|Home | About | Journals | Submit | Contact Us | Français|
Virus neutralization is governed by the number of antibodies that bind a virion during the cellular entry process. Cellular and serum factors that interact with antibodies have the potential to modulate neutralization potency. Although the addition of serum complement can increase the neutralizing activity of antiviral antibodies in vitro, the mechanism and significance of this augmented potency in vivo remains uncertain. Herein, we show that the complement component C1q increases the potency of antibodies against West Nile virus by modulating the stoichiometric requirements for neutralization. The addition of C1q does not result in virolysis, but instead reduces the number of antibodies that must bind the virion to neutralize infectivity. For IgG subclasses that bind C1q avidly, this reduced stoichiometric threshold falls below the minimal number of antibodies required for antibody-dependent enhancement (ADE) of infection of K562 cells expressing Fc-γ receptors (CD32), and explains how C1q restricts the ADE of flavivirus infection.
The development of antiviral antibodies is a critical aspect of protection against viral infections. The mechanisms of antibody-mediated neutralization have been investigated for many animal viruses, and can be characterized as a “multiple-hit” phenomenon that requires engagement of a virion with a stoichiometry that exceeds a required threshold number of antibodies (Burnet et al., 1937; Burton et al., 2001; Della-Porta and Westaway, 1978). The factors that define the stoichiometric requirements for neutralization of different classes of viruses are unknown, although the size of a virion correlates with estimates of the number of antibodies required for neutralization (Burton et al., 2001). The mechanisms by which antibodies promote viral clearance and protection from disease in vivo often extend beyond their capacity to directly neutralize virus infectivity, and include effector mechanisms mediated by the crystallizable fragment (Fc) portion of the antibody molecule (Burton, 2002; Nimmerjahn and Ravetch, 2008). These Fc-dependent effector functions include the ability to trigger antibody-mediated cellular cytotoxicity by Fc-γ-receptor (Fc-γR) bearing cells, facilitate viral clearance by phagocytic cells, and fix complement (Nimmerjahn and Ravetch, 2008; Ravetch and Bolland, 2001). Serum complement has been hypothesized to increase the potency of antibodies by promoting more efficient targeting of viruses for phagocytic destruction following opsonization, generating membrane attack complexes on the virion that lead to lysis in solution, and directly enhancing the neutralizing activity of antibodies (Volanakis, 2002; Zinkernagel et al., 2001). How complement augments the neutralization potential of antibodies has not been established, nor whether this translates into increased potency in vivo. Indeed, a recent study of the host factors required for protection from experimental simian human immunodeficiency virus (SHIV) infection following passive transfer of antibody has challenged the role of complement in the antiviral properties of neutralizing antibody in vivo (Hessell et al., 2007).
Flaviviruses are a group of positive-strand RNA viruses of global significance that cause severe encephalitic or hemorrhagic disease in humans (Mackenzie et al., 2004). Among medically relevant flaviviruses, West Nile virus (WNV) is now the primary cause of epidemic encephalitis in the United States (Sejvar, 2007) and dengue virus (DENV) is the most common mosquito-borne viral disease in the world (Kyle and Harris, 2008; Mackenzie et al., 2004). Flavivirus virions incorporate 180 envelope (E) proteins that orchestrate several steps of the virus lifecycle including virus assembly and egress, attachment and entry of target cells, and the low pH-dependent fusion between viral and endosomal membranes (Mukhopadhyay et al., 2005). The E protein is also a major target of antiviral antibodies elicited after flavivirus infection (Roehrig, 2003). Indeed, passive prophylaxis of anti-E protein antibodies confers protection in animal models of flavivirus infection (Ben-Nathan et al., 2003; Diamond et al., 2003; Roehrig et al., 2001). Furthermore, some anti-E protein antibodies have significant therapeutic potential; administration of a potently neutralizing monoclonal antibody can protect WNV-infected mice from death even after virus has spread into the central nervous system (Gould et al., 2005; Morrey et al., 2006; Morrey et al., 2007; Oliphant et al., 2005; Samuel et al., 2007). Thus, the induction of a potent antiviral humoral response is a primary goal for the development of vaccines against flaviviruses (Whitehead et al., 2007).
The presence of virus-specific antibodies, however, under certain conditions may adversely impact the outcome of flavivirus infection (Halstead, 2003). Infants with low circulating amounts of maternal anti-DENV antibodies are at an increased risk of severe disease following DENV infection (Chau et al., 2008; Kliks et al., 1988). In addition, the immune response elicited by primary DENV infection not only fails to protect from a secondary infection with a heterologous serotype of DENV, but may exacerbate disease (reviewed by (Halstead, 2003)). While the underlying mechanisms and circumstances that promote more severe clinical manifestations of infection have not yet been established in vivo, one prevailing hypothesis is that antibodies increase viral burden by increasing the efficiency of infection of Fc-γR bearing cells. This phenomenon has been studied extensively and is termed antibody-dependent enhancement of infection (ADE) (Halstead, 2003).
