The emergence of pathogenic viruses like severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) highlights the urgent need for the development of novel preventive and therapeutic antiviral therapies to foster pandemic preparedness. SARS-CoV-2 has evolved continuously which resulted in the emergence of several virus variants that show increased infectivity. However, all known SARS-CoV-2 variants utilize angiotensin-converting enzyme 2 (ACE2) as host cell receptor (Gawish et al., 2022; Zhu et al., 2022) which makes the ACE2-spike protein interaction a central target for the development of antiviral interventions. Human ACE2 receptor decoys are therefore promising recombinant therapeutics to treat SARS-CoV-2 infections by preventing the interaction of the viral spike protein with ACE2 molecules residing on the surface of target cells. In contrast to neutralizing antibodies which are affected by escape mutations, ACE2 receptor decoys are active against all known SARS-CoV-2 variants of concern (Arimori et al., 2022; Lin et al., 2022; Monteil et al., 2022; Tada et al., 2023). In comparison to the original SARS-CoV-2 isolate, evolved variants like Omicron or Delta display enhanced and prolonged binding to ACE2 (Zhou et al., 2022), and are therefore even more susceptible to ACE2-based therapeutics (Monteil et al., 2022). Of note, administration of human recombinant soluble ACE2 (hrsACE2) was well tolerated by healthy humans in a clinical trial (Haschke et al., 2013) and 0.4 mg/kg hrsACE2 administered twice daily was successfully utilized for the compassionate-use treatment of a SARS-CoV-2 infected patient with life-threatening disease symptoms (Zoufaly et al., 2020). Monomeric soluble ACE2 has a short in vivo half-life, but fusion to the Fc domain of human IgG1 has been shown to increase its half-life in the circulation to approximately 30 h due to binding to the neonatal Fc receptor (FcRn) (Liu et al., 2018). Moreover, the use of Fc fusion proteins enables the tuning of effector functions either via glycoengineering of the conserved N-glycan in the Fc domain or by the introduction of point mutations to modulate Fc receptor engagement (Liu et al., 2020). Different ACE2-Fc decoys have been designed and shown to prevent infections by the parental SARS-CoV-2 strain and a broad range of variants like Omicron in animal and human tissue models, and different recombinant ACE2 variants are currently in clinical trials (Lei et al., 2020; Higuchi et al., 2021; Huang et al., 2021; Ikemura et al., 2022; Zhang et al., 2023).

In a previous study, we have shown that plant-produced recombinant soluble ACE2-Fc containing ACE2 amino acids 18-615 is functional regarding enzymatic activity, affinity to the SARS-CoV-2 receptor-binding domain (RBD) and virus neutralization (Wuhan-Hu-1 isolate) (Castilho et al., 2021). Moreover, in vitro enzymatic removal of N-glycans from mammalian cell-produced soluble ACE2 resulted in a SARS-CoV-2 decoy with enhanced neutralizing activity (Capraz et al., 2021). To examine whether a functionally superior ACE2-Fc decoy carrying only N-glycans consisting of a single GlcNAc residue attached to asparagine instead of canonical plant or human-like N-glycans can be produced in planta we carried out a transient glycoengineering study in N. benthamiana. Such an in planta produced ACE2-Fc decoy may be useful to prevent and treat coronavirus infections including those caused by emerging variants.

