Single-virus assay reveals membrane determinants and mechanistic features of Sendai virus binding

Abstract

Sendai virus (SeV, formally murine respirovirus) is a membrane-enveloped, negative-sense RNA virus in the Paramyxoviridae family and is closely related to human parainfluenza viruses. SeV has long been utilized as a model paramyxovirus and has recently gained attention as a viral vector candidate for both laboratory and clinical applications. To infect host cells, SeV must first bind to sialic acid glycolipid or glycoprotein receptors on the host cell surface via its hemagglutinin-neuraminidase (HN) protein. Receptor binding induces a conformational change in HN, which allosterically triggers the viral fusion (F) protein to catalyze membrane fusion. While it is known that SeV binds to α2,3-linked sialic acid receptors, and there has been some study into the chemical requirements of those receptors, key mechanistic features of SeV binding remain unknown, in part because traditional approaches often convolve binding and fusion. Here, we develop and employ a fluorescence microscopy-based assay to observe SeV binding to supported lipid bilayers (SLBs) at the single-particle level, which easily disentangles binding from fusion. Using this assay, we investigate mechanistic questions of SeV binding. We identify chemical structural features of ganglioside receptors that influence viral binding and demonstrate that binding is cooperative with respect to receptor density. We measure the characteristic decay time of unbinding and provide evidence supporting a “rolling” mechanism of viral mobility following receptor binding. We also study the dependence of binding on target cholesterol concentration. Interestingly, we find that although SeV binding shows striking parallels in cooperative binding with a prior report of Influenza A virus, it does not demonstrate a similar sensitivity to cholesterol concentration and receptor nanocluster formation.

Significance

Paramyxoviruses are a family of membrane-enveloped viruses with many notable human and animal pathogens. In this study, we develop and use an assay to observe the initial step of infection—virus binding to the host membrane—for Sendai virus, the prototypical paramyxovirus, at the single-virus level. This assay uses cell membrane mimics—supported lipid bilayers—as targets for virus binding to enable easy control of the membrane components with which the virus interacts. Using our assay, we gain insight into basic biophysical questions about Sendai virus binding, including the chemical characteristics of the receptor, the cooperative nature of binding, the influence of cholesterol, and the mechanism of viral mobility following binding.

Introduction

Paramyxoviruses are a family of enveloped, non-segmented negative-sense RNA viruses that include the human parainfluenza viruses, measles virus, Newcastle disease virus, and various veterinary viruses and are responsible for significant human and animal disease worldwide. Sendai virus (SeV), formally murine respirovirus, has been a prototypical and well-studied model virus of the Paramyxoviridae family and is quite interesting in its own right (1,2). It primarily causes disease in rodents and other animals, and although it can infect human cells, it does not cause disease in humans or lead to oncogenic transformation (2,3). As such, it has been developed, among other uses, as a gene therapy vector (2), as a CRISPR/Cas9 delivery system (3), and as a commercial vector (ThermoFisher) to induce pluripotent stem cells (4), as well as a vaccine candidate backbone for HIV (5), RSV (6), SARS-CoV-2 (7), and others (8).

For all enveloped viruses, the first step in infection is binding of the virus to a receptor or attachment factor on the host cell membrane surface. SeV, as with other closely related paramyxoviruses, utilizes α2,3-linked sialic acid glycolipids or glycoproteins as receptors for cell entry (9, 10, 11, 12). Receptor binding is mediated by the hemagglutinin-neuraminidase (HN) receptor-binding protein (formerly called the HN attachment protein), a transmembrane protein embedded in the viral envelope. HN is commonly understood to be arranged in a dimer of dimers and is of a helical stalk domain with a globular head that contains the receptor-binding pocket (13,14). Upon receptor binding, a conformational change is induced in HN, which then allosterically triggers the viral F protein to catalyze membrane fusion with the host cell plasma membrane, initiating infection (13).

Although it is known that SeV utilizes α2,3-linked sialic acid receptors for cell entry, and there has been some study into the chemical requirements of those receptors in non-membrane systems (9,10,12,15,16), key features of SeV binding and behavior following binding remain unknown, including the cooperativity of viral binding, the influence of receptor localization in the target membrane, and viral mobility following receptor binding.

In this report, we examine these key features of SeV binding using a fluorescence microscopy-based single-virus-binding assay. Single-virus assays have been employed fruitfully for other families of viruses to study binding and membrane fusion (17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33). These assays are commonly implemented using synthetic target membranes such as supported lipid bilayers (SLBs) or tethered liposomes, which afford the researcher precise control over membrane topology, composition, and environment to ask specific biophysical questions. To our knowledge, such assays have not been employed for any paramyxovirus. Here, we present the development and validation of a single-virus-binding assay for SeV. We then utilize this assay to ask targeted questions about viral binding and viral behavior following binding. We investigate the role of sialic acid density and positioning for receptor binding, the cooperative dependence of binding on receptor density, the influence of cholesterol composition, and the mechanism of viral mobility following binding.

