Since all three variants showed similar binding kinetics in BLI experiments, we hypothesized that the differences in EC50 may result from rCMG2-Fc stability differences between variants during the 8-h cell incubation period at 37°C. With a less stable variant, the fraction of functional rCMG2-Fc decays over time, together with the fact that CMG2 and PA binding is reversible , resulting in a higher EC50 compared to more stable variants. The observed binding to PA was the same before and after incubation at 37°C for APO and Agly variants . For the ER variant, the fraction of functional rCMG2-Fc decayed over time, and a significant drop was observed for the overnight incubation sample , which explains the higher EC50 for the ER variant in the TNA . GnGnXF, MAN8, and Agly rCMG2-Fc glycoforms were each independently simulated for 100 ns. Figure 8 shows the initial and final conformations of all glycoforms from the simulations. The images at t = 0 are all aligned by the Fc CH2 domain; the images at t = 100 ns are each independently aligned to best illustrate macro-structural orientation. For all simulations, we see the core structure of the protein remains intact. However, there is significant variability in the macro-structural orientation of the final structures’ CMG2 and Fc domains. The final GnGnXF and Agly structures exhibit significantly contracted linkers with respect to the final MAN8 structure, where the linker is fully extended. All glycoforms retain accessibility of the PA binding site after simulation. This is in agreement with the BLI , where all three variants have similar ka and kd, and the functional ELISA , where all three variants have control curves with similar absorbance level. To quantitatively characterize the macro-structural differences between the GnGnXF, MAN8, and Agly glycoforms, we report the center of mass distance between the CMG2 and Fc domains in Figure 9. Among the glycoforms, MAN8 has the highest COM distance with the narrowest spread at around 7.2 nm, GnGnXF and Agly have lower COM distances with wider spreads, roughly centered around 6 nm,vertical plant rack with the Agly COM spread being the widest.

The significantly higher COM of MAN8 than GnGnXF and Agly is visually consistent with the final conformations in Figure 8. The large spread in Agly rCMG2-Fc COM distance is largely due to refolding of the flexible residues in and near the LNK region. The COM distance temporal profile in Supplementary Figure S10 displays a gradual increase in Agly COM distance, which tapers off around 80 ns. Thus, the larger spread in Agly COM is due to conformational changes during the simulation, not a single conformation with increased flexibility. The root mean square fluctuation data in Supplementary Figure S11 show increased Agly RMSF in and around the LNK region during the first 50 ns and a reduction in Agly RMSF during the latter 50 ns, which is consistent with the trending exhibited in the Agly COM distance.The backbone root mean square deviation of ordered domains referenced from the initial and final conformations of the CMG2 and Fc regions for all three simulated glycoforms is shown in Figure 10. A low RMSD indicates protein folding transition is minor and the protein structure is stable. For each monomer, the ordered domain of the CMG2 region were defined as residues 10–181, while the ordered domains of the Fc region were defined as the CH2 and CH3 regions: residues 210–308 and 315–414, respectively. Each domain’s RMSD was fit and referenced independently, and subsequently averaged to produce the plots in Figure 10. Two RMSD profiles were averaged to generate the CMG2 RMSD, and four RMSDs were averaged to generate the Fc RMSD for each glycoform. All RMSDs are below 1.6 Å, indicating a generally conserved fold for all the ordered domains of each glycoform. This is further confirmed by the secondary structure data, provided in Supplementary Figure S12. In general, the RMSD from the final structure is lower than that of the initial, indicating conformational convergence is progressing throughout the simulations. The low RMSD and conserved secondary structure in all three simulated glycoforms indicate that there were no major refolding events; thus, the reduced activity in the TNA and the functional ELISA of the ER variant is likely not due to reduced fold stability. The Fc RMSD is noticeably lower than the CMG2 RMSD for the Agly and GnGnXF glycoforms, indicating minor fold transitions in the CMG2 domains of these glycoforms.The hydrophobic solvent accessible surface area distributions are reported in Figure 11.

The MAN8 glycoform has the highest amount of hydrophobic SASA in the simulation, while the aglycosylated has the lowest. Increased hydrophobic SASA likely yields a greater aggregation propensity . To elucidate which residues contribute most to the hydrophobic SASA difference between MAN8 and Agly, the average per-residue hydrophobic SASA was calculated, and the five residues with the greatest positive difference of MAN8 minus Agly hydrophobic SASA were obtained. These residues were Pro199, Pro201, Leu205, and Pro300 of one monomer, and Leu206 of the other monomer. All of these residues are located in close proximity to the N-terminal region of the Fc domain, shown in Figure 12. We see that these residues are highly spread out in the MAN8 structure, moderately spread out in the GnGnXF structure, and closely associating in the Agly structure. The increased SASA in the MAN8 glycoform is consistent with the existence of the high molecular weight band in the ER variant at ~250 kDa in the SDS-PAGE results .In this paper, experimental and computational techniques were employed to study the effects of N-glycosylation on the expression, structure, function, and stability for anthrax decoy protein rCMG2-Fc.N-glycosylation was found to strongly stabilize rCMG2-Fc in planta as both APO and ER variants have over two-fold higher expression level than the Agly variant, shown in Figure 1. The increase in expression level of glycosylated variants could be attributed to a decreased susceptibility of the APO and ER variants to proteases, where the steric hindrance of oligosaccharides inhibits proteolytic degradation of glycosylated rCMG2-Fc. From a manufacturing standpoint, producing glycosylated rCMG2-Fc would require less than half the production capacity of the aglycosylated form. Thus, when glycosylation is not detrimental, preserving natural N-glycosylation sites can enhance protein production. Alternatively, retaining the aglycosylated protein in the ER can also help improve yield, since the ER has fewer types of proteases than the apoplast . Furthermore, unlike the apoplast, the ER has chaperone proteins to provide folding support . It is also possible to enhance a glycoprotein’s function via modification of the N-glycosylation profile during expression. This can be achieved using subcellular targeting and/or the co-expression of glycan-processing enzymes or addition of enzyme inhibitors in the agroinfiltration buffer. For example, high mannose Fc N-glycosylation has been shown to enhance antibody-dependent cell-mediated cytotoxicity , which can be achieved by targeting the protein to the ER or addition of mannosidase I inhibitor to the agroinfiltration buffer or cell culture medium .

