CH2 Domain of Mouse IgG3 Governs Antibody Oligomerization, Increases Functional Affinity to Multivalent Antigens and Enhances Hemagglutination

Mouse IgG3 is highly protective against several life-threatening bacteria. This isotype is the only one among mouse IgGs that forms non-covalent oligomers, has increased functional affinity to polyvalent antigens, and efficiently agglutinates erythrocytes. IgG3 also triggers the complement cascade. The high efficacy of protection after passive immunization with IgG3 is correlated with the unique properties of this isotype. Although the features of IgG3 are well documented, their molecular basis remains elusive. Based on functional analyses of IgG1/IgG3 hybrid molecules with swapped constant domains, we identified IgG3-derived CH2 domain as a major determinant of antibody oligomerization and increased functional affinity to a multivalent antigen. The CH2 domain was also crucial for efficient hemagglutination triggered by IgG3 and was indispensable for complement cascade activation. This domain is glycosylated and atypically charged. A mutational analysis based on molecular models of CH2 domain charge distribution indicated that the functional affinity was influenced by the specific charge location. N-glycans were essential for CH2-dependent enhancement of hemagglutination and complement activation. Oligomerization was independent of CH2 charge and glycosylation. We also verified that known C1q-binding motifs are functional in mouse IgG3 but not in IgG1 framework. We generated for the first time a gain-of-function antibody with properties transferred from IgG3 into IgG1 by replacing the CH2 domain. Finding that the CH2 domain of IgG3 governs unique properties of this isotype is likely to open an avenue toward the generation of IgG3-inspired antibodies that will be protective against existing or emerging lethal pathogens.

Mouse IgG3 is highly protective against several life-threatening bacteria. This isotype is the only one among mouse IgGs that forms non-covalent oligomers, has increased functional affinity to polyvalent antigens, and efficiently agglutinates erythrocytes. IgG3 also triggers the complement cascade. The high efficacy of protection after passive immunization with IgG3 is correlated with the unique properties of this isotype. Although the features of IgG3 are well documented, their molecular basis remains elusive. Based on functional analyses of IgG1/IgG3 hybrid molecules with swapped constant domains, we identified IgG3-derived CH2 domain as a major determinant of antibody oligomerization and increased functional affinity to a multivalent antigen. The CH2 domain was also crucial for efficient hemagglutination triggered by IgG3 and was indispensable for complement cascade activation. This domain is glycosylated and atypically charged. A mutational analysis based on molecular models of CH2 domain charge distribution indicated that the functional affinity was influenced by the specific charge location. N-glycans were essential for CH2-dependent enhancement of hemagglutination and complement activation. Oligomerization was independent of CH2 charge and glycosylation. We also verified that known C1q-binding motifs are functional in mouse IgG3 but not in IgG1 framework. We generated for the first time a gain-of-function antibody with properties transferred from IgG3 into IgG1 by replacing the CH2 domain. Finding that the CH2 domain of IgG3 governs unique properties of this isotype is likely to open an avenue toward the generation of IgG3-inspired antibodies that will be protective against existing or emerging lethal pathogens.
Keywords: igg3, oligomerization, multivalent antigen, polyvalent antigen, hemagglutination, ig constant region inTrODUcTiOn There are four subclasses of mouse IgGs: IgG1, IgG2a, IgG2b, and IgG3. Although structurally very similar, they significantly differ in their functions (1). Mouse IgG3s are particularly interesting, because they are able to form oligomers, which strongly influences their biological activities (2).
Mouse IgG3 was described for the first time almost 50 years ago (3) and different aspects of its biology have been investigated by several groups. The propensity of IgG3 oligomerization was noticed already by its discoverers (3). Then, other researchers reported cooperative binding of IgG3 to a multivalent antigen (4,5). Although the initial report on IgG3 oligomerization concerned molecules in solution (3), later studies revealed that binding to multivalent antigens promoted IgG3 intermolecular interactions, which in turn resulted in its increased functional affinity to the antigen (4). This phenomenon depended on Fc, because IgG3 F(ab')2 fragments did not bind to the antigen cooperatively (4). However, the exact molecular mechanism of IgG3 oligomerization remains unknown.
IgG3 is a major component of cryoglobulins in mice (2). Cryo globulins are plasma proteins that reversibly precipitate at low temperatures or at high concentrations (6). Cryogenic activity of IgG3 was shown to correlate with its ability to oligomerize, with the presence of charged residues in the variable region and the level of sialylation (7).
Also, IgG3 was reported as exceptionally effective in preventing or fighting several lifethreatening microbial infections, e.g., with Neisseria meningitidis (8) or Bacillus anthracis (9). A comp arison between the four mouse IgG subclasses with the same variable region specific to B. anthracis capsule proved that only IgG3 is protective against pulmonary anthrax in a mouse model (9). Importantly, mouse-human chimeric antibodies containing a con stant region of any human IgG subclass were not effective, although they had the same variable region as the protective mouse IgG3 (10). These reports indicated that mouse IgG3 constant region has unique properties, but the authors only speculated about possible molecular mechanisms behind the observed phenomenon.
The exceptional characteristic of IgG3 was also confirmed by our previous report that among mouse IgGs recognizing a surface antigen of erythrocytes only IgG3s are able to agglutinate the cells (11). We rejected the hypothesis that IgG3mediated hemagglutination results from oligomerization of the antibodies, because IgG3 F(ab')2 was sufficient to trigger hemagglutination. Molecular modeling indicated that IgG3 has a larger span of Fab arms than other IgG subclasses. IgG3 has a longupper hinge that may extend the Fab range, but whether this may account for the ability to hemagglutinate was not verified experimentally.
Here, we present the results of our attempts to find molecular determinants of the properties of mouse IgG3. Our experimental model comprised two antibodies (M18 and O10) specific to antigen B of the ABO blood group system (11). Antigen B is a pen tasaccharide Oglycan attached to numerous proteins and lipids on the erythrocyte surface. The large quantity and high density of the antigen, as well as a strong negative charge of the erythrocyte surface, could be considered as a safe and easytohandle model of a pathogen surface. We generated many muteins of IgG3 molecules and analyzed their functional affinity to the antigen as well as their ability to hemagglutinate and oligomerize. The results showed that IgG3derived CH2 domain determines antibody oligomeri zation and increases its affinity to the antigen. This domain also strongly enhances agglutination of erythrocytes bearing B antigen. Moreover, we investigated complement activation by the muteins, and we confirmed that known C1qbinding motifs are functional in mouse IgG3 but not in the mouse IgG1 framework.

