key: cord-1044163-iitkq67w
authors: Kainulainen, M. H.; Bergeron, E.; Chatterjee, P.; Chapman, A. P.; Lee, J.; Chida, A.; Tang, X.; Wharton, R. E.; Mercer, K. B.; Petway, M.; Jenks, H. M.; Flietstra, T. D.; Schuh, A. J.; Satheshkumar, P. S.; Chaitram, J. M.; Owen, S. M.; Finn, M. G.; Goldstein, J. M.; Montgomery, J. M.; Spiropoulou, C. F.
title: High-throughput quantitation of SARS-CoV-2 antibodies in a single-dilution homogeneous assay
date: 2020-09-18
journal: nan
DOI: 10.1101/2020.09.16.20195446
sha: ee48772cd3dd88f79cb4c0b246aa04ce83582dc9
doc_id: 1044163
cord_uid: iitkq67w

SARS-CoV-2 emerged in late 2019 and has since spread around the world, causing a pandemic of the respiratory disease COVID-19. Detecting antibodies against the virus is an essential tool for tracking infections and developing vaccines. Such tests, primarily utilizing the enzyme-linked immunosorbent assay (ELISA) principle, can be either qualitative (reporting positive/negative results) or quantitative (reporting a value representing the quantity of specific antibodies). Quantitation is vital for determining stability or decline of antibody titers in convalescence, efficacy of different vaccination regimens, and detection of asymptomatic infections. Quantitation typically requires two-step ELISA testing, in which samples are first screened in a qualitative assay and positive samples are subsequently analyzed as a dilution series. To overcome the throughput limitations of this approach, we developed a simpler and faster system that is highly automatable and achieves quantitation in a single-dilution screening format with sensitivity and specificity comparable to those of ELISA.

This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint Introduction 49

Detecting viral nucleic acids, primarily by reverse-transcription quantitative polymerase chain reaction (RT-qPCR), 50 is the method of choice for diagnosing acute viral infections. However, diagnosing acute infections does not account 51 for asymptomatic infections or bias resulting from availability of care or individuals not seeking care (1) . Therefore, 52 serology plays an important part in completing the epidemiological picture when a new pathogen, such as severe 53 acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (2, 3), emerges. Analyzing humoral responses in 54 convenience samples is a quick way to estimate how many individuals have already been exposed to a pathogen. 55

On the other hand, designing community-level and special population studies offers the opportunity to solicit 56 tailored participant information, such as symptom history, for correlation analyses (1) . Apart from surveys that aim 57 to produce generalizable information, serology has also been used to link together individual COVID-19 clusters 58

where nucleic acid evidence was absent (4). 59

While the serological studies mentioned above could potentially be conducted with qualitative serology assays 60 (which report presence or absence of SARS-CoV-2 antibodies), answering other pertinent questions requires 61 quantitative determination of antibody levels. One timely question is the possibility of waning immunity to SARS-62

CoV-2 and reinfection with this pathogen. Some studies have called into question the long-term stability of the 63 humoral immune response against this virus (5-7), and small-scale human challenge studies with a benign human 64 coronavirus have shown that at least asymptomatic reinfection with the same strain (8) or a related one (9) is possible 65 as soon as a year after initial infection. Meanwhile, other studies did not observe rapid waning of the humoral 66 response (10, 11) and yet others have evaluated the immune signature of mild SARS-CoV-2 infections and predicted 67 that this immune memory is likely to be protective (12). Longitudinal sampling combined with accurate serological 68 quantitation is necessary to monitor variations in antibody levels associated with waning immunity or reinfection 69 (13). Vaccine trials also require quantitative serology to establish immune correlates and to decide on optimal 70 vaccine dose and regimen. 71

Another timely question is the role that other animal hosts may play in the future of the SARS-CoV-2 pandemic. 72