Materials and methods

Materials

Dioleoylphosphatidylethanolamine (DOPE), palmitoyloleoylphosphatidylcholine (POPC), and cholesterol (Chol), as well as gangliosides GD1a, GM3, GM1, and GQ1b were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Oregon green-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (OG-DHPE) and Texas red-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (TR-DHPE) were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Polydimethylsiloxane (PDMS) Sylgard 184 elastomer base and curing agent were purchased from Ellsworth Adhesives (Germantown, WI, USA). Octadecyl rhodamine B chloride (R18) dye was purchased from Biotium, Inc. (Fremont, CA, USA). Sendai virus (purified Sendai Cantell Strain, batch 960216) was obtained from Charles River Laboratories (Wilmington, MA, USA) and handled according to a BSL-2 protocol at Williams College. Chloroform, methanol, and buffer salts were obtained from Fisher Scientific (Pittsburgh, PA, USA) and Sigma-Aldrich (St. Louis, MO, USA). Goat anti-mouse IgG-Alexa 488 antibody was purchased from Abcam (Cambridge, UK).

Anti-SeV antibodies

The mouse anti-HN (1A6) IgG2a antibody and the mouse anti-HN (3G12) IgG1 antibody were ordered from Kerafast Inc. (Boston, MA, USA) and produced in the laboratory of Prof. Benhur Lee (Mt. Sinai). Details of antibody production and characterization of 3G12 have not been published previously. Information on characterization of both monoclonal antibodies is included in the supporting material (Fig. S1). The anti-F (11H12) monoclonal antibody was a gift from Prof. Benhur Lee. 1A6 and 11H12 have been characterized previously (34).

Buffer definitions

Reaction buffer (RB) consisted of 10 mM NaH₂PO₄, 90 mM sodium citrate, 150 mM NaCl, pH 7.4. HEPES buffer (HB) = 20 mM HEPES, 150 mM NaCl, pH 7.2.

Sendai virus labeling and concentration estimation

SeV (Cantell strain) was egg-grown and stringently purified by the manufacturer (Charles River Laboratories) using standard purification techniques (ultracentrifugation through sucrose gradient and tangential flow filtration). Viral particles were characterized by transmission electron microscopy (TEM; see details in supporting material). They were observed to be pleiomorphic, roughly spherical particles, with outer diameters ranging from 80 to 400 nm, mean 165 nm (Fig. S2), consistent with more limited cryo-EM data in a prior report (35). To label virus, 0.5 μL (0.5×), 1 μL (1×), 2 μL (2×), or 4 μL (4×) of R18 (1.5 g/L in ethanol) or 2 μL (0.5×), 4 μL (1×), 8 μL (2×), 12 μL (3×), or 16 μL (4×) of TR-DHPE (0.75 g/L in ethanol), was mixed with 240 μL of HB buffer; 15 μL of SeV (2 mg/mL) was combined with 60 μL of the appropriate dye buffer mixture, and this solution was incubated at room temperature for 2 h Then, 1,300 μL of HB was added, and the solution was centrifuged for 50 min at 21K × g at 4°C to remove unincorporated dye. The supernatant was discarded, and the pellet was resuspended in 100 μL of HEPES buffer. Unless noted otherwise, the standard virus labeling was performed using 1× concentration of TR-DHPE.

Labeled particle concentration was estimated either by viral protein content via BCA assay or by spot counting on the microscope (see details in supporting material). Unless noted otherwise, values reported are calculated from the BCA assay.

Single-virus binding and unbinding assay

Single-virus binding assays were performed to SLBs formed by the vesicle fusion method inside microfluidic devices. SLB formation and microfluidic device construction have been described previously and are outlined briefly in the supporting material. Unless otherwise noted, the standard SLB composition was 20 mol% DOPE, 10% Chol, 0.05% Oregon green-DHPE, 0–5% ganglioside receptor (always listed), and the remaining amount (69.95%–64.95%) POPC. Following SLB formation, 4 μL of labeled virus (typically at 0.1 nM concentration) was added to the inlet hole and then drawn through to the outlet hole by pipette. Virus was incubated above the SLB for 15 min at room temperature (RT; ∼22°C). The flow cell was then rinsed with 1 mL of RB by a syringe pump at 0.8 mL/min. Images of bound virions and SLB were then collected in 5–15 separate areas in the center region of the flow cell. Little variation in viral binding numbers along the length of the flow cell was observed, indicating that volume displacement during viral introduction was quite even. Viral spots in each field of view were quantified using previously published custom-built Matlab scripts (20,21). These scripts automatically exclude from the analysis large or irregularly shaped spots, presumably clusters of viral particles (see examples in Fig. S3). Our version of these scripts, including parameters and options optimized for our SeV data is available at https://github.com/rawlelab/SendaiBindingAnalysis. To measure viral unbinding, time-lapse fluorescence micrographs were taken every 30 s for 60 min following buffer rinse. For antibody-mediated unbinding measurements, 4 μL of 20 μg/mL antibody solution in RB buffer was introduced into the flow cell following buffer rinse, and then viral unbinding was observed as above.

