Genes encoding variable regions of Varlilumab H and L chains (clone 1F5, US 2011/0274685 A1) were synthesized by Invitrogen (Carlsbad, CA; USA). The coding regions of H and L chains were introduced to pQCXIP (TaKaRa Bio, Otsu, Japan) and pQCXIH (TaKaRa Bio) while retaining the human H and L, respectively, to form the full-length Varlilumab H and L chains. The modified pPK1-BAR binary plasmid was previously constructed (Sariyatun et al., 2021). Inserts of full-length H and L chains were amplified from pQCXIP-HC and pQCXIH-LC with the addition of Xba I sites, respectively. VarHC-XbaI-Fw 5’-TAC TCT AGA ATG GAG TTT GGG CTG AGC TGG G-3’ and VarHC-XbaI-Rv 5’-TAC TCT AGA CTA TTT ACC CGG AGA CAG GGA GA-3’ were used for the amplification of Varlilumab H, while Varlilumab L was amplified by the primer pair VarLC-XbaI-Fw 5’ TAC TCT AGA ATG AGG GTC CTC GCT CAG CT-3’ and VarLC-XbaI-Rv 5’- TAC TCT AGA CTA ACA CTC TCC CCT GTT GAA GCT-3’ (the Xba I restriction site is underlined). pFK1-BAR fragments were digested by SpeI and the inserts were cleaved by the Xba I restriction enzyme (New England Biolabs, Beverly, MA) and subsequently ligated using Ligation High Ver. 2.0 (Toyobo, Osaka, Japan). Both Spe I and Xba I produced compatible sticky ends (CTAG), so the digested products can be ligated. This resulted in circular binary plasmids containing H and L chain-expression cassettes, which were designated pFK1-BAR-VarHC and pFK1-BAR-VarLC. cDNAs encoding for A. thaliana β1,3-GALT (At1g26810) and murine β1,4-GALT were isolated and then introduced in the Nde I/Sac I restriction site of the binary plasmid pRI201-AN (TaKaRa Bio).

The vectors expressing the Varlilumab H and L chains and GALTs were inserted into Agrobacterium tumefaciens strain LBA4404 by electroporation (Bio-Rad, Hercules, CA) with voltage, resistance, and capacitance at 2.4 kV, 200 Ω, and 25 µF, respectively. The RNA silencing suppressor 19 (p19) vector was kindly provided by Prof. Atsushi Takeda (Ritsumeikan University). A. tumefaciens transformants were selected using a combination of kanamycin (50 mg/L), rifampicin (50 mg/L), and streptomycin (50 mg/L). A single colony of Agrobacterium was first inoculated into 5 ml of 2xYT liquid medium with the above-mentioned antibiotics and cultivated at 28°C overnight, followed by a continuous cultivation of 200 ml of the thus-prepared medium. Cells were collected by centrifugation at 4,000 x g. Agrobacterium harboring different vectors (H, L, RNA silencing p19 or GALTs) was resuspended and mixed in modified infiltration buffer (10 mM MgSO₄, 10 mM MES, 0.56 mM ascorbic acid, 0.03% Tween-20, 100 mM acetosyringone, pH 5.8) at OD₆₀₀ 0.5 following the method of Zhao et al. (2017). The Agrobacterium containing H or L was mixed at a 1:1 ratio for expression of Varlilumab. For co-expression with the RNA silencing suppressor p19, the ratio was 1:1:2 for H, L, and p19, respectively. In GALT co-expression experiments, the ratio was 1:1:1:1 for four different constructs (H, L, p19 and GALT). N. benthamiana plants were infiltrated by vacuum infiltration (Chen et al., 2013). Infiltrated plants were grown in a controlled plant room with a 16 h light/8 h dark cycle at 28°C for 10 days. The hydroponic-based N. benthamiana wild type plants used in this study were prepared by the GreenLand-Kidaya Group (Fukui, Japan). Briefly, N. benthamiana seeds were sown into a urethan board in the dark at 21°C. Then, the germinated seedlings were grown at 21°C under a 16 h LED-photoperiod of 420 mmol/m²s (Yumex Co., Hyogo, Japan) for 10 days, followed by a 16 h yellow LED-photoperiod of 220 mmol/m²s (RAYS Co.) for 20 days. The nutrient solution used was a mixture of OAT House 1 (71%), OAT House 8 (26%) and OAT House 5 (3%) (all from OAT Agrio Co., Tokyo, Japan). The system was washed with detergent and rinsed twice with tap water between two independent experiments.