The atomic structure of the E protein and its pseudo-icosahedral arrangement on the virion has been determined for several flaviviruses (Mukhopadhyay et al., 2005), including WNV (Kanai et al., 2006; Nybakken et al., 2006). These insights, coupled with the availability of a large number of well-characterized monoclonal antibodies (mAbs) (Oliphant et al., 2005; Oliphant et al., 2006; Sanchez et al., 2005; Throsby et al., 2006), and a robust pathogenesis model in mice (reviewed by (Samuel and Diamond, 2006)), make WNV an excellent system for investigating principles that determine the potency of neutralizing antibodies (Pierson et al., 2008). Studies of the mechanisms of antibody-mediated neutralization of WNV support a requirement for the engagement of flaviviruses by multiple antibodies (Della-Porta and Westaway, 1978; Pierson et al., 2007). A stoichiometric threshold of ~30 antibodies is required to neutralize WNV (Pierson et al., 2007). Neutralization potency is governed, in part, by antibody affinity and the number of accessible epitopes displayed on the virion. Because not all epitopes recognized by antibodies are displayed equivalently on the virion (Oliphant et al., 2006; Stiasny et al., 2006), some antibodies may not bind virus particles with a stoichiometry that exceeds the neutralization threshold, even at saturating concentrations (Nelson et al., 2008; Pierson et al., 2007). Engagement of the virion with a stoichiometry below the neutralization threshold can promote ADE (Morens et al., 1987; Pierson et al., 2007). Of interest, a single antibody bound to the virion is not sufficient to augment infection of cells expressing Fc-γRIIA receptors. Instead, ADE requires the binding of ~15 antibodies to the virion (Pierson et al., 2007). Thus, ADE occurs in a rather narrow window of antibody concentrations defined at the upper limit by the number of antibodies required to neutralize infection, and at a lower limit by the number of antibodies required for stable attachment to cells.
Herein, we sought to determine how complement modulates the neutralizing activity of antibodies, and whether this translates into increased protection against WNV disease in vivo. We found that binding of the complement component C1q is sufficient to decrease the stoichiometric requirements for antibody neutralization. By reducing the threshold number of antibodies required for neutralization, C1q allows virus inactivation at lower concentrations of antibody and also has significant implications for the ability of an antibody to promote ADE. An improved understanding of the fundamental biochemical and molecular basis of antibody neutralization may facilitate the development of vaccination strategies that elicit a protective humoral response with a reduced potential for ADE in vivo.
To investigate how complement augments the potency of antibodies, we employed a validated, quantitative approach for measuring antibody-mediated neutralization in the presence and absence of complement (Pierson et al., 2006; Pierson et al., 2007). We found that the neutralization potency of the murine IgG2b monoclonal antibody (mAb) E16 (Nybakken et al., 2005; Oliphant et al., 2005), which recognizes an epitope on the lateral ridge of domain III (DIII-LR) of WNV E, was significantly improved (~8.8 fold, p=0.001, n=5) following the addition of fresh, but not heat-inactivated, mouse serum (Fig. 1A). Experiments using C1q-deficient mouse serum or purified human C1q protein demonstrate that this component of the classical complement pathway was necessary and sufficient for the augmented neutralization potency of anti-WNV antibodies in vitro (Fig. 1B), results that agree with prior studies with influenza virus (Feng et al., 2002; Mozdzanowska et al., 2006). As expected, the addition of other purified complement components (e.g., human C5) had no effect on neutralization by E16 (Fig. S1). Increased neutralization in the presence of purified C1q was also observed using mouse and human WNV-immune polyclonal sera (Fig. 1C and D, S2). Increased antibody potency was not dependent upon antibody-dependent complement-mediated virolysis as fresh C3-/- and C5-/- mouse serum equivalently increased neutralization potency (Fig. 1B, data not shown).
Previous studies of the stoichiometric requirements for WNV neutralization suggested that ~30 E16 mAbs are required to neutralize virus infectivity (Pierson et al., 2007). In this context, at least two mechanisms for C1q-augmented neutralization are possible. (a) Hexameric C1q might increase antibody avidity by virtue of its ability to simultaneously bind multiple antibodies via their Fc domain. A higher avidity would increase the number of antibodies docked onto a virion at any given concentration of antibody, which in turn would allow engagement of the virion with a stoichiometry that exceeds the required neutralization threshold at lower concentrations of antibody. (b) Alternatively, C1q could directly modulate the stoichiometric threshold required for neutralization, thereby changing the number of antibodies required to block infection. This would allow neutralization following engagement of a smaller fraction of epitopes on the virion. The neutralization potency of several classes of anti-WNV antibodies, including those commonly elicited in humans (Oliphant et al., 2007; Throsby et al., 2006), is limited by the accessibility of their epitopes on the mature virion (Nelson et al., 2008; Stiasny et al., 2006). A C1q-mediated reduction in the number of antibodies required for protection provides a mechanism for increasing the potency in vivo of antibodies that recognize cryptic determinants and are poorly inhibitory in vitro because they fail to bind the virion enough times to allow for neutralization.