Antibody inhibition assay

For antibody inhibition measurements, labeled SeV was pre-mixed with antibody (Ab) solution; final concentrations of Ab as indicated. The SeV-Ab mixture was incubated on ice for 30 min prior to introduction into the flow cell. Single-virus binding measurements were then performed as above.

Immunofluorescence measurements

Following SeV binding to SLB as above, 4 μL of anti-HN (1A6) at 20 μg/mL was introduced into the flow cell and incubated for 15 min. The flow cell was rinsed by syringe pump with 2 mL of HB, and then 4 μL of goat anti-mouse IgG-Alexa 488 at 2 μg/mL was introduced and incubated for 15 min. The flow cell was rinsed by syringe pump with 3 mL of HB, and then sequential images of TR-labeled viral particles and Alexa 488 antibody were observed by fluorescence microscopy. Co-localization was quantified using previously published Matlab scripts (20,21).

Single particle tracking

For single particle tracking (SPT), time-lapse fluorescence micrographs of Sendai virions bound to SLBs were taken every 5 s for 25 min. SPT was performed using the TrackMate (36) plug-in of ImageJ. Image stacks were analyzed using a Laplacian of Gaussian (LoG) detector, with a sigma suited to the estimated blob size, 3.0 μm. In the median filtered image, the maxima were detected, and those too close together were suppressed. Finally, a quadratic fitting scheme allowed for subpixel localization. The spots and tracks were then displayed using a HyperStack displayer, allowing for manual editing and filtering of spots by mean intensity (>420–445). Next, the virus particles were tracked using the simple Linear Assignment Problem (LAP) tracker with a linking max distance of 8.0 μm, a gap-closing max distance of 8.0 μm, and a gap-closing max frame gap of 3.0 frames. These parameters were determined by empirical optimization to successfully track both stationary and fast-diffusing particles and were verified by manually observing the resulting tracks. Subsequently, the tracks were filtered by the number of spots per track (>100–200 spots per track) before manual editing. Manual editing was used to identify remaining erroneous tracks (primarily those that the algorithm incorrectly assigned to jump back and forth between two adjacent immobile particles). The resulting mean squared displacement (MSD) versus Δt curves were fitted to a linear model to extract the diffusion coefficients, following a published algorithm to determine the optimal number of MSD points to include in the fit (37). The Matlab code of our implementation of this algorithm is available at https://github.com/rawlelab/SendaiBindingAnalysis. The lower limit of detectable diffusion (1.24 × 10⁻⁵ μm²/s) was determined by measuring the “diffusion coefficients” of virions bound directly to a glass surface with the SPT method described.

Fluorescence microscopy

Fluorescence microscopy images were acquired with an Axio Observer 3 microscope using a 63× oil immersion objective, NA = 1.4 (Carl Zeiss Microscopy, LLC., White Plains, NY), and a Lumencor Spectra III, LED Light Engine. Images were recorded with an ORCA Flash 4.0 V2 Digital CMOS camera (Hamamatsu Photonics K.K., Hamamatsu City, Japan) using a 16-bit image setting and were captured with Micromanager software (38). Images and video micrographs were captured at 100 ms/frame with 2 × 2 binning. Additional microscopy details are given in the supporting material.

Results and discussion

Assay overview and development

Our single-virus-binding assay for SeV (Fig. 1 A) was adapted from a previously reported influenza A virus (IAV) assay (23) but with some important modifications and considerations. Viral particles were fluorescently labeled with a lipophilic dye and were introduced at a known concentration (typically 0.11 nM; see supporting material for concentration determination) into a polydimethylsiloxane (PDMS) microfluidic chamber housing a glass-supported bilayer. The SLB was prepared with one of several sialic acid glycolipids, which served as receptors for viral binding. Following incubation, unbound viruses were removed via buffer rinse, and bound viruses were then imaged by fluorescence microscopy.

Figure 1.

