Codon-optimized (Thermofisher^®) DNA sequences coding for heavy chain (HC) of human IgG3-WT (i.e. G3m5* allotype), as well as -KVH, and -H mutants thereof were each inserted into magnICON^® vector pICH26211α ( Supplementary Figure 1A ) using BsaI restriction sites (17), resulting in IgG3-WT, KVH, and -H, respectively. Codon-optimized HC variable fragment (Fv) sequences of H4 (379 bp) and P5C3 (369 bp) were grafted onto respective HC constructs resulting in H4-IgG3HC-WT/KVH/H (1512 bp) and P5C3-IgG3HC-WT/KVH/H (1503 bp). magnICON^® vectors carrying H4 and P5C3 kappa light chains were described recently (18–20). All constructs carry barley α-amylase signal sequence for peptide secretion. Sequence information is available in the supporting information ( Supplementary Figure 1B .). All constructs were transformed into Agrobacterium fabrum (formerly tumefaciens) str. GV3101(pMP90), and resulting strains were used for agroinfiltration. The construction of H4 and P5C3 IgG1 vectors have been described previously (19, 20).

Nicotiana benthamiana plants (ΔXTFT line (21) were grown under the conditions of 24°C, 60% humidity with a 16 h light/8 h dark photoperiod. To produce Abs, recombinant agrobacterial strains carrying the heavy chains (H4-IgG3-WT/KVH/H or P5C3-IgG3-WT/KVH/H; H4 and P5C3-IgG1) and corresponding light chains (H4-κLC or P5C3-κLC) were grown at 29°C for 24 h, centrifuged and resuspended in infiltration buffer (10 mM MES, 10 mM MgSO4; pH 5.6). Optical density of each strain was measured by extinction at 600 nm (OD600). Final OD600 of agroinfiltration mixes were set to 0.1 by dilution with infiltration buffer and delivered into leaves of 4-5 weeks old plants using a syringe.

Four days post infiltration (dpi) total soluble proteins (TSPs) were extracted from infiltrated leaves (extraction buffer: 0.5 M NaCl, 0.1 M Tris, 1 mM EDTA, 40 mM ascorbic acid; pH 7.4) in a ratio of 1:2 w/v (fresh leaf weight/buffer) for 90 min at 4°C on an orbital shaker. Subsequently, the solution was centrifuged twice at 14,000 g for 20 min at 4°C and the supernatant vacuum filtrated using 8-12 µm and 2-3 µm filters (ROTILABO^® Typ 12A and 15A). Recombinant IgG3-WT was purified by using protein G (Protein a, G SepharoseTM Fast Flow, GE Healthcare) and IgG3-KVH/H mutants were purified using Protein A (rProA Amicogen, Cat no: 1080025). TSP extracts were loaded at a flow rate of 1.5 mL/min on a manually packed column which was pre-equilibrated with 10 column volumes (CV) PBS (137 mM NaCl, 3 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4; pH 7.4). Washing was done with 20 CV PBS. Antibodies were eluted in 1 mL fractions with 0.1 M Glycine/HCl (pH 3.0), eluates were immediately neutralized with 1 M Tris (pH 9.0) and dialyzed overnight against PBS (pH 7.0).

The monomeric forms of IgG3s were separated by preparative size exclusion chromatography using HiLoad Sephadex 200/10/300 GL column (GE Healthcare). The column was equilibrated and eluted with each 1.5 CV elution buffer (PBS; 200 mM NaCl; pH 7.2) at a flow rate of 0.4 mL/min. The fractions corresponding to full length monomeric antibodies were isolated and concentrated with Amicon centrifugal filters, MWCO 10,000 kDa (Merck Millipore, UFC5010). SDS-PAGE analyses were performed in 12% gels under reducing and non-reducing conditions. Gels were stained with Coomassie Brilliant Blue R 250 staining (Carl Roth GmbH + Co. KG). Concentrations were determined by spectrophotometer (NanoDrop™ 2000, Thermo Scientific).

