Qualitative Changes in the SARS-CoV-2 Antibody Response in the Post-Infection Phase Impact the estimates of infections in Population-Based Seroprevalence Studies

In the present study, we have determined SARS-CoV-2-specific antibody responses in a cohort of 96 individuals during the acute phase of infection and in a cohort of 578 individuals enrolled in a seroprevalence population study in Switzerland including three groups, i.e. subjects with previous RT-PCR confirmed SARS-CoV-2 infections (n=90), positive patient contact (n=177) and random selected subjects (n=311). Six serological assays detecting predominantly IgG antibodies targeting either the Spike (S) and/or the nucleocapsid (N) proteins were used including also a Luminex based assay using an S protein in its native trimeric form. Antibody responses against the S and/or the N proteins were equally sensitive in the acute infection phase although differences in sensitivity (range 83 to 97% 16-33 days post-initial symptoms) were observed between the different assays and the Luminex S protein trimer assay was the most sensitive. Interestingly, antibody responses against the N protein appear to wane in the post-infection phase of the infection while those against the S protein persist over time, as indicated by the drop in sensitivity of the assays targeting the N protein (sensitivity range 71-77%). Assays detecting anti-N IgG antibodies may substantially underestimate the proportion of SARS-CoV-2 infections in the groups of patient positive contacts, i.e. 10.9 to 32.2% reduction (P<0.05-<0.0001) and in the random selected general population, i.e. up to 45% reduction (P<0.05). The overall reduction in seroprevalence for the total cohort ranged from 9.4 to 31% (P<0.0009-<0.0001). Of note, the assay using the S protein in its native trimer form was more sensitive as compared to those using monomeric S proteins. These results indicate that assays targeting the S protein, ideally the trimeric form, should be implemented as reference test to estimate SARS-CoV-2 infections in seroprevalence population studies.


79 80
The SARS-CoV-2 is currently causing a devastating pandemic with more than 12.7 million 81 documented infections and more than 566'000 deaths, according to the latest WHO situation 82 report from July 13 th , 2020 1 . However, the true incidence of the infection is largely 83 underestimated, since in most countries asymptomatic and paucisymptomatic people are 84 tested only if they came in direct contact with sick patients or belong to at-risk subgroups. One of two proteins serve as baits in the majority of serological assays: the nucleocapsid 100 protein (N) and the Spike protein (S). It has been suggested that the serological assays 101 targeting the S protein are more specific while those targeting at N may be more sensitive, 102 particularly in the early phase of infection 7 .

104
A comparison of 17 different serological assays using sera from hospitalized patients with 105 moderate to severe COVID-19 disease confirmed the overall higher sensitivity of assays 106 targeting the N protein investigating 8 . In this study, the DiaSorin CLIA that targets the S 107 protein only exhibited 92% sensitivity whereas the Epitope Diagnostics ELISA and the Snibe 108 assay that both target the N protein exhibited a sensitivity of 96% 8 . However, this increased sensitivity of N-based assays might be at the expense of specificity, given the relatively high 110 protein sequence similarity of the N protein of SARS-CoV-2 with nucleocapsid proteins of 111 other Coronaviridae and other viruses. Moreover, during the SARS outbreak (2002)(2003)(2004), 112 Chia et al observed that anti-N antibodies waned earlier than anti-S antibodies 9 . Thus, assays 113 targeting the S proteins might be more specific and circumvent a possible decrease of 114 antibodies, previously observed with N protein of the SARS virus. Thus, assays targeting the S 115 protein might be more specific and circumvent a possible decrease in antibody titers, as the 116 one previously observed after SARS. Furthermore, it is still unknown the durability of SARS-  In a recent study, the only test scoring negative on 404 sera taken before the pandemics (in 122 2018 & 2019) was the DiaSorin test, which targets the S protein; however, this high specificity 123 was at the expense of a lower sensitivity (92%). We hypothesized that the lower sensitivity of 124 most S-based assays reported so far might be explained due to their use of a monomeric form 125 of S moieties, including the S1 or RBD domains, and that using S in its native trimeric 126 conformation might increase sensitivity while preserving specificity.