Overview of single-virus-binding assay design and validation data. (A) shows a schematic of the assay design. Supported lipid bilayers (SLBs) are self-assembled inside a microfluidic device. Sendai virions, membrane labeled with Texas red-DHPE (TR) or R18, are introduced into the flow cell, where they can bind to receptors in the SLB. Binding and, in limited cases, fusion are observed and quantified by fluorescence microscopy. (B) shows a linear relationship between the viral concentration added to the flow cell, and the relative number of virions bound. In these measurements, the SLB contained 2% GD1a receptor and the viral concentration ranged from 0.025 to 0.1 nM. Error bars are ±standard error of ≥3 sample replicates, with propagated relative error as described in the supporting material. (C) depicts the relative number of virions bound to SLBs containing either 2% GD1a or no receptor. Very little binding is observed in SLBs without receptor. Error bars are ±standard error of ≥3 sample replicates, with propagated relative error as described in the supporting material. (D) shows the fraction of spots bound to various SLBs, which show positive immunofluorescence (IF) labeling by anti-hemagglutinin-neuraminidase (HN; 1A6) antibody, followed by Alexa 488-labeled secondary antibody. Colocalization between Alexa 488 and the membrane label in the particles (TR or R18) was used to determine whether a particle was IF positive. “No SLB” indicates that particles were attached non-specifically to the glass coverslip instead of a SLB, followed by surface passivation by 30 g/L bovine serum albumin. “0.5× R18” indicates that viral particles were labeled with R18 at 0.5× concentration (see Materials and methods for labeling details); all other data in this panel were collected with TR-labeled particles. Error bars are ±standard error of ≥3 sample replicates, with ≥10 separate image locations in each sample; the “0.5× R18” sample had 1 sample replicate, and error shown is standard deviation of 10 separate image locations. (E) shows the sensitivity of Sendai virus binding to 2% GD1a SLBs following pre-treatment of the virus with 1A6 antibody. Virus labeled with R18 (1×) was incubated with 1A6 antibody at the concentrations shown for 30 min on ice prior to injection into the flow cell. The blue line shows the best fit to a sigmoid curve; IC₅₀ = 1.3 ± 0.1 μg/mL. Error bars are ±standard deviation of 6–10 separate image locations within each sample. To see this figure in color, go online.

We found that three critical aspects of the assay design were (1) the method of introduction of the virus into the microfluidic chamber, (2) the extent of labeling and identity of the fluorescent dye incorporated into the viral envelope, and (3) the minimization of receptor-triggered membrane fusion by temperature control. Some of these issues are particular to paramyxoviruses and similar viruses, such as the minimization of fusion triggered by receptor binding, but others are relevant to viral binding and fusion studies more broadly. If not carefully controlled, each can lead to unwanted artifacts (such as non-specific binding or fusion) and/or difficulties in data collection; this is discussed in detail in the supporting material (see Figs. S4–S8, Video S1, and associated text). We highlight that proper fluorescence labeling is an issue of particular note, as many previous reports studying SeV fusion have utilized fluorescent labeling of the SeV envelope (see, for example, (39, 40, 41, 42, 43)). Depending on the extent of labeling, non-specific fusion artifacts may have been occurring.

Video S1. Example lipid mixing event of a Texas Red-labeled Sendai virion fusing with an SLB

In the video, several virions bound to an SLB are observed. Some are mobile, some are not. One virion in the center of the field of view is observed to undergo lipid mixing in an "explosion", transferring its membrane dye to the SLB, followed by outward radial diffusion of the dye in the SLB. As discussed in the main text, such lipid mixing events were extremely rare (<<1%) in our assay, which was designed to study viral binding in the absence of fusion. For ease of viewing, the video has been sped up 3.3X from the original data collection (collected at 100 ms per frame). The dimensions of the field of view are 37 x 38 μm.

Assay validation

After optimizing the issues identified above, we performed various validation experiments. First, we observed that viral binding to SLBs with sialic acid receptor (2 mol% GD1a) was dose dependent with concentration of virus added (Fig. 1 B). Second, we verified that SeV showed little binding to SLBs without receptor, compared with SLBs with 2% GD1a (Fig. 1 C). Third, using immunofluorescence (IF) imaging of HN proteins on stably bound particles, we observed a much higher percentage of IF-positive particles bound to SLBs with 2 mol% GD1a (88.5% ± 3.8%) compared with particles bound to SLBs without receptor (32.3% ± 9.9%) or bound non-specifically to a glass surface (31.9% ± 5.5%) (Fig. 1 D; representative images shown in Fig. S9). Control experiments using secondary antibody showed only negligible IF-positive results (Fig. S10). This indicates that receptor binding to SLBs efficiently selects binding-active virions from non-viral or quasi-viral particles that are co-purified, a recently identified challenge in viral purification (44, 45, 46), and a potential yet often ignored confounding factor in single-virus experiments (see supporting material for further discussion).

