key: cord-321854-cy8vyb6j
authors: Ripperger, Tyler J.; Uhrlaub, Jennifer L.; Watanabe, Makiko; Wong, Rachel; Castaneda, Yvonne; Pizzato, Hannah A.; Thompson, Mallory R.; Bradshaw, Christine; Weinkauf, Craig C.; Bime, Christian; Erickson, Heidi L.; Knox, Kenneth; Bixby, Billie; Parthasarathy, Sairam; Chaudhary, Sachin; Natt, Bhupinder; Cristan, Elaine; El Aini, Tammer; Rischard, Franz; Campion, Janet; Chopra, Madhav; Insel, Michael; Sam, Afshin; Knepler, James L.; Capaldi, Andrew P.; Spier, Catherine M.; Dake, Michael D.; Edwards, Taylor; Kaplan, Matthew E.; Scott, Serena Jain; Hypes, Cameron; Mosier, Jarrod; Harris, David T.; LaFleur, Bonnie J.; Sprissler, Ryan; Nikolich-Žugich, Janko; Bhattacharya, Deepta
title: Orthogonal SARS-CoV-2 Serological Assays Enable Surveillance of Low Prevalence Communities and Reveal Durable Humoral Immunity.
date: 2020-10-14
journal: Immunity
DOI: 10.1016/j.immuni.2020.10.004
sha: 
doc_id: 321854
cord_uid: cy8vyb6j

We conducted a serological study to define correlates of immunity against SARS-CoV-2. Relative to mild COVID-19 cases, individuals with severe disease exhibited elevated virus-neutralizing titers and antibodies against nucleocapsid (N) and the receptor binding domain (RBD) of spike protein. Age and sex played lesser roles. All cases, including asymptomatic individuals, seroconverted by 2 weeks post-PCR confirmation. Spike RBD and S2 and neutralizing antibodies remained detectable through 5-7 months post-onset, whereas α-N titers diminished. Testing of 5882 members of the local community revealed only 1 sample with seroreactivity to both RBD and S2 that lacked neutralizing antibodies. This fidelity could not be achieved with either RBD or S2 alone. Thus, inclusion of multiple independent assays improved the accuracy of antibody tests in low seroprevalence communities and revealed differences in antibody kinetics depending on the antigen. We conclude that neutralizing antibodies are stably produced for at least 5-7 months after SARS-CoV-2 infection.

reduction neutralization test (PRNT) titers, which we quantified as the final dilution at which 90% 134 viral neutralization occurred (PRNT 90 ) ( Figure 1A) . RBD To determine if RBD was capable of distinguishing between SARS-CoV-2 exposed and 141 uninfected individuals and to set preliminary thresholds for positive calls, we initially tested 1:40 142 serum dilutions of samples from 30 PCR+ SARS-CoV-2 infected individuals and 32 samples 143 collected prior to September, 2019, well before the onset of the current pandemic ( Figure S1D) . 144

Using this test data set, we established a preliminary positive cutoff OD 450 value of 0.12, equal 145 to 3 standard deviations above the mean values of the negative controls. We next used this 146 preliminary threshold to test an expanded cohort of 320 negative control samples collected prior 147 to 2020. ( Figure 1B ). Reactivity to RBD was clearly distinguishable for the majority of positive 148 samples from negative controls ( Figure 1B) . However, 6.5% of the expanded negative control 149 group displayed RBD reactivity that overlapped with PCR+ individuals (Figure 1B, blue shade) , 150 some of whom may have been early into disease and had not yet generated high levels of 151 antibodies. To quantify the sensitivity of the assay relative to time of diagnosis, we measured 152 antibody levels to RBD and plotted these values against time following SARS-CoV-2 PCR+ 153 confirmation. Whereas the sensitivity was modest within the first two weeks, after 2 weeks, 42 of 154 43 samples showed high ELISA signal ( Figure 1C ). Based on these data, samples were 155 considered seropositive at OD 450 numbers above 0.39, a value slightly above the highest OD 156 obtained from the 352 subjects in the negative control group (Figure 1B) . Sera were considered 157 negative at OD 450 values below 0.12. Finally, we created an indeterminate call at OD 450 values between 0.12-0.39, as we observed some overlap between negative controls and PCR-159 confirmed samples in this range (Figure 1B, blue shade) . 160 We next applied this assay to community testing and obtained serum samples from 5882 To improve the positive predictive value, we considered the use of an orthogonal 174 antigenically distinct test. Previous studies have used full length S protein as a secondary 175 screen following RBD ELISAs (Amanat et al., 2020). While this improves the sensitivity of the 176 assay and is perfectly reasonable in high seroprevalence communities such as New York City, 177 RBD is part of S and is not antigenically distinct. Thus, a false positive for RBD would 178

