal force scale, which is given by the barrier width, f_β ∼ x_β⁻¹. γ = 0.577 is Euler’s constant.

Fig. 6 b shows that the Friddle-De Yoreo model matches both force spectra over the whole range of force loading and yields k_off = (18 ± 8) s⁻¹, x_β = (345 ± 87) pm, and ΔG = (32 ± 8) k_BT for the NA-SAPEGLA bond and k_off = (66 ± 13) s⁻¹, x_β = (106 ± 26) pm, and ΔG = (12 ± 2) k_BT for the HA-SAPEGLA bond, respectively. A summary of all extracted parameters is shown in Table 1. The values for k_off, x_β, and ΔG are close to the ones found for the interaction of the influenza virus with cells (19).

Table 1.

Kinetic Parameters Extracted from Fitting the Dynamic Force Spectra with the Friddle-De Yoreo Model and with the Bell-Evans Model, Using for the Latter One Energy Barrier to Describe the NA-SAPEGLA Bond and Two Distinct Energy Barriers to Describe the HA-SAPEGLA Bond

Protein and Fit-Model	feq [pN]	fβ [pN]	koff(feq) [s−1]	koff [s−1]	xβ [pm]	τoff [s]	ΔG [kBT]
HA Bell-Evans slow pulling	N/Aa	N/A	N/A	0.7 ± 1.4	(1.1 ± 0.5) × 103	1.4	30 ± 2
HA Bell-Evans fast pulling	N/A	N/A	N/A	31 ± 5	206 ± 25	0.03	26.1 ± 0.2
NA Bell-Evans	N/A	N/A	N/A	1.3 ± 0.7	657 ± 68	0.8	29 ± 1
NA Friddle-De Yoreo	22.3 ± 1.6	11.9 ± 3.0	112 ± 52	18 ± 8	345 ± 87	0.06	32 ± 8
HA Friddle-De Yoreo	15.1 ± 0.8	38 ± 9	97 ± 19	66 ± 13	106 ± 26	0.02	12 ± 2

f_eq is the equilibrium force in which the dissociation of the protein-receptor complex transitions from the equilibrium to the kinetic regime. f_β is the thermal force scale given by the malleability of the bond x_β. k_off(f) describes the force-induced dissociation rate. k_off is the dissociation rate in the absence of an external force. τ_off is the bond lifetime, τ_off = k_off⁻¹. ΔG is the height of the energy barrier between the bound and the unbound state of the protein-receptor complex.

^aN/A, not applicable.

The values determined for ΔG, ΔG(NA) = (32 ± 8) k_BT and ΔG(HA) = (12 ± 2) k_BT, are also within the range of those determined by quantitative microarray binding assays (50), in which ΔG is found to be ΔG ≈ (10–20) k_BT, and computational studies (51), in which ΔG is found to be ΔG ≈ (15–200) k_BT.

To gain deeper insight into the protein-receptor complex, the on rate k_on was investigated. Because it is expected that the probability of bond formation increases with increasing duration of the contact between SFM tip and surface, the half-maximal binding probability ${\text{p}}_{{\text{b}}_{\text{max}}}/2$ allows calculation of the on rate through

$${\text{k}}_{\text{on}}=\frac{{\text{N}}_{\text{A}}{\text{V}}_{\text{eff }}}{n_b\mathit{\tau }}$$ (5)

(52). Here, N_A is the Avogadro’s number, V_eff is the effective volume of the interacting binding partners, n_b is the number of binding partners involved and τ is the interaction time. The interaction time τ, defined as the time needed to reach ${\text{p}}_{{\text{b}}_{\text{max}}}/2$, is obtained from an exponential fit to the experimental data as shown in Fig. 7.

Figure 7.

Dependence of the probability of bond formation on contact time. The binding probability (p_b) increases with contact time. The increase rate is similar for NA and HA. Using an exponential fit function, the interaction time τ is calculated. The values represent mean values obtained from independently determining p_b in at least five samples. Error bars show SDs. To see this figure in color, go online.

The number of binding partners n_b within the effective volume V_eff is set to n_b = 1 because most curves show only a single binding event. V_eff is calculated assuming an effective radius r_eff = 4 nm, given by the gyration radius of the 5 kDa PEG spacer. With these parameters, the association constants were calculated for HA-SAPEGLA and NA-SAPEGLA and, using k_off obtained with the Friddle-De Yoreo model, allowed the determination of dissociation constants.

$${\text{K}}_{\text{D}}=\frac{{\text{k}}_{\text{off}}}{{\text{k}}_{\text{on}}}$$ (6)

For NA-SAPEGLA, K_D = (290 ± 130) mM, whereas for HA-SAPEGLA, K_D = (950 ± 190) mM were determined. These high K_D values (>200 mM) indicate very weak binding and so far have been reported only for the binding of β-methylsialic acid to HA (9), and they will be discussed in the next section.

