YTE or LS mutations were introduced into the constant region of human γ-chain by site-directed mutagenesis. The respective variable regions of mAbs VRC01, 3BNC117 or 10-1074 were then introduced upstream of human light and heavy chain constant region genes. Three different versions of the heavy chain constant region were prepared: the non-modified (HC), YTE modified (YTE) or LS modified (LS) constant region. Heavy and light chain gene combinations were then assembled in MIDAS-p, a modular plant expression vector (Pinneh et al., 2021). All bNAbs were expressed transiently in ΔXF Nicotiana benthamiana (Strasser et al., 2008) using Agrobacterium tumefaciens harbouring the respective T-DNA constructs. The bNAb variants will be referred to as 10-1074 HC ΔXF, 10-1074 YTE ΔXF, 10-1074 LS ΔXF and compared with CHO-derived 10-1074 CHO.

N. benthamiana leaves were harvested 6 days post-infiltration (dpi). Average crude extract yields for 10-1074 HC ΔXF and 10-1074 YTE ΔXF were comparable (44.0 mg/kg ± 5.3 fresh weight (FW) vs 47.7 mg/kg ± 24.3 FW), whereas the average yield of 10-1074 LS was much lower (3.2 mg/kg 0.1 FW). The latter yields were considered below the threshold to generate reasonable purified amounts of this bNAb variant for further analysis, so this variant was abandoned.

A standard mAb purification protocol using Protein A affinity chromatography (Teh et al., 2014) resulted in recovery of 10-1074 HC ΔXF (13.3 ± 3.1 mg/kg FW) and of 10-1074 YTE ΔXF (9.2 ± 3.9 mg/kg FW), representing only 30% and 19% recovery respectively. Addition of 0.01% polysorbate 80 during affinity purification restored the recovery yield for 10-1074 HC ΔXF (47.0 ± 10.8 mg/kg FW) but had no impact on the yield of 10-1074 YTE ΔXF (12.6 ± 0.2 mg/kg FW).

The quality and purity of 10-1074 HC ΔXF and YTE ΔXF were confirmed by SDS-PAGE ( Figure 1A ). Under non-reducing conditions, bands corresponding to the fully assembled antibody (150-180 kDa) and minor degradation products were observed. The pattern of bands was similar to that of a commercial myeloma-derived human IgG mAb control. Under reducing conditions, bands consistent with intact heavy and light chains were detected. This was confirmed by western blot, which showed specific immunoreactivity for each band, with either anti-gamma or anti-lambda chain antisera ( Figure 1B , right).

bNAbs produced in ΔXF Nicotiana benthamiana were expected to lack α1,3-fucose and β1,2-xylose. PNGase F is an enzyme that cleaves all N-linked oligosaccharides from proteins unless the core-GlcNAc carries α1,3-fucose. bNAbs were digested with PNGase F and resolved using SDS-PAGE alongside undigested bNAb controls. For all digested samples expressed in ΔXF N. benthamiana, a size shift in the heavy chain compared to the undigested control could be observed ( Figure 2A – black triangle), supporting the absence of core fucose glycosylation.

This was confirmed by mass spectrometry ( Figure 2B ), which not only revealed GnGn as the prevalent glycoform (~80%) of 10-1074 HC ΔXF and YTE ΔXF, but also confirmed the absence of β1,2-xylose and only low-levels of α1,3-fucose (<2%). Up to 15% of bNAb were aglycoslyated ( Figure 2C ).

Surface plasmon resonance (SPR) was used to compare the HIV-1 UG37 gp140 binding affinity of the plant-produced bNAb variants and their mammalian-produced non-modified counterparts ( Supplementary Table 4 ). No significant differences were observed (p>0.05, Tukey's multiple comparison test).

Viral neutralisation of bNAb variants was measured against a panel of HIV-1 pseudoviruses ( Figure 3 ). First, we confirmed that the plant expressed 10-1074 HC ΔXF had equivalent neutralising activity compared to 10-1074 CHO, using a panel of seven HIV isolates ( Figure 3A ). A more extensive comparison was then performed, comparing 10-1074 HC ΔXF with 10-1074 YTE ΔXF, using a panel of 13 HIV isolates ( Figure 3B ). No differences were observed. Comparative neutralisation curves are also shown for the three antibodies against HIV BaL.26 ( Figure 3C ).

