Evidence of variable human Fcγ receptor-Fc affinities across differentially-complexed IgG

ABSTRACT

Antibody-mediated effector functions are widely considered to unfold according to an associative model of IgG-Fcγ receptor (FcγR) interactions. The associative model presupposes that Fc receptors cannot discriminate antigen-bound IgG from free IgG in solution and have equivalent affinities for each. Therefore, the clustering of Fcγ receptors (FcγR) in the cell membrane, cross-activation of intracellular signaling domains, and the formation of the immune synapse are all the result of avid interactions between the Fc region of IgG and FcγRs that collectively overcome the individually weak, transient interactions between binding partners. Antibody allostery, specifically conformational allostery, is a competing model in which antigen-bound antibody molecules undergo a physical rearrangement that causes them to stand out from the background of free IgG by virtue of greater FcγR affinity. Various evidence exists in support of this model of antibody allostery, but it remains controversial. We report observations from multiplexed, label-free kinetic experiments in which the affinity values of FcγR were characterized for covalently immobilized, captured, and antigen-bound IgG. Across the strategies tested, receptors had greater affinity for the antigen-bound mode of IgG presentation. This phenomenon was observed across multiple FcγRs and generalized to multiple antigens, antibody specificities, and subclasses. Furthermore, the thermodynamic signatures of FcγR binding to free or immune-complexed IgG in solution differed when measured by an orthogonal label-free method, but the failure to recapitulate the trend in overall affinity leaves open questions as to what additional factors may be at play.

KEYWORDS:

• Affinity
• allostery
• antibody
• Fc receptor
• immunoglobulin G

Introduction

The separation of antibody molecules into highly variable antigen-binding fragments (Fab) and the comparatively constant crystallizable fragment (Fc) by a flexible hinge allows for the linkage of the adaptive immune system, with its ability to generate diversity in antigen recognition, to the effector activity of the innate immune system. In the latter half of the 20^th century, at least two models were advanced to explain what mechanisms linked antigen binding to effector function.¹ Historically, the associative model, in which an immune complex consisting of multiple antibodies crosslinks the receptors of an innate immune cell, has achieved widespread acceptance. These collective associative interactions overcome the weak binding of individual antibody molecules to individual receptors while also clustering the receptors within the cell membrane, allowing them to crosslink and initiate the signaling process required for effector activity. The process of receptor clustering and crosslinking is generally considered to be the result of many brief, transient interactions between antibody and receptor, both antigen-bound and those still in solution, that occur until the contemporaneous binding of two or more receptors results in an avid interaction, increasing the tendency of the immune complex to remain anchored to the cell surface. One underlying assumption of this model is that antigen-bound and free antibody are functionally identical from the perspective of the cell surface receptors. The likelihood of an avid interaction taking place could be improved if the receptors were already proximally localized rather than evenly distributed across the cell membrane, but even so, the binding interactions of an antibody to antigen via its variable (Fv) domain and to a receptor via its Fc domain are considered wholly independent.

The associative model leaves several questions unanswered, however. Among them is how antigen-bound IgG molecules can overcome the enormous excess of nonspecific antibody to occupy receptors simultaneously. Circulating IgG can reach concentrations as high as 100 μM,^2,3 orders of magnitude greater than the equilibrium dissociation constants (K_D) of the low-affinity FcγRs. As a result, FcγR are expected to be saturated by serum IgG at steady state. Compounding this is the fact that each effector cell may carry as few as 50,000 copies of an FcγR on its membrane,^4,5 although these may be clustered in lipid rafts⁶ rather than evenly distributed across the surface of the cell. The ability of high levels of antibody in solution to impede effector cell activity is well documented in assays performed in vitro, where it is known as the hook or prozone effect. After a certain optimal concentration, the system becomes saturated: all molecules of the antigen and all FcγR are bound by antibody, but progressively fewer are bridged by the same antibody; immune-complexed IgG can no longer locate vacant receptors on the effector cells and antibody effector activity wanes. Experimentally, waning activity is observed far below the levels of antibody present in serum.