generation of Vectors coding for antibody Muteins
Expression vectors coding for heavy chain muteins with swapped domains or mutated CH2 were generated using synthetic nucleic acids (Gene Art, Germany) cloned into pFUSEssCHIgmG1_ M18 (Addgene #82357) or pFUSEssCHIgmG3_M18 (Addgene #82356) plasmids (11). Only endogenous restriction sites present in the ORFs were used for cloning. Vectors coding for other muteins were prepared with Q5based site directed mutagenesis kit (NEB). All plasmids coding for heavy chain variants of M18 antibody are available via Addgene repository along with their full sequences and maps, accession numbers: 105849-105863. Plasmids encoding O10 antibody variants were obtained by replacing the sequence coding for M18 variable fragment with a corresponding O10derived cDNA using EcoRI and AfeI restric tion sites in the vectors coding for M18 heavy chain variants. The sequence of O10 antibody is proprietary and cannot be disclosed. All plasmids were verified using Sanger sequencing (Genomed, Poland) or NGS (Addgene).

Production of antibodies
The recombinant antibodies were transiently expressed in HEK293T cells cultured in DMEM with 4.5 g/l glucose (Lonza) supplemented with 10% FBS (Biowest). The cells were cotransfected with plasmids coding for a heavy chain and a cognate light chain using Lipofectamine2000 (Thermo) or PEI MAX (Polysciences, MW 40,000). In the case of M18 variants, the plasmid pFUSE2ssCLIgmK_M18 (Addgene, #82358) (11) coding for M18 light chain was used. A similar plasmid coding for O10 light chain was used for expression of O10 variants. Hybridomaderived M18 IgG3 was produced as described pre viously (11). The antibodies were purified using CaptureSelect LCkappa (mur) affinity matrix (Thermo) according to instruc tions of the supplier. Glycine-HCl (100 mM, pH 2.0) was used for elution.

Production of igg3 F(ab') 2
A sequence encoding IgG2b core hinge, HA tag and a stop codon was cloned into pFUSEssCHIgmG3_M18 downstream of the sequence coding for the upper hinge of M18. Sequences of the upper and core hinges of the antibodies are described by Dangl et al (12). Mouse IgG2b core hinge contains four cysteine residues that allow an efficient association of Fab' fragments into F(ab')2 (13). Recombinant F(ab')2 was expressed as described for other antibodies. Alternatively, F(ab')2 was produced by enzy matic digestion with IdeZ (NEB) according to the manufacturer's protocol.

