key: cord-0869107-ayadf0vr
authors: Klausberger, M.; Duerkop, M.; Haslacher, H.; Wozniak-Knopp, G.; Cserjan-Puschmann, M.; Perkmann, T.; Lingg, N.; Aguilar, P. P.; Laurent, E.; De Vos, J.; Hofer, M.; Holzer, B.; Stadler, M.; Manhart, G.; Vierlinger, K.; Egger, M.; Milchram, L.; Gludovacz, E.; Marx, N.; Koeppl, C.; Tauer, C.; Beck, J.; Maresch, D.; Gruenwald-Gruber, C.; Strobl, F.; Satzer, P.; Stadlmayr, G.; Vavra, U.; Huber, J.; Wahrmann, M.; Eskandary, F.; Breyer, M.-K.; Sieghart, D.; Quehenberger, P.; Leitner, G.; Strassl, R.; Egger, A. E.; Irsara, C.; Griesmacher, A.; Hoermann, G.; Weiss, G.; Bellmann-Weiler, R.; Loeffler-Rag,
title: A comprehensive antigen production and characterization study for easy-to-implement, highly specific and quantitative SARS-CoV-2 antibody assays
date: 2021-01-20
journal: nan
DOI: 10.1101/2021.01.19.21249921
sha: bd2e9782fcac96f80bddebd0d61342f8d854a1b9
doc_id: 869107
cord_uid: ayadf0vr

Antibody tests are essential tools to investigate humoral immunity following SARS-CoV-2 infection. While first-generation antibody tests have primarily provided qualitative results with low specificity, accurate seroprevalence studies and tracking of antibody levels over time require highly specific, sensitive and quantitative test setups. Here, we describe two quantitative ELISA antibody tests based on the SARS-CoV-2 spike receptor-binding domain and the nucleocapsid protein. Comparative expression in bacterial, insect, mammalian and plant-based platforms enabled the identification of new antigen designs with superior quality and high suitability as diagnostic reagents. Both tests scored excellently in clinical validations with multi-centric specificity and sensitivity cohorts and showed unprecedented correlation with SARS-CoV-2 neutralization titers. Orthogonal testing increased assay specificity to 99.8%, thereby enabling robust serodiagnosis in low-prevalence settings. The inclusion of a calibrator permits accurate quantitative monitoring of antibody concentrations in samples collected at different time points during the acute and convalescent phase of COVID-19.

purpose, we compared several animal cell lines and plant-based expression platforms for their ability to 90 support high-quantity and quality RBD production and assessed whether the employed production host 91 influences antigen performance. We extensively validated the tests for clinical utility featuring sera from 92 individuals covering the full spectrum of disease presentations at different time points post infection and 93 a large specificity cohort including samples with antibodies towards human coronaviruses (hCoVs) and 94 those from individuals with underlying non-infectious diseases. Moreover, we validated the tests with 95 time-resolved acute and early convalescent samples from hospitalized patients and showed that only 96 RBD-specific antibodies demonstrate excellent correlation with neutralization assays already in the early 97 phase of infection. 98

To assess the performance of the antigens for discrimination between sera from SARS-CoV-2-exposed 123 (n=124) and uninfected individuals (n=210), we applied a high-throughput (HTP) automated bead-based 124 multiplex assay (Fig. 1B, C) . The performance of diagnostic tests are commonly assessed through receiver 125 operating characteristic (ROC) curves and the analysis of area under the ROC curve (AUC-ROC). ROC curves 126 are simple graphical representations of the relationship between sensitivity and specificity of a test over 127 all possible diagnostic cut-off values and AUCs give the overall ability of a test to discriminate between 128 two populations (5). We used theses analyses to assess potential differences in the diagnostic 129 performance of RBD from different expression hosts. Almost all antigens at this high purity demonstrated 130 AUC values of >0.99, demonstrating the high suitability of the RBD from any source as diagnostic antigen. 131

The AUC value of insect-derived RBD was slightly lower (AUC: 0.978 [0.964-0.992]); the differences, 132 however, were not significant (Fig. 1B) . We then applied antigen-specific cut-offs to compare the 133 performance of the antigens at a pre-defined consensus specificity of 99.1%. At this criterion, we obtained 134 high sensitivities (range 94.4%-96.0%) with all antigens, except for insect-derived RBD. There, 135 seroreactivity with pre-COVID-19 sera was about 2 2 -(4)-fold higher than observed for CHO-expressed 136

RBDs. This resulted in 26% of COVID-19 sera to fall below the threshold, increasing the rate of false-137 negatives (Fig. 1C) . The tRBD displayed a comparable seroreactivity profile to the RBD. 138

