Probing Affinity, Avidity, Anti-Cooperativity, and Competition in Antibody and Receptor Binding to the SARS-CoV-2 Spike by Single Particle Mass Analyses

Determining how antibodies interact with the spike (S) protein of the SARS-CoV-2 virus is critical for combating COVID-19. Structural studies typically employ simplified, truncated constructs that may not fully recapitulate the behaviour of the original complexes. Here, we combine two single particle mass analysis techniques (mass photometry and charge-detection mass spectrometry) to enable measurement of full IgG binding to the trimeric SARS-CoV-2 S ectodomain. Our experiments reveal that antibodies targeting the S-trimer typically prefer stoichiometries lower than the symmetry-predicted 3:1 binding. We determine that this behaviour arises from the interplay of steric clashes and avidity effects that are not reflected in common antibody constructs (i.e. Fabs). Surprisingly, these sub-stoichiometric complexes are fully effective at blocking ACE2 binding despite containing free receptor binding sites. Our results highlight the importance of studying antibody/antigen interactions using complete, multimeric constructs and showcase the utility of single particle mass analyses in unraveling these complex interactions.


Introduction
The emergence of the SARS-CoV-2 coronavirus and subsequent onset of the coronavirus disease 2019 (COVID- 19) pandemic has necessitated the rapid development of vaccines and other treatments. [1][2][3] The primary focus of these countermeasures is the SARS-CoV-2 spike (S) protein present on the viral surface, which is responsible for initiating host infection via complexation to the human ACE2 receptor and subsequent fusion of the viral and host cell membranes. 4 The majority of vaccines developed against SARS-CoV-2 use the S protein (e.g. genetically encoded via either mRNA/DNA cargo [5][6][7] or displayed on a nanoparticle surface 8 ) to elicit an immune response. Understanding how exactly antibodies (Abs) interact with the SARS-CoV-2 S protein is a crucial component for both continuing vaccine development as well as the rational design of target biotherapeutics (e.g. monoclonal Abs). 9,10 Like the spike proteins of many other viruses, the SARS-CoV-2 S protein is present in a trimeric, membrane-embedded state. 11 Effective neutralizing Abs for SARS-CoV-2 often target the receptor binding domain (RBD) of the S protein. [12][13][14][15] As the RBD is the site of initial ACE2 receptor binding, these Abs are thought to achieve neutralization largely by sterically preventing interactions between the S protein and host receptor. 16 Due to its trimeric nature, each individual spike contains three copies of the RBD.
Given the central role of Ab binding for the successful neutralization of antigens, a seemingly simple question is: how many copies of an Ab can bind to one spike? And relatedly, how many Ab copies need to bind to induce neutralization? Since each S-trimer contains three identical copies of the S protomer, one may expect that Abs bind the S-trimer with a 3:1 stoichiometry. However, this prediction may be somewhat naïve, and the true Ab binding stoichiometry will be complicated by several factors. Firstly, the RBD is dynamic and can occupy either an "up" or "down" state, defined by its position relative to the remainder of the complex. 11 Only the up RBD state is capable of binding the ACE2 receptor. 17 As each RBD is related in the S-trimer by 3-fold symmetry, there exists a total of 4 possible conformational states of the RBDs in the S-trimer (with up : down ratios of 0:3, 1:2, 2:1, and 3:0). Certain Abs against the RBD may only recognize one of the two states, which can interconvert. 12,18,19 Therefore, any RBD-targeting Ab could conceivably bind a particular S-trimer with any stoichiometry between 0 and 3, depending on the exact conformational status of the complex. Secondly, full Abs (IgGs) possess two equivalent Fab arms, of which one or both may be involved in binding (i.e. avidity). Avidity effects are well-known to play key roles in the potency of neutralizing Abs and could manifest as an apparent decrease in binding stoichiometry. 20,21 Thirdly, anti-cooperative binding effects arising from steric conflicts between multiple binding Abs may also play a role, hampering the amount of concurrent binding allowed.