To date, the natural reservoir or intermediate hosts leading to the emergence of this virus in humans have not been 73 definitively identified. However, cats (14), deer mice (15), ferrets, fruit bats (16), and golden hamsters (17, 18) can 74 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Dose-responsive assay activation by an RBD-specific monoclonal antibody or corresponding F(abʹ) 2 or Fab 116 fragments (orange) or control antibody (gray). C) Activation by specific IgG and IgM antibodies. Human IgG and 117 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint IgM fractions were purified from a SARS-CoV-2 specific serum pool or a negative control pool and tested for 118 activity in the assay. B and C: technical triplicates with averages and standard deviations presented. 119

Training set and cut-off determination 120 Assay performance was optimized by varying factors, including antigen concentrations and ratios, buffer and 121 blocker choices, plastic materials, incubation times, and substrates. Additionally, we observed that using RBD 122 domain consisting of SARS-CoV-2 residues 319-591 (25) offered lower background luminescence signal, and 123 therefore higher sensitivity, than using residues 319-541 (26) (optimization data not shown). 124

In order to establish the assay in human diagnostics, we opted to use the two-step strategy, in which the assay cut- and-read assay and with ELISA against RBD with the same antigen length (spike residues 319-591). The mix-and-131 read assay signals obtained from single serum dilution varied by 3 orders of magnitude ( Fig. 2A) . The values were 132 categorized for receiver operating characteristics (ROC) analysis based on presence or absence of prior PCR-133 confirmed infection ( Fig. 2B and 2C ). We decided to prioritize assay specificity and chose the cut-off value of 2.5, 134 which enabled the highest sensitivity (antibodies detected in 92.9% of the samples with an earlier SARS-CoV-2 135 nucleic acid finding) while maintaining 100% specificity. We noticed that the average negative result was somewhat 136 below 1, indicating modest signal inhibition by serum. The high cut-off value (>6 standard deviations above the 137 average of negative sample signal) also minimizes impact of user error in case sample is omitted. Agreement 138 between ELISA and the mix-and-read assay was very high (kappa value 0.949; 95% confidence interval 0.908-139 0.989). The mix-and-read assay found no antibodies in 6 of 85 samples with confirmed SARS-CoV-2 infection; 4 140 of these samples were also negative by RBD ELISA; 1 was ELISA positive with a weak signal and 1 was ELISA 141 positive with a moderate signal. 142 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. Summary of the training set data. C) Receiver operating characteristics (ROC) analysis of the data and cut-off 151 selection. Cut-off value of 2.5 (as fold over blank samples) was chosen by ROC analysis of the training set data so 152 that 100% specificity could be maintained. 153 154 While serum is the standard sample matrix for serology, we wanted to establish whether plasma could be tested as 155 well. Paired human serum and EDTA plasma samples gave similar signals in the mix-and-read assay, suggesting 156 that EDTA plasma samples could also be analyzed ( Supplementary Fig. 3A ). Furthermore, testing mouse sera from 157 animals immunized with SARS-CoV-2 spike antigens formally showed that the assay works with non-human serum 158 as well ( Supplementary Fig. 3B ). 159 for use under a CC0 license.

This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint highlighted by red asterisks). B) Summary of the independent set results. C) Day-to-day reproducibility of the mix-179 and-read assay. Samples from the independent validation set were tested on 2 consecutive days and correlation 180 analysis performed on log-transformed values. Samples that gave a discordant qualitative result between the days 181 (near the cut-off, above one day and below the other) were judged negative for purposes of sensitivity and specificity 182 calculations. 183 184 Linear range 185

The assay principle used sets a theoretical upper limit to antibody quantitation. If specific antibodies (or, more 186 precisely, high-affinity antigen binding sites) exceed RBDs, then the number of antibody molecules binding 2 RBDs 187 will decline whereas the number of antibodies binding only 1 or no RBDs will increase. Therefore, increasing the 188 antibody concentration excessively may actually result in declining signal. To test the theory in practice, monoclonal 189 antibodies were spiked into non-reactive human serum at concentrations between 1 ng/mL and 100 µg/mL. Dose 190 response curves showed that maximal signal was obtained at mAb concentrations 10-30 µg/mL, corresponding to 191 1:1 to 1:3 stoichiometry between the RBDs and the antigen binding sites (Fig. 4A ). Increasing the mAb 192 concentration to 100 µg/mL resulted in lower, though strong signal, suggesting that correct qualitative results would 193 be obtained with even the most reactive clinical samples. 194