Binding kinetics of the plant-produced 10-1074 variants to soluble FcγRIIIa V158 were compared to their mammalian counterpart using SPR. Both bNAb variants produced in glycomodified ΔXF plants showed improved binding (9.4 and 6.3-fold respectively) when compared to their glycan unmodified, fucosylated mammalian counterpart ( Figure 4A )

To determine whether increased affinity of the plant-made bNAbs to FcγRIIIa V158 results in a functional improvement in antibody-dependent cellular cytotoxicity (ADCC), a Reporter Assay was performed with effector cells carrying FcγRIIIa V158 ( Figure 4B ). For both ΔXF plant-made bNAbs, a concentration-dependent induction was observed. At 4 and 10μg/ml, 10-1074 HC ΔXF and YTE ΔXF induced significantly higher ADCC activation compared with the CHO produced 10-1074 mAb (p<0.05). At these concentrations, 10-1074 HC ΔXF was also significantly more potent than 10-1074 YTE ΔXF (p<0.05).

Binding of bNAbs bearing the YTE mutation to human neonatal Fc receptor (hFcRn) was determined by SPR. CM5 chips were coated directly with the 10-1074 bNAb variants and a series of concentrations of recombinant hFcRn were flowed over the chip. The sensorgrams show much higher binding of recombinant hFcRn to 10-1074 YTE ΔXF when compared to the HC ΔXF non-modified control ( Figure 5A ). Whilst an affinity of 1.021 x 10^-7 M could be calculated for 10-1074 YTE ΔXF, the unusual nature of the dissociation curve for HC ΔXF made it impossible to calculate a comparative value.

To demonstrate that this change in binding affinity to FcRn had a functional consequence, a transcytosis assay was performed using transgenic MDCK.2 cells constitutively expressing hFcRn/hβ₂m. MDCK.2 hβ₂m-expressing cells served as control. Transepithelial resistance was measured to verify the formation of tight junctions and ranged between 250-350 Ω*cm². bNAbs were added to the apical side (pH 6.0) and the basolateral supernatant was removed after 2 hours of incubation to determine the amount of transcytosed bNAb by ELISA. Increased transcytosis was observed at all concentrations for 10-1074 YTE ΔXF. Minimal or no antibody was detected in the basolateral supernatant when 10-1074 HC ΔXF or 10-1074 CHO was tested. ( Figure 5B ).

In vivo pharmacokinetic studies were performed using B6.Cg-Fcgrt^tm1Dcr Tg(CAG-FCGRT)276Dcr/DcrJ mice transgenic for hFcRn, that were given a single intravenous 2 mg/kg dose of bNAb. Results from four replicate experiments are shown ( Figure 6 ). 10-1074 CHO and 10-1074 HC ΔXF followed first-order pharmacokinetics ( Figure 6 , Expt. 4 and Expts. 1-3 top graphs, Supplementary Figure 7 ). 10-1074 YTE ΔXF antibody depletion was delayed, following a nonlinear pattern most likely attributed to target mediated drug disposition kinetics (Dua et al., 2015). In Expts 1-3, 5/12 mice retained an average of 84.4% of the bNAb at 144 hours Figure 6 lower graphs), compared with an average of 24.7% in the 10-1074 HC ΔXF inoculated mice.

The area-under-the-curve (AUC, %*h) was calculated ( Supplementary Table 5 ), as basic determination of half-life cannot be performed due to the mixed linear and nonlinear kinetics of 10-1074 YTE. The AUC showed that the introduction of the YTE mutation led to a 39% increase in AUC and therefore a reduction in clearance of 10-1074.

The average post-purification yield for VRC01 and 3BNC117 was in both cases highest for the non-modified (HC) version of the bNAb at 65 ± 8 mg/kg FW and 51 ± 21 mg/kg FW, respectively ( Supplementary Figure 1A ). The post-purification yield of VRC01 YTE ΔXF was 39mg/Kg, VRC01 LS ΔXF was 33mg/Kg, 3BNC117 YTE ΔXF was 31mg/Kg and 3BNC117 LS ΔXF was 38mg/Kg. Batch-to-batch variability was higher in plants infiltrated with YTE or LS constructs.