The shortcomings of the associative model, in which high avidity serves as a simple substitute for high affinity, were addressed by a competing allosteric model, which proposed that perturbations of the antibody such as the act of binding to antigen could transmit intramolecular changes throughout the antibody, even across the hinge and into the opposing fragment, and result in changes to the intrinsic monovalent affinity for binding partners.^7,8 Substantial evidence exists for a subset of antibody allostery, termed configurational allostery, which results from permanent modifications to the amino acid sequence. For example, point mutations within the Fc, even at the very C terminus, can affect the kinetics of FcγR binding at the upper C_H2 domain.^9,10 Similarly, IgG glycan changes at a conserved site in the C_H2 domain can change the propensity of IgG to hexamerize through its C_H3 domain.¹¹ Configurational changes can also span the hinge region, as in the case of a change in isotype altering an antibody’s affinity for its antigen.^12–15

In comparison, under the conformational allostery model, temporary changes that alter IgG interactions with FcγR occur upon antigen binding. While evidence of conformational allostery is limited, glimpses of it can be seen in the literature dating back decades and continue to accumulate.^16–20 To this record, we contribute observations using HIV as a model system, for which particularly extensive and detailed information regarding target viral envelope antigen and cognate antibodies is available. The resultant measurements of FcγR affinity for differentially liganded IgG offer further insights into what might be considered immunological “spooky action at a distance”.

Results

Improved FcγR affinity for antigen-bound IgG measured by surface plasmon resonance

To investigate the possibility that binding events at one site of an antibody affect FcγR binding affinity at a second site, we adopted a multiplexed surface plasmon resonance (SPR) approach. Using a microfluidic printing system, we produced single SPR sensors containing covalently immobilized antibodies of interest, antigen, and capture reagents in discrete spots that could be simultaneously loaded with the antibody of interest within the SPR flow cell (Figure S1). This platform allowed simultaneous evaluation of FcγR affinity for IgG immobilized in a variety of modes and with numerous replicates. For initial experiments, we used C11,²¹ a monoclonal IgG1 antibody that targets an epitope on the HIV-1 envelope glycoprotein that is induced upon binding to the human CD4 receptor (CD4i epitope). Such epitopes can be replicated in vitro using a CD4-envelope fusion protein (full-length single chain – FLSC)²² as antigen. As shown in representative sensorgrams (Figure 1), the soluble extracellular domains of human FcγRs rapidly associate with each of the configurations of antibody found on the sensor and dissociate just as quickly when the flow of receptor is replaced with buffer. This fast-on, fast-off profile is typical of these low-affinity interactions but can complicate kinetic fitting, so despite not observing saturation under all conditions, the equilibrium dissociation constants (K_D) of the receptors were calculated from the magnitude of responses at equilibrium across a titration of FcγR. It is readily apparent even prior to fitting that the interaction of FcγRs with antigen-bound IgG is qualitatively different from other conditions by virtue of the response reaching a plateau within the range of FcγR concentrations tested. This profile demonstrates that FcγR binds C11 with higher affinity when C11 is bound to antigen than either when it is directly conjugated to the sensor or when it is captured by its Fab region using an anti-Fab capture reagent.

Figure 1. Antibody-FcγR association and dissociation profiles by SPR across assay configurations.

Binding and dissociation isotherms demonstrating higher affinity binding of FcγR to antibody that is complexed with antigen as compared to printed or otherwise captured on the sensor surface.

Representative SPR traces for binding of FcγRIIa-R131 (top) and FcγRIIIa-V158 (bottom) to C11 IgG that was covalently coupled (left), captured with an anti-Fab reagent (center, blue), or by cognate antigen (right, green). For each assay configuration and receptor pair, raw data (left) and equilibrium fits (right) are shown.

While the overall level of response varied based on how densely each region was printed with ligand during the conjugation process, replicate regions yielded consistent measurements of K_D. Across all tested FcγRs (FcγRIIa, FcγRIIb, and FcγRIIIa), antigen-bound C11 showed an improvement in binding affinity as compared to when it was covalently printed on the sensor surface without forming complexes (Figure 2a). Binding profiles of N5-i5, another CD4i-specific IgG1 antibody, showed a similarly broad improvement in FcγR binding affinity in the context of antigen-recognition, generalizing this effect to multiple variable domains (Figure 2b).