To distinguish between these possibilities, we investigated how C1q modulates neutralization potency using a genetic complementation approach that allows the number of epitopes on individual virions to be experimentally manipulated (Pierson et al., 2007). A series of WNV reporter virus particles (RVPs) composed of wild type (WT) E proteins, and a variant containing a single mutation (T332K) in the DIII-LR epitope that abolishes recognition by E16, were produced as described previously (Pierson et al., 2007). The T332K mutant E protein incorporated into these virions is not detectably bound by DIII-LR-specific neutralizing antibodies; infection by virions composed solely of this variant were not inhibited by any concentration of E16 in the presence or absence of C1q (Fig. 2A). Reducing the number of intact DIII-LR epitopes on the virion resulted in the appearance of a population of virus particles resistant to neutralization even at saturating concentrations (~200-fold greater than the apparent KD (1.5 × 10-10 M) of E16 for the virion) (Fig. 2A, left panel and Fig. S3). For example, when neutralization studies were performed using a population of virions that display, on average, ~25% of the epitopes present on a WT virion (25% WT RVPs) approximately half of the virions (54%, n=10) were resistant to neutralization by E16, in agreement with published results with the DIII-LR-specific mAb E24 (Fig. 2A, left panel)(Pierson et al., 2007). This resistant fraction corresponds to the proportion of virions in the population that do not incorporate a sufficient number of epitopes to allow for neutralization at any concentration of antibody (Fig. S3). For the population of 25% WT RVPs (which corresponds to an average of 30 epitopes/virion), roughly half the virus particles incorporate ~30 or more intact epitopes and were neutralized, whereas half the virions incorporate fewer epitopes and were not neutralized even when all the epitopes on the virion were fully engaged by antibody (Fig 2A, left panel). Strikingly, the addition of C1q reduced the resistant fraction roughly 8-fold (p<0.001, n=10) for RVPs with ~25% WT E proteins (Fig. 2A, right panel and Fig. 2B). Similar results were observed with the DIII-LR-specific mAb E24 (Fig. S4). Because all of the available epitopes on virions in the resistant fraction are already bound at high concentrations of E16 in the absence of C1q, the ability to neutralize this population of virions in the presence of C1q cannot be explained by a change in antibody avidity. Instead, these results suggest the number of antibodies required for neutralization is reduced in the presence of C1q (Fig. S3). To test this hypothesis further, we produced RVPs that incorporate on average only a tenth of the epitopes present on a WT virion (10% WT RVP; Fig. 2A, corresponding to ~12 E16 epitopes/virion) and examined their sensitivity to neutralization by E16. The majority of virions in the 10% WT RVP population were sensitive to neutralization by E16 in the presence but not absence of C1q. If engagement of 12 epitopes were required for neutralization in the presence of C1q, a resistant fraction of ~50% would be predicted. Because the resistant fraction in the presence of C1q is smaller (28%, n=4; Fig. 2B), the stoichiometric threshold of neutralization in the presence of C1q must be fewer than 12 antibodies per virion.
Different subclasses of IgG bind C1q with distinct affinities (Bindon et al., 1988). Four IgG subclasses of a humanized version of the E16 mAb (hu-E16) were engineered and tested for their capacity to neutralize WNV in the presence and absence of C1q. In the absence of C1q, all four subclasses of hu-E16 neutralized WNV at similar concentrations (Fig. S5A-D). In contrast, the potency of hu-E16 of different subclasses was markedly different in the presence of human C1q. Subclasses of hu-E16 that bind C1q avidly (hu-IgG1 and hu-IgG3) augmented neutralization to a greater degree than subclasses that bind C1q poorly (hu-IgG2 and hu-IgG4). C1q also increased the capacity of both hu-IgG1 and hu-IgG3 subclasses of E16 to neutralize the resistant fraction of 25% WT RVPs, albeit to differing degrees (Fig. 2C). In the presence of C1q, the fraction of virions resistant to neutralization by saturating concentrations of the hu-IgG1 and hu-IgG3 subclasses of E16 was reduced by ~2 and ~8-fold, respectively (p=0.001, n=5 and p=0.002, n=5, respectively).