All bNAbs were assembled correctly ( Supplementary Figures 1B–D ) and PNGase digest verified the absence of α1,3-fucose in all variants of both bNAbs ( Supplementary Figures 2A and B ). As both VRC01 and 3BNC117 contain a glycosylation site in the light chain, a size shift in the light chain could also be observed for these samples ( Supplementary Figures 2A and B ).

A slight reduction in HIV gp140 binding affinity was determined for plant produced VRC01 but not for 3BNC117 mAb ( Supplementary Table 2 ). This was not however, supported by decreased neutralisation potency of the plant-produced bNAbs ( Supplementary Table 3 ), as observed in previous studies (Rosenberg et al., 2013; Teh et al., 2014). As seen with 10-1074, modified variants of both bNAbs performed equally in neutralisation assay when compared to the respective non-modified version ( Supplementary Table 3 ).

Binding kinetics of all variants to FcγRIIIa V158 showed a 2.7 – 15-fold increase in affinity of the glycomodified plant-produced bNAbs compared to the non-modified, fucosylated mammalian versions. Like 10-1074, this increase is highest for the HC (Fc-unmodified) version of both bNAbs (1.6-4.9x) when compared to the Fc-modified bNAbs ( Supplementary Figure 3A ), but all variants demonstrated increased ADCC pathway activation ( Supplementary Figures 3B and C ).

Binding kinetics of VRC01 and 3BNC117 bNAbs to hFcRn was measured using CM5 chips coated with an anti-Human IgG Fab reagent. Equilibrium dissociation constants (K_Ds) were in the 10^-8M range ( Supplementary Figure 4A ). The YTE and LS versions of VRC01 and 3BNC117 had the highest affinities to hFcRn ( Supplementary Figure 4A ). Transcytosis assays showed that YTE and LS versions of the bNAbs was more efficiently transcytosed through hFcRn expressing cells ( Supplementary Figures 4B and C ).

In vivo, the decline in antibody after intravenous administration followed first-order kinetics ( Supplementary Figure 5A ). As with 10-1074, the modified versions demonstrated a favourable pharmacokinetic profile when compared to the non-modified versions. However, for VRC01 the mammalian produced version showed a more favourable profile than any of the plant-produced versions in the initial experiment. The AUC of VRC01 YTE ΔXF and VRC01 LS ΔXF was 35% and 28% higher respectively than VRC1 HC ΔXF. It was not possible to determine pharmacokinetics for 3BNC117 in the Tg276 mouse model, as bNAb levels were almost depleted after 96 h ( Supplementary Figure 5B ).

Mammalian cell-produced 10-1074 (ARP12477), 3BNC117 (ARP12474) and VRC01 (ARP3291) were obtained from the AIDS Reagent Programme through the Centre for AIDS Reagents, NIBSC, UK.

Site-directed mutagenesis was performed using QuikChange II site-directed mutagenesis kit (Agilent, USA) along with primers ( Supplementary Table 1 ) designed to introduce the M252Y/S254T/T256E (YTE) or M428L/N434S (LS) mutations into human IgG1 Fc region with BsaI sites to insert variable regions. 10-1074 (Mouquet et al., 2012), VRC01 (Wu et al., 2010) and 3BNC117 (Scheid et al., 2011) light and heavy chain variable regions (V_L and V_H) were synthesised with BsaI sites and cloned into vectors containing modified or non-modified constant regions of human IgG1 heavy chain (C_H), human IgG lambda chain (C_L; for 10-1074), or human IgG kappa chain (C_L; for VRC01, 3BNC117). The MIDAS-P system (Pinneh et al., 2021) was used for tandem expression of the heavy and light chain genes in a single binary vector. Briefly, the heavy and light chain genes were cloned into the MIDAS-p entry vectors pWhite and pBlue (Pinneh et al., 2021) respectively. Then, pWhite or pBlue vectors containing the genes of interests were assembled into the destination vector pMIDAS using Golden Gate Assembly with BsaI (for pWhite) and then BsmBI (for pBlue). All plant expression vectors were then electroporated into Agrobacterium tumefaciens strain GV3101 PMP90/RK.

