key: cord-0877520-ntfb5160
authors: Barlev-Gross, M.; Weiss, S.; Ben-Shmuel, A.; Sittner, A.; Eden, K.; Mazuz, N.; Glinert, I.; Bar-David, E.; Puni, R.; Amit, S.; Kriger, O.; Schuster, O.; Alcalay, R.; Makdasi, E.; Epstein, E.; Noy-Porat, T.; Rosenfeld, R.; Achdout, H.; Mazor, O.; Israely, T.; Levy, H.; Mechaly, A.
title: Spike vs nucleocapsid SARS-CoV-2 antigen detection: application in nasopharyngeal swab specimens
date: 2021-03-11
journal: nan
DOI: 10.1101/2021.03.08.21253148
sha: 661fd5bc33ea929938629bc2ba4b7d07397de021
doc_id: 877520
cord_uid: ntfb5160

Public health experts emphasize the need for quick, point-of-care SARS-CoV-2 detection as an effective strategy for controlling virus spread. To this end, many antigen detection devices were developed and commercialized. These devices are mostly based on detecting SARS-CoV-2 nucleocapsid protein. Recently, alerts issued by both the FDA and the CDC raised concerns regarding the devices tendency to exhibit false positive results. In this work we developed a novel alternative spike-based antigen assay, comprised of four high-affinity, specific monoclonal antibodies, directed against different epitopes on the spike S1 subunit. The assay performance was evaluated for COVID-19 detection from nasopharyngeal swabs, compared to an in-house nucleocapsid-based assay, composed of antibodies directed against the nucleocapsid. Detection of COVID-19 was carried out in a cohort of 284 qRT-PCR positive and negative nasopharyngeal swab samples. The time resolved fluorescence (TRF) ELISA spike-assay displayed very high specificity (99%) accompanied with a somewhat lower sensitivity (66% for Ct<25), compared to the nucleocapsid ELISA assay which was more sensitive (85% for Ct<25) while less specific (87% specificity). Despite being out-performed by qRT-PCR, we suggest that there is room for such tests in the clinical setting, as cheap and rapid pre-screening tools. Our results further suggest that when applying antigen detection, one must consider its intended application (sensitivity vs specificity), taking into consideration that the nucleocapsid might not be the optimal target. In this regard, we propose that a combination of both antigens might contribute to the validity of the results.

In December 2019, a novel zoonotic coronavirus (SARS-CoV-2), identified initially in Wuhan China, led to 51 the pandemic known as . While some affected individuals are asymptomatic, others exhibit 52 mild symptoms such as fever and cough with some deteriorating to pneumonia, severe acute respiratory 53 syndrome and death (2, 3). According to recent updates (February 2021) 

Antibodies: Polyclonal anti NC protein antibodies were prepared by immunizing rabbits with 100g of 108 purified protein in incomplete freund's adjuvant by subcutaneous injection. The vaccination regime included 109 a prime and two boost doses, given in 4-week intervals. Serum was collected 10 days post the second boost.

Monoclonal anti SARS-CoV-2 RBD, NTD and NC antibodies were isolated from a phage-display library, 111 constructed from COVID-19 patients showing substantial illness and panned against plate adhered spike and 112 nucleocapsid as previously described (18, 19, 22) . Antibodies were expressed and purified as described (18). 

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Fluorescence (TRF). A 50µl solution of the capture antibodies (in 50mM Na2CO3, pH 9.6) was added to the 132 microplate wells and incubated at 4 o C O.N. The plates were then washed (3xPBS+0.05% Tween20) and 133 blocked (100µl/well, with PBS containing 2%BSA and 0.05% Tween20). After one hour of incubation at 134 37 o C the plates were washed (3xPBS + 0.05% Tween 20) and stored at -20 o C or used directly. Samples 135 (proteins or virus) were diluted in PBS or nasopharyngeal swabs buffer, and loaded on the plates.