Antibody-dependent enhancement of infection and virus neutralization are two phenomena related by the number of antibodies bound to the virion (Morens et al., 1987; Pierson et al., 2007). Sub-neutralizing concentrations of antibody augment infection of Fc-γR-expressing cells principally by enhancing the efficiency of virus attachment (Gollins and Porterfield, 1984; Halstead, 2003)(C. Jost and T. Pierson, unpublished data). ADE is most readily observed on cells that bind virus poorly; expression of attachment factors (e.g., DC-SIGN) that increase the efficiency of virus binding reduce the magnitude of ADE (Boonnak et al., 2008; Pierson et al., 2007). Our previous work suggests that in the absence of serum complement, ADE of the Fc-γRII-expressing cell K562 is possible at concentrations of antibody that allow engagement of WNV virions by ~15 to 30 Abs (Pierson et al., 2007). Recent studies suggest that this relationship is markedly different in the presence of complement. C1q restricts ADE in an isotype-dependent fashion; antibodies that bind C1q avidly do not support robust ADE of Fc-γRII-expressing K562 cells at any concentration (Mehlhop et al., 2007)(Fig. 2D).
How C1q modulates ADE has not been investigated in mechanistic terms. Because C1q and Fc-γR bind overlapping regions of the antibody heavy chain (Duncan and Winter, 1988; Sondermann et al., 2000), C1q may block ADE by competing with Fc-γR for binding to virion-antibody complexes and preventing the enhanced cellular attachment. Competition studies confirmed that soluble Fc-γR could displace C1q bound to E protein immune complexes on a biosensor chip (Fig. S6). Despite the potential competition between C1q and Fc-γR for antibody binding, the lack of ADE in the presence of C1q was not due to an inability of virions to attach to cells. In the presence of C1q, binding of RVPs to Fc-γRII expressing K562 cells was paradoxically enhanced over a relatively broad range of antibody concentrations, despite the lack of enhanced infection (Fig. S7). This binding pattern was also observed with Raji and Vero cells (Fig. S7, data not shown), indicating the phenomenon was Fc-γR independent and possibly mediated by one or more of the panoply of cellular factors that function as C1q receptors (Ghebrehiwet et al., 1994). Exploring whether C1q restricts entry of virion-antibody complexes through the Fc-γR pathway, and if this is a requirement for ADE, will be of interest as a more detailed picture of the virus entry pathway in Fc-γR-bearing cells emerges.
Because of the established relationship between the concentrations of antibody that neutralize and enhance flavivirus infection (Morens et al., 1987; Pierson et al., 2007; Yamanaka et al., 2008), insight into the mechanism by which C1q blocks ADE can be inferred from changes in the requirements for neutralization. A reduction in the number of antibodies required for neutralization impacts the potential for ADE; lowering the stoichiometric threshold for neutralization below the ~15 mAbs required for ADE would prevent enhancement of Fc-γRII-expressing cells at any concentration of antibody. If neutralization of WNV in the presence of C1q occurs when fewer than 15 mAbs are bound to the virion, as our data suggest, the number of antibodies bound to the virus at non-neutralizing concentrations of antibody would never be sufficient to support ADE (Fig. S3). Because virions decorated with otherwise enhancing concentrations of antibody still bind Fc-γRII-expressing cells in the presence of C1q, the significant reduction in ADE in the presence of complement is likely explained by a reduction in the stoichiometric threshold for neutralization below the minimal requirements for enhanced infection rather than competition between C1q and Fc-γRs for binding to antibodies bound to the virus particle.
Not all mAbs recognize epitopes that are displayed on the average WNV virion with a frequency that allows for neutralization even when fully engaged (Nelson et al., 2008; Stiasny et al., 2006). As an additional layer of complexity, populations of flavivirus virions are heterogeneous with respect to the efficiency of virus maturation (Davis et al., 2006; Guirakhoo et al., 1992). The conformational changes in the E protein during virus maturation impact epitope accessibility and neutralization sensitivity (Guirakhoo et al., 1992; Nelson et al., 2008). Indeed, several groups of mAbs that recognize structurally distinct epitopes fail to neutralize mature WNV because virion maturation reduces epitope accessibility to levels that do not support neutralization (Nelson et al., 2008). Dose-response experiments performed with domain II-fusion loop (DII-FL)-specific mAbs revealed maturation state-dependent neutralization (Fig. 3 and Fig. S8); increasing the efficiency of virion maturation identifies a population of virions resistant to neutralization by saturating concentrations of antibody as described previously (Nelson et al., 2008). Notably, the addition of C1q increased E53-mediated neutralization of this resistant fraction of virions (Fig. 3), analogous to our experiments that genetically manipulate the number of E16 epitopes on the virion. These data are consistent with a model in which C1q augments the inhibitory potency of antibodies by reducing the stoichiometric requirements for virus neutralization.