Single-dilution quantitation with the mix-and-read assay 195

Titration results with monoclonal antibodies suggested that the assay signal saturates and declines when excessive 196 antibody concentrations are used. To test if the same phenomenon could impact SARS-CoV-2 RBD antibody 197 quantitation in clinical serum samples, 5 samples with high signal-to-background ratios were serially diluted in non-198 reactive human serum and dose-response curves were generated. All samples gave the strongest signal when 199 undiluted, showing that even the most reactive samples among the >180 positives identified here do not contain 200 high enough amounts of specific antibodies to saturate the assay. However, the shape of the dose-response curves 201 near 1:1 dilution suggest that these strong samples are approaching the higher quantitation limit (Fig. 4B) . 202 for use under a CC0 license.

This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. To investigate if SARS-CoV-2 RBD antibody titers could be inferred from a single screening dilution, end-point 203 ELISA titers were determined for 89 positive samples of the validation set. Plotting ELISA optical density (OD) 204 values from the screening dilution of 1:100 against the end-point titer showed the expected correlation at low 205 dilutions but loss of quantitation in samples that reached saturation OD in the screening (Fig. 4C) . In contrast, no 206 apparent saturation was seen in the mix-and-read assay, and those values and ELISA end-point titers correlated at 207 Pearson r 0.87 (95% confidence interval 0.80-0.91, p < 0.0001) and R 2 0.75 (p < 0.0001), indicating that sample 208 screening at a single dilution is predictive of ELISA endpoint titer in a semi-quantitative manner (Fig. 4D) . 209 This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint

The mix-and-read assay identified RBD antibodies in 88.7% of the samples with history of confirmed SARS-CoV-223 2 infection, while our RBD ELISA identified antibodies in 92.5% of those samples. The time interval from symptom 224 onset to sampling was known for 9 of the 14 samples that were found seronegative by both assays and ranged from 225 14 to 30 days, suggesting that time of sampling was not the cause of negative results. Since the mix-and-read assay 226 can detect IgM antibodies as well as IgG antibodies, class switching should not be required for detecting 227 seroconversion. However, in general, the lower physiological concentration of IgM antibodies may render them the 228 less potent assay activator as compared to IgG antibodies. In conclusion, the implied seroconversion rate of ~90% 229 is comparable to rates reported by others (5, 10, 29) , although universal seroconversion has also been reported (30). 230

The mix-and-read assay showed 100% specificity in this sample set while the specificity of ELISA was 98%. The 231 wide dynamic range of the mix-and-read assay allowed us to keep the cut-off value over 6 standard deviations above 232 the mean of the negative controls. Although apparent sensitivity could be marginally improved and the correlation 233 of the two assays further increased by lowering the threshold and accepting false positives at low frequency, we 234 chose to use a cut-off that emphasizes specificity. It is important to remember that, especially in populations in 235 which antibody prevalence is low, the positive predictive value (likelihood that a positive result is correct) 236 deteriorates rapidly when specificity falls below 100% (31, 32). For example, a test with 99% specificity and high 237 sensitivity may be useful for many purposes in an area of high prevalence, but in an area where true prevalence is 238 1%, half of the positive test results would be false. Of note, 2 of the 6 negative control samples that produced an 239 ELISA signal were near the chosen ELISA cut-off, but 4 (collected prior to recognized SARS-CoV-2 emergence) 240 gave moderate to strong OD values, implying cross-reactivity between the RBD antigen and unknown antigens. 241

Interestingly, despite the high overall correlation between the two assays, these ELISA-reactive samples were 242 negative when read on the mix-and-read assay (3 with signal < 1, one being close to the cut-off value 2.5). The 243 reason for this phenomenon is unknown, but increased specificity may arise from factors like the strict requirement 244 for simultaneous binding of 2 antigens, better preservation of antigen structure in solution than on plate surface, or 245 steric masking of cross-reactive epitopes by the split luciferase moieties. 246