The N-glycosylation profiles of the purified Abs were determined by mass spectrometry (MS) as described previously (9, 22). Briefly, respective heavy chains were excised from an SDS-PA-gel, digested with trypsin for all IgG3 variants, and analyzed with an LC-ESI-MS system (Thermo Orbitrap Exploris 480). The possible glycopeptides were identified as sets of peaks consisting of the peptide moiety and the attached N-glycan varying in the number of HexNAc units, hexose, deoxyhexose, and pentose residues. Manual glycopeptide searches were performed using FreeStyle 1.8 (Thermo), deconvolution was done using the extract function. The peak heights roughly reflect the molar ratios of the glycoforms. Nomenclature according to (23).

The direct sandwich ELISA was carried out as described previously (19) with modifications. In short, purified IgG1 or IgG3 mutants were coated on 96-well plates (Thermo fisher maxisorp) at concentration of 2.0 µg/mL (H4) and 0.5 µg/mL (P5C3) in PBS (pH7.4). SARS-CoV-2 spike protein RBD215 (Wuhan strain, (Shin et al., 2021)) was used as antigen in two-fold dilution with staring concentration 4.0 µg/mL and 0.5 µg/mL for H4 and P5C3 respectively. For detection, HRP-conjugated mAb CR3022 (1:15,000 in blocking buffer) was used. Absorbance was measured at 450 nm (reference 620 nm) using a Tecan Spark^® spectrophotometer. All samples were analyzed at least in two technical replicates. EC₅₀ values were calculated by non-linear regression of the blank-corrected data points based on a four-parametric log model with GraphPad Prism (version 9).

ELISA for IgG stability testing after freeze thawing-cycles ( Supplementary Figure 3B ): 1 µg/ml RBD (100 µl per Well; Biomol) was coated in 50 mM carbonate-bicarbonate buffer pH 9.6. After incubation and washing, 200 µl of blocking buffer (PBS containing 1% bovine serum albumin) were added. Subsequently, increasing concentrations of mAb diluted (1:5) in blocking buffer were added (from 10 µg/ml to 0.64 ng/ml, 100µl per Well). Incubation steps were performed at room temperature for 1 h and followed by three washing steps with PBS/0.05% Tween-20. HRP-conjugated goat anti-human IgG Fc (Bethyl) was diluted 1:10,000 in blocking buffer. After washing, TMB solution (Invitrogen) was added to detect anti-hIgG-Fc-HRP. The reaction was stopped with 6% orthophosphoric acid (Roth). The color reaction was monitored with the SpectraMax ix3 ELISA reader (optical density (oD) at 450 nm upon 650 nm background subtraction).

Neutralizing activity of P5C3 and H4 antibodies, was assessed by CPE assays, as described previously (24). Briefly, VeroE6 cells (VC-FTV6, Biomedica, Vienna, Austria) were seeded in 96-well plates at a density of 1x10E+4. Antibodies were serially diluted (2-fold) and incubated with SARS-CoV-2 (Delta variant GK/478K.V1 (B.1.617. 2+AY.x), GISAID name: hCoV-19/Austria/Graz-MUG21/2021) for 30 min at 37°C. Subsequently, antibody-virus mixtures were added to VeroE6 cells in sextuplicate and incubated for 96 h at 37°C. Cells were fixed with 4% neutral buffered formalin and cytopathic effects were measured by crystal violet staining. After staining, washing and drying, stain was dissolved in 10% glacial acetic acid and OD was read at 595 nm. Data were normalized to a no-antibody control (0% Inhibition) and a no-virus control (100% inhibition). IC50 values were calculated by nonlinear regression analysis with variable slopes in GraphPad PRISM Version 9.

Streptavidin-coated fluorescent microspheres (Dragon Green, 200nm, Bangs Laboratories) were blocked in Blockaid solution (Invitrogen) for one hour at 50rpm and RT. Per sample, 2x1E+9 microspheres were incubated with 1.5µg biotin-labeled SARS-CoV-2 RBD (Biomol) for two hours at 50rpm RT followed by storage at 4°C overnight. The microspheres were then washed twice in Blockaid solution spinning down at 20,000xg for 3min. Subsequently, 5.0µg/ml of P5C3 or 20.0µg/ml of H4 antibody variants were added to the microspheres and incubated for two hours at 50rpm RT. Microspheres were washed with PBS by centrifugation for 3min at 20,000xg before resuspension in PBS.