During our pre-validation experiments we observed a strong effect of residual host cell proteins on assay 139 performance (Fig. S4) , even in formulations derived from human cell lines. Therefore, RBD/tRBDs were 140 purified via an Immobilized metal affinity chromatography (IMAC) capture followed by a scalable and fast 141 flow-through Anion exchange (AIEX) chromatography step, leading to purities of up to 99%. Owing to 142 reproducible highest production yields of functional protein with adequate diagnostic performance and 143 less batch-to-batch variation, we decided to pursue with HEK-expressed tRBD for our further validations. 144 145 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01. 19 .21249921 doi: medRxiv preprint range (IQR) of the total cohort seroreactivity towards the SARS-CoV-2 NP or tRBD (outlier NP: n=17; tRBD: 237 n=4). Above these cutoffs, all sera from our specificity cohorts reacted strongly with the spike proteins of 238 circulating human coronaviruses (hCoVs) HKU-1, OC43, 229E, and NL63, confirming widespread 239 seroprevalence in the general population (Fig. 4A, B) . To further characterize the identified outliers among 240 the pre-COVID-19 sera, we calculated their relative IgG signals, set them in relation to a roughly equal 241 number of sera located at the other extreme on the seroreactivity scale (sera with readouts <25 th 242 percentile toward the respective antigen) and compared the differences in relative IgG levels to that 243 towards hCoV antigens. Among our pre-validation cohort, sera with highest relative reactivity towards NP 244 (mean difference: 0.88, p>0.0001) also demonstrated significantly elevated relative median IgG levels 245 towards the spike protein of HKU-1 (mean difference: 0.13, p=0.0113, Fig. 4B ). The specificity cohort we 246 used for clinical validation included 8 sera from individuals with PCR-confirmed hCoV infection. None of 247 these yielded false-positive readouts at a cutoff of 5.000 U/mL (Fig. 4C) at comparably low specificities of 248 95.3% (tRBD) and 96.1% (NP) (see Fig. 3A, B) . Table S2 ). Yet, sera displayed a great heterogeneity in antibody levels 259 throughout the observation period ( Table S2) . None of the false-negative results among the samples were 260 obtained with both assays. Astonishingly, 85.7% of the sera already contained neutralizing antibodies 261 (median titer: 1:24; range 1:4 -1:128, Table S2 ) as soon as by day 5 after symptom onset. Of these, 262 however, only a total of 18% of the sera demonstrated seroreactivity above the cut-off for either the NP 263 or tRBD antigen (Fig. 5) . Yet, the quantitative nature of the assay allowed us to correlate antibody levels 264 below the cut-off for seropositivity and we could demonstrate excellent correlation of tRBD-specific 265 Fig. 5) . 267

Superb assay specificity is of utmost importance for the assessment of antibodies directed against SARS-269

CoV-2, as a substantial proportion of infected individuals escapes identification due to the frequent 270 asymptomatic course of the disease, thereby distorting the true humoral seroprevalence in any given 271 population (11). The biological basis for false-positives is multifactorial, but the influence of the 272 production platform and process-related peculiarities or impurities on protein performance are factors 273 that are often underestimated. While the viral NP is almost always produced in bacteria (12, 13), we 274 expressed the spike receptor binding domain in HEK cells, CHO cells, insect cells and plants (4, 14-16). To 275 find out which of these systems leads to the highest quality and manufacturability of the RBD diagnostic 276 antigen of potentially high demand, we evaluated these production platforms and pre-validated the 277 proteins based on diagnostic performance with a large set of pre-COVID-19 and COVID-19 sera using the 278 Luminex platform. All five expression platforms demonstrated suitability for the production of functional 279 protein, proven by a binding assay with the SARS-CoV-2-RBD-specific mAb CR3022. Yet, in part due to the 280 different transfection methods used, RBD yields from CHO-K1, CHO-S as well as from Tnms42 insect cells 281 and tobacco plants were insufficient for sustainable commercial antigen production (< 1mg/L, Fig. 1) . In 282 contrast, HEK cells readily produced overall yields of 40 mg/L using PEI-transfection. Yields of 30 mg/mL 283 per liter have also been described for CHO-expressed RBD. However, this can be traced back to optimized 284 design of expression constructs and improved production processes for stable RBD-expressing CHO cells 285 together with less extensive purification protocols (17). We observed higher basal seroreactivity of control 286 sera with insect-derived RBD than with RBD from human and non-human mammalian cell lines; which is 287 in line with other reports (4). Host-related impurities do not account for that, as insect-cell produced RBD 288 demonstrated the highest purity among all our RBD samples (99%, Fig. S2) . While there was a common 289 set of false-positive samples shared by RBD from non-human and human mammalian cell lines as well as 290 plants, false-positives reactive with the insect material were entirely insect-RBD-specific (Fig. S6) . A 291 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint possible reason may be platform-specific protein modifications, such as glycosylation, that provide the 292 protein with a unique process-derived signature. Indeed, T. ni-derived insect cells were demonstrated to 293 generate core α1,3-fucose structures with allergenic potential in humans (18), which might be associated 294 with this peculiar seroreactivity profile. 295