Figure 2. Antibody affinity for FcγR across assay configurations.

Dot plot showing the affinity of each receptor-antibody pair across repeated experiments. Antigen-complexed antibody often displays a distinct improvement in affinity for FcγR.

Equilibrium binding affinities of C11 (a), N5-i5 (b), N49p7.2 (c), N49p9 (d), and VRC01 IgG3 (e) specific for two different antigens (FLSC – top; BaL – bottom) and two IgG subclasses (IgG1 and IgG3) across up to five low affinity human FcγR when covalently coupled (black), captured with an anti-Fab reagent (blue), or by antigen (green) on the SPR chip. Each dot represents an affinity calculated from an independent SPR spot across an FcγR titration series. The statistical significance of the differences between log₁₀ transformed affinities measured between each capture method and the covalently printed antibody were defined using a mixed-effect model adjusted using Dunnett’s multiple comparisons test (*p < 0.05, **p < 0.005, ***p < 0.0005).

Affinity enhancements generalize across distinct antigen targets and activating IgG subclasses

Although we observed antigen binding to correspond with an improvement in FcγR binding affinity for two distinct antibodies, we could not rule out whether some unknown property specific to the FLSC antigen accounted for these differences. We therefore conducted follow-up experiments using gp120 envelope glycoprotein from the BaL strain of HIV-1 with clonally related IgG1 monoclonal antibodies (mAbs) N49P7.2 and N49P9 (Figure 2c, d), which are broadly neutralizing antibodies (bnAbs) recognizing the CD4 binding site (CD4bs) on gp120.²³ While fold-changes in affinity were less striking and the effect observed less broadly, once again both antibodies showed improved affinity toward most FcγRs when complexed with antigen.

Finally, given this apparent linkage between Fv and Fc domains across the IgG1 hinge, we tested a subclass switched IgG3 variant of the antibody VRC01, another bnAb specific for the CD4 binding site. Human IgG3 has a substantially longer hinge than that of IgG1 and it can contain up to 11 paired cysteine residues, also having the effect of making it rigid.^24,25 Intriguingly, despite its extended and more rigid hinge, modest affinity increases up to 2-fold were observed for antigen-complexed VRC01-IgG3 as compared to unbound, and the largest effect was observed for the higher affinity FcγRIIIa (Figure 2e).

An anti-Fab capture reagent spot was included as a comparator to define whether changes in receptor affinity observed across five antibodies, two subclasses, and against two antigens might also be observed when other epitopes in the Fab are bound. In theory, this Fab capture condition could exclude a damaging effect of direct conjugation via lysine side chains as the source of altered FcγR affinity. Alternatively, if binding of the Fab domain of the antibody at any site, rather than just via complementarity-determining regions (CDRs), drove stabilization of a favored conformation, improved the orientation of presented Fc, altered surface charge or other biophysical properties, or captured some other consistent effect of general ligand binding, then this capture strategy should reproduce this effect.

However, in contrast to antigen binding, the effect of binding to an anti-Fab capture reagent was considerably more variable (Figure 2). None of the FLSC-specific antibodies showed a statistically significant difference in FcγR binding when captured by anti-Fab as compared to direct conjugation to the sensor surface (Figure 2a–b). BaL-specific antibodies N49P7.2 and N49P9 showed several statistically significant differences, although these tended to be more modest in magnitude (Figure 2c–d). The affinity of VRC01 IgG3 for FcγR when bound to an anti-Fab reagent was more dramatically modified than the IgG1 antibodies tested (Figure 2e). However, these were still not as pronounced as when VRC01 IgG3 was bound to antigen.

Alongside the Fab-captured and antigen-bound formats, the sensor chips also included staphylococcal protein A as an alternative means of immobilizing IgG. The binding site of this bacterial Fc receptor, which does not compete with endogenous FcγRs, sits at the interface of the heavy chain C_H2 and C_H3 domains²⁶ and as such it is not separated from the FcγR binding site by the hinge. The profile of the IgGs tested, including IgG3, that had been immobilized by protein A bore the greatest resemblance to the version captured by the Fab arms (Figure 3); in fact, the only statistically significant differences occurred in cases where protein A-bound IgG had slightly weaker interactions with FcγRs. The alternate interactions used to capture and present each IgG, whether by recognition of the Fab or alternative epitopes in the Fc domain, failed to consistently recapitulate the antigen-binding effect, which could be broadly reproduced across multiple antibodies, antigens, and receptors, though it was not uniform in magnitude (Figure S2 and Table S1); the most striking improvement in affinity for antigen-bound IgG exceeded 10-fold.

