key: cord-0257288-7e0mhbb9
authors: Zwarthoff, Seline A.; Widmer, Kevin; Kuipers, Annemarie; Strasser, Jürgen; Ruyken, Maartje; Aerts, Piet C.; de Haas, Carla J.C.; Ugurlar, Deniz; Vidarsson, Gestur; Van Strijp, Jos A. G.; Gros, Piet; Parren, Paul W.H.I.; van Kessel, Kok P.M.; Preiner, Johannes; Beurskens, Frank J.; Schuurman, Janine; Ricklin, Daniel; Rooijakkers, Suzan H.M.
title: C1q binding to surface-bound IgG is stabilized by C1r2s2 proteases
date: 2021-02-08
journal: bioRxiv
DOI: 10.1101/2021.02.08.430229
sha: fd0650f88a87101b695ce8c531bc931d0b71e47f
doc_id: 257288
cord_uid: 7e0mhbb9

Complement is an important effector mechanism for antibody-mediated clearance of infections and tumor cells. Upon binding to target cells, the antibody’s constant (Fc) domain recruits complement component C1 to initiate a proteolytic cascade that generates lytic pores and stimulates phagocytosis. The C1 complex (C1qr2s2) consists of the large recognition protein C1q and a heterotetramer of proteases C1r and C1s (C1r2s2). While interactions between C1 and IgG-Fc’s are believed to be mediated by the globular heads of C1q, we here find that C1r2s2 proteases affect the capacity of C1q to form an avid complex with surface-bound IgG molecules (on various DNP-coated surfaces and pathogenic Staphylococcus aureus). The extent to which C1r2s2 contribute to C1q-IgG stability strongly differs between human IgG subclasses. Using antibody engineering of monoclonal IgG we reveal that hexamer-enhancing mutations improve C1q-IgG stability, both in absence and presence of C1r2s2. In addition, hexamer-enhanced IgGs targeting S. aureus mediate improved complement-dependent phagocytosis by human neutrophils. Altogether, these molecular insights into complement binding to surface-bound IgGs could be important for optimal design of antibody therapies.

When we took a closer look at different complement activation steps, we noticed an unexpected disparity between surface detection of C3b and C1q for IgG2. While recruitment of C1q is a prerequisite to initiate antibody-dependent complement activation, we observed that detection of C1q molecules on IgG2-coated bacteria was low compared to IgG1 and IgG3 (Fig. 1B) . This was surprising because C3b deposition via these subclasses was similar (Fig. 1A) . Using a monoclonal antibody that prevents C1q-IgG interactions 46 , we showed that C3b deposition via IgG2 was driven by C1 (SFig. 3) . Also, inhibition of C1 via a bacterial protein that blocks C1r (BBK32 47-49 ) confirmed that C1 is required to deposit C3b onto IgG2-coated bacteria (SFig. 3). Furthermore, we showed that C3b molecules deposited by anti-WTA IgG2 antibodies are functional. We studied phagocytosis of bacteria by human neutrophils, the primary mechanism for elimination of S. aureus 50 . Although anti-WTA IgG2 did not potently induce Fc receptor-mediated phagocytosis of S. aureus (as expected from the predicted low affinity of IgG2 for FcγR 32 ), we found that addition of complement strongly promoted the phagocytic uptake of IgG2-labelled S. aureus (Fig. 1C) .

We wondered whether these findings could be translated to IgGs recognizing a different antigenic surface. To study this, we used an assay system in which beads are coupled with the model antigen 2,4-dinitrophenol (DNP) 51 and coated with human IgGs specific for DNP (SFig. 4A-C) 52 . To mimic the highly abundant nature of WTA, beads were coated with saturating levels of DNP (quantified by measuring IgG binding using flow cytometry) (SFig. 4B). In accordance with our findings on S. aureus, we found that anti-DNP antibodies of the IgG1, IgG2 and IgG3 subclasses all potently induced a complement response and deposit C3b molecules onto the surface of DNP-beads (Fig. 1D) . Again, whereas C3b opsonization could be correlated with the presence of C1q on beads coated with IgG1 and IgG3, almost no C1q could be detected on the IgG2-coated surface (Fig. 1E ).

In conclusion, on two independent surfaces, we showed that IgG1, IgG2 and IgG3 can all potently elicit C1-dependent deposition of C3b molecules. However, for IgG2 our data revealed an unexpected disparity between the detection of C1q and downstream deposition of C3b molecules.

To better understand the above findings, we more closely examined the molecular interactions between C1q and target-bound IgGs by using purified C1 complexes. We included two forms of C1q in our analyses: 1) C1q in complex with C1r2s2 proteases (denoted C1), which is representative for circulating C1 complexes in human blood 53,54 ; or 2) recognition molecule C1q without proteases (denoted C1q) ( Fig. 2A) . When we studied binding of different forms of C1q to anti-DNP antibodies on DNP-beads, we noticed a discrepancy between the binding of non-complexed C1q molecules versus C1 (Fig. 2B, SFig. 4D) . Particularly in the case of IgG2, we observed that binding of C1 was more efficient than binding of non-complexed C1q, as quantified by flow cytometric detection of surface-bound C1q (Fig. 2B) . Western blotting was used to confirm that the different levels of C1q detected in flow cytometry actually represent different quantities of surface-bound C1q (SFig. 4E). When C1 complexes were dissociated by EDTA, which disrupts the calcium-dependent attachment of C1r2s2 to C1q 55,56 , binding was similar to C1q alone (Fig. 2B) . On IgG1-coated beads, we observed that C1r2s2 proteases had subtle effect on binding of C1q (Fig. 2B) . Furthermore, C1r2s2 proteases had a very limited effect on the binding of C1q to IgG3 (Fig. 2B) . Similar to these results on beads, we observed that EDTA reduced binding of purified C1 to S. aureus labelled with IgG1 and IgG2, while binding to IgG3 was much less affected (SFig. 5). Altogether, these studies suggest that attached C1r2s2 proteases affect C1q-IgG interactions in a subclass-dependent manner. This finding is unexpected when considering that the direct interactions between C1 and IgG were shown to solely depend on the gC1q domains 16 .

