Recombinant neutralizing secretory IgA antibodies for preventing mucosal carriage and transmission of SARS-CoV-2

Here, we investigated the feasibility of producing neutralising monoclonal IgA antibodies against SARS-COV-2. We identified two class-switched mAbs that express well as monomeric and secretory IgA variants with retained antigen binding affinities and increased stability in mucosal secretions compared to their IgG counterparts. SIgAs had stronger virus neutralisation activities than IgG mAbs and were able to reduce SARS-CoV-2 infection in an in vivo murine model. Our findings provide a persuasive case for developing recombinant SIgAs for mucosal application as a new tool in the fight against COVID-19.


Summary 26
Passive delivery of antibodies to mucosal sites might be a valuable adjunct to  vaccination to prevent infection, treat viral carriage, or block transmission. However, 28 monoclonal IgG antibody therapies, currently used for treatment of severe infections, are 29 unlikely to prove useful in mucosal sites where SARS-CoV-2 resides and replicates in early 30 infection. Here, we investigated the feasibility of producing neutralising monoclonal IgA 31 antibodies against SARS-COV-2. We identified two class-switched mAbs that express well as 32 monomeric and secretory IgA variants with retained antigen binding affinities and increased 33 stability in mucosal secretions compared to their IgG counterparts. SIgAs had stronger virus 34 Introduction 38 COVID-19 is a mucosal infection caused by SARS-CoV-2. The virus replicates in the 39 respiratory tract and is transmitted through respiratory droplets produced when an infected 40 person coughs, sneezes, or talks. The most prominent symptoms of COVID-19 affect the 41 respiratory system (continuous cough, shortness of breath), but in some cases sensory tissues 42 in the upper respiratory tract are involved causing anosmia and loss of taste 1 . Additionally, 43 gastro-intestinal symptoms (nausea, vomiting and diarrhea) are reported in 6% of adults and 44 up to 20% in children 2 . Virus can be detected at all these sites as well as in urine 3 . 45 Infection with SARS-CoV-2 elicits systemic and mucosal immune responses 4 . Whilst attention 46 has been focused on serum antibody responses which are dominated by IgG, at mucosal sites 47 such as the respiratory, gastrointestinal, and genitourinary tracts, immunoglobulin A (IgA) in 48 the external secretions that bathe mucosal surfaces is the predominant antibody class 5 . Mucosal 49 IgA in SARS-CoV-2 can be neutralising and long-lasting 4 . 50 Various co-morbidities have been associated with diminished immune responses to  CoV-2, including immunosuppressive drugs to prevent transplant failure and diabetes 6 . 52 Seroconversion following COVID-19 vaccination can also be compromised in these and 53 similar patients 7, 8 . In such circumstances, passive delivery of antibodies might be a valuable 54 adjunct to COVID-19 vaccination, in which neutralising antibodies could be delivered directly 55 to mucosal sites either to prevent infection, treat viral carriage or block transmission. 56 Furthermore, topical delivery of antibodies could be useful to prevent carriage of virus in 57 asymptomatic individuals. 58 Neutralising monoclonal IgG antibodies 9 are already approved for systemic use in early SARS-59 CoV-2 treatment, but are unlikely to prove useful in mucosal fluids which are non-sterile and 60 rich in endogenous and exogenous proteases 10 . For mucosal sites, IgA in the form of secretory 61 16/600 Superdex 200 pg column (GE Healthcare, USA) equilibrated with PBS pH 7.4 124 connected to an ÄKTA pure (GE Healthcare, USA) FPLC system. 125