Figure 3. Evaluation of an alternative Fc domain capture strategy.

Dot plot showing the affinity of each receptor-antibody pair across repeated experiments. Antibodies captured by Protein A or an anti-Fab capture reagent generally do not show differences in affinity for FcγR.

Equilibrium binding affinities of C11 (a), N5-i5 (b), N49p7.2 (c), N49p9 (d), and VRC01 IgG3 (e) specific for two different antigens (FLSC – top; BaL – bottom) and two IgG subclasses (IgG1 and IgG3) across up to five low affinity human FcγR when captured with an anti-Fab reagent (blue), or by Protein A (purple). Each dot represents an affinity calculated from an independent SPR spot across an FcgR titration series. Statistical significance of differences between log₁₀ transformed affinities observed between capture methods were defined using a mixed-effect model adjusted using Dunnett’s multiple comparisons test (*p < 0.05, **p < 0.005, ***p < 0.0005).

To better compare the site-specificity of antibody ligands on FcγR affinity and eliminate antibody damage during chip conjugation as a confounding factor, comparisons were made among the three distinct capture strategies. When comparing antigen, protein A, or Fab capture, the antigen-bound condition produced the largest and most general enhancement of FcγR binding affinity (Figure 4). Apart from the lower affinity of VRC01 IgG3 interactions with FcγRIIa-R131 and FcγRIIIa-V158 when bound to Pro A, statistically significant differences between protein A and Fab capture were not observed. In contrast, most antigen-antibody-receptor combinations showed statistically significant increases in affinity, and the magnitude of these increases still exceeded a 10-fold enhancement (Figure 4).

Figure 4. Improved affinity for FcγR is unique to antigen-bound IgG.

Volcano plot depicting the statistical significance and change in affinity to FcγR for antibody bound to antigen or to Protein A. Antibodies bound to antigen show statistically significant improvements in affinity to each FcγR tested.

Volcano plot depicting the fold change in equilibrium binding affinity for FcγR versus statistical significance observed between each Fc capture strategy (antigen – green; protein A – purple) and the anti-Fab capture reagent.

Affinity differences are recapitulated for immobilized Fcγ receptor in biolayer interferometry

To further evaluate this phenomenon and mitigate the potential of method-specific artifacts, biolayer interferometry (BLI) experiments were conducted in which FcγR rather than antibody was immobilized on the solid substrate, and free or complexed antibody was in the solution phase. FcγR immobilization was achieved using site-specific biotinylation. Free VRC01 antibody in both IgG1 and IgG3 forms showed the fast on and fast off interactions characteristic of binding to FcγR (Figure 5). In contrast, when VRC01 IgG1 and IgG3 were first complexed with either antigen or the same anti-Fab capture reagent used in SPR experiments, both on and off rates were noticeably slowed, and up to 10-fold improvements in affinity for FcγR were observed (Table S2). While decreased on rates might be anticipated from the slower diffusion and larger hydrodynamic radius of complexed IgG, the off rates were remarkably slowed, revealing apparently stable interactions with FcγR. In an effort to exclude the possibility that the equimolar combinations of antibody and antigen or (in particular) anti-Fab used might have allowed the formation of larger “daisy chains” of complex, the experiment was repeated with 10-fold molar excesses of antigen and capture reagent and the same trend was observed (Figure S3). Thus, the improved affinity for FcγR observed for antigen-bound antibody was recapitulated using an additional method and with an inverted orientation of the antigen-IgG-FcγR complex. Notably, despite the inverted format of the analyte and stationary ligand, K_D values observed for IgG1 in BLI and SPR methods were within 2.5-fold of each other (Table S2).

Figure 5. Antibody-FcγR association and dissociation profiles by BLI across assay configurations.

Binding and dissociation isotherms demonstrating higher affinity binding and dramatically slower dissociation of antigen-complexed IgG to FcγR on biosensor tips as compared to IgG in solution.