C1q is a hexameric protein that binds most efficiently to antibodies bound to repetitive arrays of antigens; the strength of these multivalent interactions is significantly greater (~ 1,000 fold) than monomeric binding (Burton, 1985; Duncan and Winter, 1988; Kishore and Reid, 2000). This raises the possibility that C1q could augment neutralization by cross-linking E proteins on a single virion, which could prevent conformational changes in E required for virus entry and fusion. While a methodology to manipulate the number of antibodies that an individual C1q protein can bind has not been described, we assessed whether cross-linking antibodies on the virion is sufficient to reduce the stoichiometric requirements for neutralization. Incubation of E16-virion complexes with intact antibodies or Fab2 fragments of anti-murine IgG completely neutralized the resistant fraction observed using RVPs composed of 25% WT E proteins, whereas Fab fragments of anti-murine Ig had only a modest effect (Fig. 4). As C1q enhanced antibody neutralization using virions with a defined and limited number of epitopes, our results cannot be explained by a simple increase in antibody avidity. Instead, our results suggest that the capacity to cross-link antibodies on the virion is sufficient to reduce the stoichiometric requirements for neutralization.
To determine if changes in the stoichiometric requirements for neutralization translate into an increased capacity to protect against lethal WNV challenge in vivo, wild type or C1q-/- mice were passively administered limiting doses of IgG1, IgG2, or IgG3 subclass variants of hu-E16 one day prior to WNV infection and monitored for survival. Although transfer of all three hu-E16 variants protected mice from lethal encephalitis, differences in the concentration of antibody that protects 50% of animals (IC50) were observed (Fig. 4). Much of the disparity in IC50 values among IgG subclasses (e.g. hu-E16 IgG2) is likely explained by differential binding to specific activating Fc-γR, which contribute to E16-mediated protection against WNV infection in vivo (Oliphant et al., 2005).
However, we also observed a contribution of C1q to protection by specific human IgG subclasses in mice. Protection by the hu-E16 IgG3, which most avidly binds C1q, was observed at increased concentrations in C1q-/- compared to WT mice (p = 0.001, n ≥ 14 mice/dose). The 5.5-fold decrease in antibody potency in the absence of C1q in vivo was of similar magnitude to the 6-fold increase in neutralization observed in vitro with the IgG3 variant of E16 in the presence of fresh mouse sera (Fig. S5). In contrast, no significant difference in the IC50 of hu-E16 IgG2 was observed between wild type and C1q-/- mice (p = 0.8, n ≥ 15 mice/dose). Interestingly, and consistent with data from a SHIV-challenge experiment (Hessell et al., 2007), no statistical difference in the IC50 of hu-E16 IgG1 was observed in wild type and C1q-/- mice (p = 0.4, n ≥ 13 mice/dose; Fig. 4).
To gain further support for an in vivo role for C1q in antibody-mediated protection, we generated a point mutation (A330L) in the CH2 domain of hu-E16 IgG3 that abolishes C1q binding but retains Fc-γR binding (Fig. S9). In vitro studies with the A330L hu-E16 IgG3 variant demonstrated an improvement in neutralizing activity (EC50 = 1.2 × 10-11 M and 1.9 × 10-10 M for hu-IgG3 A330L and hu-IgG3, respectively; n = 3). While the improved neutralization activity of the A330L variant was unexpected, studies with anti-Cryptococcus antibodies indicate that changes in the immunoglobulin constant region can impact affinity and neutralizing activity (Torres and Casadevall, 2008). Thus, a direct comparison of the A330L and the parent hu-E16 IgG3 was not possible because of their inherent difference in neutralizing activity. However, and in contrast to the parent hu-E16 IgG3, which showed a difference in IC50 in wild type and C1q-/- mice, passive transfer studies with the A330L variant in wild type and C1q-/- mice demonstrated no significant difference in the IC50 (p = 0.1, n ≥ 11 mice/dose; Fig. 4). Taken together, these data suggest C1q improves the protective activity of WNV neutralizing mAbs in vivo in an IgG-subclass-restricted manner.
In this study, we explored how complement augments the neutralization potency of antiviral antibodies. The addition of purified human C1q protein in neutralization experiments was sufficient to augment the potency of several classes of mAbs, as well as antibodies present in WNV-immune sera. These results agree with previously described cell culture experiments with WNV (Della-Porta and Westaway, 1977, 1978), yellow fever virus (Spector and Tauraso, 1969) and DENV (Yamanaka et al., 2008) infections, which described augmentation of antibody neutralization by fresh serum or complement. Because increased neutralization did not require lysis of the virion, we investigated the basis for this phenomenon with respect to the stoichiometric requirements for antibody-mediated neutralization. “Multiple hit” models of neutralization suggest the number of antibodies bound to a virion determines infectivity (Burnet et al., 1937; Burton et al., 2001; Della-Porta and Westaway, 1978). In this regard, at least two factors determine how many antibodies are bound to a virion at any given concentration of antibody, and whether this will exceed the stoichiometric requirements for virus inactivation: antibody affinity and epitope accessibility. Our results demonstrate that increases in antibody affinity mediated by C1q are not sufficient to explain increases in neutralization potency in the presence of complement. Instead, our data support a model by which C1q also reduces the threshold number of antibodies required for neutralization of WNV.