Human peripheral blood was collected from healthy donors under approval from the Ethics Committee of FAU Erlangen (Head: Kerstin Amann, Krankenhausstr. 12, D-91054 Erlangen, Germany) and upon volunteers providing their written informed consent to participate in this study.

Leucocytes were isolated by erythrocyte lysis (25). 5x1E+5 peripheral blood leukocytes (PBL) were subsequently incubated with fluorescent Dragon Green Microspheres opsonized with biotinylated RBD and P5C3 or H4 antibodies as indicated. Microspheres only coated with antigen (RBD) and PBS were used as negative controls. Cells were incubated for 2 hours on ice gently shaking before staining of surface markers to detect CD14⁺ classical monocytes, CD14^-CD33⁺ non-classical monocytes, CD19⁺ B cells, CD56⁺CD3^- NK cells and CD33⁺SSC^high neutrophils within the living (DAPI^-) CD45⁺ cell population upon exclusion of duplets. All antibodies were acquired from Biolegend: Brilliant Violet 510 anti-human CD14 (clone: M5E2, APC anti-human CD33 (clone: WM53), PE/Cy7 anti-human CD19 (clone: SJ25C1), PE anti-human CD3 (clone: UCHT), PE-Cyanine5 anti-human CD56 (clone: MEM-188), APC/Fire™ 750 anti-human CD45 (clone: 2D1). Data was acquired on a BD FACSCantoII and analyzed with FlowJo software. The percentage of bead-positive cells was calculated within each subpopulation. Experiments were performed with PBL from five different donors.

For the ADCVI assay 1x1E+4 VeroE6 cells per well were seeded into 96-well plates the day before the assay. The next day cells were incubated with SARS-CoV-2 WA-1/2020 (Microbiologics) at an MOI of 0.05 for 60 minutes. Unbound virus was washed off and replaced with the antibody/peripheral blood mononuclear cell (PBMC) mixture (diluted in RPMI supplemented with 1% Pen/Strep, 5% FBS). The P5C3, and H4 antibodies were serially diluted starting at 10 µg/mL and 40 µg/mL respectively. Each condition was run in duplicate. The target cell (VeroE6) to effector cell (PBMC) ratio was 1:10. PBMCs from three different healthy donors were tested. After a 72-hour incubation period at 37°C, the cells were collected, mixed with DNA/RNA shield (Zymo Research), and analyzed for nucleocapsid-specific copy numbers by quantitative PCR.

For qPCR, RNA was isolated from the cell lysate with a viral RNA isolation kit according to the manufacturer’s instructions (Zymo Research) and tested for nucleocapsid-specific copy numbers. Viral RNA was used in combination with 0.1 µM of each SARS-CoV-2 N-specific primer (5’-GGGGAACTTCTCCTGCTAGAAT-3’ and 5’-CAGACATTTTGCTCTCAAGCTG-3’) as well as the reagents from the qPCRBIO SyGreen 1-step Go Hi-ROX kit (PCRBIOSYSTEMS). The qPCR was carried out on an Applied Biosystems StepOne Plus real-time PCR instrument with the following program: 1 cycle at 50°C (30 min), 1 cycle at 95°C (15 min), and 45 cycles at 94°C (15 sec) and 60°C (20 sec). SARS-CoV-2 copy numbers were quantified and reported based on a standard curve generated by serial dilutions of a SARS-CoV-2 nucleocapsid-containing plasmid using the Applied Biosystems StepOne Plus real-time PCR software. Percent reduction in viral copy numbers was calculated as follows: Reduction = 1 – (copies + Ab + PBMCs)/(copies – Ab + PBMCs) x 100.