BLI traces for binding (0-180s) and dissociation (180-360s) of VRC01 IgG1 (left) and IgG3 (right) in solution (top), complexed with an anti-Fab reagent (center), or with cognate antigen (bottom) to immobilized FcγRIIIa-V158. For each assay configuration and receptor pair, raw data (left) and equilibrium fits (right) are shown. Dotted vertical lines indicate equilibrium binding affinity (K_D).

Thermodynamic characterization by isothermal titration calorimetry

SPR and BLI suffer from several well-established limitations, including mass transport and sensitivity to bulk shifts caused by buffer formulation that can obscure the true dynamics of the analyte–ligand interaction.^27,28 Consequently, despite robustness across diverse spot densities, antibodies, and FcγR, we were unable to rule out a technical explanation for the increases in FcγR affinity for antigen-bound IgG that we observed. We turned to isothermal titration calorimetry (ITC) as a completely solution-based, label-free alternate method for evaluating the affinity of FcγRs for differently presented versions of IgG (Figure S4). We elected to solely test FcγRIIIa-V158 because its greater overall affinity made the method more tractable by requiring a smaller amount of material to achieve an appropriate ITC C value, and because this receptor typically showed the greatest change in affinity throughout the ITC and BLI experiments.

The baseline affinity of FcγRIIIa for free VRC01 IgG1 measured approximately 150 nM (Figure 6). Although this is a 20-fold improvement relative to what was measured for the covalently immobilized antibody that reflects the closest condition in SPR, it was perhaps unsurprising that some discrepancy should exist. After all, the process of covalently linking an IgG to the hydrogel matrix involves the formation of amide bonds between lysine residues and the substrate, which introduces opportunities for conformations that do not mimic IgG in solution. The rank orders of the interactions also differed from the surface-based methods however, and the antigen-bound format had the weakest binding interaction with the receptor, measuring 547 nM ± 41 nM (Figure 6).

Figure 6. Thermodynamic properties as measured by isothermal titration calorimetry.

Bar graph of thermodynamic properties of IgG: FcγR complexes. Antigen-bound IgG exhibits decreased enthalpy.

The thermodynamic parameters of the interaction between the high-affinity Val158 allele of FcγRIIIa and free IgG1 (black), anti-Fab captured (blue), or the antigen-bound configuration (green). The calculated equilibrium dissociation constants (K_D) are included parenthetically, with standard errors of the mean, at right.

The thermodynamic signatures used to calculate the K_D values for each mode of interaction may offer insight into the mechanics underlying Fc-FcγR interactions. Broadly, there were no major qualitative differences in thermodynamic signatures between the three presentations of IgG tested. Each was characterized by favorable binding enthalpy (ΔH) and unfavorable entropic contributions (TΔS). This profile suggests that binding is driven largely by hydrogen bonds and van der Waal forces with limited hydrophobic contributions, all of which is consistent with the structural basis for the IgG1-FcγRIIIa complex.²⁹ What is interesting to note, however, is that the interaction involving the antigen-bound IgG had the greatest release of enthalpy and simultaneously the least favorable entropic component. In comparison, free and C_H1-captured IgG had similar values for ΔH and TΔS despite a greater than 2-fold difference in affinity separating them.

Discussion

In the past decade, a number of studies have provided biophysical and biochemical evidence in support of the allosteric conformational model, in which protein binding or sequence changes at one end of the antibody molecule are propagated to the other and result in changes in phenotypes. Such studies of antibodies are part of a broader literature with increasing methodological sophistication. For example, distal phenotypic effects have been observed from application of magnetic fields to drive surface charge redistribution,³⁰ or based on linkage to surfaces through enantiomeric forms of chiral molecules, or due to binding of charged ligands in pockets distal to but propagated to flexible tags.³¹ Likewise, activation of photosensitive small-molecule groups that change electron polarization has been observed to induce phenotypic changes in the protein, including affinity for binding partners and catalytic properties.³² A number of these experimental frameworks demonstrate how molecular charge distributions can exert long-range effects on molecular properties such as binding affinity.³⁰