Since the surface density of antigens was earlier proposed to be a critical parameter for IgGmediated complement activation 57-59 , we wondered whether C1r2s2 proteases also affect C1q-IgG interactions at lower antigen concentrations. In the bead system, we lowered the DNP concentration by ~300-fold. At this lower DNP concentrations, we found that IgG3 was the most potent subclass triggering complement activation in human serum (SFig. 6) and that C3b deposition by IgG1 and IgG2 was inefficient. This corresponds with the idea that IgG3 more potently drives complement activation on low abundant antigens (presumably because it has a longer hinge region 21, 57, 58, 60 ). Upon studying binding of purified C1 complexes, we observed that C1r2s2 proteases can also affect C1q binding to IgG3 on beads with a lower DNP concentration. First, we observed that C1 bound more efficiently to IgG3 than non-complexed C1q (Fig. 2C) . Also, disruption of C1 with EDTA caused an almost complete reduction of C1q binding to IgG3 (Fig. 2C) .

Altogether this suggests that C1r2s2 proteases can enhance binding of C1q to all IgG subclasses. However, the extent to which C1r2s2 proteases contribute to C1q binding depends both on the IgG subclass and antigen concentration.

To corroborate these results, we employed surface plasmon resonance (SPR) as an orthogonal technique to confirm the binding profiles and obtain dynamic information about the formation and stability of C1q-IgG complexes (Fig. 3A, SFig. 7) . Using a flat DNP-labelled surface, prepared by coupling DNP-PEG-NHS to an activated carboxyl sensor chip (SFig. 7A), monoclonal anti-DNP antibodies could be captured at high density and stability (SFig. 7B). No binding was observed on a MeO-PEG-NHS surface that was used as a reference (SFig. 7C). By injecting either C1q or C1 for 60 seconds, the assembly (during injection) and stability of C1q-IgG/C1-IgG complexes (during the dissociation period) could be monitored over time. SPR analysis indeed confirmed that C1r2s2 proteases affect the stability of C1q-IgG complexes, especially for IgG1 and IgG2 (Fig. 3A) . Whereas both C1q and C1 bound to the IgG1-and IgG2-coated sensor chips during protein injection, C1 dissociated much slower than C1q after the injection was stopped, which suggests that C1 forms more stable interactions with the antibody-coated surface. Additionally, we observed a weaker association of C1q in absence of C1r2s2 to IgG2-coated chips during the injection (Fig. 3A) . In line with our studies on IgG3labelled beads, C1r2s2 proteases did not notably affect C1q binding to IgG3-coated chips. However, SPR experiments showed that C1r2s2 proteases had a more pronounced impact on binding of C1q when the IgG3 capturing concentration was lowered 5-fold (SFig. 7D). As expected, neither C1q nor C1 showed detectable binding to IgG4. To directly visualize C1q-IgG complexes, we performed high-speed atomic force microscopy (HS-AFM) experiments. To enable reliable identification of C1q-IgG complexes in HS-AFM, we employed the anti-DNP triple mutant (IgG1-E345R, E430G, and S440Y, denoted IgG1-RGY), which was shown to efficiently associate into IgG1 hexamers in solution 19, 61 . DNPlabelled supported lipid bilayers were therefore pre-incubated with anti-DNP IgG1-RGY 17 . Then, C1q or C1 was added and the resulting complexes where visualized via HS-AFM (Fig.  3B , Movies S1 and S2). While the complexes of C1 that bound to the IgG1-RGY hexamers were not significantly disturbed by the minimal forces exerted by the HS-AFM tip, C1q alone was frequently removed from the IgG1-RGY hexamers as a result of the tip-sample interaction under the same experimental settings. Altogether, the above data suggest that attached C1r2s2 proteases improve the stability of C1q-IgG complexes on target surfaces. We propose that C1r2s2 proteases affect the conformation of C1q in a way that facilitates stable docking to surface-bound IgGs. Earlier studies reported that the solution structure of C1q shows a high degree of flexibility 62-64 . While the six collagen arms firmly bundle via disulphide bonds in the upper stalk, such interactions are lacking below the stalk and enable a rather flexible arrangement of the gC1q domains (Fig. 3C) . A recent in situ 3D structure of C1qr2s2, bound to surface-antigen-bound IgM, showed that the C1r2s2 proteases are packed inside the C1q molecule as two antiparallel C1rs heterodimers 65 . Each C1rs dimer binds three of the six C1q collagen helices, which limits the flexibility of the collagen arms and results in a near-hexagonal arrangement of the six gC1q domains. Based on recent observations that IgG hexamers are the ideal docking platform for C1q 15,17 , we propose that this hexagon-like arrangement of the six gC1q domains induced by the attachment of C1r2s2 to C1q favors multivalent, high-avidity binding of C1q with clustered IgG (Fig. 3C) .

Next, we wondered how the observed differences in stability between C1-IgG and C1q-IgG could explain why we observed a discrepancy between C1q and C3b detection on IgG2-coated surfaces in serum (Fig. 1) . Because C1q in human serum circulates as a complex with C1r2s2 proteases, we assume that IgG2-coated S. aureus and DNP-beads (1 µg/ml DNP) can recruit C1 and activate complement. However, since human serum contains an inhibitor that removes C1r2s2 proteases from C1q, we hypothesized that subsequent removal of proteases results in dissociation of C1q-IgG complexes. To test this, we incubated IgG-covered beads with purified C1 and, after washing, incubated the C1-bound beads with human C1-esterase inhibitor (C1-INH) 66 . C1-INH is a human serpin that inactivates the proteases by forming a covalent bond with the catalytic site of both C1r and C1s. As expected, incubation of C1-IgG complexes on beads with C1-INH led to removal of C1r2s2 proteases from C1q on IgG1-, IgG2-and IgG3covered beads as evidenced by the detection of C1-INH−C1r and C1-INH−C1s complexes in the sample supernatant using western blotting (SFig. 8). In line with our hypothesis, we found that the dissociation of C1r2s2 by C1-INH had differential effects on the stability of the beadbound C1q-IgG complexes (Fig. 4A) . Whereas dissociation of C1r2s2 proteases by C1-INH did not affect C1q-IgG3 complexes, it caused a strong (80%) reduction of C1q binding to IgG2beads and a 10% reduction on IgG1-beads (Fig. 4A) . Similar results were obtained when C1r2s2 proteases were removed from C1-IgG complexes using EDTA (Fig. 4A) .