Infiltrated leaves were harvested on 2, 4, 6, 8 and 10 days post-infiltration (dpi) to evaluate the transient expression level of antibody. Leaf samples were homogenized in liquid nitrogen using a mortar and pestle, followed by protein extraction in 100 mM sodium phosphate, 100 mM sodium chloride and 40 mM ascorbic acid at pH 6.0 (Husk et al., 2014). The CHO-derived human Varlilumab (Thermo Fisher Scientific, Waltham, MA) and plant crude extract were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) under reducing and non-reducing conditions. For the non-reducing condition, the samples were mixed with non-reducing loading buffer (125 mM Tris-Cl, pH 6.8, 20% glycerol, SDS, 0.1% bromophenol blue) and separated on 10% SDS-PAGE. Under the reducing condition, the buffer contained 5% β-mercaptoethanol. For protein detection, the gels were stained with Coomassie brilliant blue (CBB) (Ready to Use; Nacalai Tesque, Kyoto, Japan) following the manufacturer’s instructions. For western blot analysis, the separated proteins in polyacrylamide gel were transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, Bedford, MA). The membrane was blocked with skim milk (5%) in PBS-T (1.47 mM KH₂PO₄, 10 mM Na₂HPO₄, 2.7 mM KCl, 137 mM NaCl, 0.05% Tween-20, pH 7.4) for 1 h. The membrane was then washed with PBS-T and incubated with primary antibody anti-human IgG (H+L chain) rabbit IgG (lot WI3374611; Invitrogen) at a dilution of 1:5000, followed by secondary antibody anti-rabbit-HRP IgG (Sigma, St. Louis, Missouri, United States) at a dilution of 1:10,000. Finally, the Varlilumab signals were visualized by adding Immobilon Forte Western HRP substrate (Millipore), and the chemiluminescence was imaged by using an iBright Imaging System (Invitrogen).

Leaves were harvested and extracted in a cold extraction buffer (2 ml buffer for 1 g of leaves). The plant debris was removed by centrifugation at 15,000 x g for 10 min at 4°C. Then the starches and small plant debris were pelleted by adjusting the pH to 7.0 using sodium hydroxide and centrifuged at 15,000 x g for 10 min at 4°C. Finally, the clarified extract was passed through filter paper using a vacuum, and the plant-derived antibody was purified by using rProtein A Sepharose fast flow (Pharmacia Biotech AB, Uppsala, Sweden). The column was equilibrated with a 10 column volume (CV) of equilibration buffer (20 mM sodium phosphate, pH 7.0). The plant extract was loaded at a speed of 2 ml/min. The unbound proteins were washed out of the column using 10 CV of wash buffer (20 mM sodium phosphate, pH 7.0). The purified antibody was eluted from the column by using 10 CV of elution buffer (100 mM glycine, 200 mM L-arginine, pH 3.0) and neutralized immediately with neutralizing buffer (1M Tris buffer, pH 8.0).

A 96-well plate was coated with 50 μL of goat anti-human IgG (H+L chain) (MBL, Woburn, UK) 1:10000 in 1x PBS buffer and incubated overnight at 4°C. The plate was washed three times with 0.05% Tween-20 in PBS (PBS-T) buffer and blocked with 5% BSA in 1x PBS buffer for 1 h at room temperature (RT). After three additional washings with PBS-T, human plasma IgG (Calbiochem, La Jolla, CA) or plant crude extracts were added into the wells and the plate was incubated at RT for 1 h. The plate was then washed, supplemented with 50 μL of anti-human IgG (H+L chain) conjugated with HRP and incubated for 1 h at RT. The plate was developed using SIGMAFAST™ OPD (Sigma) substrate solution (200 μL/well). The reaction was stopped by addition of 1 M HCl (50 μL/well) and the antibodies were measured at 450 nm with a Bio-Rad imark™ microplate reader.