Considering the known impacts these various effects can have on Ab efficacy, the stoichiometries of Ab binding to the SARS-CoV-2 S protein is surprisingly poorly characterized. This is likely due in part to the lack of biochemical and biophysical methods to effectively probe such heterogeneous interactions effectively and efficiently. For example, surface plasmon resonance (SPR) and biolayer interferometry (BLI) are highly effective at rapidly quantifying antigen binding, but provide only an ensemble-averaged overview and yield limited structural information. [22][23][24] Single particle electron microscopy (EM) can often provide near-atomic details of protein structure and protein-protein interactions (allowing direct mapping of Ab epitopes on the full SARS-CoV-2 S ectodomain 11,[25][26][27], but due to the extended flexibility of full-length IgGs is typically (with some exceptions 28 ) only able to visualize binding of antibody fragments (i.e. truncated Fab domains), and thus may not directly capture any effects of avidity or steric interactions that would occur in the full IgG. Nuclear magnetic resonance spectroscopy and X-ray crystallography can yield atomic protein structures, but due to limitations with size and conformational/glycosylation-induced heterogeneity, respectively, have been largely restrained to studies on truncated single RBD constructs, and thus remain relatively blind to both the up : down dynamics of the full trimer as well as potential avidity effects. 29,30 Native mass spectrometry (MS) is an analytical technique capable of measuring the mass of proteins and protein complexes. 31 As any binding event leads to a corresponding increase in mass, native MS offers a convenient readout of ligand binding and can readily distinguish different binding stoichiometries and different ligands by their unique masses. In the context of monitoring interactions to the SARS-CoV-2 S protein, the feasibility of these experiments is greatly hindered by the extreme heterogeneity caused by the high degree of glycosylation present on the S protein (the so-called glycan shield). [32][33][34] This heterogeneity leads to a normally untenable degree of spectral complexity that obfuscates the charge state assignments required for correct mass determination. 35 While some success has been reported in the conventional native MS analysis of SARS-CoV-2 S and other viral spike proteins (e.g. by metabolic glycan engineering 36 or limited charge reduction 37 of truncated constructs), these modified constructs may not exhibit the same binding behaviour as the real viral spike protein, given the known importance of glycan structure in these interactions. 32 Here, we report the application of two single particle approaches for mass analysis, mass photometry 38 and charge-detection native mass spectrometry 39,40 , to circumvent the need of conventional charge assignment and allow successful measurement of the full SARS-CoV-2 Strimer ectodomain, as well as the binding stoichiometries to full-length neutralizing IgGs. Our measurements reveal that IgG binding to the SARS-CoV-2 S-trimer can exhibit a diversity of binding behaviours that is not captured when studying the truncated Fabs or RBD constructs alone.
We also demonstrate that these techniques can be used to monitor binding of the ACE2 receptor, as well as the S proteins from other variants of concern of the SARS-CoV-2 virus. These ultrasensitive single particle approaches (requiring only ~femtomoles of sample) thus offer a powerful addition to the toolkit of contemporary biophysical tools by providing a "one-shot" method for determining Ab affinity, anti-cooperativity, and avidity simultaneously. Our findings highlight the biophysical complexity of the multimeric interactions that occur between Abs and the SARS-CoV-2 S protein.

Results
Single Particle Mass Analysis of the SARS-CoV-2 S-trimer. Originally introduced as interferometric scattering mass spectrometry (iSCAMS) 38 , mass photometry (MP) is a light scattering-based, label-free, mass analysis technique that determines the mass of a single particle from its scattering intensity. 41 Since MP does not rely on any charge state determination, the masses of extensively glycosylated proteins can be readily measured. Advantages of MP include its rapid analysis time and a minimal need of sample preparation. A representative MP histogram of the SARS-CoV-2 S-trimer is depicted in Figure 1A. The S-trimer exhibits a large primary distribution at 474 kDa, while a minor low-mass distribution is also observed and can be assigned as residual S-monomer. Of note, no species corresponding to higher-order aggregates (i.e. dimers of S-trimers 25 ) are observed.  Supplemental Table S1.

B
Alternatively, charge detection mass spectrometry (CD-MS) can be used to overcome the charge inference problem in native MS by directly detecting both the charge and mass-to-charge (m/z) ratio of an ion. 42 Due to this two-dimensional detection method, peaks that are unresolved in the m/z dimension may still be resolvable in the charge dimension, aiding in the assignment of complex spectra. A representative Orbitrap-based CD-MS histogram of the SARS-CoV-2 S-trimer is depicted in Figure 1B. Again, a single major distribution of particles corresponding to the Strimer with is observed, with a minor distribution corresponding to the S-monomer also detected.
The higher mass resolution achievable by CD-MS (as exhibited by the narrower mass distributions of the S-trimer relative to MP) highlight an important advantage of CD-MS. The trimer mass measured by CD-MS (477 kDa) is within ~1% of the mass determined by MP. The close agreement in the results of these two disparate single particle methods underscores the robustness and complementarity of these approaches.