Measurements of antibody concentration
Antibody concentrations in cell culture media were measured using a standard sandwich ELISA on plates coated with sheep antimouse Fab polyclonal antibody (Jackson Laboratory, cat. #515005072, lot #105461). Goat antimouse kappa polyclonal antibody (1:3,000, BioRad, cat. #105008, batch #160617), HRP labeled streptavidin (1:40,000, Sigma), and the TMB substrate for HRP (BD Bioscience) were used for detection. The HRP dependent colorogenic reaction was stopped with 1 M HCl, and the absorbance at 450 nm was read using the microplate spectrophotometer Synergy H1 operated by Gen5 2.00 Software (BioTek). Purified M18 (IgG3) and MCP21 (IgG1, Sigma) were used as standards. Concentrations of muteins were calculated based on their Fab type (IgG1 or IgG3type Fab). BCA assay (Sigma) was used for measurements of purified antibody concen trations with bovine γglobulin (Thermo) as a reference.

antibody Binding to immobilized erythrocytes
Polystyrene plates were coated with 50 µg/ml polylLys (Sigma) for 1 h at room temperature (RT). Then, 100 µl of 0.1% (hematocrit) suspension of red blood cells in PBS were added to the wells and the cells were allowed to settle for 1 h at RT. After gentle aspiration of the solution, the cells were fixed using 0.025% glutaraldehyde for 40 min. Endogenous peroxidase activity was blocked with 3% H2O2 for 1 h. The plates were blocked overnight with 0.2% gelatin in PBS containing 0.05% Tween20 at 4°C. Then, the cells were incubated with serial dilutions of cell culture media containing analyzed antibodies followed by detection with antimouse kappa polyclonal antibody conjugated with biotin (1:3,000, BioRad) and HRPlabeled streptavidin (1:40,000, Sigma). All reagents were diluted in the blocking buffer. The colorogenic reaction was performed, and the absorbance at 450 nm was measured as described above.

c1q Binding
Duplicates of polystyrene plates were coated with 6 µg/ml of BSA conjugated with the discriminating trisaccharide of the B antigen (Dextra Laboratories, cat #NGP6323, batch #ATDX232039) overnight at 4°C, blocked with 1% BSA in PBS (1 h) and incubated with serial dilutions of cell culture media containing O10 muteins (2 h). Then, one set of plates was used for evaluation of C1q binding by analyzed antibodies; and the plates were incubated with 2 µg/ml of C1q purified from human serum (Biorad, cat. #22215504, batch #130815; 2 h) and next with HRPlabeled antihuman C1q polyclonal antibody (1:400, Abcam, cat. #ab46191, lot #GR2054365; 1 h). The second set of plates was used to analyze quantities of the muteins bound to the immobilized antigen; and the plates were incubated with antimouse kappa polyclonal antibody conjugated with biotin (1:3,000, BioRad, 2 h) and HRPlabeled streptavidin (1:40,000, Sigma; 1 h).
Absorbance of HRP product was measured as described above. Each incubation step was preceded by extensive washing with 0.05% Tween20 in PBS. The antibodies, streptavidin, and C1q were diluted in 0.1% BSA in PBS. All incubations, except plate coating, were at RT.
The normalized C1q binding (a) was calculated by dividing the signal corresponding to C1q binding (b) by the signal cor responding to bound antibody (c). Washed human red blood cells suspended to a hematocrit of 2% were coated with 3 µg/ml or 1.5 µg/ml of the analyzed antibodies for 1.5 h at RT, washed twice with PBS and resuspended in PBS with Ca 2+ and Mg 2+ . Then, the same volume of human comple ment serum (Sigma, cat. #S1764, lot #SLBS5471V, #SLBQ0752V or #SLBP0461V) diluted to 7 CH50 U/ml in PBS with Ca 2+ and Mg 2+ was added to the coated red blood cells. The samples were centrifuged after 2 h of incubation at 37°C, and the absorbance of released hemoglobin was measured in the supernatants at 540 nm.