presumably also be apparent in S ELISAs. We therefore first tested nucleocapsid (N) protein, as 179 several other commercial serological tests quantify antibodies to this antigen (Bryan et al., 2020; 180 Burbelo et al., 2020) . IgG antibody titers to N protein in our collected sample cohort showed a 181 strong correlation to PRNT 90 titers (Figure 2A) . A weaker correlation was observed between N-182 reactive IgM levels and PRNT 90 titers ( Figure S2A) . We next assayed reactivity to N antigen overlapped substantially between negative and positive controls ( Figure 2B) . Moreover, 5 185 confirmed COVID-19 samples showed very weak reactivity to N ( Figure 2B) . Because of the 186 relatively poor performance of N protein as an antigen in our hands, we next tested the S2 187 domain of S protein as another candidate to determine seropositivity. RBD is located on the S1 188 domain, rendering S2 antigenically distinct (Bosch et al., 2003; Li, 2016; Wrapp et al., 2020) . 189

IgG antibody titers to S2 correlated well with PRNT 90 titers ( Figure 2C) , consistent with reports 190 of S2-specific neutralizing antibodies to SARS-CoV-1 and SARS-CoV-2 (Duan et al., 2005; 191 Song et al., 2020). Assessment of S2 serum reactivity in the pre-2019 cohort revealed that 192 approximately 3.3% of these samples overlapped with signals in PCR-confirmed COVID-19 193 samples ( Figure 2D) . We thereafter employed a threshold of OD 450 >0.35, as our cutoff for S2 194 positivity, which was 5 standard deviations above the average seroreactivity from the original 195 32-samples from the negative control cohort. Specificity control testing using 272 negative 196 control sera showed that reactivities of negative samples against RBD and S2 were largely 197 independent of one another, as samples with high signal for one antigen rarely showed similar 198 background for the other (Figure 2E ). Based on these data, we chose to rely on combined RBD 199 and S2-reactivities as accurate indicators of prior SARS-CoV-2 exposure. 200

With this improved combinatorial RBD and S2 assay to exclude false positives, we re-201 examined the original samples from the cohort of 5882 subjects that displayed RBD OD 450 202 values greater than 0.12 ( Figure 1D-E) . Of the 13 non-neutralizing samples that displayed high 203 (OD 450 >0.39) RBD reactivity, 12 lacked S2 reactivity ( Figure 2F ). In contrast, the remaining 60 204 RBD+ neutralizing samples all displayed substantial reactivity to S2 ( Figure 2F) . Five of the 9 205 samples that fell below the RBD cutoff, yet still neutralized virus, displayed strong reactivity to 206 S2 ( Figure 2F ). Based on these data, we established a scoring criterion of RBD OD 450 >0.39, S2 207 and all other samples as seronegative. Applying these criteria to 320 samples obtained prior to 209 2020 would lead to 317 negative, 3 indeterminate, and 0 positive calls. Using these same (Atyeo et al., 2020). We therefore examined our data for these trends. First, in our PCR-223 confirmed cohort, we plotted IgG titers relative to the time of disease onset, stratified by disease 224 severity. Severe disease (hospital admission) correlated with significantly higher antibody titers 225 against RBD and N than those with mild disease, who were symptomatic but did not require 226 hospital admission, whereas S2 titers were not statistically significantly different (Figure 3A-C) . 227