The backbone sequence-predicted mass of the S-trimer construct used here (390.349 kDa) underestimates the observed mass measured by both techniques by ~90 kDa, reflecting the extensive glycosylation profile of the S protein. To estimate the expected mass contribution of the glycan shield, we calculated the average N-glycan masses derived from the glycoproteomic data of Allen and coworkers. 43 The calculated glycan (92.0 kDa) and resultant total S-trimer (482.4 kDa) masses agree quite well (within 2%) with the masses measured by both MP and CD-MS. The glycan mass contribution measured here is somewhat lower than the recent results of Miller and co-workers 44 who reported large mass discrepancies of ~40% from similar glycoproteomic experiments. However, it should be noted the constructs used in that study differ from the one employed here in several key aspects (e.g. absence of stabilizing 2P mutations, different expression systems, etc.), as well as differing substantially in experimental setup (electrostatic linear ion trap vs. Orbitrap), which all may be factors accounting for this apparent discrepancy.
Abs targeting the S-trimer can exhibit diverse binding characteristics. To establish the capability of single particle mass measurements to resolve the binding of Abs to the S-trimer, we initially screened the binding of a representative panel of twelve monoclonal anti-S-trimer IgGs using MP (Figure 2). These previously reported Abs, originally isolated from the sera of convalescent COVID-19 patients, target a variety of epitopes and exhibit varying neutralization potencies (Supplemental Table S3). 12 Upon incubation of the S-trimer with the IgGs, new species of larger mass in the MP histograms are readily observed (Figure 2A-B, Figure S1). The evenly spaced, successive mass shifts of ~150 kDa correspond to the binding of 1, 2, and 3 intact IgGs to the S-trimer. The particle distributions for each of the Abs is summarized as a heat map in Figure   2C.
Our measurements reveal that Abs targeting the S-trimer can bind with a variety of preferred stoichiometries. Interestingly, none of the tested Abs exhibited a preference for the "complete" 3:1 (IgG:S-trimer) stoichiometry given the symmetry of the S-trimer. One may predict that these binding differences simply reflect different affinities of each Abs. Indeed, the two tested Abs with the lowest observed binding stoichiometries (COVA1-25 and COVA1-21) also have the weakest  The data clearly reveal that the "complete" 3:1 binding is not achieved for either Ab. COVA2-15 preferably binds two IgGs, whereas just one COVA1-18 binds to the S-trimer. Increasing concentrations of Ab do not change the preferred binding stoichiometries ( Figure S3). Binding of both Abs to the S-trimer was also measured by CD-MS and very similar binding behaviour was observed, further illustrating the complementarity between MP and CD-MS ( Figure S4). The low-abundance signals observed between 1200 and 1600 kDa originate from Ab-binding induced S-trimer dimers. (C) Fractional occupancies of each IgG-bound S-trimer species for a panel of twelve monoclonal Abs. A large diversity of binding stoichiometries are observed, ranging from 0 to 2. None of the tested Abs exhibited a preference for 3:1 binding. Additional representative MP histograms are depicted in Figure S1. A tabulation of binding stoichiometries related to these data are provided in Supplemental Table S2.
To help delineate other factors that may be modulating these stoichiometries, we next produced and evaluated Fab fragments and measured their binding to the S-trimer. Unlike the IgGs of each Ab, Fabs are only capable of binding one copy of an antigen (i.e. no avidity effects are possible), and due to their smaller size the contributions of steric clashes on the observed binding behaviour is minimal. These Fab experiments closely mimic previously reported analyses performed by single particle EM, where binding of Fab fragments was monitored. 11,[25][26][27] It is important to emphasize that while Fab fragments can clearly serve as a useful in vitro analogue, it is the intact IgG that is the biologically relevant species during the human immune response.

COVA2-15 and COVA1-18.
For these subsequent investigations, we focus specifically on two Abs: COVA2-15 and COVA1-18. These Abs, which both target epitopes on the RBD, were chosen firstly for their clinical relevance as both are among the most highly potent amongst the tested Abs in neutralizing the Wuhan SARS-CoV-2 strain, possessing near-identical neutralization potencies (IC50 ~0.008 μg/mL). 12 COVA1-18 has also been shown to protect cynomolgus macaques from high dose SARS-CoV-2 challenge. 10 Secondly, despite these similar efficacies, our results indicate that these two Abs exhibit quite distinct (and representative) binding stoichiometries: COVA2-15 exhibits a preference for a 2:1 stoichiometry (with particles corresponding to 1, 2 or 3 bound IgGs, Figure 2A), whereas COVA1-18 displays a preference for 1:1 binding (with particles corresponding to 0, 1 or 2 bound IgGs, Figure 2B). In other words, the binding stoichiometries of these two Abs appear uncorrelated to both affinity and neutralization potency.