The potency of several classes of antibodies is limited by the accessibility of the epitope they recognize on the surface of the virus particle (Nelson et al., 2008; Oliphant et al., 2006; Stiasny et al., 2006). When epitopes are displayed on the virion in small numbers, antibodies that recognize them fail to neutralize infectivity at any concentration because the total number of antibodies that can dock on the virion does not exceed the requirements for neutralization. The neutralization potency of antibodies specific for the DII-FL is limited by the poor accessibility of this epitope on mature flavivirus virions (Nelson et al., 2008; Oliphant et al., 2006; Stiasny et al., 2006). In this context, viruses resistant to neutralization at saturating concentrations of antibody represent virus particles that do not display enough epitopes to allow for neutralization by antibodies of that specificity. For WNV, the efficiency of the maturation process strongly impacts neutralization sensitivity by modulating epitope accessibility (Nelson et al., 2008). In a similar fashion, genetic methods that reduce the number of epitopes on the average virion also allow for the production of viruses resistant to neutralization. In this system, the size of the fraction of virions resistant to neutralization by saturating concentrations of antibody is inversely related to the number of intact epitopes on the average virion in the population (Pierson et al., 2007). In these two complementary experimental models, the ability of C1q to promote neutralization of the resistant fraction of viruses cannot be explained by increases in antibody affinity because in both cases, it is the number of epitopes on these virus particles that is limiting. In the presence of saturating concentrations of antibody, there are no additional sites for antibodies to bind.
The complement-mediated reduction in the number of antibodies required for neutralization has significant implications for humoral immunity to flavivirus infection. First, a lower stoichiometric threshold allows inactivation of virus particles that would otherwise be resistant to neutralization by antibodies specific for poorly accessible epitopes. A significant proportion of human antibodies elicited by flavivirus infection recognize the DII-FL (Crill et al., 2007; Lai et al., 2008; Oliphant et al., 2006; Throsby et al., 2006). Antibodies specific for this poorly accessible epitope exhibit limited potency in vitro (Oliphant et al., 2006; Stiasny et al., 2006), and may be sensitive to the maturation state of the virus particle (Fig. 3). In the absence of complement, the protective capacity of these antibodies may depend in part on the composition of the virus. Mature WNV has been shown to be significantly less sensitive to neutralization by DII-FL-specific mAbs and polyclonal WNV-immune sera obtained from roughly 50% of the recipients of two candidate WNV vaccines (Nelson et al., 2008). A reduced stoichiometric requirement for neutralization in the presence of complement increases the fraction of virus particles sensitive to neutralization by DII-FL-specific antibodies, and may explain why polyclonal responses composed of antibodies with this specificity still can be protective in vivo.
Changes in the neutralization potency of antibodies also impact their potential to enhance flavivirus infection. Recent studies have shown that complement significantly reduces the capacity for antibodies to promote ADE via undefined mechanisms (Mehlhop et al., 2007; Yamanaka et al., 2008). Because antibody-mediated neutralization of infection and ADE are two outcomes related by the number of antibodies bound to the virion (Morens et al., 1987; Pierson et al., 2007), a reduction in the number of antibodies required for neutralization should impact ADE. Surprisingly, despite significant changes in neutralization potency (EC50) in the presence of C1q (Fig. 1A and and2A),2A), a corresponding change in the concentrations at which ADE occurs was not observed (Fig. 2D and (Mehlhop et al., 2007)). Instead, C1q restricted ADE almost completely. Mechanistically, this is explained as follows: C1q enhances neutralization by changing the number of antibodies required for virus inactivation, rather than the number of antibodies bound to the virion at a given concentration (e.g. avidity). For K562 cells expressing activating Fc-γRIIA receptors, ADE occurs when roughly 15-30 antibodies engage WNV. For IgG subclasses (e.g. murine IgG2b or hu-IgG3) that bind C1q avidly, the neutralization threshold in the presence of C1q falls below the minimal stoichiometric requirements for ADE, and explains why C1q restricts ADE in an IgG-subclass-specific manner (Fig. S3).
An ability to directly enhance neutralization is consistent with a protective role for complement after infection by WNV, and likely other flaviviruses. Mice deficient in components of the classical, lectin, or alternative complement activation pathways all showed increased WNV replication and lethality after infection (Mehlhop and Diamond, 2006; Mehlhop et al., 2005). Immunological analysis revealed that in many of the complement-deficient strains of mice used in these studies, the phenotype was related to complement's function in priming of adaptive B and T cell responses. Nonetheless, C1q-/- mice showed increased vulnerability to WNV infection despite relatively intact B and T cell responses, suggesting an independent protective effect, possibly through interaction with complement-fixing antibodies (Mehlhop and Diamond, 2006). The neutralization and protection studies with deficient serum and mice establish a C1q-dependent antibody protective mechanism. More recent studies have confirmed that C1q-dependent protection by complement-fixing anti-WNV antibodies does not require membrane attack complex deposition or virolysis in vivo, as a similar IC50 of E16 is observed in C5-/- and WT mice (Mehlhop et al., 2009). Thus, C1q is the critical complement component that modulates antibody potency against WNV; as such, its inclusion in flavivirus serum neutralization titration assays may improve the correlation between in vitro neutralizing activity and in vivo protection.