The experimental techniques showing distal effects include circular dichroism,³³ nuclear magnetic resonance,³⁴ fluorescence correlation microscopy,²⁰ hydrogen-deuterium exchange mass spectrometry (HDX),^20,35 electron microscopy,³⁵ and crystallography,³⁶ and are further supported by extensive molecular dynamics simulations.^37,38 Relevant to observations in this study, crystallography studies have also demonstrated that distal loops in the CH1 domain exist in different conformations when antibodies are free versus bound to antigen,³⁹ and can be transmitted through linkers thought to be passive.⁴⁰ Similarly, changes in both crystal structures and the relative population of distinct conformational states of IgG related to factors such as post-translational modification have been widely reported.^41–46

Our observations in diverse modes and designs of label-free kinetic experiments contribute to this rich but still unorthodox body of literature by demonstrating that increases in the affinity of Fcγ receptors for antigen-bound IgG compared to the same antibody immobilized by alternative means appear to be a common phenomenon that cannot be adequately explained by antibody specificity, target antigen, or even the subclass of IgG involved. However, the absence of such a trend when using isothermal titration calorimetry as an orthogonal method leaves open the possibility that these observations reflect a technical artifact related to surface charge, protein density as related to antibody hexamerization driven by high local concentrations,⁴⁷ or other unknown factors. Differences in sensors’ surface characteristics have been associated with >10-fold shifts in the observed affinity values,^48–50 and an analogous situation may indeed account for the discordant K_D values measured between methods (up to 16-fold) in this work, but the disparities within a method are not so readily explained. Caveats to interpretation include: 1) the use of equilibrium fits in the calculation of affinity values given that saturation of binding was not reliably reached as a somewhat general challenge of characterizing relatively low-affinity interactions, and 2) the dramatically different kinetics observed between SPR and BLI data sets. Whether this distinction relates to changes in instrumentation, the inversion of analyte and surface ligand, or other factors is unknown, but is worth considering when interpreting results from different experimental techniques.

The work presented here has placed an emphasis on generalizing the conduct of the antibodies studied rather than comprehensively isolating how each feature of an IgG molecule might potentially contribute to allostery. For example, only one IgG3 molecule was evaluated, and it included a different variable domain than the IgG1s tested, preventing resolution of the impact of each of these concomitant changes. Tests were conducted with broadly neutralizing HIV-specific antibodies, which are known to have unusual features, such as more extensive degrees of mutation,⁵¹ and the antigens tested were recombinant rather than natively presented on the surface of a viral particle. The low affinity of the Fc-FcγR interaction led to increased ambiguity in data fitting quality for modeling affinity, and the magnitude of affinity enhancements was not always large, though results from affinity engineering studies have suggested that the 2–10-fold improvements typically observed here may well be biologically relevant.

The selection of ITC as an orthogonal method was motivated by the desire to measure all the protein constituents in solution, though ITC is not without drawbacks of its own. Immobilizing one binding partner in a hydrogel matrix or other surface, as is the case for SPR and BLI, could better reflect the situation in vivo wherein the receptors, and in many cases antigen, are anchored in a cell membrane. In comparison, ITC permits the tumbling of the molecules, which is particularly a concern in the case of the bivalent anti-Fab reagent since this could lead to the formation of “daisy chains” of antibody and antigen, something that cannot occur when the capture reagent is affixed to a substrate. What ITC does offer though is the ability to evaluate more than just the strength of the association and dissociation of the protein–protein complexes. The relative contributions of enthalpic and entropic terms across the various modes of binding tested here suggest that there is more at work than the comparatively narrow range of K_D values may initially suggest. The interaction between FcγR and antigen-bound IgG consisted of the largest release of enthalpy and the smallest contribution of entropy of the formats tested. Differences in enthalpy and entropy parameters raise the possibility that the Fc domain can exist in conformations that are differentially well-suited or “pre-organized” to interact with FcγR, or alternatively, may exhibit differential entropic costs. The distinctions in results between and methods add a layer of experimental complexity to interpretation of in vitro binding assays that extends beyond IgG complex state, the experimental variable that we sought to focus on, between methods. SPR, BLI, and ITC approaches present different strengths and limitations. ITC requires considerable sample, but provides thermodynamic insight. BLI and SPR offered superior throughput but require immobilization. While insights from each method have been shown to reliably relate the behavior of protein–protein complexes in more complex settings ranging from in vitro activities in cell-based assays to in vivo outcomes in human and animal models, the distinctions observed here and elsewhere among them raise questions as to both quantitative and qualitative comparability.