The binding behavior of the COVA1-18 and COVA2-15 Fabs differ substantially from that of their corresponding IgGs. When added in excess, the clearly observed mass shift reveals a preference for 3:1 binding for the COVA2-15 Fab by both MP (Figure 3C) and CD-MS ( Figure  3G-H)greater than the 2:1 seen for the full IgG. This stoichiometry agrees well with recent EM structures of the S-trimer bound to COVA2-15 Fabs, in which electron density for three bound Fabs was reported, and is in line with all three RBD copies of the S-trimer being occupied. 12 Titration of COVA2-15 Fab at lower concentrations produce species of intermediate mass, corresponding to binding stoichiometries lower than 3:1 (Figure 3A-B). Interestingly, the COVA1-18 Fab exhibited essentially no binding to the S-trimer even when added in excess ( Figure 3F), in contrast to the COVA1-18 IgG that revealed 1:1 binding (Figure 2B). This poor binding may explain why previous attempts to obtain a cryo-EM structure of COVA1-18 with the S-trimer using Fabs were unsuccessful (Andrew Ward, personal communication).  In the case of COVA1-18, the Fab displays substantially less binding than its corresponding IgG. This dramatic affinity loss going from intact IgG to Fab fragment is a hallmark of avidity (bivalent interactions). 21,26,45 The possibility of avidity in the neutralization potency of COVA1-18 has recently been suggested, with measured KD and pseudovirus IC50 values of the Fab more than 1 and 2 orders of magnitude worse, respectively, when compared to the full IgG. 10 In the context of viral spike proteins, the bivalent IgGs can theoretically bind in two distinct modes: interspike (bridging between two different spike trimers) or intra-spike (binding two domains on the same spike). 21 Distinguishing between different binding modes using standard biochemical assays that only monitor ensemble-averaged binding behaviour is not straightforward. By comparison, the mass measurements presented here readily allow differentiation of the two scenarios by their unique stoichiometries: intra-spike binding will produce Ab-bound species containing only one S-trimer, whereas inter-spike binding will produce species that will contain two S-trimers. Returning to Figure 2, the prominence of the [S + 1 Ab] species suggests that the intra-spike binding mode is the more prevalent mode for COVA1-18, although some signals in the 1200 to 1600 kDa range can be observed (which are absent in both the isolated S-trimer and in the presence of Fabs), suggesting a minor contribution of inter-spike binding is also possible. The lower-than-expected 1:1 binding stoichiometry seen in the COVA1-18 IgG then likely corresponds to a single Ab occupying two RBD binding sites on a single S-trimer due to bivalent binding ( Figure 4B). Higher binding stoichiometries may then be inhibited due to the single available RBD site remaining (i.e. intra-spike binding is no longer possible).
For COVA2-15, the scenario is different as its Fab shows a higher binding stoichiometry than its corresponding IgG. One possibility is that binding of an initial IgG hampers the subsequent binding of additional IgGs (i.e. anti-cooperativity, Figure 4D). Considering that the smaller COVA2-15 Fab readily binds with the full 3:1 stoichiometry, the most likely source of this behaviour in this scenario would be steric clashes arising from the full IgG(s) that occlude the COVA2-15 IgG from fully occupying all three RBD sites. An alternative possibility is that COVA2-15, like COVA1-18, may also be capable of S-trimer binding via intra-spike crosslinking.
In this scenario, one COVA2-15 IgG would bind bivalently to two RBD sites, while the remaining RBD site is occupied by a second, monovalently bound IgG ( Figure 4E). This arrangement would also appear as a 2:1 binding stoichiometry, albeit with a different spatial configuration. Unlike COVA1-18, where avidity is a prerequisite for binding, in this arrangement COVA2-15 would seemingly not depend on this avidity to maintain affinity for the S-trimer, as evidenced by the binding capability of the COVA2-15 Fab (Figure 3A-C). Given that a small population of a 3:1 stoichiometry is observed for the COVA2-15 IgG (Figure 2A), it is likely that there exists a contribution of Fab-like, "monovalent-only" binding ( Figure 4D) even if bivalent binding is the dominant binding mode (Figure 4E). Taken together, these results highlight the rich complexity inherent to IgG -S-trimer interactions, and the capacity of single particle analyses to aid in this unraveling this complexity.  Figure S2). In the absence of any Ab, the S-trimer readily binds ACE2, with a predominant 1:1 stoichiometry at low mixing ratios as detected by MP (Figure 5D). MP measurements at higher ACE2 concentrations were partially impeded by spectral interference caused by a sub-population of a tetrameric ACE2 state which is of comparable mass to the free S-trimer (~400 kDa vs. 477 kDa), although the species corresponding to ACE2-bound S-trimers remain unobstructed ( Figure S5). While these species

Virus Variants of Concern.