One limitation of our study is that the work was conducted primarily to evaluate a new diagnostic approach. Control 247 samples were sourced based on availability and not as part of a serological survey, and some associated data, 248 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint including demographics, accurate location, symptoms, underlying conditions, and hospitalization history of the 249 individuals were absent or incomplete and are not reported here. Furthermore, we note that the samples originated 250 in the U.S., and would encourage testing representative control cohorts any time a serological test is implemented 251 in a new area or population, or indeed species, to account for potential differences in baseline reactivity. Of note, 252 we found that convalescent SARS-CoV-1 sera can also activate the assay (data not shown). While this has little 253 practical relevance in human diagnostics due to the small number of convalescent SARS-CoV-1 patients, it is 254 noteworthy with regards to ecological investigations and instances where infection with a closely related virus is a 255 possibility. 256

In summary, we used SARS-CoV-2 as an example to demonstrate that modern protein complementation enables 257 design of simple yet robust serological assays that are easy to automate. Complementation detects antibodies due 258 to their binding of multiple antigens simultaneously, and thus the assay has no apparent antibody class or species 259 restrictions. The improved dynamic range means that more information can be derived from primary screening than 260 is possible with ELISA. Complementation-based serology has the power to increase the output of human SARS-261

CoV-2 serology efforts and to advance One Health investigations. 262

While this manuscript was in preparation, a qualitative commercial product utilizing the same principle was released 264 This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint 14 The DNA sequences for the above-described fusion proteins were cloned into an episomal vector (System 274 Biosciences) into which a puromycin resistance cassette had been added to enable selection of pools stably 275 expressing these proteins. To express the proteins, human Expi293 cells (Thermo Scientific), growing in Expi293 276

Expression medium, were transfected with FectoPRO reagent (PolyPlus Transfection). The supernatants were 277 clarified by slow-speed centrifugation and filtered over polyethersulfone (0.2 micron) membranes, and His-tagged 278 proteins were purified by immobilized metal affinity chromatography using HisTrap Excel nickel columns (Cytiva). 279

After dialysis in PBS, the proteins were flash frozen in liquid nitrogen and stored at -80°C. 280

The proteins for RBD/ACE2 binding assay were produced the same way as above. Human codon-optimized ACE2 281 (residues 1-615, NM_001371415) was expressed with C-terminal LgBit using the same linker and His-tag sequence 282 as described earlier. Similarly, RBD (residues 1-14 and 319-541) was expressed with C-terminal SmBit. 283

The RBD that was used as ELISA antigen consisted of residues 319-591 and was expressed with an N-terminal IL-284 6 signal peptide, 8 × His and 3C protease cleavage site. The protein was produced and His-purified as above, and 285 the epitope tag removed by digestion with 3C protease. The resulting product was further purified by size-exclusion 286 chromatography using Superdex 200 columns. 287

Mix-and-read assay procedure 288

After initial optimization in a 96-well format, the assay was moved to 384-well plates, and all data presented here 289 were produced using that setup. 20 ng of SmBit-SARS-2-RBD, and equivalent amount of LgBit-SARS-2-RBD in 290 1:1 molar ratio were added per well of 384-well plates in 20 µL 1% BSA/PBS (w/V) diluent. 10 µL sample was 291 then added, and the mix was incubated for 1 h at room temperature. 30 µL Promega NanoGlo assay reagent was 292 then added and the luminescence resulting from SmBit and LgBit proximity was quantified 10 min later using a 293 Synergy Neo2 instrument (BioTek) with 0.5 s integration time and with gain set to 200 and read height to 7.5 mm. 294

Wells with diluent instead of sample were used to determine the assay baseline. Samples with strong signal (up to 295 2-3 orders of magnitude above baseline) were removed and the plate was read immediately again to minimize error 296 from signal leak between wells. Assay performance was monitored using 4 standards: normal human serum (from 297 plasma) as a negative control, and RBD-specific mouse monoclonal antibody 3A2 spiked into the same serum at 3 298 different concentrations (10, 2.5, and 0.5 µg/mL). 299 for use under a CC0 license.