red Blood cells and agglutination
Standard human red blood cells were purchased from Regional Centre of Blood Donation and Blood Treatment in Katowice, Poland. Agglutination tests were performed in 96flat bottom plates. Serially diluted solutions of analyzed antibodies (100 µl) were gently mixed with 0.45% (hematocrit) suspension of red blood cells. The level of agglutination was analyzed using a phase contrast microscope after 20 min of moderate shaking. A sixpoint scale was used for evaluation of agglutination intensity: from 4+ (complete cell aggregation), to 3+, 2+, 1+, ± to a negative score. The agglutination score reflects both the size of aggregates and quantity of nonagglutinated cells.
Concentrations of IgG3 and F(ab')2 in the media were measured using ELISA and equalized for the comparative tests. We verified the quality of analyzed proteins and confirmed that recombinant IdeZ protease, which cleaves IgG3 at a single site in its hinge region, generates homogeneous F(ab')2 (Figures 1A,B). Based on this analysis, we estimated that IgG3 was about 32 to 64times more potent than its F(ab')2 in hemagglutination ( Table 1; Figure 1B). The results indicate that the Fc of IgG3 strongly enhances hemagglutination induced by this isotype.
The ch2 Domain Derived From igg3 enhanced hemagglutination efficacy of an antibody Our previous attempts to explain the mechanism of IgG3dependent hemagglutination brought us to the hypothesis that the elongated hinge of IgG3 determines its hemagglutination ability (11). In light of the new results, the hypothesis required revision. To elucidate which domains of IgG3 are crucial for its hemagglutination abil ity, we generated a panel of domain muteins of agglutinating IgG3 and nonagglutinating IgG1 isotypes (Figure 2A). We generated pairs of IgG1 and IgG3 molecules with the same variable regions and with swapped: (i) hinge regions; (ii) hinge regions + CH1 domains; (iii) CH2 domains, and (iv) CH3 domains and searched for muteins of two types: lossoffunction in the case of IgG3 and gainoffunction in the case of IgG1. Preliminary experiments showed that hinge swapping between IgG1 and IgG3 hinders disulfide bonds formation between chains of the muteins ( Figure S1 in Supplementary Material). However, the IgG1_h3 and IgG3_h1 variants were functional and their affinity to the antigen was similar to that of the parental molecules (shown in Figure 3). As demonstrated by Dall' Acqua et al., immunoglobulins with modified hinges frequently form functional heterotetramers (HC)2(LC)2 despite the lack of disulfide bonds between the chains (15). Gel filtration confirmed that IgG1_h3 and IgG3_h1 have molecular mass greater than 150 kDa and form stable (HC)2(LC)2 heterotetramers (data not shown). To make sure that the results of the following experiments are not the consequence of incorrect assemblies of the hinge swapped muteins, we also generated muteins with swapped frag ments comprising hinge regions and CH1 domains. All muteins were successfully expressed and their integrity was verified using SDSPAGE and Western blotting ( Figure S1 in Supplementary Material).
Then, we compared hemagglutination induced by the muteins (Tables 2 and 3; Figures 2B,C). The results showed that neither Results of two independent experiments designated as I and II.  (Tables 2 and 3). Moreover, the paired IgG3 mutein (IgG3_CH21) had about 16times reduced hemaggluti nation score in comparison to the parental IgG3. IgG1_CH23 as a gainoffunction mutein was particularly interesting, because it indicated that the CH2 domain of IgG3 is the one critical for hemagglutination. However, IgG1_CH23 aggluti nated erythrocytes with considerably lower score than native IgG3.
We also compared hemagglutination efficacy of native M18 IgG3 and its deglycosylated form. Deglycosylated IgG3 agglu tinated erythrocytes about 16times weaker than the native molecule (Table S1 in Supplementary Material).
To sum up, the ability of IgG3 to agglutinate erythrocytes results from its unique structure, in which the CH2 domain is especially important and strongly enhances the efficacy of the process. Although the IgG3 F(ab')2 is sufficient to trigger hemagglutina tion, its efficacy is much lower in comparison to fulllength IgG3, probably just due to the lack of the CH2 domain. The hinge region seems to have little influence on agglutination ability, because IgG3 with the IgG1derived hinge agglutinated erythrocytes only slightly less effectively than the parental molecule.

igg3 constant Domains Modify Functional affinity to an antigen
There is a general agreement that the increased functional affinity of IgG3 results from an avidity effect caused by the interactions between the Fc fragments of the molecules (16). In line with that, we observed that IgG3 binds to erythrocytes much more effi ciently than IgG1 with the same variable region and much more efficiently than IgG3derived F(ab')2 ( Figure S2 in Supplementary Material). Some authors discussed also the potential role of Nglycans in IgG3 unique properties (17,18), but we did not observe any differences in antigen binding between control and deglycosylated antibody ( Figure S2 in Supplementary Material).
Aiming to understand why IgG3 has increased functional affinity, we analyzed antigen binding by the domain muteins ( Figure 3). The results showed that the hinge region of IgG3 does not influence the functional affinity of the antibody, but muteins with the swapped CH1 + hinge, CH2, or CH3 domains had changed functional affinity. The introduction of IgG3derived CH1 + hinge or CH2 domain into the IgG1 framework enhanced antigen binding in comparison to the parental IgG1. Conversely, the paired IgG3 muteins with IgG1 derived CH1 + hinge or CH2 had reduced functional affinity. The swapping of the CH3 domains resulted in the IgG3 mutein with decreased affinity, but in the paired IgG1 mutein, the effect was not substantial. The calculated EC50 values of antigen bind ing for IgG3 muteins indicated that the CH2 domain had the strongest influence on IgG3antigen interaction ( Table 4). CH2 swapping resulted in IgG3 muteins with 3-12 times decreased functional affinity.
Overall, the results indicate that the higher (in comparison to IgG1) functional affinity of IgG3 to its antigen does not depend on a separate constant domain of this isotype, but rather is an additive result of discrete properties of the all three constant domains CH1, CH2, and CH3, but not the hinge region. Of all the constant domains, CH2 contributes the most to the high functional affinity of IgG3.