Altogether, these data demonstrate that C1-INH can dislodge C1q from IgG-coated surfaces by removing C1r2s2 from surface-bound C1-IgG complexes. While C1-INH binds and removes C1r2s2 regardless of the IgG subclass, subsequent C1q dissociation depends on the stability of C1q-IgG complexes (Fig. 4B) . It seems likely that the dislodgement of C1q as result of C1r2s2 dissociation explains why we could not detect C1q on IgG2-beads in a serum environment (Fig. 1E) . The observed deposition of C3b molecules on IgG2-beads in serum (Fig. 1D ) furthermore suggests that C1 inhibitory mechanisms lag behind on C1-mediated cleavage of complement proteins.

Finally, we determined how C1q-IgG interactions are influenced by IgG oligomerization. Recent studies demonstrated that oligomerization of target-bound IgG into hexamers can be enhanced by specific point mutations that strengthen Fc-Fc contacts 15, 18 . In contrast to the RGY mutation used above, these so-called HexaBody® mutations do not enhance hexamer formation in solution but specifically enhance hexamerization on target surfaces. Here, we modified anti-DNP IgGs using hexabody mutations E430G or E345K 18 and verified that these mutations did not affect IgG binding to beads (SFig. 9A-B). Upon studying interactions with purified forms of C1/C1q, we observed that both Fc mutations strongly improved binding of non-complexed C1q (i.e. C1q or C1-EDTA) to IgG1 and IgG2 on DNP-beads (Fig. 5A , SFig. 9C). For IgG3, hexabody mutations did not affect binding of non-complexed C1q (Fig. 5A , SFig. 9C). SPR experiments corroborated the results in the bead assay, showing that both the E430G and E345K mutation increase complex stability of IgG1 and IgG2 with C1q and have little effect on the association of C1q with IgG3 (SFig. 10). Similar results were obtained for S. aureus, where introduction of E430G into anti-WTA IgG, strongly enhanced binding of noncomplexed C1q to IgG1 and IgG2, while not affecting IgG3 (SFig. 11A-B). Interestingly, we observed that hexabody mutations did not have a strong impact on the binding of fully assembled C1 to IgG1 and IgG2 on DNP-coated beads (Fig. 5B ) or SPR chips (SFig. 10). On S. aureus we observed that the E430G mutation induced a subtle improvement of C1 binding (SFig. 11C). These data suggest that unstable C1q-IgG interactions between non-complexed C1q and IgG can be overcome by promoting formation of high avidity multimeric IgG platforms on the surface. In situations where C1q-IgG complexes are already stable (e.g. for C1 binding to IgG1, IgG2 or IgG3 or C1q binding to IgG3), hexabody mutations do not or only slightly improve C1q-IgG binding. The observation that Fc mutation E430G enhances C1q binding to IgG1 and IgG2 in human ΔIgGΔIgM serum (both on beads (Fig. 5C ) and S. aureus (SFig. 12)) suggests that hexabody mutations also improve the ability of IgGs to retain C1q on target surfaces in a serum environment.

We also examined whether enhanced hexamerization could affect complement binding and activation via IgG4, the IgG subclass that is considered incapable of reacting with C1q 39 . Previous studies showed that the Ser at position 331 in the heavy chain of IgG4 is critical for determining its inability to bind C1q 38,67 . Structural modelling has furthermore shown that the two Fab arms of IgG4 obstruct its C1q binding site and that the short IgG4 hinge region allows only limited flexibility 35-37 . As the IgG4 used in our study also contains the essential Ser331 residue, we expected no binding to C1q or C1. However, when we introduced the E430G or E345K mutation in anti-DNP IgG4, we observed dose-dependent binding of fully assembled C1, but not C1q ( Fig. 5A-B, SFig. 9C) . Also on S. aureus (SFig. 11) and in SPR assays (SFig. 10), we observed that enhancing Fc-Fc interactions in IgG4 enabled binding to the fully assembled C1, and negligible binding to non-complexed C1q. On S. aureus in human serum, we could detect only low levels of C1q when introducing the hexabody mutation E430G in IgG4 (SFig. 12B). This observation is not unexpected considering our earlier hypothesis that C1r2s2 removal by C1-INH results in dislodgement of C1q when C1q-IgG interactions are unstable. Taken together, these data suggest that the poor interaction of surface-bound IgG4 with C1q can be overcome in the presence of both hexabody mutations and C1r2s2 proteases which confirms a role for the proteases in C1q binding.