This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint ELISA 300

ELISAs were conducted against in-house produced SARS-CoV-2 RBD antigen (spike residues 319-591). 50 ng of 301 antigen was bound to Immulon 2 HB plates by overnight incubation in 100 µL PBS at 4°C. Wells without antigen 302 were similarly prepared. Wells were washed 2× with 300 µL 0.1% Tween-20 in PBS (PBS-T) and blocked with 303 300 µL 5% (w/V) non-fat dry milk (CellSignaling Technologies) in PBS-T for 1 h at room temperature (RT). After 304 3 washes, samples diluted 1:100 in 100 µL 5% milk in PBS-T were added and incubated for 1 h at RT. The plates 305 were washed 3×, and mouse secondary anti-human IgG antibody conjugated to HRP (Accurate Chemical 306 #JMH035098) was added at 1:10.000 dilution in 100 µL 5% milk in PBS-T and incubated for 30 min at RT. After 307 3 washes, 100 µL TMB Ultra ELISA substrate (Thermo Fisher Scientific) was added and reactions were stopped 308 after 20 min at RT with addition of 100 µL 1 M hydrochloric acid. Optical density was read at 450 nm and densities 309 from antigen-containing wells were corrected by reducing the values obtained from respective no-antigen wells. 310 RBD-ACE2 binding assay procedure 311 3.2 ng RBD-Smbit (residues 319-541, C-terminal tag) was diluted in 20 µL 0.02% BSA/PBS and mixed with 20 312 µL mAb dilution (17 µg/mL) or mAb fragment dilution (corresponding molarity) in 96-wells. After 1 h incubation 313 at RT, ACE2-LgBit was added in volume of 20 µL so that 10× molar excess over the RBD was achieved. After 1 h 314 incubation at RT, 60 µL NanoGlo reagent (Promega) was added and luminescence quantified as with the mix-and-315 read assay. Wells where antibody dilution was substituted with the diluent served to determine the baseline signal 316 of RBD-ACE2 binding. 317

Monoclonal antibodies and mouse sera 318

The monoclonal antibodies were generated by repeated mouse immunizations and cloned using standard methods. 319

Polyclonal mouse serum from such immunizations was used to test the assay performance. Thorough description 320 of these antibodies will be presented elsewhere (Chapman et al., in preparation). All animal procedures were 321 approved by the Institutional Animal Care and Use Committee of the Georgia Institute of Technology. 322

Monoclonal antibody 3A2 was cleaved into F(abʹ) 2 and Fab fragments using Pierce Mouse IgG1 Fab and F(abʹ)2 323

Preparation Kit (Thermo Fisher Scientific) according to manufacturer's instructions. Briefly, the two fragments 324 were generated by controlling the specificity of immobilized ficin by cysteine-HCl concentration and incubation 325 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint time. Immobilized protein A was used to remove uncleaved mAb and Fc fragments. The fragments were 326 concentrated with 500× buffer exchange against PBS using spin columns with 10 kDa molecular weight cut-off 327 (Amicon, Millipore Sigma). mAbs were diluted to 10/3/1/0.3 µg/mL and the fragments were diluted to 328 corresponding antigen-binding molarities and analyzed as above. 329 mAb 3A2 was also used as quantification control. For this purpose, purified mAb was spiked into normal human 330 serum at concentrations of 10, 2.5, or 0.5 µg/mL; normal serum was left unspiked as the negative control. The 331 preparations were aliquoted and frozen, and freshly thawed aliquots were used to monitor run-to-run and day-to-332 day assay performance. 333

Positive and negative human serum and plasma samples were sourced via BioIVT, iSpecimen, StemExpress and 335 Emory University. Time to collection from symptom onset was available for 129 of 186 positive serum controls 336 and varied from 6 to 71 days, with 10% and 90% percentiles at 15 and 53 days, respectively. Additionally, residual 337 serum samples collected for hantavirus and Zika virus serology prior to emergence of SARS-CoV-2 were tested as 338 negative controls (as part of the training set and the validation set, respectively). The hantavirus samples were 339 inactivated by 5 × 10 6 RAD gamma irradiation from a 60 Co source. All samples were heated to 56°C for 10 min 340 before use. The serum samples from commercial sources were analyzed in a blinded fashion. Approximately 6% of 341 initially analyzed samples were excluded from final analysis after identifying multiple samples from the same donor 342 (pre-outbreak residual samples) or due to inconclusive information regarding previous SARS-CoV-2 nucleic acid 343 testing results (commercially sourced samples after de-blinding). Residual specimen materials were used for 344 diagnostics development under a protocol that was reviewed and approved by the CDC Institutional Review Board § . 345