Purified Varlilumab was separated on SDS-PAGE under reducing conditions and stained with CBB solution. The heavy chain was excised from the gel and destained with 50 mM NH₄HCO₃: acetonitrile (1:1 v/v) with overnight intermittent vortex mixing. Then, the heavy chains of Varlilumab were in-gel digested using Trypsin Gold (Promega, Madison, WI) in ProteaseMAX™ Surfactant (Promega) at 50°C for 1 h. The digested peptides were collected from the gel and the reaction was terminated by adding trifluoacetic acid to a final concentration of 0.5%. Digested glycopeptides were applied to a nanoLC-MS/MS system with an ESI-Qq-TOF mass spectrometer (micrOTOF-Q II; Bruker Daltonics, Bremer, Germany) and a nanoLC system (1,200 series; Agilent Technologies, Palo Alto, CA) incorporating a trap column (5 µm, 0.3 x 5 mm) and an analytical column (3.5 µm, 0.075 x 150 mm), both packed with Zorbax 300SB C-18 (Agilent Technologies).

The CD27 ELISA was carried out using recombinant human CD27 (R&D Systems, Minneapolis, MN). Briefly, human CD27 was coated to ELISA plates overnight at 4°C, followed by a blocking with 5% BSA. Each of the following was incubated for 1 h at RT: purified soil-derived Varlilumab, hydroponic-derived Varlilumab, CHO-derived Varlilumab (Thermo Fisher Scientific) as a positive control, and a human plasma IgG (Calbiochem) as a negative control. A goat anti-human IgG (H+L)-HRP (horseradish peroxidase) antibody and substrate SIGMAFAST™ OPD (Sigma) were used for detection. Samples were analyzed at 450 nm using a Bio-Rad imark™ microtiter plate reader.

All experiments were performed in three independent replicates. Statistical evaluations were performed using the Minitab software package. Values of p<0.05 were considered to indicate statistical significance.

To produce recombinant Varlilumab in N. benthamiana plants, a modified plant expression vector was used (Sariyatun et al., 2021). The genes encoding the heavy and light chain were individually inserted into the plant expression vector ( Figure 1A ). The plasmids were transferred into Agrobacterium tumefaciens LBA4404 using electroporation. Transient production was performed using vacuum infiltration, followed by protein extraction and protein A affinity chromatography purification. The expression and purity of plant-derived Varlilumab were analyzed by SDS-PAGE to confirm the size and integrity of the antibody ( Figures 1B, C ). CBB staining and WB using rabbit anti-human IgG (H+L) and anti-rabbit IgG-horseradish peroxidase (HRP) were used to confirm the production of Varlilumab fragments in N. benthamiana and successful purification from the Varlilumab-expressing leaves. As shown in Figure 1B , plant-derived Varlilumab was detected at 55 kDa (H) and 26 kDa (L) under reducing conditions. A band corresponding to the assembled antibody was present at approximately 150 kDa under non-reducing conditions, indicating that the antibody was in fully assembled form (H₂L₂). There were no differences in size between the purified soil-based and hydroponic-based Varlilumab in CBB staining or WB ( Figure 1C ). These results confirmed that H and L were co-expressed, resulting in the expression of fully assembled Varlilumab in N. benthamiana.