There is ongoing concern that newly emerging strains of the SARS-CoV-2 virus harboring additional mutations in the S protein may negatively impact the potency of already-existing anti-SARS-CoV-2 monoclonal Abs. 46

Discussion
We demonstrate here the unique application of two single particle approaches, MP and CD-MS, for interrogating the interaction stoichiometries between full Abs, the ACE2 receptor, and the SARS-CoV-2 S protein ectodomain. We find that different Abs can exhibit surprisingly distinct binding behavior. In the case of the potent neutralizing Abs COVA2-15 and COVA1-18, different binding stoichiometries can arise despite commonly targeting the RBD and having identical neutralization potencies. This behaviour is not fully recapitulated when analyzing the binding of Fab fragments, stressing the necessity of studying Ab-antigen interactions in the context of the full, non-truncated IgG. Our results highlight the complex interplay of affinity, avidity, and anticooperativity effects in these interactions, and the capability of single particle mass analysis to shed light on these co-occurring phenomena.
Our analyses here focus primarily on the binding behaviour of the two representative neutralizing Abs COVA2-15 and COVA1-18. One may wonder if the determinants of the 1:1 and 2:1 binding behaviour that we uncovered for these Abs can be generalized to other anti-S-trimer Abs (e.g. Figure 2C). While it is tempting to speculate, for example, that all 1:1 binding IgGs bind in a manner analogous to COVA1-18 (i.e. bivalently), in reality the situation may be more complex. Other factors such as steric blockage, incompatible angles of approach, or the location of the epitope cannot be dismissed a priori. As such, the binding determinants of each Ab should be determined on a case-by-case basis. Nevertheless, the experimental approaches outlined in this work, especially in combination with already-established methods such as single particle EM, are well-suited to address these questions.
Our investigations were enabled by the capacity of recently developed single particle approaches to overcome the high degree of mass spectral complexity normally brought by the extensive glycosylation of the SARS-CoV-2 S protein. We expect that these technologies will open the door for studies into similarly complex biological systems, such as glycoproteins from other viruses and biological agents. We foresee that these techniques will be especially useful in the characterization and rational design of biotherapeutics, e.g. monoclonal Ab cocktails or multivalent nanobodies. 51,52 It is anticipated that single particle mass analysis will provide a powerful addition to the toolbox of contemporary biophysical methods to study protein-protein interactions.

Acknowledgements.
This and A701V. Both S constructs were produced in HEK293F suspension cells (ThermoFisher) and purified as previously described. 12 For the human ACE2 receptor, soluble ACE2 was generated as described previously 12 by using a gene encoding amino acids 18-740 of ACE2. The IgGs and Fab fragments used in this study were produced as previously described. 12 Nitrogen was used as collision gas. The in-source-trapping voltage and HCD voltage were optimized for maximal ion transmission. After multi scan acquisition, .RAW files were centroided and converted into mzXML format for processing as previously described. 39    The data clearly reveal that full stoichiometric binding is not achieved for either Ab, but also that COVA2-15 preferably binds two Abs, whereas even at excess preferably just one COVA1-18 binds to the S-trimer. As expected, lower mixing ratios result in lower observed binding stoichiometries. The low-abundance signals observed between 1200 and 1600 kDa originate from Ab-binding induced S-trimer dimers. The measured masses and abundances related to these data are provided in Supplemental Table S4.  Figure 2, and also reveal that full stoichiometric binding is not achieved for either Ab, but also that COVA2-15 binds more readily two Abs, whereas even at access preferably just one COVA1-18 binds to the S-trimer. The measured masses and abundances related to these data and provided in Supplemental Table S4.
COVA2-15 (IgG) COVA1-18 (IgG) Figure S5. MP histograms of ACE2 binding to the S-trimer at increasing ratios. Even at an excess of 4:1, a substantial number of S-trimers still only bind one ACE2, and essentially no binding of 3 ACE2 is observed. Interference from a tetrameric ACE2 sub-population occlude accurate quantitation of the free S-trimer.