For all the uncertainty surrounding it, conformational allostery is a model well worth considering, particularly given the interesting questions its existence prompts. For example, how would changes be reliably maintained, lost, or evolved in the context of class switch recombination and affinity maturation and the course of an immune response? How might bivalent versus monovalent antigen binding affect conformational states? It should be noted that not all studies conducted to date that have observed changes upon antigen binding have in fact measured improved affinity. Recent structural insights into the B cell receptor (BCR)^52–54 may suggest answers as to whether similar allosteric or charge networks may be at play in effective signaling through the BCR. In the case of protein A-bound IgG, it is tempting to speculate that the lower affinities of Fcγ receptors measured here compared to the equivalent antigen-bound IgG might contribute to its role as a pathogen defense molecule; the ability to restrict an immunoglobulin molecule from assuming a pose conducive to FcγR binding could certainly serve as a fitness advantage. In sum, while we have come away from this work with experimental data supporting the conformational allostery model for IgGs in the context of simplified experimental systems, we also find this work raises many new questions, while broadly supporting what others have hypothesized – namely that allostery may be a feature of essentially all dynamic proteins.⁵⁵

Materials and methods

Recombinant protein expression and purification

VRC01 (IgG1) was produced by transient co-transfection of 293F human embryonic kidney cells by 25 kDa linear PEI (Polysciences, 23966–1) using the heavy chain and associated kappa light chain provided by the AIDS Reagent Program of the NIH (12035 and 12,036, respectively). For production in an IgG3 backbone, the heavy chain variable region was grafted to a human pFUSE-CHIg construct (Invivogen, pfuse-hchg301) and expressed as above. IgG was affinity purified from cell culture supernatant using an Äkta pure FPLC system (Cytiva) and either MabSelect protein A (Cytiva) in the case of IgG1 or protein G (ThermoFisher, 101242) for IgG3. Following affinity purification, the antibodies were processed further using a phosphate-buffered saline (PBS)-exchanged HiPrep 16/60 Sephacryl S-200 HR size exclusion chromatography column (Cytiva) to purify monomers. C11, N5-i5, N49P7.2, and N49P9 were produced by transient co-transfection of 293F human embryonic kidney cells using the heavy chain and associated kappa light chain before purifying as previously described.^20,23 Human Fcγ receptors were produced as previously described.⁵⁶

Surface plasmon resonance

The ligand conditions depicted in the schematic representation of the SPR assay (Figure S1) were prepared by covalent linkage to a carboxymethyldextran-functionalized biosensor (CMD200M, Xantec Bioanalytics) using a continuous flow microspotter (Carterra) capable of producing 96 discrete regions of interest. The microfluidic pathways were primed with 10 mM sodium acetate (pH 5.0) and activated for 5 minutes with 100 µL of 10.4 mM EDC and 1.4 mM sulfo-NHS (ThermoFisher) prepared in 10 mM MES (pH 5.0). VRC01, BaL gp120 (NIH AIDS Reagent Resource), and goat F(ab′)2 anti-human F(ab′)2 capture reagent (Jackson ImmunoResearch, 109-006-097) were formulated in 10 mM sodium acetate (pH 5.0) at concentrations ranging from 25 nM up to 400 nM and along a 2-fold dilution series of at least two points. Ligand was flowed over the activated regions for 10 minutes, followed by 5 minutes of washing with sodium acetate. The sensorchip was loaded on a PBS + 0.05% Tween 20-primed imaging-based SPR (MX96, IBIS Technologies) and quenched with a 150 µL injection of 1 M ethanolamine (3.0). Five rounds of conditioning that alternated between 25 µg/mL anti-human Fab in sodium acetate and 10 mM glycine (pH 2.0) then followed to remove any non-covalently associated material and provide an estimate as to the binding capacity of the compatible spots.