Finally, to verify that specific HN-receptor interactions were responsible for viral binding in our assay, we examined antibody inhibition of SeV binding by pre-treating the virus with a neutralizing anti-HN monoclonal antibody (1A6) prior to introduction into the flow cell. We determined the IC₅₀ value in our assay at 1.3 ± 0.1 μg/mL (Fig. 1 E) and at saturating antibody concentrations we observed an ∼10-fold decrease in binding to SLBs with 2% GD1a compared with virus that had been mock treated with buffer only. This antibody inhibition was observed both with TR-labeled virus and with R18-labeled virus (at appropriate R18 concentrations), indicating a robust antibody response in our assay independent of dye identity (Fig. S11). Conversely, no inhibition of apparent binding was observed to SLBs without GD1a, supporting the IF results above and indicating that particles bound to membranes without receptor are largely bound non-specifically or are non-viral particles. Similar inhibition was also observed with a separate neutralizing anti-HN antibody (3G12) (see Fig. S12). Interestingly, we also observed moderate inhibition of binding with an anti-F monoclonal antibody (11H12) (34) (see Fig. S12). This suggests that F and HN proteins may be closely intermixed on the viral surface, such that antibody binding to F leads to partial inhibition of HN receptor binding by steric hindrance. Together, these antibody inhibition results underscore that HN-receptor interactions are responsible for viral binding in our assay. More broadly, they also demonstrate the capability of assays such as these to cleanly monitor antibody inhibition of binding to specific receptors without convolving with other steps in the infectious cycle such as membrane fusion.

Characterization of viral behavior and studies of viral-binding mechanisms

To our knowledge, single-virus measurements of this type have not been conducted for SeV or other paramyxoviruses. Therefore, we performed various experiments/measurements to characterize viral behavior and to examine specific questions relating to viral binding, including the influence of receptor chemical structure, cooperativity of binding, dependence on cholesterol concentration, and mechanism of mobility following binding. These are each described sequentially below.

Number and positioning of α2,3-linked sialic acids on receptors directly modulates viral binding

Number, positioning, and chemical linkage of sialic acid (SA) receptors can be an important modulator of viral attachment for SA-binding viruses and can play a role in both host and tissue tropism (47). Previous work has indicated that SeV and other SA-binding paramyxoviruses such as mumps virus and HPIV3 have a preference for α2,3-linked SA residues on either glycolipids or glycoproteins (13). To study the influence of the number and positioning of α2,3-linked SA residues on SeV binding in our assay, we studied binding to membranes containing 2 mol% of different SA-gangliosides: GM3, GM1, GD1a, or GQ1b (Fig. 2). GM1, GD1a, and GQ1b all possess the same ganglio-tetrose backbone with one, two, or four SA residues either branched internally off the second backbone sugar and/or in the terminal position, as depicted (Fig. 2A). GM3 is similar in structure to GM1, with one SA residue, but is lacking the final two sugars of the tetrose backbone, which transforms its SA into a terminal residue rather than an internal branching residue.

Figure 2.

Chemical structure of gangliosides in SLBs directly modulates Sendai virus (SeV) binding. Single-virus-binding measurements were performed to SLBs containing 2% GM3, GM1, GD1a, GQ1b, or no receptor. (A) shows the chemical structure of each receptor used. Sialic acids are highlighted in blue. Note that GM1, GD1a, and GQ1b each possesses the same ganglio-tetrose backbone, whereas GM3 lacks the final 2 sugars of the tetrose backbone. (B) shows the relative number of virions bound to SLBs with the different ganglioside receptors. Binding measurements are shown relative to 2% GD1a. Error bars are ±standard error of ≥4 sample replicates, with propagated error as described in the supporting material. To see this figure in color, go online.

We observed that whereas binding to 2% GM1 membranes occurred at nearly identical levels to membranes containing no receptor at all, membranes containing 2% GM3 elicited a moderate amount of viral binding. This indicates that the final two sugars of the tetrose backbone in GM1 likely inhibit access to the internally branched SA, preventing effective binding by HN. On the other hand, GD1a, with an identical internally branched SA to GM1 but also containing an additional terminal SA, exhibited approximately two times greater viral binding than GM3. This suggests that either the terminal SA of GD1a is a much more favorable binding target for HN or the addition of the terminal SA alters the ganglioside head group structure such that the internally branched SA becomes more accessible than in GM1. GQ1b exhibited the highest level of binding, approximately two times greater than GD1a, suggesting that the additional SA residues (4 total for GQ1b), as well as the greater distance of the terminal SAs from the tetrose core facilitate increased binding by HN (Fig. 2 B).

SeV exhibits different cooperative binding to different ganglioside receptors

Monomeric receptor-binding affinities of HN for the closely related HPIVs are rather weak (K_d ∼0.1 mM), (48), suggesting that avidity rather than affinity likely governs SeV binding, and opens the possibility that binding may be cooperative. To directly study the cooperativity of SeV binding, we observed virus binding to SLBs with varying GD1a concentrations (Fig. 3 A). We observed a characteristic sigmoidal response to GD1a concentration, indicative of cooperativity. To quantify the extent of cooperativity, we fit the data to a Hill model binding curve:

$$\mathit{Re}lBinding=\frac{R{B}_{\mathit{max}}}{1+{\left(\frac{K}{\left[receptor\right]}\right)}^{C_F}}$$ (1)

where RB_max is the maximum relative binding, [receptor] is the mole percent of receptor in the target SLB, K is equal to [receptor] when RelBinding = 0.5 x RB_max, and C_F is the cooperativity factor. Best-fit values are shown in Table 1.