Fc-Dependent Oligomerization of the Domain Muteins
The hallmark of mouse IgG3 is its ability to oligomerize. The process depends on Fc fragment, but its exact molecular mecha nism is unknown. We analyzed whether the domain muteins form noncovalent complexes using polyethylene glycol (PEG) precipitation with a labeled IgG3 probe (2). In comparison to the original method, we used biotinylated IgG3 instead of radiolabeled IgG3. The IgG3biotin interacted with oligomerizing muteins and the complexes comprised the mutein and the probe. The com plexes were precipitated using PEG, and then, IgG3biotin was quantified in precipitates and supernatants using ELISA. A high precipitate/supernatant ratio of IgG3biotin quantities indicates that the mutein forms oligomers.
The experiment showed that 5 out of 10 analyzed molecules form PEGprecipitable oligomers-IgG3 (control) and all IgG3 muteins but the one containing IgG1derived CH2 domain and none of IgG1 muteins but the one with IgG3derived CH2 domain FigUre 4 | Oligomerization of the domain muteins. The antibodies (M18 variants, 150 µg/ml) were incubated at 4°C and oligomers were precipitated using PEG. Biotinylated IgG3 was used as a probe that oligomerized with the muteins and became a part of the complexes. The charts present results from two independent experiments. The results obtained for 100 and 20 µg/ ml of the muteins are shown in Figure S3 in Supplementary Material. A percentage of the total IgG3-biotin detected in the precipitates and the supernatants is presented in Figure S4 in Supplementary Material.  (Figure 4). Oligomerization did not depend on CH2 glycosyla tion ( Figure S5 in Supplementary Material). The results indicate that the CH2 domain is crucial for oligomerization of IgG3.

complement activation by the igg1 and igg3 Muteins
Similar to human antibodies, there are pronounced differences between mouse IgG subclasses in their ability to trigger comple ment cascade. Mouse IgG3 activates complement efficiently, whereas mouse IgG1 does not. Although there are many excel lent reports concerning correlation between human antibody structure and its ability of complement activation, the structural determinants of mouse antibodies that allow to trigger the cas cade are not precisely known. The best characterized Ig with respect to complement activa tion is human IgG1, in which several amino acid residues were identified as crucial for the initiation of the complement cascade ( Figure S6 in Supplementary Material) (19)(20)(21). The sequence alignment of human IgG1, mouse IgG1, and mouse IgG3 indicated that the majority of human IgG1 amino acid residues involved in complement activation are conserved in both mouse isotypes ( Figure S6 in Supplementary Material). However, it revealed two differences between mouse IgG1 and IgG3 within the regions cor responding to those involved in C1q binding by human IgG1-in the Nterminal fragment of the CH2 domain (Val231Ser238 in IgG1 and Ile234Pro238 in IgG3, EU numbering (22)) and in the residue 322 ( Figure S6 in Supplementary Material).
To verify whether these motifs are involved in complement activation by mouse IgG3, we generated additional muteins in which we swapped them between IgG1 and IgG3-IgG1_ILGGP (Val231Ile Pro232Leu Glu236Gly Val237Gly Ser238Pro); IgG3_ VPEVS (Ile234Val Leu235Pro Gly236Glu Gly237Val Pro238Ser); IgG1_Arg322Lys; IgG3_Lys322Arg, and a double mutein IgG1_ ILGGP_Arg322Lys. The IgG3 heavy chain containing VPEVS did not associate with a light chain and was not secreted (Figure 5A). Lys322Arg replacement completely abolished complement acti vation by IgG3 indicating that Lys322 is crucial for this process (Figures 5B,C). The three muteins of IgG1 did not bind C1q nor activated complement cascade indicating that the IgG1 framework prevents activation of complement (Figures 5B,C). The results showed that the known C1qbinding motifs are functional in the mouse IgG3 but not in the mouse IgG1 framework.
Some authors observed a correlation between hingedepend ent segmental flexibility of an antibody and its ability to activate complement (12). Thus, the differences between activity of mouse IgG1 and IgG3 are frequently explained on the basis of the length of their hinges. We decided to empirically verify this hypothesis using the domain muteins.
First, we analyzed the binding of the complement cascade ini tiator (C1q) to the muteins (Figure 6A). The results showed that the hinge modification does not affect C1q binding. The swapping of the CH2 domain from IgG1 into IgG3 abolished C1q binding by the latter. Interestingly, the paired mutein (IgG1_CH23) did not gain the ability to strongly interact with C1q; its binding of C1q reached ~12% of that characteristic for native IgG3. Swapping of the CH1 + hinge domains or the CH3 domains between IgG1 and IgG3 moderately diminished C1q binding by IgG3 and did not increase its binding by IgG1.
We also analyzed complement activation in serum triggered by erythrocytes coated with the domain muteins ( Figure 6B). The levels of erythrocyte lysis indicated that all muteins contain ing the CH2 domain derived from IgG3 activate complement cascade. The observed differences in C1q binding were not reflected by the different efficacy of the cascade triggering. The muteins with low (IgG1_CH23) or moderate (IgG3_CH1h1, IgG3_CH31) ability of C1qbinding activated complement cascade with efficacy similar to that of the parental IgG3. It seems that in the case of the antibodies comprising IgG3derived CH2 domain, even weak interaction with C1q was sufficient to effectively activate the whole complement cascade.
The results showed that both IgG1 and IgG3derived CH1, hinge, and CH3 domains are permissive for C1q binding and complement activation. The CH2 domain of IgG1 is a non permissive framework for the known C1qbinding motifs.
Overall, the results pointed to the CH2 domain as the major determinant of mouse IgG3 functions and unique properties of this isotype. In the last part of our work, we sought for features of the IgG3derived CH2 domain that may account for IgG3 distinctive characteristic.