Having demonstrated that enhanced IgG hexamerization improves the stability of C1q-IgG complexes on S. aureus, we wondered whether hexabody mutations also have an impact on the downstream complement effector mechanisms. In previous studies, we and others have shown that phagocytosis of S. aureus, which is eminent for human immune defense against these bacteria 68 , critically depends on the labelling of S. aureus with complement-derived opsonins [69] [70] [71] . While antibacterial IgGs may trigger phagocytosis of S. aureus via Fc receptors, additional opsonization of the bacterial surface with C3b and iC3b enhances the efficacy of particle uptake via engagement of complement receptors. To study the effect of hexamerenhancing mutations on phagocytosis, we first determined the efficiency by which anti-WTA mutant IgGs triggered covalent deposition of opsonic C3b on S. aureus (Fig. 6A) . Upon incubation of S. aureus with anti-WTA IgGs and ΔIgGΔIgM serum, we observed that the E430G mutation enhanced C3b deposition mediated by IgG1 and IgG2, but not IgG3 (Fig.  6A ). Next, we determined whether enhanced IgG oligomerization on S. aureus improved the phagocytosis of bacteria by human neutrophils in serum (Fig. 6B) . In full correspondence with the observed improvements at the level of C1q (SFig. 12B) and C3b (Fig. 6A) , we observed that introduction of the E430G mutation in anti-WTA IgG1 and IgG2, but not IgG3, enhanced the phagocytic uptake of fluorescent S. aureus in human serum. Consistent with the finding that anti-WTA IgG4-E430G antibodies can induce C3b deposition, we observed that this antibody could induce phagocytosis of S. aureus (Fig. 6B) . As expected, the E430G mutation did not affect the IgG-dependent phagocytosis via Fc receptors in the absence of complement (SFig. 13). Altogether these data demonstrate that Fc-engineering of anti-S. aureus IgGs can be a useful strategy to improve a complement response against this important pathogen and subsequent uptake by phagocytes.

The classical pathway of complement activation is a major contributor to pathophysiological processes in our body, including infection, inflammation, autoimmunity and transplant rejection. Here we demonstrate in great detail how the initiating step of this pathway, namely the binding of the large C1 complex to the surface of IgG-labelled target cells, is influenced by its attached proteases C1r2s2 and antibody oligomerization. Using well-defined monoclonal human antibodies and highly purified model systems to study C1q-IgG interactions on target surfaces, we here demonstrate that C1r2s2 proteases affect the capacity of C1q to form an avid complex with surface-bound IgG molecules. Although biophysical studies demonstrated that C1r2s2 proteases modify the highly flexible structure of C1q in solution 62-64 , it was not known how this would affect C1q binding to surface-bound IgG. Okada et al. (1985) proposed a potential role for C1r2s2 proteases in C1q binding to polyclonal rabbit IgG, but the mechanism and relevance for human IgG remained unclear 72 . Since our data now show that associated C1r2s2 are important for generating stable C1q-IgG complexes on surfaces, we propose that the associated proteases limit the conformational flexibility of the C1q arms and stabilize a spatial near-hexagonal arrangement of the six gC1q domains that facilitates multivalent binding of C1q to antibody clusters on the surface (Fig. 3C) .

Understanding how C1q molecules form a stable interaction with surface-bound IgGs is crucial for future design of therapeutic antibodies. Antibodies against bacteria and tumor cells cannot directly neutralize an infection but require activation of the immune system. For Gram-positive bacteria, the functionality of antibodies in human immune protection critically depends on the antibody's capacity to induce phagocytic killing 68, 73 , either directly or via complement. Here we show that Fc-engineering could be a useful strategy to combat Gram-positive bacteria, as both enhancement of C1q and C1 binding to S. aureus can improve the opsonisation and phagocytosis of these bacteria in human serum. Recent studies on tumor cells and Neisseria gonorrhoeae have already shown that enhanced hexamerisation of IgG by Fc-engineering can potentiate the capacity of monoclonal antibodies to induce complement-mediated lysis via MAC pores 15, 18, 19, 23 . Here, we demonstrate for the first time that the same mutations can be used to improve complement-dependent phagocytosis. Furthermore, since complement activation also appears crucial to induce protective antibodies against viral infections such as HIV or SARS-CoV2 74,75 , designing effective complement-triggering antibodies may also be valuable for therapeutic development of anti-viral antibodies 24 . In addition to its role in complement activation, increasing evidence suggests that C1q also enhances the efficacy of antibodies in the absence of other complement proteins. On tumor cells, it was shown that C1q can act as a structural component in potentiating the efficacy of anti-tumor antibodies by inducing outside-in signaling of death receptors on tumor cells 76 . Furthermore, C1q directly enhances the neutralizing activity of anti-viral antibodies in vitro and in vivo 77 .

Our study highlights that the IgG-Fc domain has a large impact on the antibody's capacity to form stable complexes with C1q. Detailed comparison of various monoclonal antibodies indicate that the difference in stability between C1-IgG and C1q-IgG is most apparent under conditions where IgG oligomerization is less favorable. Based on our results with hexamerenhancing mutations, we think that effective IgG oligomerization can overcome the need for proteases to establish stable C1q-IgG complexes, likely because six antibodies can also form a stable, high avidity complex with six C1q arms. Furthermore, hexamer-enhancing mutations in IgG2 and IgG1 also increased C1 binding to S. aureus, indicating that even when C1r2s2 is associated, efficient IgG oligomerization further stabilizes C1-IgG complexes. Interestingly, the finding that hexamer-enhanced IgG4 enabled binding to fully assembled C1, but less strongly to non-complexed C1q, could suggest that IgG4-E430G does not efficiently form hexamers. In all, we hypothesize that differences between human IgG subclasses to bind C1q (in the absence of proteases) could be linked to their ability to form hexameric platforms. However, also the binding affinity of IgG subclasses for gC1q may play a role. It remains to be determined whether our experiments on beads and Gram-positive bacteria (both rigid surfaces) can be translated to highly fluidic eukaryotic or Gram-negative membranes.

Our data also shed light on the mechanisms by which C1-INH inactivates C1, which are not yet fully defined. In 1998, Chen and Boackle already showed that purified C1-INH, besides removing C1r2s2, can dislodge C1q molecules from an IgG-coated surface. Although they used pooled human IgG in their C1-INH studies, they suggested that the IgG subclasses might be differentially affected by C1-INH 78, 79 . Our data now provide a better insight into the subclassspecific effects of C1-INH. They show that C1-INH inhibits the C1r2s2 proteases in surfacebound C1-IgG independent of the IgG subclass, but that it mediates C1q dissociation only in conditions where C1q-IgG complexes are not stable. This suggests that protease dissociation by C1-INH can result in either I) complete removal of both C1q and C1r2s2 from the IgG-coated surface, or II) removal of only C1r2s2 while C1q remains stably bound to the IgG platform (Fig.  4B) . Although C1-INH can inactivate C1r and C1s in solution too, it is important to note that we here only investigated the effect of C1-INH on C1 that is activated on an IgG-labelled surface.