Based on our observation that RBD tends to form homodimers in an unpredictable manner among 296 different production batches of the same expression host, we used an optimized, truncated version of an 297 RBD as diagnostic antigen -tRBD-, enabling the production of large amounts of RBD with consistent quality 298 ( Fig. S2) . For tRBD performance, antigen purity was of utmost importance, even when expressed in human 299 cell lines. At a consensus specificity of 99.1%, a reduction in tRBD purity by 10% (pure: 97.5%, impure: 300 87.5% purity) resulted in a drastic reduction in sensitivity by 83.9% (pure: 95.2% versus impure: 11.3%, 301 respectively) in the Luminex pre-validation assays (Fig. S4) . Since purity after an IMAC capture step was 302 highly batch-dependent and resulted in inconsistent seroreactivity profiles, our standard downstream 303 process included a scalable AIEX chromatography polishing step to account for these inconsistencies and 304 to improve the diagnostic performance of the antigens. 305

The two test antigens, tRBD from HEK cells and NP from E. coli, were further used for ELISA assay 306 development. We configured the assays with a number of sera taken from SARS-CoV-2-infected 307 individuals with weak antibody responses to ensure high assay sensitivity. In contrast to available 308 literature, we used high antigen coating concentrations (6 µg/mL) to yield satisfactory readouts (4, 19, 20) 309 and to achieve a high dynamic measurement range. A caveat of many assay validation studies is that 310 performance characteristics are skewed by the exclusive inclusion of samples from hospitalized 311 individuals, where robust antibody levels are to be expected (21). Likewise, the sole consideration of 312 healthy donors in control groups may lead to overestimated assay specificity, as the impact of potential 313 cross-reactive factors present in the general population is largely ignored. In this respect, auto-antibodies 314 commonly found in individuals with inflammatory diseases (22) were already described to cross-react with 315 SARS-CoV-1 antigens (23). To challenge our tests systems, we biased our large specificity cohort (n=1,126) 316 by including samples with an increased propensity for cross-reactivity, including sera from individuals with 317 inflammatory illnesses (n=359), sera from PCR-confirmed hCoV infections (n=8) and sera drawn during 318 winter months to increase the likelihood of respiratory infections (n=494). Similarly, our sensitivity cohort 319 (n=244) included convalescent sera from SARS-CoV-2-infected individuals covering the full spectrum of 320 clinical manifestations (from asymptomatic to ICU patients). Among them 21% of the sera were collected 321 from asymptomatic individuals or from individuals with mild to moderate illness, who may mount less 322 robust and durable antibody responses after an infection (24). Based on these cohorts, we defined 323 adequate test parameters to enable highly specific detection of SARS-CoV-2-specific antibodies. A cutoff 324 deduced by the 99 th percentile method (8.000 U/mL) allowed for high specific serodiagnosis with 99.2% 325 for the Technozym RBD Test and 99.1% for the Technozym NP Test (at sensitivities of 86.3% and 76.7%, 326 respectively). This is a remarkable result for a tetramethylbenzidine-based manual test system, 327

considering the highly diverse nature of our study cohorts. While some automated systems were 328 described to achieve specificities approximating 100% (25, 26), assay performance is highly cohort-329 specific. The use of diverse study cohorts was also associated with performance deteriorations in such 330 test platforms (i.e. Abbot, Specificity: 97.5%)(27). For the Meduni Wien Biobank cohort we had 331 performance data with CE-marked automated test systems available (10) to directly compare with our 332

ELISAs at the high specificity cut-off criterion (8.000 U/mL is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; specific antibodies and SARS-CoV-2 antigens, they appeared to have a limited effect on assay 339 performances (Fig. 4C) . 340

Yet, for an estimated seroprevalence of 5% in the general European population (28, 29), a test with a 341 specificity and sensitivity of 99.2% and 86.3%, respectively, only scores a PPV of 85.0% resulting in 15 342 false-positive results out of 100, which is still insufficient. In line with previous results from us and others 343 (30-32), we demonstrate that false-positive results are largely antigen-dependent (Fig. 2B, Fig. 4C) . 344