Enhancement of antibody-mediated neutralization by complement is not unique to flaviviruses. The addition of complement to in vitro assays has been shown to augment neutralization by antiviral antibodies of several families of RNA and DNA viruses (reviewed in (Carroll, 2004; Zinkernagel et al., 2001)). However, few studies have defined the complement components that mediate this effect and the mechanism of action. For example, the first four components of the classical pathway were shown as necessary for the increasing antibody-mediated neutralization of Newcastle disease virus (Linscott and Levinson, 1969), hepatitis C virus (Meyer et al., 2002), HIV (Sullivan et al., 1998), and HSV (Daniels et al., 1969, 1970). Complement-dependent antibody neutralization of vaccinia virus was augmented in C5- but not C3-depleted serum; thus, coating and opsonization rather than virolysis have been postulated as an explanation for the enhanced neutralizing activity (Rafii-El-Idrissi Benhnia et al., 2008). In contrast, complement-triggered virolysis contributes to the augmented antibody-dependent neutralization of HIV and parainfluenza virus in vitro (Spear et al., 1993; Vasantha et al., 1988). Complement-enhanced antibody neutralization of influenza virus in vitro appears to depend solely on C1q (Feng et al., 2002; Mozdzanowska et al., 2006).
Our studies demonstrate a strict isotype requirement for improved antibody neutralization of WNV, results that agree with in vitro investigations with influenza virus (Feng et al., 2002; Mozdzanowska et al., 2006), and in vivo studies with neutralizing yellow fever virus antibodies that showed enhanced protection correlated with IgG subclasses that efficiently fix complement (Schlesinger and Chapman, 1995). Antibody isotypes that bind C1q poorly (e.g., mIgG1, hu-IgG2, and hu-IgG4) showed no enhanced neutralization in the presence of C1q or serum in vitro. Moreover, the experiments with the C1q avid hu-E16 IgG3 in wild type and C1q-/- mice show that C1q affects a ~5.5-fold change in the IC50 of antibody protection, similar in magnitude to the enhanced neutralizing activity observed in vitro with murine C1q (6-fold) or human C1q (19-fold) (Fig. S5). To our knowledge, this is the first direct demonstration of C1q changing the potency of an antibody in vivo. Indeed, despite a prominent effect in vitro (Feng et al., 2002; Mozdzanowska et al., 2006), C1q did not contribute to the prophylactic activity of anti-influenza neutralizing mAbs in vivo (Mozdzanowska et al., 2006). The isotype-dependent improvement of antibody potency in vivo may also, in part, clarify why complement does not augment the protection conferred by passive transfer of an HIV-specific human IgG1 mAb (Hessell et al., 2007), as this subclass binds C1q less efficiently than human IgG3 mAbs (Bindon et al., 1988). Similarly, we did not observe a change in protection of the hu-E16 IgG1 in C1q-/- mice. One caveat to this analysis is that given the small shift in neutralization potency in vitro of hu-E16 IgG1 in the presence of C1q (Fig S5; 2.5- and 1.1-fold for human and murine C1q, respectively), a larger number of animals (n= 21 mice per each antibody dose for a 2-fold effect) would be required to obtain sufficient statistical power to conclude definitively that C1q has no effect on the protective activity hu-E16 IgG1 in vivo. Finally, structural differences between retro- and flaviviruses, coupled with the ability of the latter to incorporate complement-modulating cellular factors may add to the complexity of this comparison.
In summary, we have established a biochemical mechanism for how the complement component C1q increases the neutralizing activity of antibodies. In addition, we suggest these mechanistic insights have practical consequences with broad applications. In general, a goal of vaccination is to produce high affinity neutralizing antibodies. However, the cryptic properties of epitopes commonly elicited by infection and the heterogeneous nature of flaviviruses may limit the efficacy of antibodies regardless of the strength with which they bind virions. Our recent studies indicate that humans infected with WNV or vaccinated with WNV non-infectious particles dominantly develop IgG1 anti-WNV responses (E. Mehlhop and M. Diamond, unpublished data). Moreover, some patients associated with severe DENV infection have high levels of non-complement-fixing IgG4 anti-DENV antibodies (Koraka et al., 2001). As C1q can alter the stoichiometric requirements for neutralization, vaccine strategies that use adjuvants that favorably skew IgG-subclass responses to C1q-fixing IgG subclasses may elicit more protective antiviral antibodies. Moreover, engineering therapeutic antiviral antibodies against flaviviruses with improved C1q-binding properties should enhance neutralization potency and limit the potential for ADE.