Vacuum infiltration of glycomodified ΔXF line (Strasser et al., 2008) or wild-type (WT) N. benthamiana was performed as described by Kapila et al. (1997). Agrobacterium tumefaciens GV3101/PMPRK carrying the heavy and light chain in MIDAS-p for each bNAb was grown for 24 h in Luria-Bertani broth supplemented with 50 µg/mL rifampicin, kanamycin and carbenicillin. After centrifugation the bacterial pellet was resuspended in infiltration solution containing 0.1 mM acetosyringone, 0.01 mM MES and 0.01 mM MgCl₂. The final OD₆₀₀ of the suspension was adjusted to 0.2. 7-week-old ΔXF Nicotiana benthamiana plants were immersed into the bacterial suspension in a desiccator attached to a vacuum pump, a vacuum was applied for 2 minutes at 150 mbar before releasing, allowing the bacterial solution into the interstitial spaces of the plant leaves. The plants were then further grown in a controlled environment at 26°C for 6 days with a 18/6 hr light/dark cycle.

Antibodies were extracted from infiltrated leaves 6 days post infiltration (dpi) in 3 volumes (w/v) of PBS pH 7.4 with or without addition of 0.01% of polysorbate 80. Plant debris was removed by filtration through miracloth and centrifugation at 15008 x g for 50 minutes at 4°C. The crude extract was passed through a 0.45 µm filter, followed by a 0.22 µm filter.

mAbs were purified using Protein A-agarose (Sigma Aldrich, USA). The column was calibrated with 5 column volumes (CVs) of Binding Buffer (Sodium phosphate pH 7.0), before the introduction of the clarified plant crude extract. Washing steps were performed using 10 CVs of Binding Buffer and elution was performed using 0.1 M Glycine-HCl (pH 2.7). The eluate was neutralised using 1 in 10 dilution of 1 M Tris-HCl (pH 9.0) with or without addition of 0.1% polysorbate 80. Buffer-exchange to PBS (pH 7.4) was carried out using a Slide-A-Lyzer Dialysis Cassette (3500 MWCO). An Amicon Ultra-15 Centrifugal filter was used to concentrate the sample approximately 10-fold. Purified antibody concentrations were determined using Surface Plasmon Resonance.

The purity and assembly of antibodies was verified using SDS-PAGE and Western blot respectively. Samples were separated using NuPAGE 4-12% Bis-Tris gels (Novex, Invitrogen, USA) with 1% MOPS buffer (Novex, Invitrogen, USA). Gels were stained with Instant Blue (Expedeon, UK). For Western blotting, the separated proteins were blotted onto a nitrocellulose membrane (Amersham, GE Healthcare, UK) by semi-dry transfer. Blocking was performed with 5% (w/v) non-fat dried milk (NFDM) powder (Marvel) in TBST [1% (v/v) TBS + 0.1% (v/v) Tween 20]. Anti-hIgG-γ or anti-hIgG-κ/λ (both from The Binding Site, UK) antiserum was used for detection of heavy and light chains, respectively. The blots were developed using Pierce ECL Plus Western Blotting detection system following the manufacture's protocol.

PNGase F digest was carried out following the manufacturer's instructions (New England Biolabs, USA). Digested and non-digested glycoproteins were then separated via SDS-PAGE using a NuPAGE 10% Bis-Tris gel (Novex, Invitrogen, USA).

Neutralisation potency of the generated bNAbs was determined using a TZM-bl assay and HIV-1 pseudovirus as previously described (Wei et al., 2003; Montefiori, 2004; Sarzotti-Kelsoe et al., 2014). All cells were grown in and dilutions performed using Growth Medium [DMEM with 10% Fetal Calf Serum (FCS), Penicillin-Streptomycin (100 units/mL and 100µg/mL respectively)]. Pseudoviruses were generated by transfecting HEK 293T cells with an Env-deficient backbone plasmid and the relevant Env-plasmid.