Orthogonal testing is suggested by the Centers of Disease Control and Prevention (CDC) to remedy 345 specificity problems in low transmission settings (33). Previous studies have used RBD as screening antigen 346 and the trimeric spike protein or the spike S2 domain in second-line tests to confirm initial positive results 347 (4, 32). Such conventional orthogonal test strategies, however, increase specificity often at the expense 348 of sensitivity. We therefore established an adaptive orthogonal test algorithm where positive sera were 349 first identified with the tRBD ELISA allowing for highly sensitive testing (at the expense of specificity) and 350 samples within a predefined area of uncertainty then underwent confirmatory testing with the NP ELISA 351 (10). This two-test algorithm resulted in a cumulative specificity of 99.8% and an even higher sensitivity 352 of 88.1% (+0.037, p<0.050),yielded a PPV of 96.3% [86.7-99.1] (Fig. 3) . This is an excellent result for a 353 manual test format and its specificity is on par with other orthogonal tests relying on automated systems 354 (10). 355

As the Technozym NP and RBD ELISAs provide a five-point calibrator, set ELISA antibody levels can be 356 quantified, compared and followed over time. For such an application, we chose a cut-off of 5.000 U/mL 357 that allowed for more sensitive analysis of antibody levels at acceptable specificity, adapted from the cut-358 off given by the Youden index. With convalescent sera taken at median 43-54 days post-symptom onset, 359 the tRBD ELISA allowed for a more sensitive detection of antibodies than the NP ELISA (Fig. 3A, B) . Yet, 360 time-resolved analysis of seroconversion demonstrated that NP-specific antibodies develop earlier after 361 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. an infection and true positive rates were consistently higher with the NP ELISA for samples collected 362 within the first 15 days post-symptom onset (Fig. 5, Table S2 ). This phenomenon has already been 363 described in patients infected with SARS-CoV-1(34, 35) and was associated with higher sensitivities of 364 other SARS-CoV-2 test systems, relying on the NP, in the early phase after an infection (36). Determining 365 the neutralizing capacity of SARS-CoV-2 anti-RBD antibodies is critical to elucidate possible protective 366 effects of the immune response. Considering all neutralizing activity above background as positive, we 367 observed neutralizing antibodies in 85% of the sera already by day five after symptom onset (Fig. 5) , which 368 is in line with previous studies (37, 38). Of note, RBD-seroconversion, defined by antibody levels above a 369 threshold of 5.000 U/mL, was observed for only 6% of the sera at this time point. Yet, despite 33 out of 370 35 samples demonstrating reactivity below our pre-defined cutoff, neutralizing titers correlated well with 371 RBD-specific IgG responses. A recent study demonstrated that the early neutralizing response is 372 dominated by RBD-specific IgA antibodies (39). As we exclusively measured RBD-specific IgG responses 373 we cannot rule out that part of the early neutralizing activity we observe derive from neutralizing IgA or 374 even earlier IgM responses. 375

Tests for the screening of reconvalescent COVID-19 patients for the presence of anti-SARS-CoV-2 376 antibodies are of great interest for identifying suitable donors for convalescent plasma therapy (40). A 377 retrospective, propensity score-matched case-control study performed at the Mount Sinai hospital (New 378 York, NY) provides evidence for a survival benefit in patients receiving convalescent plasma transfusion as 379 an effective intervention in COVID-19 (40). In August 2020, the FDA issued a new guidance on the 380 Emergency Use Authorization (EUA) for COVID-19 convalescent plasma, recommending plasma donations 381 to be qualified by either the Mount Sinai COVID-19 ELISA IgG Antibody Test or Ortho VITROS IgG assay 382 (41). Prior to this guidance, NTs of at least 1:160 were considered acceptable in the absence of high-titer 383 samples (42). As we did not have the beforementioned tests available, we qualified plasma donors 384 according to the NT 1:160 criterion. The fraction of samples exceeding this threshold gradually increased 385 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint over time and by day 15 after symptom onset, 53% of the sera and by day 22, 72% of sera had titers higher 386 than 1:160 (Fig. 5, Table S2 ). The geometric mean RBD titers in these sera corresponded to 159.1 U/mL 387 and 183.7 U/mL, respectively. Since correlates of protection from infection remain to be determined we 388 cannot deduce whether these titers are clinically relevant in prophylaxis, at this point. 389

In conclusion, we have developed two highly specific, quantitative, easy-to-implement and now 391 commercially available SARS-CoV-2 antibody tests and defined optimal thresholds for their application in 392 different aspects of clinical use. In addition to their simple format, the two tests are equally well suited as 393 most automated CE-marked systems for high specificity applications, such as seroprevalence studies. 394

Moreover, the RBD ELISA allows for the identification of donors for convalescent plasma therapy as RBD-395 specific antibody levels correlate well with the induction of functional neutralization responses. Both tests 396 allow to comprehensively monitor the dynamics of antibody responses after infection. Yet, our data 397 disclose different kinetics for antigen-specific antibody responses, which affect their performance at 398 different time points after an infection. These findings are essential for ongoing efforts to establish 399 serological tests for clinical diagnostics. In this respect, also test performance with convalescent sera 400 collected more than 2 months after infection and the effect of antigen-specific antibody waning should 401 be carefully addressed in future studies and compared to the comprehensive findings of this study. 402 403 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. 