Infiltrated leaves were harvested at 2, 4, 6, 8, and 10 days post-infiltration (dpi). Varlilumab was quantified in total protein extracts from each sample by an enzyme-linked immunosorbent assay (ELISA). Transient expression of RNA silencing suppressor p19 was tested to determine its efficiency for increasing the expression of Varlilumab. The highest antibody production was observed at 8 dpi (54 ± 4 µg/g of fresh weight), and co-expression with p19 increased the levels of accumulated antibodies by 3.2-fold (174 ± 5 µg/g of fresh weight) in soil-grown plants ( Figure 1D ). Surprisingly, the expression of Varlilumab in hydroponic-based N. benthamiana reached 618 ± 50 µg/gram of fresh weight, 3.5-fold higher than the expression in soil-based N. benthamiana. Even at 12 dpi, the hydroponic systems showed high-level Varlilumab production of 151 ± 12 µg/gram of fresh weight, whereas soil-based agroinfiltrated N. benthamiana died after 10 dpi. Therefore, co-expression with p19 was necessary and 8 dpi was chosen as the collection day for antibody purification.

We next performed an antigen-binding assay using human CD27 marker ( Figure 1E ). Both soil- and hydroponic-derived Varlilumab showed human CD27-binding affinity similar to that of commercial Chinese hamster ovary (CHO)-derived Varlilumab. A human plasma antibody that cannot bind to the human CD27 protein was used as a negative control. The results indicated that both soil- and hydroponic-based Varlilumab were successfully produced as functional antibodies in N. benthamiana.

To investigate the N-glycan profiles of Varlilumab produced in CHO cells and N. benthamiana, we conducted an N-glycan analysis of purified Varlilumab produced in CHO cells, soil-based N. benthamiana, and hydroponic-based N. benthamiana ( Figure 2 ). The trypsin-digested N-glycopeptides were analyzed using nanoLC-MS/MS. The percentages of each structure in the total N-glycan profile were also determined ( Tables 1 , 2 ). In CHO-derived Varlilumab, five N-glycoforms were observed. The most abundant N-glycopeptide was GN2F, accounting for 51.4% of the total, followed by single-branched β1,4-galactosylated IgG at 21.5% ( Table 1 ). As expected, the N-glycan analysis of the soil-based and hydroponic-based Varlilumab revealed high levels of β1,2-xylosylated and α1,3-fucosylated N-glycan variants, accounting for 72.3% and 76.8% of the total profile, respectively. GN2XF was also a predominant structure, accounting for 54.5% and 58.5% of the profile in soil- and hydroponic-based Varlilumab, respectively ( Table 2 ). The Lewis a (Le^a) structure was not observed in the un-modified setups of either soil- or hydroponic-grown N. benthamiana plants.

To improve the effectiveness of plant-derived Varlilumab, N-glycan engineering was performed with the aim of producing Varlilumab with Gal residues on its N-glycan. Previous research indicated that Gal residues stabilize the antibody’s configuration, improving the binding between the Fc region and Fc receptors (Kiyoshi et al., 2018). In the previous N-glycan analysis, single-branched β1,4-galactosylated IgG accounted for 21.5% of CHO-derived Varlilumab, and in un-modified plants, more than half of plant-derived Varlilumab bore a GNGN core structure, which could be extended via β1,4-linked galactose by β1,4-GALT (Bohlender et al., 2022). In this experiment, both soil- and hydroponic-based N. benthamiana plants were used for the co-expression of Varlilumab with murine β1,4-GALT. The constituent N-glycans of each modification in Varlilumab were determined by de novo sequencing of the trypsin-digested N-glycopeptides using nanoLC-MS/MS analysis. The N-glycopeptides bearing the sequence of N297 were successfully detected. The majority of the N-glycans of Varlilumab co-expressed with β1,4-GALT were galactose-terminated, with the galactose-terminated N-glycans accounting for 42.5% and 55.3% of total N-glycan variants in soil-based and hydroponic-based plants, respectively ( Figure 3 and Table 3 ). However, the modified product formed by β1,4-GALT co-expression was a hybrid structure having a single Galβ1,2-1,4-GlcNAc (42.5% of total N-glycan variants in soil-based and 55.3% of that in hydroponic-based plants). Moreover, a non-mature ER-specific glycans (Man9) and high mannose structures were also observed in soil-based plants. Transient expression of β1,4-GALT in N. benthamiana results in a partial galactosylation in plant-derived Varlilumab.