Figure 3.

SeV exhibits cooperative binding to SLBs with either GD1a or GQ1b. Single-virus-binding measurements were performed to SLBs with varying concentrations of either (A) GD1a or (B) GQ1b. Solid lines show fits of the data to a Hill model binding curve (Eq. 1) to quantify the extent of cooperativity. The relative number of virions bound in both panels is calculated relative to 2% GD1a, and for ease of comparison are shown overlaid in (C). Error bars are ±standard error of ≥3 sample replicates, with propagated relative error as described in the supporting material. Best-fit parameters are shown in Table 1. To see this figure in color, go online.

Table 1.

Hill model fit values for GD1a and GQ1b cooperative binding data

Ganglioside Receptor	RBmax	K	CF
GD1a	0.95 ± 0.08a	0.86 ± 0.07a	3.3 ± 0.4a
GQ1b	3.6 ± 0.80a	1.5 ± 0.5a	2.1 ± 0.4a

^aAll error estimates are determined by bootstrap resampling of individual samples and all images taken within each sample.

Interestingly, this best fit is quite similar to data collected for influenza virus A/H3N2/X-31 (IAV^X31), which also can use GD1a as a receptor. In a similar binding experiment using SLBs with GD1a, best-fit values for IAV^X31 were K = 0.85, and C_F = 2.36 ± 0.85 (23), suggesting that both IAV^X31 and SeV have evolved a similar cooperative binding response to SA receptors despite numerous structural and functional differences between the hemagglutinin of IAV^X31 and the HN SeV. In protein enzymology, it has been noted that enzymes often have evolved a binding affinity that matches the physiological concentration of the ligand, presumably driven by selection to maximize their dynamic range (49). A similar driving force may be at play in the evolution of these two viruses.

We also studied the cooperative binding response of SeV, using GQ1b as a receptor (Fig. 3 B). We observed a more pronounced curve relative to GD1a, with increased viral binding at all GQ1b concentrations >0.25 mol%, consistent with the higher potency of GQ1b as a receptor compared with GD1a in bulk infectivity measurements (16). Best-fit values to the Hill binding curve are reported in Table 1. The cooperativity factor for GQ1b was significantly lower than the cooperativity factor for GD1a (p value = 0.03 by bootstrap resampling), whereas both RB_max and K (half-max concentration) were significantly larger for GQ1b than for GD1a (p values = 0.006 and <0.002, respectively). This suggests that fewer HN-GQ1b complexes may be needed for stable binding compared with HN-GD1a complexes. This is likely due to the increased number of both internal and terminal SA residues per ganglioside of GQ1b compared with GD1a (see Fig. 2 A).

Cholesterol concentration in target SLB does not strongly influence viral binding

In studies of SeV and closely related paramyxovirus, fusion and infectivity are sensitive to cholesterol concentration in both the viral and target membranes, generally with less cholesterol resulting in less fusion (50, 51, 52, 53, 54). However, it is not clear whether this effect is due to modulation of binding, fusion, or both. In similar binding measurements to those performed here for SeV, it has been shown that cholesterol in the target SLBs can modulate binding of influenza virus (IAV^X31) to GD1a (23), with an ∼2-fold linear increase in binding avidity with increasing cholesterol concentrations from 0% to 30%. By using molecular dynamics simulations, it was suggested that this was due to the formation of receptor nanoclusters in the target SLB, which were stabilized by increasing concentrations of cholesterol. In the context of multivalent, cooperative binding, nanoclusters could function as binding “hotspots” to increase binding avidity.

Therefore, to examine whether cholesterol concentration in the target membrane would modulate SeV binding, we studied SeV binding to SLBs with either 1% or 2% GD1a and varying concentrations of cholesterol from 0% to 30%. The remaining lipid composition of the SLB was matched to what had been used for IAV^X31. Interestingly, we observed that there was no strong dependence of SeV binding on SLB cholesterol concentration (see Fig. 4). This suggests that sensitivity of SeV fusion and infection to cholesterol concentration is likely not due to viral binding with a similar mechanism IAV^X31.

Figure 4.

SeV binding to SLBs with GD1a is insensitive to cholesterol concentration. Single-virus-binding measurements were performed on SLBs with varying concentrations of cholesterol and either 1% GD1a (yellow bars) or 2% GD1a (red bars). The SLBs also contained 20% DOPE, 0.05% Oregon green-DHPE, and the remainder POPC (47.95%–78.95%). The relative number of virions bound was calculated relative to 2% GD1a, 10% cholesterol. Within a set concentration of receptor, little difference in binding was observed across the range of cholesterol concentrations tested. By comparison, influenza A virus in Ref. (23) shows an ∼2-fold linear increase in binding to similar SLBs across the same range of cholesterol. Error bars are ±standard error of ≥3 sample replicates, with propagated relative error as described in the supporting material. To see this figure in color, go online.