The pET30acer-CASPON-NP expression vector was transformed into E. coli enGenes-X-press for growth-493 decoupled recombinant protein production as described elsewhere (8) Germany). All purified proteins were quantified by measuring their absorbance at A280 with a Nanodrop 529 instrument and stored at -80°C until further use. 530

The purification of NP was optimized and performed as described by De Vos and colleagues (9). In brief, 532 NP was produced by using the CASPON platform process(45) with modifications. The process consisted of 533 an IMAC capture step (WorkBeads 40 Ni NTA, Bio-Works, Uppsala, SE) of the clarified cell lysate. A 534 nuclease treatment (Salt Active Nuclease High Quality, ArcticZymes Technologies ASA, Tromsø, NO) was 535 required to reduce CASPON-NP nucleic acid binding. Imidazole was removed from the IMAC eluate using 536 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint For NP, full-length content was defined as FL[%]=FL/total area. 576

Interaction studies of RBD, tRBD and NP with in-house produced anti-RBD mAb CR3022 and a commercial 578 anti-SARS-CoV-2 nucleocapsid protein antibody (ab272852, Abcam, Cambridge, UK) were performed on 579 an Octet RED96e system using high precision streptavidin (SAX) biosensors (both from FortéBio, Fremont, 580 CA). Antibodies were biotinylated using the EZ-Link Sulfo-NHS-LC-Biotin kit (Thermo Fisher Scientific, 581

Waltham, MA). Excess sulfo-NHS-LC-biotin was quenched by adding Tris-HCl buffer (800 mM, pH 7.4) to a 582 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. However, the interaction between the CR3022 mAb and the final tRBD batches were also evaluated 605 kinetically by fitting the BLI data to a 2:1 heterogeneous ligand binding model. Note, although the CR3022 606 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint mAb has two identical binding sites, the second binding event is dependent on the first binding since 607 allosteric effects or sterical hindrance can ultimately lead to a positive or negative cooperative binding 608 behavior (51, 61, 62) . However, in case of the reported interaction, the affinity constant (KD) values are 609 very close to one other in the low nanomolar range. 610

The interaction between the NP protein and the anti-NP mAb is difficult to characterize due to avidity 611 effects that arise from the dimeric nature of both interaction partners. Kinetic evaluation of the BLI data 612 is problematic since the dissociation curves are heterogenic. Additionally, if the dissociation phase shows 613 less than 5% decrease in signal during the defined dissociation phase, as observed for the lower 614 concentration range of NP protein, a precise determination of the dissociation rate constants (kd) is not 615 possible (63, 64). However, it is feasible to calculate an upper limit for the kd (s -1 ) which is given by is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint mode for eluting peaks). MS-scans were recorded (range: 150-2,200 Da) and the six highest peaks were 631 selected for fragmentation. Instrument calibration was performed using ESI calibration mixture (Agilent, 632 Santa Clara, CA). The analysis files were converted (using Data Analysis, Bruker) to mgf files, which are 633 suitable for performing a MS/MS ion search with MASCOT. The files were searched against a database 634 containing the target sequences. In addition, manual glycopeptide searches were done. Glycopeptides 635 were identified as sets of peaks consisting of the peptide moiety and the attached N-glycan varying in the 636 number of HexNAc, hexose, deoxyhexose and pentose residues. Theoretical masses of these peptides 637 were determined using the monoisotopic masses for the respective amino acids and monosaccharides. 638