128
We thus developed an IgA and IgG serological assay using the trimeric form of the S protein 129 and the Luminex technology. This new assay was first evaluated using 96 sera obtained during The newly developed Luminex assay had the highest sensitivity in the sera of the post-140 infection phase thus providing more accurate estimates of SARS-CoV-2 infections in the 141 general population.

142
Taken together, these results indicate that S protein based assays should be used in 143 population-based seroprevalence studies and that the trimeric form is associated with greater 144 sensitivity than the monomeric form. CC-BY-NC-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.

148
Novel Serological Binding Assay using the S protein trimer 149 A stabilized trimer of the full-length S protein, encompassing both its S1 and S2 moieties, was 150 coupled to beads for capturing antibodies in a new Luminex assay. We hypothesized that 151 conformational epitopes would be preserved in the trimeric S protein, providing a greater 152 sensitivity to detect IgG antibodies (Supplementary Figure 1A and B). First, the specificity for 153 IgG antibody binding was established with sera from 256 pre-COVID-19 pandemic healthy 154 adults from 18 to 81 years of age and an additional set of 108 patients (Figure 1A) adults and for the diverse panel of 108 subjects. A cut-off for positivity was set at 4-fold above 160 a negative control standard, which is slightly more than four standard deviation above the 161 mean of all negative control samples (mean MFI ratio 0.84 + 4×0.75 SD). Using this threshold, 162 only one sera of the 256 pre-COVID-19 people and two patients with acute HIV or CMV viral 163 infections gave a positive signal (Figure 1 A). As such, the Luminex assay using the stable 164 trimeric S protein gave a high overall specificity of 99.2% and no cross-reactive antibodies 165 were detected in sera from people infected with pre-pandemic coronaviruses or from 166 patients with autoimmune diseases that can produce polyreactive antibodies. Considering 167 that two of the three false positives from the 364 SARS-CoV-2 negative donors had MFI signals 168 less than 6 ( Figure 1A), an additional criteria for positivity was established for large general 169 population screens, including the post-infection cohort. Here, sera with signal intensities 170 between 4 and 6 were defined as being at the limit of positivity, which increases the assay 171 sensitivity to 99.7% with only one acute HIV infected subject having a 6.8 MFI signal.

173
The sensitivity of the assay was next evaluated using sera from 96 acutely infected SARS-CoV-174 2 PCR-positive patients with blood sampling at 0-5 days, 6-10 days, 11-15 days and 16-33 days 175 post-onset of symptoms (POS). As anticipated, sera collected during the early stage of the 176 infection (0-5 days POS) had low or undetectable levels of anti-S protein IgG antibodies, with 177 a rate of positivity of 12.5% (1 in 8 subjects; Figure 1B). Seropositivity increased to 42.1% 178 . CC-BY-NC-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 July 17, 2020. mean MFI signal ratio of 85 was obtained with the Luminex assay with sera collected from 181 these patients 16-33 days POS, which was significantly higher than the signal obtained with 182 sera from all other time categories (P =0.05 relative to 11-15 days and P < 0.0001 relative to 183 0-5 days and 6-10 days POS; Mann-Whitney test). Interestingly, the only subject that was 184 negative in the S protein trimer assay at day 25 post-onset of symptoms became seropositive 185 when re-tested seven days later. Since signal intensity often comes at the expense of 186 specificity, it is all the more important that the Luminex assay using the S protein trimer  binding curve with the commercially available CR3022 anti-S1 IgG1 antibody ( Figure 1C). This 196 clone was initially identified as a neutralizing antibody for SARS-CoV-1, but was subsequently 197 shown to bind specifically to the SARS-CoV-2 Spike protein 15 . Evaluation of serum dilutions 198 from 1/300 down to 1/24'000 allowed to precisely determine the anti S protein IgG antibody 199 concentrations that ranged from 54 µg/ml to 820 µg/ml in a panel of sera tested from patients  CC-BY-NC-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.