Discussion

In this study, the forces required to rupture the molecular bonds between SA and HA as well as NA were measured for rates of force loading between 100 and 50,000 pN/s. The most probable rupture forces were between 10 and 100 pN. The use of a thioglycosidic bond between SA and PEG ensures that the bond dissociation could be measured without interference of the enzymatic activity. Although the rupture forces for HA and NA were similar up to loading rates of ∼10,000 pN/s, the NA-SA bond ruptured at lower forces for loading rates above 10,000 pN/s. The similarity of the rupture forces for the HA and NA binding, especially in the slow regime of force loading, supported earlier studies (14, 15, 53) claiming that the activity of the proteins is precisely balanced but also that the true binding potential might be obscured, as predicted by the Friddle-De Yoreo model (29, 30), if the force spectra are not fully known. We also will show that NA can contribute to the initial cell attachment of the virions. The rather short timescale for spontaneous dissociation in comparison to the enzymatic activity proved that typically multiple NA-SA rebinding events are necessary for SA cleavage. This might be fostered by the arrangement of NA in patches on the virus surface (2).

As outlined in the Results, the Bell-Evans model does not provide an appropriate and comprehensive description of our results. The integration of the Bell-Evans formalism into the Friddle-De Yoreo model has been described previously (30, 48). Considering x_β as a measure of malleability, the large values (>1 nm) found in the regime of slow force loading indicate that receptors need to be separated from the protein by large distances (z axis) to prohibit rebinding. The assumption of rebinding in the slow regime is also affirmed by the small values for k_off related to very long average bond lifetimes.

The K_D values, calculated for the dissociation of NA-SAPEGLA and HA-SAPEGLA, might seem too large by three to four orders of magnitude, but the strong influence of the effective radius for the volume of receptor-ligand interaction, K_D ∼ r_eff⁻³, has to be taken into account. In this study, r_eff was chosen to be the gyration radius of the PEG linker (∼4 nm). Even though this is a reasonable assumption, it is worthwhile to consider bond formation while the linker is partially stretched and r_eff tends to larger values. E.g., from r_eff = 20 nm, which is about half the contour length of the linker, it follows that K_D(HA) = 7.5 mM and K_D(NA) = 2.4 mM. This is close to values reported in previous studies (9, 54, 55). Details can be found in Supporting Materials and Methods, Section 4.3. Independently of r_eff, it holds that K_D(NA) ≈ 1/3K_D(HA). Therefore, the values for K_D suggest a stronger binding of SA to NA in comparison to HA. A second cause for the overestimation of K_D might be the number n_b of involved protein-SA bonds because K_D depends linearly on n_b. Even though care was taken to use only single-bond rupture events for the analysis, multibinding could not be excluded completely. Also, it has been shown that flexible PEG linkage can cause an increase in K_D (56).

A comparison of k_off, x_β, ΔG, and K_D determined by the Friddle-De Yoreo model shows that at least for H1N1, the binding of SA to NA is more stable than that of HA (Video S1). Thus, NA could also potentially contribute to binding to the host cell. NA’s involvement in binding to the target cell and the associated increased likelihood of infection was already considered (57). However, it should be noted that the parameters measured in our study for NA may be somewhat different in its native environment, i.e., on the virus surface. Even if the relation of the parameters between HA and NA determined in our model system should remain at the virus surface, the binding of the virus to the surface of the host cell will mainly be realized by HA because the ratio of HA/NA is ∼5:1 (2).

Video S1. Energy Landscapes of Protein-Receptor Bonds Derived from Single Molecule Force Spectroscopy

The video provides a brief introduction to dynamic single molecule force spectroscopy and the determination of kinetic parameters of molecular bonds. Rupture forces, characteristic for the bonds between sialic acid and the influenza A virus proteins HA and NA, are shown and the energy landscapes of the protein-receptor bonds are obtained. The energy landscapes indicate the higher stability of the NA-SA bond.