Nasopharyngeal swabs samples were added to the plate following addition of NP-40 (to a final concentration 137 of 0.1%) and centrifugation (2 min, 15000 rpm, 4 o C). After 30 minutes incubation the plates were washed 138 and biotinylated reporter antibodies were loaded for additional 30 min. After an additional wash step, 139 streptavidin-europium diluted 1:1000 was loaded for 20 min. After a final wash enhancement solution 140 (Delfia, Perkin elmar) was added and the resulting signal was measured using a microplate reader (excitation 141 340nm, emission 612nm).

Clinical samples: Nasopharyngeal swab specimens were collected as part of routine nursing home sampling 143 in Israel (tenants and staff) and from the emergency department of SHEBA Medical Centre (SMC, Ramat 144 Gan, Israel). qRT-PCR testing of samples collected from SMC was carried out in SMC while the nursing 145 home samples were tested in-house. Samples from SMC were accompanied with clinical details regarding 146 symptomology, as well as time elapsed since symptom onset. A total of the 46 positive and 23 negative qPCR 147 samples were obtained from the emergency department and a total of 100 positive and 115 negative qPCR 148 samples were acquired from the nursing homes. Samples were analyzed by the in-house developed antigen 149 tests (Not all samples were analyzed by both tests due to insufficient sample volume). The study was 150 approved by the SMC institutional review board committee (approval number -7769-20-SMC).

Signal analysis: TRF signals were calculated as signal-to-noise (S/N) ratios between the fluorescence values 152 of antigen containing samples ('S' for signal) -or PBS/swab buffer containing samples ('N' for noise). To 153 determine the limit of detection (LOD) for each test, the average background fluorescence was calculated as 154 the mean of at least six "noise" replicates (PBS/swab buffer). The LOD was defined as 3 standard deviations 155 (SD) above the average background with a coefficient of variation (CV) < 20%. These values were used to 156 calculate the signal-to-noise ratio that was used as a cutoff value (S/N ratios ≥ 1.7) and was considered to be 157 . CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 the LOD threshold in accordance with ICH guidelines for validation of analytical procedures (23). This 158 calculation enabled the normalization of multiple experiments. Sensitivity was defined as the proportion of 159 the PCR positive samples that tasted positive in the ELISA tests while specificity was defined as the 160 proportion of PCR negative samples that tasted negative in the ELISA tests.

panel of high-affinity, anti-spike human monoclonal antibodies from phage-display libraries generated from 164 sera of 19) . The performance of these antibodies was evaluated for SARS-CoV-2 165 detection in a TRF ELISA-based format. The optimized spike assay (S-assay) includes a combination of four 166 monoclonal antibodies that bind separate epitopes and are thus able to bind simultaneously to the spike 167 protein with no apparent structural hindrance as determined using BioLayer interferometry -BLI (Fig 1a) . CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 immunized with SARS-CoV-2 nucleocapsid, are implemented as capture antibodies. A monoclonal antibody 185 (KD=0.7nM, kon=5.6x10 5 1/Ms; koff=3.8x10 -4 1/s), isolated from the same COVID-19 patient's sera phage-186 display library, is applied as the reporter antibody. The NC-assay's specificity and sensitivity were analyzed 187 as described for the S-assay. Results (Fig 2a and b , red line) indicate that this assay specifically recognizes 188 its intended target (nucleocapsid) with no detectable activity on the reciprocal antigen (spike). Similar to the 189 S-assay, this assay also demonstrates a wide dynamic range (three orders of magnitude) and enables detection 190 of 1 ng/ml of the recombinant nucleocapsid. 