Interaction of anti-RBD-IgGs with C1q was analyzed by enzyme-linked immunosorbent assay (ELISA). All incubation steps were performed at room temperature for 1h and followed by three washing steps with PBS/0.05% Tween-20. 1µg/ml of mAbs were coated in 50mM Carbonate-Bicarbonate buffer pH 9.6 (Sigma). After incubation and washing, 200µl of blocking buffer (PBS containing 3% bovine serum albumin, 0.1% gelatine and 0.05% Tween-20) were added. Afterwards, increasing concentrations of native human C1q (Sigma Aldrich) diluted in blocking buffer were added (two-fold dilutions from 65ng/µl to 2.03ng/µl) followed by incubation with HRP-conjugated sheep anti-human complement C1q (BioRad) diluted 1:500 in blocking buffer. Bound C1q was detected upon addition of TMB substrate solution (Invitrogen) and stopping the reaction with 6% orthophosphoric acid (Roth) after 5min. Thereafter, the assay was analyzed on the SpectraMax ix3 ELISA reader, where the signal intensity was measured at 450nm. Background absorbance at 650nm was subtracted.

Previously we have established plant-based expression modules that facilitate efficient antibody subtype switch (9, 19, 20, 22, 26). In this context several monoclonal Abs (mAb) were expressed as IgG1-4 formats. While IgG 1, IgG2 and IgG4 expressed at the same range, nucleotide codon optimization was required for efficient expression of IgG3 (20). In the present study, P5C3 and H4, two broadly neutralizing anti-SARS-CoV-2 mAbs, served as a template. Both Abs are originally IgG1 Abs derived from convalescent human sera. They bind to nonoverlapping epitopes at the receptor-binding domain (RBD) of the spike protein, with significant differences in their antigen-binding potencies (19, 27, 28). While P5C3 exhibits binding affinities in the picomolar range, these values are orders of magnitude higher for H4, depending on the virus isolate (27, 28). Next to the original IgG1 format, here we generated three IgG3 variants, IgG3-WT, referring to the most prevalent human G3m5* allotype, IgG3-H (R435H), and IgG3-KVH (K392; V397; H435 present in G3m15*, 16*) (11). KVH mutations are based on previous observations on beneficial effect on IgG3 expression and purification (16) ( Supplementary Figure 1 ).

In total four IgG variants per antigen binding domains, H4 and P5C3, were recombinantly expressed in the glyco-engineered Nicotiana benthamiana line ΔXTFT (21) through agroinfiltration. Four to five days post infiltrations (dpi) IgGs were purified by affinity and subsequent size-exclusion chromatography. SEC profiles exhibited three fractions that were separately collected (F1-F3). F2, representing fully assembled mAbs, was monitored on SDS gel electrophoresis which confirmed the purity of all mAb variants ( Figure 1 ). F2 was hence used for further biochemical and functional analyses. F1 and F3 represent high molecular weight Abs (most probably aggregates) and degraded HC ( Supplementary Figure 2 ), respectively. H4 and P5C3 IgG1 have been characterized previously (9, 19).

While IgG3-WT carries an arginine at position 435, in KVH and H variants this amino acid was replaced by a histidine (R435H). This confers increased serum half-life and allows purification with protein A instead of protein G (15, 16). A direct comparison of purification performance using the two affinity matrices was performed with P5C3 IgG1 and IgG3-KVH ( Table 1 ). The purification yield of both subtypes was slightly reduced by protein G compared to protein A. Thus, protein A was used for all purifications, except IgG3-WT where only Protein G is applicable. H4 and P5C3 IgG1 variants were expressed at yields usually obtained using the magnicon expression system (9). The purification yields of the three IgG3 variants were about 20-30% lower compared to the IgG1 orthologue ( Table 1 ). Purified mAbs were stored at -20°C and stability monitored by antigen binding. All mAbs were stable at least 10 months post purification ( Supplementary Figure 3A ), however upon freeze-thawing cycles both IgG3-WT mAbs were partially degraded ( Supplementary Figure 3B ).