To produce glycoengineered ACE2-Fc with N-glycans consisting of single GlcNAc residues, the catalytic domain of endo-β-N-acetylglucosaminidase H (EndoH) from S. plicatus was fused to the trans-Golgi targeting and retention sequence of rat α2,6-sialyltransferase (ST) (Strasser et al., 2009). In addition, GFP was attached to the C-terminus of ST-EndoH to enable detection on immunoblots and analysis of its subcellular localization in plant cells (Figure 1A). Confocal microscopy showed that transiently expressed ST-EndoH-GFP was detected exclusively in puncta resembling Golgi bodies in N. benthamiana leaf epidermal cells (Figure 1B). We co-infiltrated the ACE2-Fc expression cassette encoding amino acids 18-740 of the extracellular ACE2 domain (Figure 1C) together with ST-EndoH-GFP and the α-mannosidase inhibitor kifunensine to achieve the accumulation of mannosidic N-glycans that can be cleaved by EndoH in the Golgi apparatus. Immunoblot analysis of total protein extracts obtained 3 days after infiltration showed that ACE2-Fc migrated faster after synthesis in the presence of ST-EndoH-GFP which is indicative of in vivo N-glycan removal (Figure 1D). When ST-EndoH-GFP was co-infiltrated without kifunensine, a similar shift in mobility was detectable indicating that the majority of the ACE2 N-glycans are removed, probably in the cis/medial-Golgi, before their processing into EndoH-resistant complex N-glycan structures can occur (Figure 1C). These initial experiments were carried out with ΔXT/FT plants that are glycoengineered to produce mainly human-like GlcNAc₂Man₃GlcNAc₂ (GnGn) N-glycan structures lacking plant-specific β1,2-xylose and core α1,3-fucose residues (Strasser et al., 2008). To show that the approach is more broadly applicable, we transiently co-expressed ACE2-Fc and ST-EndoH-GFP in wild-type N. benthamiana without kifunensine. Also, in wild-type plants that normally produce EndoH-resistant complex N-glycans carrying β1,2-xylose and core α1,3-fucose residues, we observed a shift in mobility upon SDS-PAGE and immunoblotting of ACE2-Fc. The shift in mobility was detectable 3, 4 and 5 days after infiltration revealing that ST-EndoH-GFP remains active throughout this incubation period which is frequently used for transient protein expression in N. benthamiana (Figure 1E).

For further characterization, we purified ACE2-Fc co-expressed with ST-EndoH-GFP and kifunensine from ΔXT/FT plants by affinity and size exclusion chromatography. The overall yield after protein A purification was 80–100 μg/g leaf material with approximately 50 μg/g recovered as dimer after SEC purification which is consistent with previously reported expression levels for ACE2-Fc (Castilho et al., 2021). As a control, we produced ACE2-Fc without co-expression of ST-EndoH-GFP and subsequently digested the purified ACE2-Fc dimer in vitro with EndoH or PNGase F. Upon SDS-PAGE, the migration position of the in vitro digested ACE2-Fc proteins was comparable to the in vivo digested ACE2-Fc implying that the majority of ACE2 N-glycans can be removed in planta by ST-EndoH-GFP co-expression (Figure 2A). To confirm this, we subjected purified ACE2-Fc to proteolytic digestion and analyzed the resulting glycopeptides by mass spectrometry (MS) (Castilho et al., 2021). ACE2-Fc produced in the glycoengineered N. benthamiana ΔXT/FT plants carried predominantly GnGn N-glycans on six of the seven ACE2 N-glycosylation sites (Table 1). On site N690, GlcNAc₁Man₃GlcNAc₂ (MGn) appeared as the most abundant N-glycan. Except for site N432 which was only glycosylated to approximately 50%, all other sites were efficiently glycosylated. By contrast, single GlcNAc residues were detected almost exclusively on all seven ACE2 N-glycosylation sites when ACE2-Fc was co-expressed with ST-EndoH-GFP in the presence of the α-mannosidase inhibitor kifunensine (Figure 2B; Table 1). In the absence of kifunensine, ACE2-Fc was also almost quantitatively deglycosylated in vivo and hence single GlcNAc residues were dominant on all seven N-glycosylation sites. However, small amounts of GnGn as well as other N-glycan forms were still detectable. This indicates that addition of kifunensine is required to produce homogeneously deglycosylated ACE2-Fc in planta (Table 1).