The present study includes work with human sera from three different sites. Acute lithium heparin plasma (1:1,200-diluted) or assay buffer (blank samples) was applied to each well. Assays were incubated for two 738 hours at RT on the plate shaker (600 rpm). Assay plates were placed on the magnetic plate holder and the 739 supernatants were poured off by inverting the plates. Microspheres were washed by removing the 740 magnetic plate holder and the addition of 100 µL Wash buffer (PBS; 0.05% (V/V) Tween-20; 0.05% (w/V) 741 NaN3; pH 7.4) per well. After two minutes of incubation at room temperature, plates were again placed 742 on the magnetic plate holder and supernatants were poured off. After three wash steps 50 µL of a 1:1 743 mixture of 2.5 μg/mL goat anti-human R-Phyco AffiniPure F(ab')2, Fcγ-specific (# 109-116-098) and F(ab')2-744 specific IgG (# 109-116-097, both Jackson ImmunoResearch Laboratories Inc., West Grove, PA) in Assay 745 buffer were added. Plates were incubated for 1 h at room temperature on the plate shaker (600 rpm) in 746 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. was applied (150 µL/well) and plates were incubated for 25 minutes at RT with shaking. Reactions were 769 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. were determined by ROC-analysis and the non-parametric 99th right-sided percentile method (CLSI C28-789 A3). Sensitivities, specificities, PPV, and negative predictive values (NPV, both at 5% estimated 790 seroprevalence) were calculated. ROC-analysis data from automated tests (including Abbott ARCHITECT 791 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; SARS-CoV-2 IgG, DiaSorin LIAISON® Anti-SARS-CoV-2 S1/S2 IgG) were available for 64 of the positive and 792 1117 of the negative samples from a previously published study (30). 793 (71) with the following alterations: the heat-treated sera were diluted 1:4 in triplicates in serum-free 805 HEPES-buffered DMEM medium. In the case neutralizing antibody titers were determined in human 806 lithium heparin plasma, no heat-treatment was applied and the medium was supplemented with 1x 807

Antibiotic/Antimycotic solution (Thermo Fisher Scientific, Waltham, MA). The heat treatment had no 808 effect on neutralizing titers, as verified in a pre-experiment on SARS-CoV-2 positive and negative plasma 809 samples. In addition, a toxicity control, which was processed the same way as plasma samples, was 810

included. Here, no virus was added, to prevent a false readout of the assay. Cytopathic effect (CPE) was 811 evaluated and scored for each well using an inverted optical microscope. To determine neutralization titer 812 the reciprocal of the highest serum dilution that protected more than 50% of the cells from the CPE was 813 used and was calculated according to Reed and Muench (72) .Briefly, assays were performed with Vero 814 . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. For the assay both sera and plasma was used. 818

819 Raw data were assessed for normality of distribution and homogeneity of variances using the D'Agostino-820

Pearson omnibus test before statistical procedures. Differences in median seroreactivities between pre-821 COVID and COVID sera were compared using the Mann-Whitney U tests on blank-corrected log2-822 transformed median fluorescence intensities (Luminex data) or OD490 absorbances (ELISA), respectively. 823

Correlation analyses of nonparametric data were performed by Spearman's rank-order correlation (rs), 824 otherwise Pearsons' correlation (r) was used. Relative IgG signals of outliers against SARS-CoV-2 and hCoV 825 antigens were compared by One-Way ANOVA followed by a Sidak test to correct for multiple comparisons. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted January 20, 2021. ; https://doi.org/10.1101/2021.01.19.21249921 doi: medRxiv preprint 

Antibody 840 responses to SARS-CoV-2 in patients with COVID-19

SARS-CoV-2 diagnostic pipeline

Diagnostic accuracy of serological 847 tests for covid-19: systematic review and meta-analysis

A serological assay to detect SARS-CoV-2 seroconversion in humans

The use of relative operating characteristic (ROC) curves in test performance 855 evaluation

Expression of SARS-857 coronavirus nucleocapsid protein in Escherichia coli and Lactococcus lactis for serodiagnosis and 858 mucosal vaccination

861 Production of Circularly Permuted Caspase-2 for Affinity Fusion-Tag Removal: Cloning, Expression in 862 Escherichia coli, Purification, and Characterization

Bacteriophage 864 Inspired Growth-Decoupled Recombinant Protein Production in Escherichia coli

868 Comprehensive characterization of highly pure SARS-CoV-2 nucleocapsid protein produced in 869 Escherichia coli by native chromatography

Increasing both specificity 873 and sensitivity of SARS-CoV-2 antibody tests by using an adaptive orthogonal testing approach

Connecting clusters of COVID-19: an epidemiological and serological 878 investigation. The Lancet Infectious Diseases

N-terminally truncated nucleocapsid 880 protein of SARS-CoV-2 as a better serological marker than whole nucleocapsid protein in evaluating 881 the immunogenicity of inactivated SARS-CoV-2

Dual ELISA using 884 SARS-CoV-2 nucleocapsid protein produced in E. coli and CHO cells reveals epitope masking by N-885 glycosylation

Rapid production of SARS-CoV-2 receptor binding domain (RBD) and spike specific 890 monoclonal antibody CR3022 in Nicotiana benthamiana