Neutralizing titers were also higher in those with severe disease relative to mild cases ( Figure  228 3D). Through campus screening efforts, we also identified 6 PCR+ individuals who either never 229 developed symptoms or had only a brief and mild headache or anosmia. Although previous 230

reports suggested that such individuals may infrequently seroconvert or frequently serorevert 231 Ko et al., 2020) . Given that older adults, as well as those of male sex, exhibit disproportional 234 morbidity and mortality from COVID-19, we also sought to test whether humoral immunity in 235 these subjects may be quantitatively reduced . Contrary to this expectation, we 236 did not observe any adverse impact of advanced age on humoral immunity (Figure 3E-H) . before settling to a more stable nadir at later timepoints, as would be expected for all acute viral 248 infections. We considered the possibility that we may have missed subjects that had 249 seroreverted prior to their antibody test, thereby incorrectly raising our estimates of the durability 250 of antibody production. Therefore, to examine the duration of IgG production in more depth, a 251 subset of seropositive individuals with relatively low titers was tested longitudinally up to 226 252 days post-onset. These data again revealed stable RBD and S2 IgG levels at later stages of 253 convalescence (Figures 4A-B) . However, N-reactive IgG levels were quite variable and most 254 samples approached the lower limit of detection at later timepoints ( Figure 4C) . A direct 255 comparison in matched subjects of the changes in RBD, S2, and N IgG titers over time 256 confirmed the variability in N responses and rapid decline in a subset of individuals ( Figure 4D ). 257 of time in all but one subject ( Figure 4E) , which showed evidence of neutralizing antibodies that 259 did not quite reach a PRNT 90 titer of 20 ( Figure S4) . These data suggest persistent neutralizing, 260 RBD, and S2-specific antibodies, but variable and often declining N-reactive titers during 261

convalescence. Together, these data are consistent with the maintenance of functionally 262 important antibody production for at least several months after infection, and caution against the 263 use of α-N antibodies to estimate immunity or seroprevalence. Here, we demonstrated that using two antigenically distinct serological tests can greatly 267 remedy specificity problems that are exacerbated in low SARS-CoV-2 seroprevalence 268 communities. RBD and S2 seroreactivity behaved independently for SARS-CoV-2-unexposed 269 individuals, thereby suggesting that the theoretical false positive rate of the overall assay is the 270 product of the two tests. Using neutralization assays to confirm these results, we found our 271 empirically determined false positive rate to be <0.02% (1/5882), consistent with the 272 independence of the RBD and S2 tests. The tight co-incidence between RBD/S2 positivity and 273 the presence of neutralizing antibodies, even in low seroprevalence populations, is especially 274 valuable for identifying individuals who likely have some degree of immunity. Surprisingly, 275 nucleocapsid (N), which is used by several commercial serological tests as an antigen, did not 276 perform as well in our assays, with high false positive and negative rates. The reasons for the differences in antibody responses across antigens are difficult to 306 explain, given the identical inflammatory environment in which these responses arose. One 307 possibility is that the avidities of germline precursors differ for N-and S-protein specificities. For 308 both memory and plasma cells, there appears to be a 'sweet spot' of antigen avidity that Taken together, we have reported a highly specific serological assay for SARS-CoV-2 323 exposure that is usable in very low seroprevalence communities, and that returns positive 324 results that are highly co-incident with virus neutralization. Using this assay, we characterize the 325 responses in different subject populations by age, sex and disease severity, we demonstrate 326 that antibody production persists for at least 3 months, and we suggest explanations for some 327 reports that concluded otherwise. 328 329 J o u r n a l P r e -p r o o f Limitations of current study: One caveat to our study is that in our community testing cohort 330 we may have missed individuals who were seropositive initially but then seroreverted by the 331 time of the antibody test. Second, the latest timepoint post-disease onset in our study is 226 332 days. It remains possible that antibody titers will wane substantially at later times. Additional 333 serial sampling of PCR-confirmed mild cases will be required to test these possibilities. Another The graphical abstract for this study was created on biorender.com. 