Surprisingly, this also suggests that SeV binding, unlike IAV^X31, is insensitive to the formation of receptor nanoclusters in the target SLB, even though both viruses exhibit a similar cooperative binding dependence on GD1a concentration (see Fig. 3 and discussion above). This indicates that although both viruses may have evolved a similar cooperative binding response to SA-glycolipids, the binding mechanisms are distinct enough to produce different sensitivities to target cholesterol composition. This may in part be driven by the fact that whereas both viruses bind to host cells at the plasma membrane, IAV fuses in the late endosome, which has a comparably lower cholesterol content (55) than the plasma membrane itself, where SeV fuses.

Estimating τ_unbind for SeV-SLB binding

To characterize SeV behavior and stability following binding, we performed time-lapse imaging of bound viral particles following solution rinse, quantifying net change in bound virions over time. We observed that bound viruses were quite stable on 2% GD1a membranes, decreasing only to ∼60%–80% of their initial bound value over 60 min (Fig. 5 A). During this time, average SLB background intensities remained quite low, indicating that viral detachment, not membrane fusion, was the primary cause of the decrease. Detached viral particles could also be observed in the solution above the SLB.

Figure 5.

Antibody treatment following SeV binding increases the rate of viral unbinding. Shown are example unbinding curves for SeV bound to SLBs with 2% GD1a. Following virus binding, antibody solution (or a no-antibody control solution) was introduced into the flow cell, and viral unbinding from the SLB was observed over 1 h. Shown are example data of the fraction of viruses bound over time in the presence of (A) no antibody, (B) anti-HN (1A6) at 20 μg/mL, and (C) anti-HN (3G12) at 20 μg/mL. Red lines represent best fits to an exponential decay curve (Equation S4) to estimate the characteristic decay time (τ_unbind). To see this figure in color, go online.

To estimate the characteristic decay time of virus unbinding τ_unbind, we fitted the curves to an exponential decay (Equation S4), obtaining a τ_unbind of 79 ± 27 min. This indicates that once virus particles are bound to the underlying SLB, they remain quite stably attached for some time. To place these results in context, we note that the SA receptor is in large excess on the SLB surface in these measurements (approximately 2 × 10⁴/μm² at 2% GD1a if we estimate 1 × 10⁶ lipids total/μm²) and that these measurements were made under conditions (high Cl⁻ concentration, neutral pH, RT) where SeV neuraminidase activity is reduced (56). Therefore, the measured τ_unbind is unlikely to be simply due to the neuraminidase action of HN globally depleting the SA density on the membrane over the course of the measurement. We anticipate investigating the interplay between neuraminidase activity and viral binding/unbinding in future work.

Virus mobility and evidence for “rolling” mechanism

We observed that many particles exhibited mobility by diffusion on the SLB following binding. To quantify the diffusion behavior, we performed single-particle tracking of virions in time-lapse micrographs over 25 min and then estimated the diffusion coefficient (D) of each particle via linear fits of mean squared displacement (MSD) versus Δt curves. We observed that ∼50% of particles were immobile (D < 1.2 × 10⁻⁵ μm²/s, the lower limit of detectable diffusion). The majority of mobile particles exhibited low diffusion coefficients (D < 0.025 μm²/s), but a small percentage did exhibit higher values, ranging up to 0.3 μm²/s (Fig. 6). By comparison, 100-nm liposomes anchored by a single lipid in a SLB with a similar composition have D ≈ 1 μm²/s (57). We also observed that the fraction of immobile particles increased from ∼50% to ∼70% after 60 min following buffer rinse (Fig. S13). As ∼20% of particles unbind over this same time window (see Fig. 5 A), these data suggest that immobile particles are less likely to unbind than mobile ones.

Figure 6.

Distribution of diffusion coefficients for mobile virions on supported lipid bilayers. Diffusion coefficients were calculated from single-particle tracking (SPT) traces of bound virions in time-lapse micrographs over 25 min. Inset shows sample SPT traces; 50% of particles were immobile (D < 1.2 × 10⁻⁵ μm²/s, the lower limit of detectable diffusion); these are not shown in the distribution. The vast majority (∼80%) of mobile virions exhibited slow diffusion D < 0.025 μm²/s. The SLB contained 2% GD1a. Total number of particles analyzed was 1,035. To see this figure in color, go online.