For neutralisation assays, bNAbs were used at a starting concentration of 5 μg/ml and titrated threefold in triplicate in clear flat-bottom 96-well plates. Pseudoviruses at a dilution 20x above the background was added to each well, excluding the cells-only control, and incubated for 1 h. 1x10⁴ TZM-bl cells (Centre for AIDS Reagents, NIBSC, UK), in media supplemented with DEAE dextran, were added to each well and plates were incubated for 48 h at 37°C, 5% CO₂. The supernatant was removed, cells were washed with PBS and lysis buffer (Promega, USA) was added and plates were kept at -80°C overnight. The thawed cell lysate was mixed 1:1 with Bright-Glo luciferase substrate (Promega Luciferase Assay System, USA) in a black flat bottom 96 well plate. Luminescence was measured using a GloMax plate reader (96 Microplate Luminometer, Promega, USA). Percentage of reduction in relative light units (RLU) was calculated relative to the RLU of the positive (virus plus cells) control. GraphPad Prism was used to plot the resulting curve and calculate IC₅₀s.

Antibody concentration was determined by capturing the purified antibodies and titrations of known concentration of human IgG kappa (Sigma, UK) on the Protein A-CM5 chip. The response units (RUs) were used to calculate antibody concentration from a standard curve using the BIAcore Evaluation software.

The binding kinetics of all variants of 10-1074, 3BNC117 and VRC01 to gp140, FcγRIIIa and FcRn were determined using SPR on a BIAcore X-100 instrument (Cytiva, USA) at 25°C. Measurements for gp140 and FcγRIIIa were performed using a Protein A capture approach. Protein A (Sigma Aldrich, USA) was immobilised onto a CM5 chip aiming for 5000 response units (RU) with standard amine coupling. HBS-EP+ (10 mM HEPES, pH7.4, 150 mM NaCl, 3 mM EDTA and 0.05% surfactant P-20) was used as running buffer. Each bNAb was captured to a R_max of 50. For gp140 measurements, UG37 gp140 (HIV Reagent Programme, USA) was applied at a concentration of 80 μg/ml and a flow rate of 40 μL/min. A contact time of 135s and a dissociation time of 3600s were used. Binding kinetics to FcγRIIIa were determined by applying a range of FcγRIIIa (R&D Systems, USA) concentrations (1, 0.5, 0.25, 0.125 and 0.0625 μM). A contact time of 40s and a flow rate of 50 μL/min were used. Chips were regenerated with 10 mM glycine-HCl, pH 1.5 for both experiments.

The binding kinetics of hFcRn to 10-1074 variants was determined by coating CM5 chips with the respective bNAb to a RU of 10000 and applying multiple concentrations (75,100,150,200,250 nM) of rhFcRn (R&D Systems, USA) at a flow rate of 30 µL/min, a contact time of 60s and dissociation time of 90s. Binding of VRC01 and 3BNC117 to FcRn was determined by using a Human Fab Capture kit (Cytiva, UK). The bNAbs were captured to the Rmax of 100RU and rhFcRn at concentrations of 10, 25, 50,75,100 nM were flowed over the chip with contact time of 60s and dissociation time of 90s, at a flow rate of 30 µL/min. In both experiments, PBS-T (PBS with 0.05% Tween-20), pH 6.0 was used as running buffer. Chips were regenerated with 10 mM glycine-HCl, pH 2.1.

All sensorgrams were corrected with appropriate blank references and fit globally with BIAcore Evaluation software using a 1:1 Langmuir model of binding or a two-state reaction model.

Human neonatal Fc receptor (hFcRn) transcytosis assay was performed as previously described (Claypool et al., 2002). Briefly, 0.75 x 10⁶ MDCK hβ₂m (vector only control) or MDCK hFcRn/hβ₂m (supplied by Professor Richard Blumberg of Harvard Medical School) cells were grown on 12mm transwell plates with a pore size of 0.4µm (CoStar, USA) in Growth Medium [DMEM with 10% Fetal Calf Serum (FCS), Penicillin-Streptomycin (100 units/mL and 100µg/mL respectively) and 2mM L-glutamine] at 37°C, 5% CO₂ for 4 days. 12 hours before the experiments, the medium was changed to serum-free medium without antibiotics. Trans-epithelial resistance was measured on the day of the experiment. The cells in the transwell were washed with Hanks' Balanced Salt Solution (HBSS), pH 7.4. Then, the input (apical) and output (basal) wells were equilibrated with HBBS, pH 6.0 and HBBS, pH 7.4 respectively, for 20 minutes, at 37°C, 5% CO₂. 10, 5 or 2.5 µg of respective bNAbs were added to apical side (pH 6). Basolateral supernatant was collected after a 2 hour further incubation. Antibody concentration was quantified using ELISA.