Properties of Muteins With reversed charge of the ch2 Domains
The most striking difference between mouse IgG3derived CH2 and CH2 domains of other IgG subclasses is their charges; only the former has a strong positive charge. For example, at pH 7.0, the net charge of the CH2 domain of IgG1 is −2.6 and of IgG3 is +2.6 (cal culated using http://protcalc.sourceforge.net/). Hovenden et al. (9) found a correlation between the charge of CH2 domains of mouse IgG subclasses and their affinity to a negatively charged polyvalent antigen (polyglutamic acid, polyGA); and the high affinity of IgG3 to polyGA was attributed to the charge of its CH2 domain.
We analyzed spatial distribution of charged residues on the CH2 surface of IgG1 and IgG3 using previously obtained molecular models (11) and data deposited in PDB record 1IGY (Figure 7A). We identified 29 residues that differ between CH2 domains of mouse IgG1 and IgG3, 9 of which have different charge ( Figure  S9 in Supplementary Material). Based on the models, we selected four basic residues (His274, Lys282, Arg315, and Lys326) that are regularly spaced on the outer surface of the CH2 domain of IgG3  Figure S7 in Supplementary Material. Error bars correspond to uncertainty calculated as described in Section "Materials and Methods." (B) Complement cascade activation by the domain muteins. Erythrocytes coated with 3 µg/ml of the muteins were incubated with complement serum. 100% lysis corresponds to water-induced lysis. The bars present mean values and standard deviation of duplicates from one experiment. Results obtained with 1.5 µg/ml of the muteins are presented in Figure S8 in Supplementary Material. (a,B) Representative results of two independent experiments.
( Figure 7A; Figure S9 in Supplementary Material). The same resi dues in IgG1 are not charged. To verify whether CH2 charge influ ences IgG3 properties, we generated two muteins in which the four residues were swapped-IgG3_CH2charge (His274Gln Lys282Val Arg315Asn Lys326Ala) and IgG1_CH2charge (Gln274His Val282Lys Asn315Arg Ala326Lys). These muteins were expressed, correctly assembled, and soluble ( Figure S1 in Supplementary Material). The introduced mutations reversed the charge of the CH2 domains. It was 0.6 and −0.7 at pH 7.0 for the CH2 domain of IgG1_CH2charge and IgG3_CH3charge, respectively.
We compared properties of the parental molecules and the muteins with modified CH2 charge. We observed that the charge influenced binding to erythrocytes ( Figure 7B). However, hemag glutination, oligomerization, C1q binding, and complement activation were not affected by this charge modification ( Table 5; Figures 7C-E). The results indicate that the four analyzed resi dues have only limited impact on the IgG3 properties. We cannot exclude that other charged residues within the CH2 domain of IgG3 may influence or determine properties of this isotype.