Finally, our findings are also relevant for the correct interpretation of functional antibodies in research and diagnostics. C1q binding is often used as an important parameter to determine the presence of complement-fixing antibodies. For instance in transplantation, the serum of organ recipients is tested for the presence of donor-specific complement-binding anti-HLA antibodies (which could induce antibody-mediated rejection) by measuring binding of purified C1q [80] [81] [82] [83] . Based on our findings, direct C1q binding is an unreliable measure for downstream complement activation, as complement-binding IgGs that bind fully assembled C1, but not non-complexed C1q, are not detected. This also holds true for analyses of antibody-dependent complement activation in a human serum environment. For several antibodies, we observed that detection of surface-bound C1q is an unreliable read-out for IgG-dependent complement activation. This was most evident for wild-type IgG2 and IgG4-E430G, which could potently trigger C1-mediated complement activation (and phagocytosis), while we could hardly detect C1q on the surface. Although many studies have addressed the complement-binding andactivating properties of antibodies, surprisingly few have differentiated between C1q and C1. Our systematic investigation of C1q versus C1 binding to all IgG subclasses in both purified and human serum environments was key for the revelation of the importance of C1r2s2 in establishing stable C1q-IgG complexes. 17 , 20.2 ± 1.5 nm (anti-DNP IgG1-RGY + C1q), and 18.5 ± 1.7 nm (anti-DNP IgG1-RGY + C1). (C) gC1q domains mediate binding to IgG. In C1q, the gC1q domains are flexible. We hypothesize that in C1 the associated C1r2s2 proteases fix the collagen arms and thereby orient the gC1q domains in a hexagon-like platform that favors binding to (hexameric) IgG clusters. in (B) ). C1q or C1 was injected for 60 seconds to allow association, after which the injection was stopped and dissociation was monitored. C1 dissociated slower than C1q alone, indicating that C1 binds more stable. (B-D) Data show one representative of two independent experiments. SPR responses were normalized to account for the molecular weight difference between C1 (766 kDa) and C1q (410 kDa). RU, reponse units. 

SPR experiments showing binding of purified C1 or C1q to sensor chips that are subsequently coated with immobilized DNP and with 20 nM anti-DNP IgG (either wild-type, E430G or E345K mutants). C1 or C1q was injected for 60 seconds to allow association, after which the injection was stopped and dissociation was monitored. Data show representative of two independent experiments. SPR responses were normalized to account for the molecular weight difference between C1 (766 kDa) and C1q (410 kDa). Data shown for wild-type IgG1-4 are identical to those shown in Fig. 3A . RU, reponse units. Initially detected C1q molecules subsequently dissociate from IgG hexamers. Image size: 400 x 400 nm 2 (100 x 100 pixel 2 ), 1 s /frame. Movie S2: HS-AFM of the IgG1-RGY -C1 complex in the absence of C1 in solution. C1 molecules remain stably bound to IgG hexamers during HS-AFM imaging. Image size: 400 x 400 nm 2 (100 x 100 pixel 2 ), 1 s /frame. Table 1 : Protein sequences used for antibody production Variable and constant heavy and light chain protein sequences used for antbody production. The residues E345 and E430 are highlighted in grey.

All complement proteins were obtained from Complement Technology Inc. (TX, USA): C1 (A098), C1q (A099), C1r (A102), C1s (A104) and C1-inhibitor (A140).

Human serum from 20 healthy donors was pooled and depleted of IgG and IgM as previously described 33 . Informed consent was obtained from all subjects in accordance with the Declaration of Helsinki. Approval from the Medical Ethics Committee of the University Medical Center Utrecht was obtained (METC protocol 07-125/C approved on March 1, 2010). Briefly, IgG and IgM were separated from serum by affinity chromatography using HiTrap Protein G High Performance column (Cytiva, GE Healthcare) and CaptureSelect IgM Affinity Matrix (Thermofisher Scientific). Complement levels and complement activity were determined after depletion, using ELISA and classical/alternative pathway hemolytic assays respectively. Since C1q was partially co-depleted during the procedure, the ΔIgGΔIgM serum was reconstituted with C1q to physiological levels (70 µg/ml).

Recombinant monoclonal anti-DNP (DNP-G2a2) IgG1 antibody with the triple RGY mutation 15 used for HS-AFM experiments was obtained from Genmab (Utrecht, the Netherlands) 16, 84 . Other monoclonal human anti-DNP and anti-WTA IgG of the four IgG subclasses were produced recombinantly in human Expi293F cells (Life Technologies) as described before 51 with minor changes. The constant regions of heavy (HC) and light chain (LC) were cloned separately into pcDNA3.4 (ThermoFisher Scientific). Subsequently, gBlocks containing codon optimized variable (VH and VL) region sequences with an upstream KOZAK and HAVT20 signal peptide (Integrated DNA Technologies, IA, USA) were cloned upstream the HC and LC regions, respectively, using Gibson assembly (New England Biolabs). NheI and BsiWI were used as the 3' cloning sites for VH and VL, respectively, in order to preserve the immunoglobulin heavy and kappa light chain amino acid sequence. Sequences of IgG1, IgG2, IgG3, IgG4 variable and constant heavy and light chains were previously described 51,85 (Supplemental Table 1 ). VH and VL sequences were derived from previously described anti-WTA GlcNAc-β-4497 (anti-WTA-4497; based on patent WO/2014/193722) 86 and anti-DNP 87 (Supplemental Table 1 ). Single point mutation E430G or E345K was introduced in the heavy chain expression vectors either by PCR-mediated site directed mutagenesis or by gene synthesis (IDT). Antibody amino acids are numbered according to EU nomenclature 88 . After expression, IgG1, IgG2 and IgG4 antibodies were isolated from cell supernatant using a HiTrap protein A column (Cytiva, GE Healthcare). IgG3 antibodies were isolated using a HiTrap Protein G High Performance column (Cytiva, GE Healthcare). Antibodies were dialyzed overnight in PBS and filter-sterilized though 0.22 µm Spin-X filters. Antibodies were analyzed by size exclusion chromatography (SEC) (Cytiva, GE Healthcare) and separated for monomeric fraction in case aggregation levels were >5%. Antibody concentration was determined by measurement of the absorbance at 280 nm and antibodies were stored at -20°C.