Detection of SARS-CoV-2 from nasopharyngeal swab specimens: Finally, we wanted to verify that the 199 tests can be used for SARS-CoV-2 detection in nasopharyngeal swab specimens. To this end we tested a total 200 of 146 qRT-PCR positive and 138 qRT-PCR negative swab samples. qRT-PCR analyzed swab samples were 201 stored at 4 o C (for up to a month), processed and tested in concert by both tests (not all samples were analyzed 202 by both tests due to insufficient sample volume). The criteria for performance assessment were sensitivity, 203 specificity and accuracy. In this respect, qRT-PCR results were used to determine "true positive" (meaning 204 COVID-19 patient) or "true negative" (meaning healthy or non-COVID-19 patient) status for each sample.

Evaluation was based on crude S/N ratios that were determined for each assay. A plot of S/N values vs Ct 206 for both assays is presented (Fig 3) . In this plot, Ct values above 41 represent negative qRT-PCR samples. 207 S/N values were than translated to "positive" and "negative" results based on a calculated cut-off value (as 208 detailed in materials and methods), resulting in contingency tables that were analyzed with GraphPad Prism 209 6 ( Table 1 ). The analysis is presented separately for all the samples, including the full spectra of Ct values 210 ("all Ct", Table 1 left panel) and for samples with high viral loads ("Ct<25", Table 1 

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint (Table 1 and Fig 3) indicate that both assays are suitable for COVID-19 diagnosis from nasopharyngeal 212 swabs. As indicated (Table 1) , the S-assay demonstrates higher specificity (99.1%) compared to that of the 213 NC-assay (87.0%). These results are in accordance with CDC's lab alert, cautioning against false positive 214 results when implementing antigen-based detection (which is mostly nucleocapsid-based). Alas, the higher 215 specificity of the S-assay is coupled with lower sensitivity: 49.3% vs 63.4% for the S-assay and the NC-assay 216 respectively (Table 1, 

We wanted to determine whether the difference in sensitivity between the tests is statistically significant ( CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint besides their qRT-PCR results. The assumption, that nursing homes samples include positive samples from 239 asymptomatic or recuperating patients (several days after symptom onset and no longer infective), is 240 reasonable. It is also logical to assume that most qRT-PCR negative samples originating from routine 241 screening were probably collected from healthy individuals. 

One can surmise that despite the statistical difference, assay performance is similar for both the ER and 264 nursing homes samples and depends solely on the assay's LOD.

. CC-BY-ND 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. 

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 In the context of the suggested scenarios presented in Fig 5, it is important to note that in our hands the 293 sensitivity of each test seemed to correlate directly with the viral load regardless of whether the sample 294 originated from a symptomatic or an asymptomatic individual in accordance with the literature (25, 27).

However, for both applications, antigen tests are suitable up to 5-10 days after disease onset while viral loads 296 are high and the test actually indicates the presence of the virus and not residual RNA.

The antigen of choice in most commercial assays i.e. the nucleocapsid, indeed demonstrated higher sensitivity 298 than that observed with the spike-based assay (Fig 3 and table 1 ). This higher sensitivity might have been 299 predicted in advance, given that according to the literature, only around 100 spike trimers are present on each 300 virion, resulting in an estimated total of 300 monomers, while around 1000 copies of the nucleocapsid are 301 expressed in each virion (31). Non-specific reactions however, are harder to predict given that sequence 302 homology is not the only indicator for possible interactions. In this regard, serological studies indicate that 303 cross-reaction of pre-existing antibodies in SARS-CoV-2 negative individuals with the spike protein of 304 endemic and seasonal corona viruses is low and is predominantly directed against S2 (32, 33). This might 305 indicate that our S-assay, comprised of a combination of anti RBD and NTD antibodies, will exhibit high 306 specificity. The fact that the spike assay is comprised of four different, high affinity, monoclonal antibodies 307 can enhance specificity while still maintaining the ability to recognize emerging mutants (34). In another 308 serological study (35), the authors examined cross-reactions of SARS-CoV-2 nucleocapsid with other 309 coronaviruses in an attempt to assess its utility for seroprevalence studies. The authors found that anti-SARS-310 CoV-2-nucleocapsid IgG also recognized N229E nucleocapsid and IgG directed against HKU1 nucleocapsid 311 recognized SARS-CoV-2 nucleocapsid. Cross reactivity was stronger with alpha-rather than beta-312 coronaviruses despite having less sequence identity, reiterating the importance of conformational 313 recognition. Other coronaviruses notwithstanding, unpredicted interactions with bacteria or viruses present 314 in healthy or symptomatic individuals might cause non-specific interactions (as is indeed the case for some 315 of the commercial nucleocapsid-based antigen tests). It seems that only sampling of a large population of 316 healthy or "corona-like" symptomatic individuals will provide quality data regarding true assay specificity 317 (in addition to cross reactivity studies with known viruses, bacteria and fungus, as was performed by some 318 . CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint that the spike-based test was significantly more specific than the nucleocapsid based test (Table 1) .