DiscUssiOn
We summarized the results of the experiments in Table 6. We observed that molecular determinants of the unique features of IgG3 are present in the CH2 domain. However, the modifications of CH2 differently affected the features suggesting that their molecular bases are different.
The prominent role of the CH2 domain in IgG3 biology was originally reported by Hovenden et al. (9). The authors inves tigated highly protective IgG3 antibodies against the capsular antigen of B. anthracis. They generated an IgG3 mutein with CH2 swapped from nonprotective IgG2b. The mutein lost pro tective activity of the parental molecule and had reduced affinity to the antigen. In contrast to the work of Hovenden et al., we generated, for the first time, an antibody mutein that gained the unique properties of IgG3. We swapped IgG3derived CH2 into IgG1, and the obtained molecule (IgG1_CH23) had properties typical for IgG3-it agglutinated erythrocytes, oligomerized, had increased functional affinity to a polyvalent antigen, and activated the complement cascade. Thus, we proved that these unique features of mouse IgG3 could be transferred into a new antibody framework.
The mechanism of IgG3dependent hemagglutination is still not completely understood. We previously reported that F(ab')2 of IgG3 is sufficient to agglutinate erythrocytes (11). Here, we show that the presence of the CH2 domain in the IgG3 molecule pro foundly diminishes the antibody concentration required for the F(ab')2mediated process. Moreover, the introduction of IgG3 derived CH2 into IgG1 framework resulted in the IgG1_CH23 mutein that agglutinates erythrocytes. The results indicate that efficient hemagglutination is triggered only by the antibodies equipped with the IgG3derived CH2 domain.
The CH2 domain of IgG3 is positively charged at neutral pH. In contrast, the CH2 domains of other IgG subclasses are negatively  Figure S4 in Supplementary Material. Results for IgG1 and IgG3 are the same as in Figure 4 because the data were collected in the same experiments. (D) C1q binding by the muteins (O10 variants, 3 µg/ml). Data used for calculation of the normalized binding are shown in Figure S7 in Supplementary Material. The chart presents representative results of two independent experiments. Error bars correspond to uncertainty calculated as presented in Section "Materials and Methods." (e) Complement cascade activation by the muteins (3 µg/ml). Erythrocytes coated with the antibodies were incubated with complement serum. Complete lysis (100%) corresponds to water-induced lysis. Representative results of two independent experiments are shown.
charged under the same condition. Considering that erythrocyte surface has a strong negative charge and high zeta potential, it was likely that a positive charge of the IgG3derived CH2 domain reduces the zeta potential and as a consequence enhances hema gglutination. Unexpectedly, net charge modification of the CH2 domains in IgG1 and IgG3 did not change hemagglutination potential of these isotypes, and we had to reject the hypothesis linking the CH2 net charge with the efficiency of hemagglutination. Alternatively, antibody oligomerization may explain hemag glutination enhancement by the CH2 domain of IgG3. We showed that this domain solely determined antibody oligomerization in solution and thus most probably also on a multiepitope surface. It is possible that oligomerization between antibodies bound to separate erythrocytes occurs parallel to a sensitization phase of hemagglutination. Thus, antibody oligomerization may lead to the formation of zipperlike structures that stabilize cell aggre gates and increase a hemagglutination score. Moreover, the CH2 domain of IgG3 increased functional affinity of an antibody to erythrocyte surface. Thus, hemagglutination enhancement may at least partially depend on the increased affinity.
However, the observed enhancement of hemagglutination by the CH2 domain of IgG3 was affected by enzymatic deglyco sylation. In contrast, oligomerization in solution and increased functional affinity to polyvalent antigen were independent of CH2 glycosylation. This difference indicates that antibody oligomerization does not fully account for the CH2 domain mediated enhancement of hemagglutination.
Mouse IgG3 has a putative site of Nglycosylation in its CH3 domain on Asn471. Panka reported that the mutation of this Asn residue into Ser diminished the selfassociation of IgG3 (17). This finding was later contradicted by Kuroki et al., who provided evidence that this putative Nglycosylation site in the CH3 domain is not occupied and the mutation Asn471Thr does not influence IgG3 selfassociation or cryoglobulin activity (18). Our observations are in line with the findings of Kuroki et al. We did not observe any differences between oligomerization of IgG3 and its enzymatically deglycosylated variant. It is impor tant to note that we and Kuroki et al. used PEGprecipitation for oligomerization analyses. Panka used different methods, ELISA and native electrophoresis, which may account for the discrepancies.
Greenspan et al. showed that Fcdependent oligomerization increases functional affinity of IgG3 to polyvalent antigens (5). Our results confirm that finding, but we showed that the rela tion between oligomerization and increased functional affinity is more complex than previously thought. First, functional affinity of IgG3 was influenced not only by Fc region (CH2 and CH3 domains) but also by the CH1 domain. Second, functional affinity to the polyvalent antigen (B antigen) was modulated by the CH2 charge. In contrast, oligomerization in solution required only the presence of the CH2 domain of IgG3 and was insensitive to the introduced charge modifications. The results showed that the mechanism behind high functional affinity may depend on more factors than oligomerization in solution does.