Magnetic Dynabeads M-270 Streptavidin (Invitrogen) (~6-7 x 10 5 beads per sample) were washed in PBS-TH (PBS, pH 7.4, 0.05% (v/v) Tween, 0.5% human serum albumin (HSA)) and incubated with 0.003 or 1 μg/ml biotinylated 2,4-dinitrophenol (DNP-PEG2-GSGSGSGK(Biotin)-NH2; 1186 Da; synthesized by Pepscan Therapeutics B.V., the Netherlands) in 0.1 ml/sample PBS-TH for 30 minutes at 4°C, shaking (±700 rpm). Next, DNPcoated beads were washed once in PBS-TH and incubated in 0.05 ml/sample PBS-TH with 20 nM anti-DNP IgG or a 3-fold serial dilution of anti-DNP IgG (starting from 90 nM) for 30 minutes at 4°C, shaking (±700 rpm). Subsequent incubations mentioned below were performed in 0.025 ml/sample PBS-TH for 30 minutes at 4°C, shaking (±700 rpm), unless otherwise stated. After each incubation, beads were washed three times with PBS-TH or VBS-TH (Veronal Buffered Saline pH 7.4, 0.25 mM MgCl2, 0.5 mM CaCl2, 0.05% (v/v) Tween, 0.5% HSA) dependent on the buffer used in the subsequent incubation step. For detection of IgG binding, beads were incubated with 1 µg/ml goat anti-human kappa-AlexaFluor647 (Southern Biotech, 2060-31). Next, beads were washed and fixed in 0.15 ml/sample 1% paraformaldehyde in PBS-TH. For C1q binding experiments, beads were incubated in 0.025 ml/sample VBS-TH with a 3-fold serial dilution (starting at 10 nM) of C1 or C1q for 30 minutes at 37°C, shaking (±700 rpm).

To determine binding of 'C1-EDTA', C1 was incubated in VBS-THE (VBS-TH, 10 mM EDTA) instead. To study the effect of C1-INH and EDTA on C1q binding, beads were first incubated with 0.025 ml/sample 3 nM C1 and then with 0.025 ml/sample VBS-THE or 0.025 ml/sample 200 nM C1-INH in VBS-TH for 30 minutes at 37°C, shaking (±700 rpm). For C1q detection, beads were incubated with 5 μg/ml FITC-conjugated rabbit anti-human C1q (Dako, F0254). To determine IgG, C1q and C3b deposition to beads in human serum, DNP-coated beads (~6-7 x 10 5 beads per sample) were incubated in 0.05 ml/sample VBS-TH with 20 nM anti-DNP IgG and a 3-fold serial dilution of ΔIgGΔIgM serum (starting from 10%) for 30 minutes at 37°C, shaking (±700 rpm). Otherwise, DNP-coated beads were incubated in VBS-TH with a 3-fold serial dilution of anti-DNP IgG (starting from 90 nM) and 1% ΔIgGΔIgM serum for 30 minutes at 37°C, shaking (±700 rpm). For IgG, C1q and C3b detection, beads were washed and incubated in 0.025 ml/sample PBS-TH with a detection antibody for either IgG, C1q or C3b (1:300 FITC-conjugated goat anti-C3 IgG, De Beer Medicals, 1100). After incubation with detection antibodies, bead samples were washed and fixed in 0.15 ml/sample 1% paraformaldehyde in PBS-TH and binding of IgG, C1q or C3b to the beads was determined by flow cytometry (BD FACSVerse). Data were analyzed based on single bead population using FlowJo software and presented as Fluorescence Intensity (FI) means ± SD or geometric means ± SD of at least three independent experiments.

To determine binding of different forms of C1 to DNP-beads by western blotting, the amount of beads (and incubation volumes) were scaled up 4-fold (~24-28 x 10 5 beads/sample). Magnetic Dynabeads M-270 were successively incubated with 1 µg/ml DNP-biotin, 20 nM anti-DNP IgG and 3 nM C1q/C1/C1-EDTA, as described above. Next, beads were washed in 

Beyond binding: antibody effector functions in infectious diseases

First of two parts

Complement driven by conformational changes

Structural Implications for the Formation and Function of the Complement Effector Protein iC3b

Complement: a key system for immune surveillance and homeostasis

Structural Basis for the Function of Complement Component C4 within the Classical and Lectin Pathways of Complement

The internal thioester and the covalent binding properties of the complement proteins C3 and C4

Bacterial killing by complement requires membrane attack complex formation via surface-bound C5 convertases

Complement defense mechanisms

Distinct localization of the complement C5b-9 complex on Gram-positive bacteria

Complement and its role in innate and adaptive immune responses

Complement is activated by IgG hexamers assembled at the cell surface

Structures of C1-IgG1 provide insights into how danger pattern recognition activates complement. Science (80-. )

Unraveling the Macromolecular Pathways of IgG Oligomerization and Complement Activation on Antigenic Surfaces

A Novel Platform for the Potentiation of Therapeutic Antibodies Based on Antigen-Dependent Formation of IgG Hexamers at the Cell Surface

Molecular Basis of Assembly and Activation of Complement Component C1 in Complex with Immunoglobulin G1 and Antigen

Reaction between the isolated globular sub-units of the complement component C1q and IgG-complexes

Immunoglobulin flexibility in complement activation

Weak Fragment Crystallizable (Fc) Domain Interactions Drive the Dynamic Assembly of IgG Oligomers upon Antigen Recognition

Complement alone drives efficacy of a chimeric antigonococcal Component

Non-linearity of the collagen triple helix in solution and implications for collagen function

Conformation and restricted segmental flexibility of C1, the first component of human complement

Insights into IgM-mediated complement activation based on in situ structures of IgM-C1-C4b

The critical concentration of C1-esterase inhibitor (C1-INH) in human serum preventing auto-activation of the first component of complement (C1)

The structural requirements for complement activation by IgG: does it hinge on the hinge?