214
Comparison of Immunoassays targeting the S and N SARS-CoV-2 proteins during the acute 215 phase of infection 216 The new Luminex S protein trimer IgG assay was compared with five commercially available 217 SARS-CoV-2 immunoassays: i) two ELISAs from EuroImmun and Epitope Diagnostics detecting 218 IgG against the S1 and N proteins, respectively, ii) two CLIA from Diasorin and Snibe detecting 219 IgG against S and S+N proteins, respectively, and iii) a pan-Ig ECLIA from Roche targeting the CC-BY-NC-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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint assays using monomeric proteins but inferior to those obtained with the trimeric S protein 243 (Supplemental Figure 2A and B).

245
We also compared the Luminex S protein trimer IgG and IgA assay with the EuroImmun ELISA 246 IgG and IgA test. We evaluated these four tests using 81 sera from the acute phase patients. 247 We observed that the two IgA tests detected anti-SARS-CoV-2 antibodies earlier than the IgG McNemar test). Although both IgA antibody tests show this same trend, it appeared that the 257 Luminex S protein trimer IgA assay was systematically more sensitive (56%) than the 258 EuroImmun ELISA IgA tests (48%) in all early-infected subjects up to 10 days POS ( Table 2). CC-BY-NC-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.

275
The two tests detecting anti-S antibodies, Diasorin and EuroImmun, both had significantly 276 higher sensitivity (88.9%) than the Epitope Diagnostics ELISA anti-N (76.7%). Finally, the Snibe 277 CLIA IgG test, which targets mainly the N protein and an undefined part of the S protein, 278 showed a much lower sensitivity, i.e. 71.1% (Table 3). significantly better and detected between 9.4% and 31% more seropositive participants than 301 the other five assays (Figure 5).

303
Taken together, these results indicate that assays detecting anti-N IgG antibodies may 304 substantially (i.e. up to 45%) underestimate the proportion of SARS-CoV-2 exposed individuals 305 . CC-BY-NC-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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint compared to those detecting anti-S IgG antibodies in population-based seroprevalence 306 studies.  underestimation ranging from 11 to 33% for the former and 30 to 45% for the latter group.

335
Of note, the new Luminex-based assay using the trimeric S protein showed a greater 336 sensitivity than two assays using monomeric S1 or RBD. The underestimation of SARS-CoV-2 337 . CC-BY-NC-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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint seropositive individuals was in the range of 18-22% in the 'positive patient contacts' and 30-338 35% in the population-based samples. The greater sensitivity of the Luminex assay using the 339 trimeric S protein likely results from the conservation of conformational epitopes. This 340 increased sensitivity was not obtained at the expense of cross-reactivity, since the specificity 341 of the novel Luminex assay was 99%, in spite of including sera typically prone to cross-

368
. CC-BY-NC-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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint In conclusion, these results provide new insights in the evolution of the SARS-CoV-2 antibody  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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint  CC-BY-NC-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 July 17, 2020. CC-BY-NC-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.

466
The purity of S protein trimer was determined to be > 99% pure by SDS-PAGE analysis.

468
Receptor binding domain (RBD) and S1 SARS-CoV-2 proteins were prepared as previously  Epitope Diagnostic IgG ELISA were done manually for the essential of the protocol at the 496 . CC-BY-NC-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 July 17, 2020. .  515   516  517  518  519  520  521  522  523  524  525  526  527  528  529  530  531  532  533  534 . CC-BY-NC-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 July 17, 2020. .  . CC-BY-NC-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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint 563 564 565 . CC-BY-NC-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 July 17, 2020.  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 July 17, 2020. . https://doi.org/10.1101/2020.07.14.20153536 doi: medRxiv preprint  . CC-BY-NC-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.