The observed influence of the CH1 domain on functional affinity is difficult to explain. The CH1 domain of IgG3 has a more positive net charge than the CH1 domain of IgG1 (9). It is likely that the net charge of the CH1 domain influences the binding of the domain muteins to erythrocytes, which have a strong negative charge. However, the IgG3derived CH1 domain also enhanced the binding of IgG1_CH13 to a surface with the immobilized trisaccharide BBSA conjugate. Thus, the results support previous observations (23) that the CH1 domain may influence a variable domain and a paratope of an antibody.
According to the general view, the Fab and Fc fragments are independent parts of an antibody (24). However, our results dem onstrate that the Fc, particularly its CH2 domain, may influence Fabmediated antigen binding. There are two possible mechanisms of this phenomenon-intramolecular signaling (25) [called by some authors as an intramolecular allostery (16)] or intermolecu lar cooperativity.
There are several examples of intramolecular signaling observed by different authors investigating how the isotype switching changes an antibody affinity to its antigen [reviewed in Ref. (16,26)]. The effects of the CH1 domains or Fc fragments on variable regions are well documented, but considered a rather unique phenomenon (16). It is more likely that the increased affinity of IgG3 to its antigen results from cooperativity of its CH2 domains. Within this domain, a specific site of selfassociation may be present, which governs oligomerization of an antibody and pre determines the increased affinity to multivalent antigens. However, we cannot exclude other scenarios-the involvement of both the CH2 and CH3 domains in IgG3 intermolecular interactions or even sole CH3CH3 interactions, assuming that the CH2 domains influence the whole molecule structure and promote reciprocal interactions of the CH3 domains of neighboring molecules.
Other factors, e.g., influence of the CH1 domain on a paratope, properties of an antigen (charge), spatial distribution of epitopes, intermolecular forces between epitope and paratope, or a variable domain framework may further modulate functional affinity of IgG3 upon multivalent antigen binding. Diebolder at al. described recently an interesting example of Fcdependent antibody oligomerization. Analyses of antibody binding to DNPlabeled liposomes (a multivalent antigen) revealed that human IgG may form hexamers through noncovalent interactions between their constant regions (27). Several muta tions that enhance these interactions and subsequent complement activation were reported (27). The Fcinteractions promoting antibody hexamerization did not change affinity to the cognate antigen. Thus, this phenomenon seems to be different from IgG3 oligomerization, and it is still an open question whether mouse antibodies are able to form such hexamers.
Currently, no structure is available for a fulllength mouse IgG3 or its Fc fragment. We performed some analyses using a molecular model of IgG3 obtained by comparative modeling, but its resolution is not sufficient for indepth studies. IgG3 crystal lization might provide a direct insight into the mechanism of its oligomerization, as was in the case of human IgG1 hexameriza tion described in the cited work (27).
Complement cascade activation, as an effector function of antibodies, constitutes a firstline of defense against microbial infections. As the cascade progresses, components of the comple ment are deposited on a pathogen surface and act as opsonins for phagocytic cells. Moreover, the complement lyses invading pathogens by forming membrane attacking complex. We con firmed that C1qbinding motifs, known from human IgG1, are functional in the mouse IgG3 framework. On the other hand, we did not observe complement activation by mouse IgG1 equipped with the motifs. The results indicate that the presence of the known C1qbinding motifs is not sufficient for complement activation by an antibody. The motifs must be surrounded by a permissive framework, provided e.g., by human IgG1 or mouse IgG3.
Our work suggests that a novel type of monoclonal antibod ies may be generated by replacing the CH2 domain of a human antibody with the homologs fragment of mouse IgG3. Human IgG1 subclass is the most feasible target framework for gen eration of such IgG3inspired hybrid mouse/human molecule (28). Our observation indicates that the generated hybrid anti body should preserve the ability to activate complement and may have increased affinity to polyvalent antigens. Since the mouse IgG3 subclass is highly protective against several life threatening microbial infections, the hybrid molecule may be very useful in preventing or fighting lethal pathogens. However, the hybrid antibody with the mouse CH2 may be immuno genic. To decrease the risk of an unwanted immune response, the mouse component should be reduced to a minimum. Thus, the properties of the CH2 domain derived from mouse IgG3 should be further investigated and efforts should be made espe cially to identify fragments of this domain that determines its properties.

aUThOr cOnTriBUTiOns
TK conceived and did all experiments. TK and JB analyzed and discussed the results. The manuscript was written by TK and JB.