The evaluated parameters of sensitivity and specificity depend solely on the cut-off values for positive vs 321 negative results for each test. As indicated in Fig 3, the gray zone, delimited by the dashed and dotted lines, 322 represents the decline in sensitivity resulting from elevating the cut-off value in order to improve assay 323 specificity. We applied receiver operating characteristic (ROC) analysis and multinomial logistic regression 324 in an attempt to define an optimal LOD for both tests. While the S-assay displays 100% specificity when 325 applying a S/N ratio>2, for the NC-assay, in order to achieve 98% specificity, assay cut-off should be above 326 S/N>7 (resulting in an increase in the assay's LOD to 4x10 5 pfu/ml). Application of different antibodies 327 might improve assay's specificity.

Another important issue when considering which antigen in more suitable for SARS-CoV-2 detection is the 329 genetic variability and mutation rate of both antigens. During the evolution and global spread of the 330 pandemic, this topic was addressed by several labs (36-38), describing several viral mutants that are now 331 circulating all over the world (34). It is logical to assume that escape mutants are more likely to manifest in 332 the spike protein (especially for mutants arising due to excessive use of plasma from recuperating individuals 333 for treatment of COVID-19 patients), however researches discovered mutations in both genes (36-38). The 334 nucleocapsid, as well as open reading frames 3a and 8, are implicated as having roles in virulence, 335 transmission and pathogenicity (39) and are therefore as likely to undergo mutations as the spike protein 336 despite not being presented on the virus surface. We therefore propose that for SARS-CoV-2 detection, both 337 antigens will be used simultaneously or consecutively as was also suggested by others (40) 

The authors declare no conflict of interest.

. CC-BY-ND 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 March 11, 2021. 

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint Fig. 1 Characterization of the antibodies incorporated in the SARS-CoV-2 TRF-based S-assay. a.

Concomitant binding of anti-SARS-CoV-2 antibodies. The ability of the assay's antibodies to bind 459 simultaneously to SARS-CoV-2 was tested using the Octet Red biolayer interferometry (BLI) system.

Biotinylated BL11 was immobilized to a streptavidin sensor and interacted with the spike's S1 subunit. The Bio-BL11

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint c.

. CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint ns ns . CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101/2021.03.08.21253148 doi: medRxiv preprint . CC-BY-ND 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 March 11, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

Author Correction: A new coronavirus 353 associated with human respiratory disease in China

The COVID-19 epidemic

Predicting infectious SARS-CoV-2 from diagnostic samples

Viral RNA load as 426 determined by cell culture as a management tool for discharge of SARS-CoV-2 patients from infectious 427 disease wards

Serologic 431 cross-reactivity of SARS-CoV-2 with endemic and seasonal Betacoronaviruses. medRxiv

Preexisting and de novo 433 humoral immunity to SARS-CoV-2 in humans

Could new COVID variants undermine vaccines? Labs scramble to find out

Immunogenicity and 437 crossreactivity of antibodies to SARS-CoV-2 nucleocapsid protein

All authors read and approved of the final manuscript 350