Copious production of SARS-CoV nucleocapsid protein employing codon 893 optimized synthetic gene

Trimeric SARS-CoV-2 Spike Proteins Produced 896 from CHO Cells in Bioreactors Are High-Quality Antigens. Processes

High-level expression of 898 the monomeric SARS-CoV-2 S protein RBD 320-537 in stably transfected CHO cells by the EEF1A1 -899 based plasmid vector

Minimizing 901 fucosylation in insect cell-derived glycoproteins reduces binding to IgE antibodies from the sera of 902 patients with allergy

SARS-CoV-2-specific 906 ELISA development

Development and Optimization of In-house ELISA for Detection of Human 909

IgG Antibody to SARS-CoV-2 Full Length Spike Protein

Serum-IgG responses to 912 SARS-CoV-2 after mild and severe COVID-19 infection and analysis of IgG non-responders

Autoantibody Biomarkers in Rheumatic Diseases

Cross-917 reaction of SARS-CoV antigen with autoantibodies in autoimmune diseases

Clinical and immunological 921 assessment of asymptomatic SARS-CoV-2 infections

Performance Characteristics of the Abbott Architect SARS-CoV-2 IgG Assay and 924

Comparison of the Clinical Performances of the Abbott Alinity IgG, Abbott 927 Architect IgM, and Roche Elecsys Total SARS-CoV-2 Antibody Assays

Performance of six SARS-CoV-2 immunoassays in comparison 931 with microneutralisation

7% of the 934 Austrian population had antibodies to SARS-CoV-2. Extrapolation of Austria-wide COVID 19 935 prevalence study

SARS-CoV-2 seroprevalence worldwide: a systematic review 941 and meta-analysis

Side-by-Side Comparison of Three Fully Automated SARS-CoV-2 Antibody 945 Assays with a Focus on Specificity

Clinical evaluation of five different automated SARS-CoV-2 949 serology assays in a cohort of hospitalized COVID-19 patients

Assays Enable Surveillance of Low-Prevalence Communities and Reveal Durable Humoral Immunity. 957 Immunity

Interim Guidelines for COVID-19 Antibody Testing 959 in Clinical and Public Health Settings

Differential Sensitivities of Severe Acute Respiratory Syndrome (SARS)

Coronavirus Spike Polypeptide Enzyme-Linked Immunosorbent Assay (ELISA) and SARS Coronavirus 964

Nucleocapsid Protein ELISA for Serodiagnosis of SARS Coronavirus Pneumonia

Profiles of Antibody Responses against Severe Acute Respiratory Syndrome 968

Coronavirus Recombinant Proteins and Their Potential Use as Diagnostic Markers. Clinical Diagnostic 969 Laboratory Immunology

972 Evaluation of Three Commercial SARS-CoV-2 Serologic Assays and Their Performance in Two-Test 973

Rapid Generation of Neutralizing Antibody Responses

Evaluating the Association of Clinical Characteristics 983 With Neutralizing Antibody Levels in Patients Who Have Recovered From Mild

IgA dominates the early neutralizing antibody response to SARS-CoV-2

994 Convalescent plasma treatment of severe COVID-19: a propensity score-matched control study. 995 Nature Medicine

41. Food and Drug Administration (FDA), Recommendations for Investigational COVID-19 Convalescent 997 Plasma

Neutralizing Antibody Responses

A new 1006 coronavirus associated with human respiratory disease in China

pEAQ: versatile expression vectors for easy and 1008 quick transient expression of heterologous proteins in plants

Advanced purification platform using circularly permuted caspase-2 for 1012 affinity fusion-tag removal to produce native fibroblast growth factor 2. submitted to

Potent binding of 1015 2019 novel coronavirus spike protein by a SARS coronavirus-specific human monoclonal antibody

Marker-free plasmids for gene therapeutic applications-Lack of antibiotic resistance gene 1019 substantially improves the manufacturing process

Automated alkaline lysis for industrial scale cGMP 1021 production of pharmaceutical grade plasmid-DNA

Application Note 28-4094-85 AA: PlasmidSelect Xtra for downstream processing of 1023 supercoiled plasmid DNA

High-level and high-throughput recombinant protein production by transient 1025 transfection of suspension-growing human 293-EBNA1 cells

Fcab-HER2 Interaction: a Ménage à Trois. Lessons from X-1029 Ray and Solution Studies

The Transcriptome of 1031 the Baculovirus Autographa californica Multiple Nucleopolyhedrovirus in Trichoplusia ni Cells

Comparative transcriptome 1034 analysis of a Trichoplusia ni cell line reveals distinct host responses to intracellular and secreted 1035 protein products expressed by recombinant baculoviruses