Expression and purification of RBDHis 126
For production of the recombinant receptor binding domain of the SARS-CoV-2 spike protein, 127 Expi293F™ cells were maintained and transfected according to the manufacturer's manual in 128 FreeStyle™ expression medium (all Thermo Fisher, US). High-quality plasmid preparations 129 were obtained using a Plasmid Midi kit (QUIAGEN, US). For the transfection of 200 mL 130 culture with a cell density of 3.0x10 6 cells/ml, a total of 200 µg plasmid DNA were mixed in 4 131 mL OptiPro™ SFM medium and combined with another 4 mL OptiPro™ containing 640 µL 132 ExpiFectamin (all Thermo Fisher, US). The mixture was incubated for 15 minutes before 133 gradual introduction to the cells. The culture was incubated for 7 days at 37°C in a humidified 134 atmosphere with 8% CO2 on an orbital shaker rotating at 125 rpm. The supernatant containing 135 the secreted soluble protein was harvested by centrifugation at 20 000 x g for 30 minutes at 136 4°C and additionally filtrated through a 0.45 um Durapore membrane filter (Merx Millipore,137 Germany). Clarified cell supernatant was diluted 1:2 in loading buffer (20 mM Tris, 500 mM 138 NaCl and 10 mM imidazole). The solution was loaded onto a 5 mL HisTrap HP column (GE 139 Healthcare, US) equilibrated with 5 column volumes of loading buffer, and bound protein was 140 eluted by applying buffer containing 20 mM Tris, 500 mM NaCl and 300 mM imidazole. 141 Fractions containing the protein of interest were pooled and dialyzed against PBS at 4°C 142 overnight using a dialyzing cassette with 10 kDa molecular-weight cut off (Slide-A-Lyzer, 143 For determination of the ratio of functional and fully assembled SIgA to total IgA in each size-166 exclusion fraction, similar ELISA assays were performed. Capture was with 150 ng/well 167 purified recombinant RBDHis or anti-alpha HC antibody (ab97211, Abcam, UK). Purified 168 mAbs were diluted to 2 µg/mL in blocking solution and added to RBD and anti-IgA coated 169 plates in normalized concentrations, and incubated for 1.5 h at 37°C. Detection of secretory 170 component or antibody kappa or lambda chains was as described above. 171 To determine the binding of the purified recombinant mAbs to SARS-CoV-2 RBD the ELISA 172 plates were coated with 150 ng/well purified recombinant RBD-His and purified mAbs were 173 added to the wells in normalized concentration as above. For detection, HRP-labeled anti-174 human kappa or lambda light chain antibody were used as above. Half-maximal concentration 175 (EC50) was calculated in GraphPad Prism 9.0. 176

Competitive ELISA 177
To determine the capability of purified mAbs to inhibit binding of RBD-His to the Ace2 178 receptor a competitive binding ELISA was performed. Purified Ace2-Fc was kindly provided 179 by Elisabeth Lobner (BOKU Vienna) and 500 ng/well were coated on an ELISA plate at 4°C 180 overnight, followed by blocking with PBS containing 2% (w/v) BSA and 0.1% Tween 20 (v/v). 181 RBD-His was incubated with varying molar ratios of the different mAbs starting with 2:1 182 [mAbs:RBD-His] reducing to 0.007:1 for 1h at 37°C before addition to the wells. Binding of 183 RBDHis to Ace2-Fc was detected using an HRP-labeled anti-His antibody (71840, Sigma, US) 184 and plates were developed as described above. 185

Surface plasmon resonance (SPR) spectroscopy 186
The binding kinetics of plant-produced IgG and IgA mAbs to SARS-CoV-2 RBD-His were 187 determined on a BIAcore X-100 instrument (GE healthcare, Chalfont St Giles, UK) at 25 °C 188 with buffer HBS-EP+ (10 mm HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.05% 189 surfactant P-20). The monoclonal mouse anti-His antibody (SAB2702220, Sigma, US) was 190 immobilized onto a CM5 chip with standard amine coupling. Purified RBD-His was diluted in 191 HBS-EP+ buffer and injected at a concentration of 1 μg/mL for 30 sec at the flow rate of 192 30 μL/min, followed by injection of 5 different concentration of each mAb with a flow rate of 193 30 μL/min. The second lowest concentration was repeated to ensure reproducibility, and 194 allowed to dissociate before regeneration with 10 mM Glycine pH 1.7 for 1 min at the flow rate 195 of 10 μL/min. Referenced and blanked sensorgrams were fitted with BIAcore Evaluation 196 software using a 1:1 Langmuir model of binding. Each assay was performed in duplicate. 197