Fcγ receptors were formulated at 25 µM in PBS + 0.05% Tween20 and serially diluted over an 8-point, 3-fold dilution series. The antibody of interest was made at 25 µg/mL. The instrument was maintained at 25°C throughout the experiment. After a run-in period of six blank buffer injections, the surface was exposed to two consecutive rounds of the antibody of interest over a 5-min association period to create the Fab-captured IgG condition on the anti-human Fab regions and antigen-bound antibody on the envelope glycoprotein spots. The antibody of interest did not interact with the regions that had already been coupled with the antibody itself and subsequently quenched. One buffer injection was used to separate the sensor-loading steps from the FcγR injections. The Fc receptor analytes were flowed from the lowest concentration of the series (approximately 11 pM) to the highest (25 µM) using 5 min of association and dissociation. The weak micromolar affinity of Fcγ receptors for IgG meant that a pair of buffer injections could readily serve as the regeneration conditions for the chip. This process, continuing from the (re)loading of antibody of interest, was repeated until each of the FcγRs had been evaluated.

Initial processing of the SPR data was performed using SprintX (IBIS Technologies). The signal from each region of interest on the sensor was referenced using the nearest unconjugated interspot to account for bulk shift and nonspecific binding. The blank injection immediately preceding each series of receptors was also subtracted from the signal of each of the eight injections within a receptor series and showed negligible dissociation of the IgG from each capture reagent. Affinity values for each receptor-complex pair were calculated in Scrubber 2 (BioLogic Software) using the signal at equilibrium occurring during a 10-s window at the end of the association phase. The maximum response (Rmax) was individually calculated for each region of interest and receptor pairing. Regions with an Rmax of less than 25 were manually discarded because of insufficient signal.

Biolayer interferometry

BirA-biotinylated human FcγRIIIa was loaded onto a streptavidin-coated biosensor (Sartorius) at 0.75 µg/mL in 1× PBS for 5 min. VRC01 was formulated at twice the intended concentration and diluted in an equal volume of an equimolar solution of antigen or anti-human Fab capture, or PBS alone. The tested concentrations ranged from 5 µM to approximately 10 nM over a series of 1:3 dilutions. Association of the complex was measured over a 3-min period followed by 3 min of dissociation. The plate containing the analytes was maintained at 30°C by the instrument over the course of the experiment. The affinity values for each interaction were calculated as EC₅₀ values in GraphPad Prism (version 9.4, Dotmatics) using the average signal measured in a 10-s window at the end of the association phase and a four-parameter log(agonist) vs response model.

Isothermal titration calorimetry

Complexes were formed from equal volumes of 20 µM solutions containing VRC01 in an IgG1 heavy chain and either antigen or anti-human C_H1 capture reagent. For the free IgG condition, VRC01 was instead diluted 1:2 in 1× PBS. FcγRIIIa-V158 at 100 μM was injected using 6 μL per injection and injections spaced 90 s apart. These concentrations of macromolecules and FcγR were chosen based on the theoretically complete formation of 10 µM complex following the mixture of 20 µM reagents. This would yield an ITC c value of 10 or greater when calculated as c = n * K_a * [macromolecule], where n is the stoichiometry of the ligand for the macromolecule and Ka is the association constant of the ligand. Temperature was maintained at 25°C and the sample cell mixed at 750 rpm. Measurements were made with a reference power of 6.0 using a MicroCal PEAQ-ITC (Malvern Panalytical).

Abbreviations

BCR	=	B cell receptor
BLI	=	biolayer interferometry
bnAb	=	broadly neutralizing antibody
CD4bs	=	CD4 binding site
CD4i	=	CD4 binding-induced epitope
CDR	=	complementarity-determining region
EC50	=	effective concentration midpoint
Fab	=	antigen-binding fragment
Fc	=	crystallizable fragment
FcγR	=	Fcγ receptor
FLSC	=	full-length single chain (HIV env – CD4 fusion)
Fv	=	variable fragment
HDX	=	hydrogen-deuterium exchange mass spectrometry
ITC	=	isothermal titration calorimetry
KD	=	equilibrium dissociation constant
mAb	=	monoclonal antibody
SPR	=	surface plasmon resonance

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2023.2231128

Additional information

Funding

This work was supported in part by NIAID P01AI162242.