Neutrophil-Mediated Phagocytosis of Staphylococcus aureus

Phagocytosis Escape by a Staphylococcus aureus Protein That Connects Complement and Coagulation Proteins at the Bacterial Surface

Immune evasion by a staphylococcal complement inhibitor that acts on C3 convertases

Anti-opsonic properties of staphylokinase

Co-operative interaction of subcomponents of the first component of complement with IgG: a functional defect of dimeric Facb from rabbit IgG

Antibody-Dependent Cellular Phagocytosis and Its Impact on Pathogen Control

HIV-specific Fc effector function early in infection predicts the development of broadly neutralizing antibodies

Distinct Early Serological Signatures Track with SARS-CoV-2 Survival

Dual Epitope Targeting and Enhanced Hexamerization by DR5 Antibodies as a Novel Approach to Induce Potent Antitumor Activity Through DR5 Agonism

Complement protein C1q reduces the stoichiometric threshold for antibody-mediated neutralization of West Nile virus

C1 inhibitor removes the entire C1qr2s2 complex from anti-C1Q monoclonal antibodies with low binding affinities

A Newly Discovered Function for C1 Inhibitor, Removal of the Entire C1qr2s2Complex from Immobilized Human IgG Subclasses

Novel C1q assay reveals a clinically relevant subset of human leukocyte antigen antibodies independent of immunoglobulin G strength on single antigen beads

The role of complement in antibody mediated transplant rejection

The Biology of IgG Subclasses and Their Clinical Relevance to Transplantation

Technical Limitations of the C1q Single-Antigen Bead Assay to Detect Complement Binding HLA-Specific Antibodies

Effect of immunoglobulin variable region structure on C3b and C4b deposition

Staphylococcal protein A inhibits complement activation by interfering with IgG hexamer formation

Novel antibody-antibiotic conjugate eliminates intracellular S. aureus

Minimizing immunogenicity of the SDR-grafted humanized antibody CC49 by genetic manipulation of the framework residues

Sequences of proteins of immunological interest. (U.S. Dept. of Health and Human Services

Synthetic Effects of secG and secY2 Mutations on Exoproteome Biogenesis in Staphylococcus aureus

The Staphylococcus aureus polysaccharide capsule and Efbdependent fibrinogen shield act in concert to protect against phagocytosis

Staphylococcal superantigen-like 5 binds PSGL-1 and inhibits Pselectin-mediated neutrophil rolling

A high-speed atomic force microscope for studying biological macromolecules

IgGs are made for walking on bacterial and viral surfaces

Efficient neutralization of primary isolates of HIV-1 by a recombinant human monoclonal antibody

50 µL sample buffer (0.5 M Tris-HCl, 2% (w/v) SDS, 20% (v/v) glycerol, 0.1% (w/v) bromophenol blue, pH 7) with 25 mg/ml dithiothreitol (DTT) for 5 minutes at 95°C. Beads were pelleted and 10 µL of the sample supernatant was run on a 10%

Blots were washed between the successive antibody incubations. To determine the removal of C1r and C1s from bead-bound C1 by C1-INH by western blotting, the amount of beads (and incubation volumes) were scaled up 3-fold (~18-21 x 10 5 beads/sample)

Next, beads were pelleted and supernatant was collected and mixed 1:1 with 2x sample buffer containing 50 mg/ml DTT for 5 minutes at 95°C. Next, 10 µL samples were run on a 10% SDS-PAGE gel, transferred to a PVDF membrane and membranes were blocked, as described above

Wood46 (with low expression of staphylococcal protein A (Spa)) or S. aureus Newman Δspa/sbi (knock-out of Protein A and the second immunoglobulin-binding protein (Sbi)) 89 was or was not fluorescently labelled by transformation with a GFP-expressing plasmid 90 . Bacteria were grown until log phase in Todd Hewitt Broth medium, washed and frozen at -20°C until use. To prepare samples, bacteria were thawed and diluted to OD=0.1 in RPMI-H (RPMI, 0.05% HSA). All subsequent incubations mentioned below were performed in RPMI-H at 4°C, shaking (±700 rpm), unless stated otherwise. After each incubation, bacteria were washed once in RPMI-H by centrifugation. For detection of IgG binding, 0.025 ml bacteria was incubated with 0.025 ml 3-fold serial dilution anti-WTA IgG (starting from 90 nM, final concentration). Next, bacteria were

To determine binding of 'C1-EDTA', C1 was incubated in RPMI-HE (RPMI-H, 10 mM EDTA) instead. For C1q detection, bacteria were incubated with 5 μg/ml FITC-conjugated rabbit anti-human C1q. To determine IgG, C1q and C3b binding to bacteria in serum, 0.050 ml bacteria were mixed with 0.025 ml 3-fold serial dilution anti-WTA IgG (starting from 90 nM, final concentration) and 0.025 ml 1% (IgG or C1q detection) or 5% (C3b detection) ΔIgGΔIgM serum (final concentration) and incubated 30 minutes at 37°C, shaking (±700 rpm). For IgG, C1q or C3b detection bacteria were next incubated with detection antibody for IgG