One-shot vaccination with an insect cell-derived low-dose influenza A H7 1039 virus-like particle preparation protects mice against H7N9 challenge

Generation of glyco-engineered Nicotiana benthamiana for the production of 1042 monoclonal antibodies with a homogeneous human-like N-glycan structure: XylT and FucT down-1043 regulation in N. benthamiana

N-Glycosylation engineering of plants for the biosynthesis of glycoproteins with bisected and 1046 branched complex N-glycans

Distinct Fcα receptor N -glycans modulate the binding affinity to immunoglobulin A (IgA) antibodies. 1049

Ligand binding assays at equilibrium: validation and interpretation: 1051 Equilibrium binding assays

Limit of blank, limit of detection and limit of quantitation

A detection and quantification label-free tool to speed up downstream processing of model 1056 mucins

Few and Far Between: How HIV May Be Evading Antibody Avidity

Antigen clasping by two antigen-binding sites of an exceptionally specific antibody for histone 1062 methylation

An approach for liposome immobilization using 1064 sterically stabilized micelles (SSMs) as a precursor for bio-layer interferometry-based interaction 1065 studies

Kinetic analysis 1069 of a high-affinity antibody/antigen interaction performed by multiple Biacore users

Probing the binding mechanism and affinity of tanezumab, a 1072 recombinant humanized anti-NGF monoclonal antibody, using a repertoire of biosensors

Evaluation of four commercial, fully automated SARS-CoV-2 antibody tests suggests 1077 a revision of the Siemens SARS-CoV-2 IgG assay

A statistically defined endpoint titer determination method for 1080 immunoassays

The LEAD (Lung, Heart, Social, Body) Study: 1083 Objectives, Methodology, and External Validity of the Population-Based Cohort Study

Usage Data and Scientific Impact of the Prospectively Established Fluid Bioresources at the Hospital-1087

SARS-CoV-2 Cell Entry Depends 1090 on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

An approach to lifting self-isolation for health care workers with prolonged shedding of SARS-CoV-2 1094 RNA

A SIMPLE METHOD OF ESTIMATING FIFTY PER CENT ENDPOINTS12

United States) for providing the 1101 constructs used for the production of RBD and CR3022. We thank George Lomonossoff

PBL) (Norwich, UK) for supplying the pEAQ-HT expression 1103 vector. The authors thank Naila Avdic

University of Natural Resources and Life Sciences (BOKU) Vienna or and/or

ACIB GmbH affiliated) for their support in the cloning, production and analysis of SARS-CoV-2 antigens 1108 described in this work. The BLI and mass spectrometry equipment was kindly provided by the

VIBT GmbH and the BOKU Core Facilities Biomolecular & Cellular Analysis (BmCA) and Mass Spectrometry

Boehringer Ingelheim RCV GmbH & Co KG fully supported the endeavor and granted access to 1111 manufacturing technologies for process development and manufacturing

The authors thank Irene Schaffner and Jakob Wallner (BOKU BmCA) for assisting in BLI measurements and 1113

ForteBio for providing SAX biosensors. The authors want to thank Maria Ozsvar-Kozma

Medical University Vienna) for 1115 outstanding technical assistance. The authors thank the following collaborators for providing biomaterial 1116 and data for ELISA evaluation: Manfred Nairz

Markus Anliker

Harald Schennach (Central Institute for Blood Transfusion & 1120

Daniel Aletaha (Department of Internal 1121

Andreas Leiherer (Vorarlberg 1123 Institute for Vascular Investigation and Treatment

BOKU Spin-off Novasign GmbH supported the sustainability concept of BOKU by creating and 1125 hosting the BOKU-COVID19 Portal (portal.boku-covid19.at) that enables researchers worldwide to obtain 1126 trial samples of the within this work described SARS-CoV-2 antigens free of charge

We thank the Vienna Science and Technology Fund (WWTF) for partial funding of this project

We thank the University of Natural Resources and Life Sciences (BOKU) Vienna, the Ludwig 1132

Boltzmann Institute for Experimental and Clinical Traumatology and the Ludwig Boltzmann Gesellschaft 1133 for financial support of the project

12J6520N and V443719N, and the OEAD (Austria) for scholarship ICM

Author Contributions

V. conducting a research and investigation process, specifically performing the 1146 experiments, or data/evidence collection

G. performed data validation and evaluated data reproducibility

Ri.S. performed formal analysis to analyze and synthesize data

were responsible for 1157 supervising research activities and for

Competing interests: 1163The authors declare that they have no competing interest, but J.M. and G.S. owns an interest in enGenes 1164Biotech GmbH, the legal entity commercializing the enGenes-X-press technology and the antigens 1165