ELISA was used to detect mAbs in transcytosis assays and for serum half-life experiments. 96-well flat-bottom immunosorbent plates were coated overnight at 4°C with 5 μg/ml of sheep α-hIgG γ-chain antibody (The Binding Site, UK). Wells were blocked with 5% NFDM (w/v) in PBS with 0.1% (v/v) Tween 20 for 1 hour at 37°C.

Serum samples were applied at 1:50 (Bleed 1) and 1:15 (all remaining bleeds) dilution in duplicate. Basolateral supernatants from transcytosis assays were applied without dilution. A threefold serial dilution row in blocking buffer was performed. Plates were incubated overnight at 4°C. Peroxidase-conjugated anti-IgG κ or IgG λ (Sigma Aldrich, USA) at 1 in 1000 in blocking buffer was applied and plates incubated for 1 h at 37°C. Plates were washed with wash buffer (water with 0.01% Tween 20) after each incubation step. Bound antibodies were detected using TMB substrate solution (Invitrogen, USA). The colour reaction was stopped using 2M sulphuric acid and absorbance was measured at 450nm using a Sunrise plate reader (Tecan). Antibody concentrations were determined using titration curves of known human IgG κ or IgG λ as standards.

To determine the ability of the bNAbs to activate ADCC, ADCC Bioreporter Assay, V Variant (Promega, USA) was used. bNAbs were diluted to an appropriate starting concentration and 3-fold dilution series performed in sterile white flat bottom 96 well plates. An equal volume of HIV-1 UG37 gp140 was added to each well and plates incubated for 1 hour at 37 (5% CO₂). ADCC effector cells (Jurkat T cells constitutively expressing surface FcγRIIIa-V158) were added to each well. Plates were incubated for 6 hours at 37°C, 5% CO₂, then, left at room temperature for 20 minutes before adding Bio-Glo Luciferase Substrate (Promega, USA). After 5 minutes, luminescence was measured using a GloMax. ADCC pathway activation was calculated as fold induction [Fold of induction= RLU (sample-substrate)/RLU(cells only-substrate)].

C57BL/6 mice with the mouse neonatal Fc receptor (FcRn) knocked out and transgenic for human FcRn [mFcRn KO hFcRn(276) ^Tg/Tg; stock number 004919, The Jackson Laboratories, USA], were housed and bred in the Biological Research Facilities (BRF) at St. George's University of London (SGUL) under Home Office Regulations. All animal experiments were reviewed and approved by the Animal Welfare and Ethical Review Body (AWERB) of SGUL.

7- to 9-week-old male and female mice were randomised based on age and weight. Animal number per group (3-5) was chosen in accordance with previously published papers (Petkova et al., 2006; Zalevsky et al., 2010). Mice were weighed two days prior to injection. bNAbs were injected intravenously via the tail vein at 2mg/kg (~30-60 µg/mouse). PBS was injected as negative control. 50 µL blood samples were taken from the tail or saphenous vein after 24, 96, 144, 192 hours post-infusion. Mice were sacrificed after 192 or 264 h and a final blood sample via cardiac puncture was taken.

Blood samples were allowed to clot for 1 hour at 37°C followed by 4°C overnight, before centrifugation at 13,000rpm for 10 minutes and collection of serum. Serum was kept at -20°C. Analysis of bNAb present in serum was performed via an α-hIgG ELISA. The percentage of antibodies remaining in the serum was calculated against that from day 1 (set as 100%). The AUC was calculated using GraphPad Prism.

All graphs were drawn and analysed using the GraphPad prism 8 software (GraphPad, USA). A One-way or two-way ANOVA with Tukey's multiple comparison test was carried out for statistical analysis as indicated in the text and/or figure legends.