Purification of IgG and IgM from human serum 346 Human serum samples from SARS-CoV-2 positive donors or control donors (N = 14 each) were pooled, and IgG 347 and IgM antibodies were purified according to manufacturer's recommendations using CaptureSelect CH1-Xl and 348

Poros CaptureSelect IgM Affinity matrices (Thermo Fisher Scientific). The samples were concentrated and the 349 buffer was changed to PBS using 10.000 Da spin columns (Amicon). The purified fractions were analyzed by SDS-350 PAGE both under non-heated, non-reduced conditions and under heated, reduced conditions. 351 for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint for use under a CC0 license. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint

The COVID-358 19 serology studies workshop: Recommendations and challenges

A new coronavirus 361 associated with human respiratory disease in China

WHO statement regarding cluster of pneumonia cases in Wuhan

Connecting clusters of COVID-19: an epidemiological and serological investigation

Clinical and immunological assessment 371 of asymptomatic SARS-CoV-2 infections

Rapid decay of anti-SARS-CoV-2 antibodies in persons with mild Covid-374 19

Longitudinal analysis of the humoral response to SARS-CoV-2 spike RBD in convalescent plasma 377 donors

The time course of the immune response to experimental 379 coronavirus infection of man

The behaviour of recent isolates of human respiratory coronavirus in vitro and in volunteers: Evidence 381 of heterogeneity among 229E-related strains

Dynamics of IgG 387 seroconversion and pathophysiology of COVID-19 infections

SARS-CoV-2 infection induces robust, neutralizing antibody responses that are stable for at least three months

Functional SARS-CoV-395 2-specific immune memory persists after mild COVID-19, medRxiv

Coronavirus protective immunity is short-lasting

Susceptibility of ferrets, cats, 401 dogs, and other domesticated animals to SARS-coronavirus 2

SARS-CoV-2 infection, neuropathogenesis and 404 transmission among deer mice: Implications for reverse zoonosis to

SARS-CoV-2 in fruit bats, ferrets, pigs, 408 and chickens: an experimental transmission study

Simulation of the clinical and pathological manifestations of

COVID-19) in golden Syrian hamster model: implications for disease pathogenesis and 414 transmissibility

Pathogenesis and transmission of SARS-CoV-2 in golden 417 hamsters

SARS-CoV-2 421 infection in farmed minks, the Netherlands

Luminescence-and fluorescence-based 424 complementation assays to screen for GPCR oligomerization: Current state of the art

Applications of protein fragment complementation assays for analyzing biomolecular 427 interactions and biochemical networks in living cells

It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted

doi: medRxiv preprint reporter optimized for accurate measurement of protein interactions in cells

Severe Acute Respiratory Syndrome

Coronavirus 2-Specific antibody responses in coronavirus disease patients

The RBD of the spike protein 439 of SARS-group Coronaviruses is a highly specific target of SARS-CoV-2 antibodies but not other pathogenic 440 human and animal Coronavirus antibodies

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

A serological 447 assay to detect SARS-CoV-2 seroconversion in humans

COVID-19: laboratory evaluations of serological assays -GOV

Comparison of different serological 451 assays for SARS-CoV-2 in real life

Navy service members -USS Theodore Roosevelt

Antibody responses to SARS-CoV-2 in 461 patients with COVID-19

Test, test, test for COVID-463 19 antibodies: the importance of sensitivity, specificity and predictive powers

Are antibodies tests accurate? Understanding predictive values and uncertainty 465 of serology tests for the novel coronavirus

It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted

This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted September 18, 2020. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted September 18, 2020. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This article is a US Government work. It is not subject to copyright under 17 USC 105 and is also made available preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this this version posted September 18, 2020. . https://doi.org/10.1101/2020.09.16.20195446 doi: medRxiv preprint