To see if the produced proteins are functional, we analyzed the binding of ACE2-Fc to RBD by SPR. Consistent with findings for mammalian cell-produced ACE2 (Capraz et al., 2021), the K_D value was lower for the in vivo trimmed ACE2-Fc variants produced with or without addition of kifunensine (Figure 3A and Supplementary Figure S1). Next, we compared ACE2-Fc with GnGn N-glycans and ACE2-Fc with “single GlcNAc” N-glycans in a SARS-CoV-2 live virus neutralization assay using the Delta variant. At all tested ACE2-Fc concentrations, plant-produced ACE2-Fc with N-glycans consisting of single GlcNAc residues displayed higher virus neutralization activity compared to plant-produced ACE2-Fc with GnGn structures (Figure 3B). Together, these data show that the in vivo “deglycosylation” approach described in this study allows the production of a potentially superior ACE2-Fc decoy receptor in plants.

Engineered decoys that closely mimic viral cell entry receptors are very promising candidate therapeutics to combat diseases that are caused by continuously mutating viruses. Various ACE2 decoys based on different designs have been described since the beginning of the SARS-CoV-2 pandemic (Chan et al., 2020; Arimori et al., 2022) and functional human ACE2 decoys were successfully produced in plants (Siriwattananon et al., 2020; Castilho et al., 2021; Daniell et al., 2022). An in planta deglycosylation approach was already previously applied to produce glycoengineered ACE2 (Mamedov et al., 2021). In this study, a poly-histidine-tagged ACE2 variant was retained in the ER together with EndoH to facilitate the cleavage of mannosidic N-glycans from ACE2. While binding of the glycosylated and deglycosylated ACE2 variants to RBD was comparable, the EndoH-coexpressed ACE2 was far less potent than its glycosylated counterpart in a virus neutralization assay. These data are contrary to our findings and the data from in vitro deglycosylation of mammalian cell-produced ACE2 variants (Capraz et al., 2021). The reason for the discrepancy is currently unknown, but we speculate that the premature removal of mannosidic N-glycans in the ER could have some adverse effect on ACE2 folding. In our approach, ST-EndoH-GFP likely acts after the majority of the protein is properly folded and released from lectin-based quality control in the ER. Our data suggest that ST-EndoH-GFP is already partially active in the cis/medial-Golgi where Golgi α-mannosidase I trims mannose residues to enable the initiation of complex N-glycan formation by N-acetylglucosaminyltransferase I (GnTI) (Strasser et al., 2021). Overexpressed ST-EndoH-GFP on its way to the trans-Golgi or upon recycling from the trans-Golgi to the cis/medial-Golgi could compete with Golgi α-mannosidase I for substrates, which would explain the considerable amounts of ACE2 “single GlcNAc” N-glycans generated in the absence of the α-mannosidase inhibitor kifunensine.

The increased virus-neutralization activity of plant-produced ACE2-Fc carrying GnGn N-glycans as compared to its counterpart derived from human cells (Castilho et al., 2021) is likely attributed to the absence of terminal sialic acid on N-glycans attached to N90 and N322, which interfere with spike binding (Capraz et al., 2021). The removal of most ACE2 N-glycans by EndoH in planta further increased the affinity to RBD. Moreover, we showed here that the ACE2-Fc with “single GlcNAc” N-glycans displays increased virus neutralization. While enzymatic deglycosylation of purified ACE2 decoys in vitro is in principle feasible (Capraz et al., 2021), the in planta N-glycan removal procedure appears more straightforward and could lead to a more affordable product.

Deglycosylated soluble ACE2 decoys with enhanced properties have also been generated in mammalian cells by mutating N-glycosylation sites like N90 or N322 or combinations thereof (Capraz et al., 2021; Isobe et al., 2022). Of note, mutation of the N53 site reduced the affinity to RBD and virus uptake and attenuated the enhancing effect of the N90 and N322 mutations suggesting that site-specific removal of N-glycans could result in an optimized ACE2-decoy. However, these approaches may interfere with protein folding and involve the introduction of non-natural mutations which raises safety concerns when recombinant ACE2 decoys are administered in high concentrations to humans (Arimori et al., 2022). Whether the “single GlcNAc” N-glycans on ACE2-Fc are preferable in this aspect or also lead to unwanted immunological side effects remains to be investigated. In summary, our data show that a potentially superior soluble ACE2-Fc decoy with homogeneous N-glycans consisting of single GlcNAc residues can be generated in planta by co-expression of a Golgi-targeted endoglycosidase.