We anticipate two possible mechanisms for the observed diffusion of SeV on SLBs. First, SeV diffusion could be dominated by the diffusion of several lipid receptors to which the virus is stably bound. However, monomeric HN-binding affinity for SA receptors is rather weak for closely related paramyxoviruses (∼0.1 mM K_d for HPIV-1, -2, and -3) and monomeric k_off values quite fast (∼0.2 s⁻¹) as measured by surface plasmon resonance (48), suggesting that individual HN-receptor attachments are unlikely to be stably bound during our observation window. Instead, a more likely mechanism is a rolling mechanism, wherein numerous HN attachment proteins engage weakly with individual lipid receptors, and random binding and unbinding (and/or neuraminidase cleavage of the SA) leads to Brownian diffusion on the SLB surface, as suggested for measurements of influenza virus (58,59).

To differentiate between these mechanisms, we performed an antibody-mediated unbinding experiment (Fig. 5 B and C). Following viral binding and buffer rinse, anti-HN antibody (1A6) at 20 μg/mL was introduced into the flow cell, and τ_unbind was measured as above. If viral diffusion were dominated by a rolling mechanism, we would expect a significant decrease in τ_unbind, as the rolling particle became sequentially bound by a higher density of antibody until it could no longer sustain sufficient HN-receptor contacts to remain bound. If, on the other hand, HN-receptor complexes were stable and viral diffusion were dominated by lipid diffusion of receptors bound by HN, we would not expect a significant change in τ_unbind, as the HN-receptor interface would be inaccessible to antibody binding; antibodies would predominantly bind to the distal viral surface. We observed that there was a significant decrease in τ_unbind following antibody treatment (Table 2), lending support to the rolling mechanism of viral diffusion.

Table 2.

τ_unbind Values^a for Sendai virus unbinding from SLBs in the presence or absence of anti-HN antibodies

Antibody Treatment	τunbind (min)b
No antibody	79 ± 27
Anti-HN (1A6)	12.0 ± 6.9c
Anti-HN (3G12)	25.6 ± 7.6c

^aThe entire set of best-fit values to the exponential decay curve is given in Table S1.

^bError estimates are the standard deviation of best fits to 3–5 sample replicates.

^cBoth the 1A6 and 3G12 τ_unbind values are significantly different from the no-antibody values (p < 0.015, 2-Sample Kolmogorov-Smirnov test).

What might be the mechanistic origins of the immobile virions? Several non-mutually exclusive explanations exist. First, immobile particles may be virions with a large enough number of HN-SA engagements that rolling diffusion becomes difficult to observe at our experimental timescale (∼25 min). Tantalizingly, these particles may represent virions with large organized patches of HN on their surface, as has been reported in a cryo-EM study of HPIV3 (60) but for which no structural data as yet exist for SeV. Second, although all SLBs used in these measurements were observed to be defect free by homogeneous fluorescence, we cannot rule out the possibility that some immobile particles may be bound to small, unobserved (sub-diffraction limit) defects. Third, immobile particles may be those with multiple activated F proteins, whose fusion peptides have been inserted into the SLB but not yet collapsed into a post-fusion conformation. However, as we observe little evidence of fusion, this would suggest that collapse of this unstable intermediate would be slow relative to the experiment timescale, which seems less likely.

Conclusion

In this report, we have presented the development of a single-virus-binding assay for the prototypical paramyxovirus SeV, using glass-supported lipid bilayers as the target membrane for binding, and have discussed key design features that are critical to avoid unwanted artifacts. Using this binding assay, we investigated various biophysical questions relating to SeV binding. First, we found that the number and positioning of α2,3-linked sialic acids on receptors can strongly influence SeV binding. Second, we investigated the cooperative nature of SeV binding, observing a strikingly similar cooperative dependence on GD1a concentration with influenza A virus (IAV^X31), suggesting a common evolutionary response toward receptor engagement despite many differences at the viral protein level. Third, we demonstrated that SeV exhibits little dependence on cholesterol concentration in the target SLB, suggesting that the sensitivity of SeV infection to cholesterol is likely due to a later step in the entry mechanism and that, unlike IAV^X31, SeV binding is relatively insensitive to the presence of receptor nanoclusters. Finally, we characterized SeV behavior following binding, including the rate of detachment and viral mobility, and provided evidence for a “rolling mechanism” of mobility on the SLB surface. Together, these results provide mechanistic insight into the biophysics of SeV binding and may be useful to suggest constraints on host cell tropism and possible inhibitor design. We look forward to the extension of this assay to study other paramyxoviruses in the future.

Author contributions

A.L. and O.K. designed experiments, collected and analyzed data, and helped write the manuscript. N.S. and D.V. collected data and contributed to single-virus assay design. P.F.A. collected bulk fusion data. P.A.T. and K.D.A. produced antibodies and collected antibody characterization data. B.L. provided anti-SeV antibodies and antibody characterization data. R.J.R. designed experiments, analyzed data, and wrote the manuscript.

Editor: Ilya Levental.