Mass spectrometry 198
A total of 20 µg purified protein was reduced, S-alkylated and digested with trypsin (Promega,199 USA). Glycopeptides were then analysed by capillary reversed-phase chromatography and 200 electron-spray mass spectrometry using an Agilent Series 6560 LC-IMS-QTOFMS instrument 201 as reported previously 29 . 202 MAb stability assays in human saliva 203 Saliva was donated by two healthy volunteers and processed immediately after collection. 204 Neither donor had a previous natural infection with SARS-CoV-2 but both had received a two-205 doses vaccination regime and their salivas had been shown to contained low levels of RBD 206 specific IgG but not SIgA antibodies (Ma, personal communication). The saliva was clarified 207 by centrifugation at 3 000 x g for 15 minutes. Supernatants were collected and aliquoted into 208 100 µL aliquots before being mixed with 10 µg of each IgG and SIgA mAb variant in a volume 209 less than 5 µL. Following the immediate collection of a 15 µL sample (0 minutes time-point), 210 antibody/saliva solutions were incubated at 37°C and sampled at 5, 60, 150, 240 and 1440 211 minutes. Samples were analyzed using a sandwich ELISA assay as described above, using 212 plates coated with 150 ng/well purified recombinant RBD-His in PBS pH 7.4. The mAbs/saliva 213 solutions were diluted in blocking buffer 1:1000, added to the wells in normalized 214 concentrations and incubated for 1.5 h at 37°C. The corresponding purified mAb in PBS buffer 215 with known concentration was used as control. IgG and SIgA mAbs were detected using HRP-216 labelled mouse anti-IgG Fc (AP113P, Merck, Germany) and mouse anti-secretory component 217 (IgA) antibody (SAB4200787, Sigma, US), followed by HRP-labelled anti-mouse antibody 218 (SAB5300168, Sigma, US), respectively. 219

Recombinant production of anti-SARS-CoV-2 mucosal antibodies in plants 285
We and all were functional in terms of binding to the SARS-CoV-2 spike protein ( Figure 1A, 296 Figure S1). However, assembly into multimeric secretory IgA when the JC and SC were co-297 expressed differed significantly and was highest for COVA2-15, followed by 2E8 and was 298 reduced for COVA1-22 and 2-15 ( Figure 1A). COVA2-15 and 2E8 variants were therefore 299 selected for further analysis and characterization. 300 After affinity purification all IgG and IgA isotypes of COVA2-15 and 2E8 were subjected to 301 size-exclusion chromatography. Both COVA2-15 and 2E8 IgG variants display single 302 monodisperse peaks at the expected retention time for proteins with a mass of ~ 150 kDa 303 ( Figure 1B, dark grey shaded area). COVA2-15 and 2E8 monomeric IgA variants also display 304 a major peak corresponding to the monomeric structural unit with additional minor peaks at 305 lower retention times representing high molecular weight aggregates ( Figure S2).  infiltration of IgA with the JC and SC result in a major peak with minor shoulders at earlier 307 retention times (Figure 1B, green/blue shaded area) as well as a second peak representing non-308 assembled monomeric IgA (Figure 1B, light shaded area). Each of the eluted fraction was 309 analyzed by ELISA to determine the ratio of fully functional and assembled secretory IgA. 310 Recombinant IgAs were captured with RBD and detected with anti-SC antibody and compared 311 to total IgA by using an anti-IgA heavy chain antibody for capture and an anti-kappa or lambda 312 light chain antibody for detection ( Figure 1B, grey bars). Thereby it was shown that the major 313 peak and its shoulder at higher retention time (green/blue shaded area) of all variants contains 314 fully assembled and functional SIgA, while the peak shoulder observed for COVA2-15 SIgA1 315 and SIgA2 at an earlier retention time likely contains non-functional high molecular weight 316 aggregates (HMWA). In general, formation of multimeric COVA2-15 and 2E8 IgA variants 317 was very efficient compared to COVA1-22 and 2-15 ( Figure S3) and previous reports of other 318 multimeric IgA variants in plants 19,21,22 , whereas COVA2-15 SIgA1 and SIgA2 displayed 319 better assembly than their 2E8 counterparts. 320 Size-exclusion chromatography fractions containing either the secretory and monomeric IgA 321 species were pooled and were further analyzed using non-reducing SDS-PAGE ( Figure 1C). We were able to detect the single glycopeptide corresponding to the JC of the secretory IgA 356 variants ( Figure 1D, Table S2