To assess the stability of the binding of SA to NA, however, the enzymatic activity of NA k_cat must also be taken into account. The ratio of the dissociation constants k_off and k_cat gives a measure of the probability that the bound substrate, i.e., SA, is split or dissociated before it is split (57) The enzymatic activity of NA played no role in the measurements presented here because the PEG-linker structure is not accessible to NA cleavage (58). However, comparing the results presented here on binding kinetics with the well-characterized enzymatic activity of NA from biochemical studies (57, 59), deeper insight into the function of NA can be gained. Xu et al. (59) measured an enzymatic activity of k_cat = 3.17/s (τ_cat = 0.3 s) and K_m = 373 μM using the artificial fluorescent SA substrate 2′-(4methylumbelliferyl)-D-N-acetylneuraminic acid. For H1N1, k_on was determined to 0.17 μM⁻¹ s⁻¹ (60). With k_off = K_m k_on − k_cat (57), it follows that k_off = 60 s⁻¹ or τ_off = 0.02 s with τ_off = k_off ⁻¹. This value corresponds very well to the value of τ_off = 0.06 s determined by SMFS. In this way, SMFS contributes to further understanding the NA function by adding direct measurements of binding kinetics and dynamic binding strength under nonequilibrium conditions.

A comparison of k_cat and k_off now shows that the probability of dissociation of SA is ∼6 (k_off = 18 s⁻¹, τ_off = 0.06 s⁻¹) to 20 (k_off = 60 s⁻¹, τ_off = 0.02 s⁻¹) times greater than for enzymatic cleavage. This was also reported for NA of other virus strains, e.g., for A/Anhui/1/2013 (57). In this case, the probability of dissociation of an SA host cell receptor typical for human viruses was ∼1000 times higher than that for cleavage of SA. It should be noted that both the binding affinity to HA and the enzymatic activity of NA are largely determined by both the structure of the SA substrate and the structure of the SA binding pockets of HA and NA, respectively (59). Interestingly, several studies have provided evidence that coordinated levels of the receptor binding activity of HA and the receptor cleavage activity of NA are important for efficient virus replication (59). I.e., a reduction in HA affinity also leads to a reduction in enzymatic NA activity. To what extent this also applies to the stability of the bond (k_off) remains to be investigated, e.g., by using SMFS.

Conclusions

Recombinant proteins HA and NA were immobilized on NTA-coated glass surfaces via a histag complexed with Ni²⁺ ions. Visualization of densely packed surfaces with oriented proteins with the active headgroup pointing upwards was checked with SFM in force spectroscopy mode. Measured sizes for height and width of single proteins were in line with previously reported data, confirming the reliability of the method. The binding interaction between the SA receptor and both major spike virus proteins was probed by SMFS in an independent manner, revealing that the profiles of the energy landscape for both bonds display a similar behavior when tested at slow pulling speeds, whereas a clear difference occurred when probed at faster pulling. Within this higher regime of loading rates >10,000 pN/s, the HA-SAPEGLA bond sustained higher forces. The widely applied Bell-Evans model was not well suited to describe the force spectra, assuming a single energy barrier. Instead, a model introduced by Friddle, De Yoreo, and co-workers that considers rebinding at low loading rates was used. It was shown that the NA-SAPEGLA bond has a deeper energy sink, resulting in a slower off rate and therefore longer lifetime. Both protein-receptor complexes had a similar interaction time. The resulting K_D for NA-SAPEGLA was threefold smaller than for HA-SAPEGLA. This confirms a more stable attachment of NA to SA compared with HA to SA. This enhanced binding might have an implication during viral entry. A longer lifetime of the NA-SAPEGLA bond could be related to a functional requirement of the NA enzyme to cleave mucins and other SAs present in the mucus barrier of the upper epithelial tract. If the bond lifetime would be too short, the cleaving capacity could decrease, and the virus could remain trapped before reaching the cell surface. The short bond lifetime of the single HA-SAPEGLA bond is compensated during virus binding on the cell by the formation of multiple connections to enhance the virus adhesion strength and also its residence time to eventually become endocytosed. The results presented here from the binding interaction of two main viral proteins with their cellular receptor showed that important differences can arise when the bond tensile strength is explored for a broader range of loading rates. Quantitative data were provided to support the fact that the individual SA-HA bond is weak and therefore the virus must use multiple protein-receptor connections to stabilize its residence time on the cell surface before endocytic uptake is possible. These protein-receptor connections are not limited to HA but also include NA. Furthermore, the nonlinear increase in rupture forces with increasing pulling speeds points to a tightening or reinforcement in the virus adhesion strength, becoming even stronger when multivalent interactions are involved. This makes the virus more resistant to overcome large shear forces upon binding.

Author Contributions

This study was conceived by J.P.R., A.H., and R.H. Experiments were planned and designed by S.B., D.L., J.L.C.-C. and V.R.-S. SAPEGLA and HOPEGLA were synthesized and characterized by S.B. SFM imaging was done by J.L.C.-C. and V.R.-S. SMFS was performed by V.R.-S. All authors contributed to writing the manuscript.

Editor: Nancy Forde.

Supporting Citations

References (61, 62, 63, 64, 65) appear in the Supporting Material.