It is worth noting that N-glycosylation of Fc is not strictly required when Fc is fused to the target protein for the purpose of increasing circulation half-life . SDS-PAGE and Western blot confirmed that intact rCMG2-Fc was produced, with bands near 50 kDa . There is also a band around 250 kDa in the non-reducing SDS-PAGE for the ER variant , which may correspond to a high molecular weight protein aggregate. The hypothesis of increased aggregation propensity of the ER variant is supported by the hydrophobic SASA predicted from our MD simulations. MAN8 was found to have significantly higher hydrophobic SASA,growing strawberries vertical system with the five strongest contributing residues in the N-terminal region of the Fc domain. Protein aggregation might reduce PA binding capacity, as it is evident in the functional ELISA , which is later discussed in the rCMG2-Fc function section.The ability for rCMG2-Fc variants to sequester anthrax PA and prevent cell death was assessed using a cell-based TNA. Results are shown in Figure 5, where the ER variant has a statistically higher EC50 value than APO and Agly variants. The possibility of FcγR dependent toxin neutralization was ruled out as the EC50 values did not change upon FcγR blocking. This is likely because previous studies used antibodies or serum against PA that bind to PA but not necessarily blocks the its binding site to the anthrax cellular receptor CMG2. Thus, when incubated with cells, the antibody-bound PA can still bind to CMG2 and form prepore, resulting in LF and EF endocytosis. In this situation, Fc and FcγR could form an immune complex that is then sorted and degraded in the lysosome . In our experiments, rCMG2-Fc competitively inhibits binding between cellular receptor CMG2 and PA. This completely eliminates LF and EF internalization. The binding kinetics between rCMG2-Fc variants and PA were determined by BLI, results shown in Figure 6. The 37°C KD values were slightly lower than the 25°C KD values, but no appreciable difference in binding kinetics as a function of glycosylation was observed at 25 or 37°C. Considering the CMG2 domain is linked to Fc through a flexible linker, it is not surprising that the glycosylation of the Fc domain has minimal impact on the binding kinetics of the CMG2 domain with PA. Moreover, the sub-nanomolar affinity reported in this work is consistent with previous work on rCMG2 and PA binding kinetics , which is direct evidence that neither the fused Fc domain nor its glycosylation interferes with CMG2/PA binding kinetics. Even though the kinetics of all three variants were unaffected by glycosylation, it is possible that the fraction of functional protein changes over time.

The BLI experiments only characterized the interaction kinetics, which are independent of the fraction of functional rCMG2-Fc on the sensor tip. The hypothesis that the fraction of functional protein at 37°C is glycosylation-dependent was confirmed with the functional rCMG2-Fc ELISA, where the ER variant lost the most activity overnight, consistent with the ER variant having the highest EC50. However, the MD simulation data exhibit high fold stability in all glycoforms, according to the RMSD of ordered domains as well as the secondary structure . Thus, we hypothesize that the reduction in activity is not due reduced fold stability. Moreover, the MAN8 had the highest hydrophobic SASA among the three simulated glycoforms, indicating a higher aggregation propensity. The residues that contributed the most to the decreased hydrophobic SASA in the simulated Agly from the MAN8 were located in the N-terminal region of the Fc domain, just after the C-terminal region of the linker . This region is also more extended in the MAN8 glycoform, as shown in the final conformation and the COM distance distribution . This exposure of hydrophobic residues in the thinly extended N-terminal region of the Fc domain could facilitate the aggregation with other rCMG2-Fc or protein fragments. This could explain the reduced activity for ER variant in the functional ELISA and the 250 kDa band observed in the non-reducing SDS-PAGE gel for the ER variant . Fusion protein aggregation has been observed in the literature for another Fc fusion protein ALK1-Fc, where a high abundance of MAN5 glycoform was found as high molecular weight aggregates . It is worth noting that a recent study on high-mannose type IgG showed a decrease in protection factors of backbone amide nitrogen in the CH2 domain , which could also contribute to the reduced activity of the ER variant. Meanwhile, Lu et al., found that highmannose glycans have no detrimental effect on antibody stability and aggregate rate . The antibodies used in their study were IgG1 and IgG2, which has a molecular weight ~150 kDa. Since both the protein size and structure can affect protein stability, it is not surprising to see diverse protein stability results performed on different molecules. Within Fc-fusion proteins, structure can still vary depending on the fusion partner size and structure. However, we do expect Fc-fusion proteins with similar structure, molecular weight and conserved glycosylation site as rCMG2-Fc to likely exhibit similar behaviors. The APO and Agly variants had no significant difference in fraction of functional protein after being incubated overnight at 37°C. This result is in agreement with a previous study where human IgG1s stored at 37°C for 21 days had no difference in aggregation or fragmentation , suggesting the absence of glycans had no major impact on stability under physiological temperature.