To target mammalian β1,4‐GALT to the trans or late Golgi of plants, the native localization domain of mammalian β1,4‐GALT was replaced by the N-terminal CTS of Arabidopsis thaliana β1,2-XYLT (Saint-Jore-Dupas et al., 2006) or rat α2,6-ST (Strasser et al., 2009). In this study, we chose β1,3-GALT because its subcellular localization was predicted to be exclusively located in the plant Golgi apparatus. A. thaliana β1,3-GALT was co-expressed with Varlilumab to observe how it modifies the N-glycan profiles of the antibody. The N-glycan analysis indicated that the formation of the Le^a structure was enhanced ( Figure 3 and Table 4 ). In previous analysis, there was no Le^a structures observed in the un-modified setups of both soil- and hydroponic-based N. benthamiana plants ( Table 2 ). In this modification, there were 46.1% and 22.7% of Le^a structures formed in the soil- and hydroponic-based plants, respectively. In contrast, the percentage of the substrate acceptor of β1,3-GALT, i.e., GN2XF decreased from 54.5% to 25.3% in the soil-based plants and from 58.5% to 27.3% in the hydroponic-based plants. Furthermore, N-glycan structures containing both β1,2-Xyl and α1,3-Fuc residues were different between the soil- and hydroponic-based N. benthamiana plants. In the soil-based plants, the combined β1,2-XYLT and ɑ1,3-fucosyltransferase (FUCT) worked cooperatively to form 100% of N-glycans contained both β1,2-Xyl and α1,3-Fuc residues whereas in hydroponic plants, 13.7% of N-glycan structures contained only β1,2-Xyl residues. In short, co-expression with A. thaliana β1,3-GALT improved galactosylation efficiency and enhanced the formation of the Le^a structure of plant-derived Varlilumab compared to the un-modified setup.

In our previous experiments, it was suggested that the original CTS region of murine β1,4-GALT acted early in the Golgi network of transiently agroinfiltrated N. benthamiana wild type leaves and single β1,4-galactosylated N-glycan structures were produced, whereas galactosylation efficiency was improved by co-expression with A. thaliana β1,3-GALT. Therefore, we hypothesized that the cytoplasmic tail (CT) and transmembrane domain (TMD) of β1,3-GALT would result in hybrid-type N-glycans in the antibody. Previous studies have also demonstrated that the N-terminal CTS domain of β1,4-GALT is important in determining the N-glycosylation profile of recombinant proteins in plants. When the original CTS of β1,4-GALT was replaced with a rat α2,6-sialytransferase CTS domain, more homogeneous biantennary galactosylated N-glycan profiles were achieved, as this swapping led to translocation of GALT to the trans-Golgi compartments (Schoberer et al., 2009; Castilho et al., 2011). In this study, to investigate whether the expression of CT and TMD of β1,3-GALT results in hybrid-type N-glycans in the antibody, the original CT and TMD regions of mouse β1,4-GALT were replaced by the first 23 amino acids encoding for CT and TMD of β1,3-GALT. The chimeric β1,3β1,4-GALT was placed in the pRI201AN vector ( Figure 4A ). This construct was co-expressed with Varlilumab and the N-glycans profile of modified Varlilumab was analyzed using nanoLC-MS/MS. Biantennary galactosylated N-glycan (m/z 3089.6) was observed ( Figure 4B ) and accounted for 4.4% of the total N-glycan profile. Moreover, non-mature ER-specific glycans (Man9) accounted for 7.3% of the total profile, high mannose structures (Man6, Man7 and Man8) made up 20.8%, and the remaining 67% are complex N-glycans carrying both β1,2-Xyl and α1,3-Fuc residues ( Table 5 ). This result suggested that CT and TMD of A. thaliana β1,3-GALT are sufficient to form biantennary β1,4-galactosylated N-glycans in plant-derived Varlilumab.