020 ml bacteria were mixed with 0.030 ml RPMI-H with a 3-fold serial dilution anti-WTA IgG (starting from 67 nM, final concentration), 1% ΔIgGΔIgM serum (final concentration) and either monoclonal anti-C1q antibody (HB8327 a4b11 46 , ATCC, 10 µg/ml final concentration) or BBK32 (kindly provided by Dr. Brandon Garcia, 10 µg/ml final concentration). Bacteria were incubated 20 minutes at 37°C, shaking (±700 rpm). For C3b detection bacteria were next incubated with FITC

After incubation with detection antibodies, bacteria were washed and fixed in 0.15 ml/sample 1% paraformaldehyde in RPMI-H and analyzed using flow cytometry (BD FACSVerse). Data were analyzed by FlowJo software and presented as Fluorescence Intensity (FI) means ± SD or geometric means ± SD of at least three independent experiments

The authors kindly thank dr. Rob de Jong, dr. Annette Stemerding, dr. Lubka Roemenia and dr. Leendert Trouw and for scientific advice. We thank dr. Brandon Garcia for providing BBK32. This work was supported by an ERC Starting Grant (639209-ComBact, to S.H.M.R) and ERC Advanced Grant (233229-Coco, to P.G.), the Utrecht University Molecular Immunology Hub (UMI-Hub, ESTIMATE), and the Swiss National Science Foundation (31003A_176104, to DR). JP acknowledges support by the European Fund for Regional Development (EFRE, Regio 13), the Federal State of Upper Austria, and the Austrian Science Fund (FWF, P33958, P34164).

AK, JAGS, PWHIP, KPMK, FJB, JS and SHMR are co-inventor on a patent describing antibody therapies against S. aureus.

Human neutrophils were isolated from blood of healthy donors by the Ficoll-Histopaque gradient method 91 . Phagocytosis was performed in a round-bottom 96-well plate. GFP-labelled Wood46 bacteria (20 µL of 3.75 x 10 7 cells/ml) were mixed with 2-fold serial dilutions of anti-WTA IgG in RPMI-H or 1% ΔIgGΔIgM serum (20 µL volume) for 15 minutes at 37°C for opsonization. Subsequently, neutrophils (10 µL of 7.5 x 10 6 cells/ml) were added giving a 10:1 bacteria to cell ratio and incubated for 15 minutes at 37°C on a shaker (750 rpm) in a final volume of 50 µL. The reaction was stopped with 1% ice-cold paraformaldehyde and neutrophil-associated fluorescent bacteria were analyzed by flow cytometry by scatter gating on neutrophils. Phagocytosis was defined by the percentage of cells with a positive fluorescent signal (% GFP-positive cells) of all neutrophils representing the overall phagocytosis efficacy.

All surface plasmon resonance experiments were performed at 25ºC using a Biacore T200 instrument (Cytiva, GE Healthcare) equipped with a flat carboxymethyldextran sensor chip (CMDP; Xantec). For DNP surface preparations, flow cells were activated for 7 minutes with 20 nM were prepared in HBST ++ and injected for 120 seconds over the DNP and control surfaces. Subsequently, C1 or C1q (20 nM) were injected for 60 seconds to associate on top of the antibodies and dissociation was observed for 300 seconds. The surfaces were regenerated with a 60 second injection of 10 mM glycine pH 2.0 at a flow rate of 20 µL/min. Data were collected at a rate of 1 Hz and analyzed with Scrubber (version 2.0c; BioLogic). Reference signals from the control surface (MeO) were subtracted from the DNP surface signals. The responses were normalized for analyte size contribution by dividing the data points of each analyte by the molecular weight of C1 (766 kDa) or C1q (410 kDa), respectively, and multiplying them by 100. The data were exported to Prism (version 8.1.2.; Graphpad). (molar ratio), and dissolved in a 2:1 mixture of chloroform and methanol. After the solvents were rotary-evaporated for 30 minutes, the lipids were again dissolved in chloroform, which was then rotary-evaporated for 30 minutes. Drying was completed at a high vacuum pump for 2 hours. The lipids were dissolved in 500 µL Milli-Q H2O while immersed in a water bath at 60°C, flooded with argon, and sonicated for 3 minutes at 60°C to create small unilamellar vesicles. These were diluted to 2 mg/ml in buffer #1 (10 mM HEPES, 150 mM NaCl, 2 mM CaCl2, pH 7.4) and frozen for storage using liquid N2.

High-speed atomic force microscopy HS-AFM 92,93 (RIBM, Japan) was conducted in tapping mode at RT in buffer, with free amplitudes of 1.5 -2.5 nm and amplitude set points larger than 90%. Silicon nitride cantilevers with electron-beam deposited tips (USC-F1.2-k0.15, Nanoworld AG), nominal spring constants of 0.15 N m -1 , resonance frequencies around 500 kHz, and a quality factor of approx. 2 in liquids were used. Imaging was performed in buffer #1. All IgGs were diluted and incubated in the same buffer. DNP-labelled supported lipid bilayers for HS-AFM were prepared on muscovite mica. The liposomes were incubated on the freshly cleaved surface (500 µg/ml in buffer #1), placed in a humidity chamber to prevent evaporation, and heated to 60 °C for 30 minutes. Then the temperature was gradually cooled down to RT within 30 minutes, followed by exchanging the solution with buffer #1. After 10 minutes of equilibration at RT and 15 more buffer exchanges, the SLB was ready for imaging. In order to passivate any exposed mica, SLBs were incubated with 333 nM IgG1-b12 (irrelevant human IgG1 control antibody against HIV-1 gp120) 94 for 10 minutes before the molecules of interest were added. Anti-DNP IgG1-RGY (20 µg/ml) was incubated for 5 minutes followed by several buffer exchanges to remove umbound IgGs from solution. After that, C1q (20 µg/ml) or C1 (15 µg/ml) was added for 10 minutes, followed by another buffer exchange. In case of C1q, imaging was performed in the presence of C1q in soluiton to enable rebinding during the experiment.

Statistical analysis was performed with Prism software (version 8.0.1; GraphPad). All data are presented as means ± SD or geometric means ± SD from three independent experiments unless otherwise stated.