All IgG subtypes carry one conserved glycosite (GS) in the CH2 domain of the heavy chain at amino acid position asparagine 297 (N297). Note, IgG3 allotype G3m5* contains an additional site at position 392 in the CH3 domain, which is not present in the KVH mutant (N392K). For IgG1 multiple studies demonstrate the importance of glycosylation in immune responses (29). Here H4 and P5C3 Abs were produced in glyco-engineered Nicotiana benthamiana line ΔXTFT, synthesizing complex N-glycans lacking plant specific core fucose and xylose (21). To determine the N-glycosylation status of the recombinant mAbs, liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) was performed. MS spectra of the two IgG1 formats displayed a single dominant glycoform at the Fc GS, namely xylose and core fucose-free GlcNAc-terminated structures (predominantly GnGn, 88% and 90%), accompanied by mannosidic structures (7, 12%, Figure 2 , Supplementary Figure 4 ). Such profiles are typical for ΔXTFT-produced IgGs (21). MS spectra of IgG3 were similar but not identical compared to IgG1. While fucose/xylose free complex N-glycans are still dominating, all IgG3 variants carry increased mannosidic structures ranging from 15-31%. In the context of glycosylation, most IgG3 allotypes exhibit two peculiarities compared to other IgG subtypes, namely a potential second N-GS at CH3 domain (N392) and an O-glycosylated hinge region. We did not detect glycans at either site ( Supplementary Figure 5 ). Collectively, expression profiling and biochemical analyses did not reveal obvious differences between the IgG3 variants. Importantly, all IgG3 variants carry identical glycosylation profiles. Thus, in the present study, this excludes a possible impact of this important post-translational modification on functional activities as reported previously (e.g., (12)).

The functional activity of anti-SARS-CoV-2 mAb variants was determined by antigen-binding assays. Direct sandwich ELISAs using SARS-CoV-2 spike protein RBD (delta variant) as antigen and HRP-labelled mAb CR3022 as secondary antibody was performed. CR3022, initially developed against SARS-CoV, broadly detects SARS-related coronaviruses (30) and does not compete with the binding of H4 and P5C3, respectively. Within mAb variants that target the identical epitope no significant differences were observed. EC₅₀ values for the low binder H4 ranged between 955 and 1212 pM ( Figures 3A, B ). Accordingly, similar antigen-binding activities were observed comparing high-affinity P5C3 variants, with EC₅₀ values between 130-153 pM ( Figures 3A, B ). Neither class switch nor Fc mutations had an impact on antigen-binding properties.

To determine the neutralization activities of P5C3 and H4 Abs, Vero cell–based SARS-CoV-2 plaque reduction assays were performed as described previously (24). All Ab variants were able to efficiently neutralize SARS-CoV-2 infection ( Figure 3B ; Supplementary Figure 6 ). However, in contrast to the antigen binding activities, clear differences in neutralization (NT) potencies were observed comparing H4 variants. All three H4-IgG3 variants had a higher NT potency compared to the IgG1 orthologue, well pronounced for IgG3-WT and -H, with 10-fold lower IC₅₀ values (6.951 versus 0.559 nM). Also, different NT potencies were observed between IgG3 allelic variants, in the order of IgG3-H ≫ IgG3-WT > IgG3-KVH (IC₅₀ 0.605 vs 2.594 and 4.578 nM, respectively). By contrast, no significant differences in NT activities for P5C3 variants were observed (NT IC₅₀ ranging between 0.10 and 0.23 nM). Note, the remarkable high neutralization potency of P5C3 that results in especially low IC₅₀ values might skew the results and activity differences between the tested variants might have been overlooked under the current settings.