Stability of anti-SARS-CoV-2 mAbs in human saliva 375
Due to its unique structural features SIgA is expected to be better suited to survive and function 376 on mucosal surfaces than IgG 32, 33 . To evaluate the stability of plant-produced anti-SARS-377 CoV-2 IgG and secretory IgA variants in human secretions, an in vitro experiment with 378 COVA2-15 and 2E8 IgG, SIgA1 and SIgA2 was performed using saliva from two donors 379 (Figure 2). Each mAb variant was incubated with saliva supernatant, incubated at 37°C and 380 sampled at the times indicated. Time-point samples were analyzed for structural integrity and 381 retained antigen binding capacity by sandwich ELISA capturing using RBD and detection with 382 HRP-conjugated anti IgG-Fc or anti-secretory component. Although the rates of degradation 383 for both IgG and IgA variants based on COVA2-15 and 2E8 varied between experiments when 384 different saliva samples were used, intact IgG was lost at a consistently faster rate over the 385 experimental time-course than secretory IgA variants. The half-life of the SARS-CoV-2 IgG 386 mAbs were calculated using a one phase decay non-linear regression model. Half-lives of 387 COVA2-15 and 2E8 IgG variants were up to 30 minutes and were increased 5 to 10-fold for 388 COVA2-15 SIgAs and 2E8 SIgA2, but were difficult to determine for 2E8 SIgA1 as they did 389 not decline to a plateau in the tested time-frame. 390 391

Binding characteristics of different antibody formats to SARS-CoV-2 RBD 392
Binding of the monomeric and secretory IgA formats to the SARS-CoV-2 receptor binding 393 domain (RBD) was tested using ELISA and the half-maximal effective concentrations (EC50) 394 was determined ( Figure 3A, Table S4). The binding behavior of monomeric and secretory 395 IgA1 and IgA2 was comparable to their IgG counterpart, whereas COVA2-15 variants 396 generally display stronger binding than 2E8 variants. In a competitive ELISA assay COVA2-397 15 and 2E8 IgG and IgA mAbs were further analyzed for their capability to inhibit RBD 398 binding to the ACE2-receptor ( Figure 3B). Plant-produced IgG and IgA antibodies were able 399 to inhibit RBD binding to ACE2-Fc using this assay, although 2E8 variants needed to be 400 administered in higher molar ratios. Generally, secretory IgAs performed better, compared with 401 monomeric IgA and IgG as expected due to their multivalency. 402 The binding kinetics of IgG, monomeric and secretory IgA variants of COVA2-15 and 2E8 to 403 RBD were further investigated using surface plasmon resonance (SPR) spectroscopy. RBD 404 was captured with a CM5 chip with immobilized anti-His antibody and different concentrations 405 of each mAb were injected in multi-cycle kinetic experiments and curves were fitted in a 1:1 406 binding model (Figure 3C). A rapid association (ka) and very low dissociation rate (kd) were 407 characteristic for all COVA2-15 mAb variants, whereas a moderate association and faster 408 dissociation rate was observed for 2E8 IgG. Secretory IgA versions, particularly in the case of 409 2E8, displayed a more rapid association and a much-reduced dissociation rate with an up to 10-410 fold increase in affinity (KD) compared to IgG (Table 1). This avidity effect was not observed 411 so clearly for the COVA2-15 variants, likely due to the already near-irreversible nature of the 412 interaction of these monomeric formats with RBD. 413 414