HEK-293T, BHK21, C6/36, K562, and Raji-DC-SIGNR cells were maintained as described previously (Pierson et al., 2005; Pierson et al., 2006; Pierson et al., 2007). WNV RVPs were produced in HEK293T or BHK21 cells as described (Pierson et al., 2006; Pierson et al., 2007). Methods to manipulate the number of DIII-LR epitopes on WNV RVPs (Pierson et al., 2007), modulate the efficiency of virion maturation (Nelson et al., 2008), and generate the human IgG subclass variants of E16 (Mehlhop et al., 2007) have been described. The A330L variant of hu-E16 IgG3 was generated by site-directed mutagenesis using the QuikChange® Mutagenesis Kit (Stratagene, La Jolla, CA).
The neutralization potency of antibodies or immune sera was measured in the presence or absence of fresh sera or purified human C1q protein (Complement Technologies, Tyler, TX) using a Raji B lymphoblastoid cell line that expresses DC-SIGNR as described (Mehlhop et al., 2007; Pierson et al., 2006; Pierson et al., 2007). Neutralization potency was calculated as a function of the concentration of antibody required to block 50% of the infection events using non-linear regression analysis (GraphPadPrism4, San Diego, CA). All neutralization studies were performed using conditions designed to satisfy the “percentage law” (Andrewes and Elford, 1933).
All mice were housed in a pathogen-free mouse facility at Washington University School of Medicine. Studies were performed in compliance with the guidelines of the Washington University School of Medicine Animal Safety Committee. All mice received affinity-purified mAb by intraperitoneal injection one day prior to footpad infection with 102 plaque forming units (PFU) of WNV isolate 3000.0259 that was propagated once in C6/36 cells. Five week-old wild type C57BL/6 mice were purchased commercially (Jackson Laboratories, Bar Harbor, ME). Congenic C1q-/- mice were bred at Washington University and infected with WNV at eight to ten weeks of age, resulting in a baseline mortality rate similar to that of 5 week-old wild-type mice. Kaplan-Meier analysis of survival data was performed using the log-rank test and Prism software (GraphPadPrism4, San Diego, CA). IC50 analyses were performed by non-linear regression and statistical significances were determined using analysis of variance (ANOVA) and F-tests.
The binding of human FcγRs and C1q to the hu-E16 IgG3 Fc was analyzed by surface plasmon resonance (SPR) using a BIAcore 3000 biosensor (BIAcore, Uppsala, Sweden). WNV DIII was immobilized on the CM-5 sensor chip by amine coupling kit as recommended by the manufacturer. A surface treated with amine coupling reagents was used as a blank. Binding experiments were performed in HBS-EP buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.005% P20 surfactant). Wild type or variant antibody was bound to the DIII surface at approximately 1000 RU, followed by injection of soluble Fc-γRs: human CD32A-131His or Arg-G2, N297Q (Stavenhagen et al., 2007)) at a concentration of 100 nM or C1q at 24 nM and a flow rate of 30 μl/min for 60 sec with dissociation time of 60 sec. Each receptor or C1q was injected in duplicate. Between injections within an experiment, bound Fc-γR was completely dissociated using 1 M ethanolamine-HCl, pH 8.5. Between experiments, the naked antigen surface was regenerated by pulse injection of 10 mM glycine pH 1.5. Binding responses were normalized to the level of captured wild type hu-E16 IgG3 antibody.
We thank J. Atkinson and members of our laboratories for helpful discussions, H. Hickman, H. Virgin, T. Kristie, and B. Moss for critical comments on the manuscript, and J. Bramson and M. Loeb for the convalescent human WNV sera. This work was supported by the Intramural Research Program of the National Institutes of Allergy and Infectious Diseases (NIAID) and grants from the Pediatric Dengue Vaccine Initiative (T.C.P and M.S.D), the NIH (grants U01 AI061373 and R01 AI073755 (M.S.D and D.H.F.)), and the Midwest Regional Center of Excellence for Biodefense and Emerging Infectious Diseases Research (U54 AI057160).
Author Contributions. S.J. generated hu-E16 IgG subclass-switch variants and characterized antibody affinity; E.M., C.J. and S.N. analyzed neutralization of mouse and human mAbs; S.N. performed the stoichiometry studies; S.J. and S. G. characterized antibody-affinity measurements by SPR; E.M. analyzed the in vivo protective capacity of E16 variants; E.M., S.N., D.H.F., M.S.D. and T.C.P. designed the studies, analyzed the data, and wrote the paper; all authors discussed the data and commented on the manuscript.
Conflict of Interest Statement. S.G. and S.J. are employees of MacroGenics, a company that has licensed the E16 antibody from Washington University for commercial use. M.S.D. is a consultant for MacroGenics.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.