In addition to virus neutralization, IgG Abs direct effector responses by binding to Fcγ receptors (FcγRs) via their Fc region. Due to the low IgG affinity of most activating FcγRs, efficient binding and subsequent FcγR activation is initiated through IgG-mediated clustering, which in turn is caused by the engagement of several antibodies on an antigen target, forming multivalent immune complex (IC). This clustering mechanism ensures that more than one IgG is present whenever effector responses occur. Recently, studies using IC and FcγR-expressing CHO cells or human peripheral blood leukocyte populations, revealed differential FcγR binding by IgG subclasses (25, 31). While FcγR-expressing cell lines are well suited to characterize specific IgG-FcγR interactions, blood leukocytes as target cells yield biologically highly relevant insights into IgG-binding in the context of appropriate receptor expression levels. Most importantly, leukocyte subsets such as CD14⁺ classical and CD14^- non-classical monocytes as well as neutrophilic granulocytes express multiple FcγR (activating FcγRI and activating FcγRIIa on classical monocytes, activating FcγRIIIa and FcγRIIa on non-classical monocytes, FcγRIIa/b and GPI-linked FcγRIIIb on neutrophils) that could cooperate in IC binding while NK cells and B cells only express FcγRIIIa or inhibitory FcγRIIb, respectively (32). Aiming to compare P5C3 and H4 Abs, ICs were generated using fluorescent microspheres that can be opsonized with any biotinylated antigen of choice due to loading of microspheres with streptavidin. Incubation with biotinylated SARS-CoV-2 spike RBD protein and RBD-specific P5C3 and H4 Abs thus leads to opsonization in correct conformation (33) ideal for FcγR-binding studies.

Generally, the results demonstrate binding of virtually all mAbs to FcγR expressing leukocytes. P5C3 mAbs generally exhibit a stronger binding compared to the H4 variants ( Figure 4 , Supplementary Figure 4 ). IC binding was most pronounced for NK cells and neutrophils both of which dominantly express FcγRIIIa/b known to interact strongly with afucosylated IgG (34). Binding to FcγRIIa/b and FcγRI expressing classical monocytes was stronger than to FcγRIIa/b and FcγRIIIa expressing non-classical monocytes and weakest for FcγRIIb expressing B cells where P5C3 IgG3-KVH and all H4 mAbs were comparable to the RBD control. Importantly, compared to IgG3-WT and the IgG1 orthologue, both IgG3-H variants displayed similar binding behavior to all cell types suggesting functional equivalence of this variant. In contrast, binding of KVH mAbs was found to be notably increased or decreased in a cell-type dependent manner indicating altered binding to especially to FcγRIIIa/b and FcγRIIb.

We further investigated antibody-mediated effector functions of the P5C3 and H4 Abs in the presence of human effector cells (PBMCs) and SARS-CoV-2 infected target cells (VeroE6) using an ADCVI assay. ADCVI is a measure of FcγR-mediated antiviral activity and occurs when antibodies bound to virus-infected target cells engage FcγR-bearing effector cells. The resulting virus inhibition is due in part to target cell death mediated through ADCC and/or ADCP and in part to non-cytolytic mechanisms such as β-chemokine release from the effector cells. ADCVI activity has been particularly observed with protection from lentivirus infections (35, 36). In this study we assessed the ADCVI activity of the P53C and H4 variants at three mAb concentrations. For P5C3 mAbs we detected a similar ADCVI activity trend: KVH > H ≥ IgG1 > WT in all three antibody concentrations tested (i.e., 10, 2, and 0.4 µg/mL) ( Figure 5A ). Interestingly, ADCVI was more pronounced in the H4 Abs (Ab concentration: 40, 8, and 1.6 µg/mL): H≥ WT ≥ IgG1 ≫KVH. ( Figure 5B ). Overall, ADCVI levels decreased in a dose-depended manner, however the ADCVI activity for H4 IgG3 H retained higher activities compared to the other variants even at a concentration of 1.6 µg/mL.

Beyond induction of FcγR-dependent effector functions, IgG3 is also a prominent activator of the complement cascade (8, 37). Activation of the classical complement pathway occurs when IgG is deposited on a target cell surface leading to recruitment of C1q. Binding of C1q initiates the C1 serin protease activity resulting in sequential cleavage of downstream complement proteins. Ultimately, complement activation results in formation of membrane attack complexes directly lysing the target cell. Alternatively, opsonization with complement fragments may induce phagocytic uptake via complement receptors expressed on myeloid effector cells (38). As activation of the complement system was also shown to be relevant for anti-viral immune responses (39) we further compared C1q binding ( Figure 6 ). Overall, all P5C3 and H4 IgG1 as well as IgG3 mAb variants were able to interact with C1q comparably confirming their functional equality for this effector function.