Neutralization activity of different antibody formats 415
The neutralization ability of COVA2-15 and 2E8 IgG and IgA antibodies was investigated 416 using a live virus neutralization assay with a clinical isolate of SARS-COV-2 (England/2/2020) 417 propagated in Vero E6 cells stably expressing ACE2 and TMPRSS-2. Plaques were counted 418 and expressed as % for non-neutralising control (Figure 4). All COVA2-15 mAb variants 419 showed high neutralization potential with 50% inhibitory dose (ID50) values ranging from 2 420 ng/mL to 10 ng/mL, which are in accordance to previously reported data of COVA2-15 IgG 421 variants produced in a mammalian expression system 23 . 422 The RBD-targeting 2E8 mAbs showed reduced capability to block RBD binding to ACE2 in 423 the competition ELISA ( Figure 3B)

Efficacy of intranasally administered anti-SARS-CoV-2 mucosal antibodies in ACE2 mice 430
To compare the prophylactic efficacy of COVA2-15 IgG, SIgA1 and SIgA2 in vivo, mAbs 431 were administrated intranasally to hACE2 transgenic mice 24 hours prior to challenge with 432 SARS-CoV-2 ( Figure 5A). High levels of viral RNA (3.4  10 6 copies/mg) were detected in 433 the lungs of control and isotype treated control mice, which were significantly reduced in the 434 prophylactic groups, particularly those treated with 250 μg (average of 10 mg/kg) COVA2-15 435 IgG, as evidenced by real-time PCR (Figure 5B). Significant reduction in viral RNA was also 436 observed in mice treated with 250 μg COVA2-25 SIgA1 or SIgA2. The results correlate with 437 clinical protection, with partial protection afforded by SIgA antibodies and full protection by 438 IgG (Figure 5C). Mice receiving COVA2-15 mAbs (IgG and SIgAs) treatment showed less 439 weight loss than the controls (Figure 5D). Histopathological analysis of lung tissues 440 demonstrated that SARS-CoV-2 induced lung lesions, focal infiltration of inflammatory cells 441 around bronchi and blood vessels (blue arrows) and alveolar septal thickening (green arrows) 442 in the control mice. There was also narrowing and collapse of the alveolar wall with creation 443 of larger cystic cavities. In the COVA2-15 IgG treated groups, there was little pathology but 444 the appearance of lungs in the SIgA treated groups resembled the PBS treated control group 445 more closely (Figure 5E). In previous studies it was shown that the JC incorporation is the limiting factor for secretory 464 IgA formation 21,22,35 . This is consistent with our finding, where we also did not observe an 465 increase of dimeric IgA when the amount of infiltrated JC was varied. Other factors that were 466 reported to contribute to dimer formation were the involvement of certain human chaperones 467 Whilst increased valency might confer increased neutralization capacity for some antibody 492 candidates, SIgA is also believed to have a longer half-life in mucosal secretions due to its 493 unique structural features making it less susceptible to proteolysis. SIgA also has unique 494 interactions with structural and functional components of the mucosa and displays non-495 inflammatory properties 32, 33 . Some of these characteristics are conferred by the extensive N-496 glycosylation of heavy chains and secretory components of IgA 39 , which in humans carries 497 mostly branched complex N-glycans with high levels of sialic acid and with seven putative 498 sites occupied in varying degrees 40 . This study confirmed that plants are capable of performing 499 these complex post-translational modifications with relatively high homogeneity compared to 500 mammalian production systems and mostly absent N-glycan modifications such as β1,2 xylose 501 and α1,3 fucose, which are commonly found in plants that have not been glycoengineered 41, 502 42 . The elongated hinge-region of plant-produced IgA1 on the other hand, exhibits plant-503 specific conversion of prolines to hydroxyprolines followed by addition of unbranched 504 arabinose chains. Hydroxyproline residues are not found on human proteins such as IgA1 and 505 concerns have been raised that the presence of arabinose chains may bear a risk of an unwanted 506 immune response. Consistent with previous observations of increased half-life for IgA in the 507 mucosa, plant-produced COVA2-15 and 2E8 SIgA variants were significantly more stable in 508 saliva compared to their IgG counterparts 19 . Interestingly, SIgA1 and SIgA2 showed similar 509 rates of degradation, although in humans SIgA1 is more prone to degradation by proteases 510 produced by oral bacteria selectively cleaving sites in the extended hinge-region of IgA1, 511 leaving SIgA2 as the predominant isotype in mucosal secretions 43,44 . Here, the activity of 512 bacterial proteases on the extended IgA1 hinge-region might be reduced due to the conversion 513 of prolines to hydroxyprolines, which lie within the recognition sequence of many bacterial 514 proteases and are partially extended with arabinose chains, thereby potentially masking the 515 cleavage site 45,46 . To date there is only limited knowledge about safety and efficacy of the 516 repeated application of recombinant mAbs let alone plant-produced IgAs to mucosal surfaces 517 or influences of plant-specific modifications. However, repeated application of a plant-made 518 SIgA to the oral cavity did not cause any side effects 12 . 519 Neutralizing antibodies against SARS-CoV-2 are increasingly used in early treatment of severe 520 COVID-19, but only administered by the systemic rout. SIgAs applied topically to mucosal 521 sites might provide a different, much earlier intervention to tackle viral carriage and 522 transmission. SARS-CoV-2 is mainly present in the nasopharynx and lungs 47, 48 , so direct 523 administration to the upper respiratory tract might provide faster and more robust antiviral 524 activity in the sites, where the virus resides and replicates 48,49 . Here, the in vivo study addressed 525 protection against SARS-CoV-2 challenge in a hACE2 mouse model. This model has the 526 advantage of being strongly informative while being technically straight forward compared to 527 carriage and transmission blocking studies in other animal models. We demonstrated partial 528 protection against SARS-CoV-2 infection in mice treated with plant-produced SIgA1 and 529 SIgA2, but also that the plant-produced IgG COVA2-15 provided complete protection in this 530 model. The apparent superiority of IgG might be due to limitations inherent with the study, 531 specifically the inability to directly compare antibody concentrations for IgG and SIgAs due to 532 different ELISA formats used to quantify concentrations. Additionally, although the murine 533 model is an invaluable system, the upper respiratory murine mucosal microbiota differs to that 534 of humans and can result in an antibody class being favored over another. 535 In summary, we demonstrated that neutralising monoclonal IgA antibodies against SARS-536 CoV-2 can be produced as monomeric and secretory formats in a plant-based expression 537 system. We showed that these antibodies are able to maintain their structure and binding 538 affinities when incubated in the harsh environment of human saliva. Importantly, we showed 539 that these plant-generated antibodies have strong virus neutralisation activity and can reduce 540 SARS-CoV-2 infection in an in vivo murine model. Therefore, our preliminary data provide 541 strong evidence of the value of secretory IgA in clinical management and/or prevention of 542  544

Declaration of interest 545
The authors declare that the research was conducted in the absence of any commercial or 546 financial relationships that could be construed as a potential conflict of interest. 547 each group in X axis was indicated according to table in A. Each dot represents one mouse. 801 The limit of detection was 2.3*10 4 copies/mg referenced to blank control which was not 802