Further information and requests for resources and reagents should be directed to and will be 442 fulfilled by the Lead Contact, Deepta Bhattacharya (deeptab@arizona.edu). 443

This study did not generate new unique reagents. 445

The data generated in this study and corresponding analyses have been described in main and 447 Table 1 , as well as below in the text. Subjects were recruited in three ways. First, targeted 456 recruitment was used to recruit confirmed positive COVID-19 PCR test subjects with severe 457 COVID-19, defined as one that needed hospitalization into the Banner-University Medical 458

Center. Second, targeted recruitment was used to recruit subjects with confirmed positive 459 COVID-19 PCR test who did not require hospitalization (mild/moderate COVID-19 cases). aliquoting, serum was used for the ELISA assay with or without freezing and thawing as 469 described below. Finally, sera from 352 subjects recruited into the above two IRB protocols prior 470

to September, 2019, served as negative controls for assay development. Based on local and 471 general prevalence, it would be expected that 96-98% of these subjects have previously 472 encountered seasonal coronaviruses (Gorse et al., 2010) . Freezing and thawing had no effect 473 on levels of antibodies detected by ELISA or PRNT. 474

Lenti-X TM 293T cells (Takaro Bio USA) were grown at 37ºC, 5% CO 2 in high glucose 476 DMEM supplemented with 10% fetal bovine serum, non-essential amino acids, 477 penicillin/streptomycin, glutamine, and sodium pyruvate. The serum/plasma dilution that contained 10 or less plaques was designated as the NT90 titer. 544

Statistical analyses were performed in GraphPad Prism (v8) and Microsoft excel (v16.40). The 548 threshold for indeterminate seropositivity to RBD was calculated as 3 standard deviations above 549 the average OD value of the pre-pandemic negative control group. RBD seropositivity was 550 established with an indeterminant range from an OD value 3 standard deviations above the 551 mean OD value of the negative control cohort (OD 450 =.12) to an OD slightly above the highest 552 OD value observed in the negative control cohort (OD 450 =0.39). Readings above OD 450 =0.39 553 were considered seropositive. The seropositive threshold to S2 was determined by calculating 554 J o u r n a l P r e -p r o o f the OD value 5 standard deviations above the average OD (OD 450 =0.35) of the pre-pandemic 555 negative control cohort. 556

Correlation r values between antibody titers and neutralizing titers were determined using a 558

Pearson correlation. P values to compare non-linear regression fits of antibody and 559 neutralization titers over time grouped by disease severity, patient age, and patient sex were 560 calculated in GraphPad Prism. Null hypothesis was set for a single curve to fit all subject 561 groups, which was rejected with less than 95% confidence. LOESS soothing splines were 562 generated in GraphPad Prism. Pseudo-R 2 values were calculated using the squared correlation 563 between the predicted outcomes and the actual outcomes from the fitted model (Efron, 1978) . 

The Coronavirus

Spike Protein Is a Class I Virus Fusion Protein: Structural and Functional Characterization of the 592

Performance Characteristics of the Abbott Architect

Detection of Nucleocapsid Antibody to

SARS-CoV-2 is More Sensitive than Antibody to Spike Protein in COVID-19 Patients

The time course of the 602 immune response to experimental coronavirus infection of man

Disappearance of 605 antibodies to SARS-associated coronavirus after recovery

Information for Laboratories about Coronavirus (COVID-19)

Early Release -Antibody Responses to SARS-CoV-2 at 8 Weeks 610

A human SARS-CoV neutralizing antibody against epitope on S2 protein

Regression and ANOVA with Zero-One Data: Measures of Residual Variation

Retroviral Vectors Pseudotyped with Severe Acute Respiratory Syndrome Coronavirus 618 S Protein

Prevalence of Antibodies to

Four Human Coronaviruses Is Lower in Nasal Secretions than in Serum

Complete mapping of mutations to the

SARS-CoV-2 spike receptor-binding domain that escape antibody recognition

Immune Response to SARS-CoV-2 in Iceland

Persistence of IgG Antibodies in SARS-CoV Infected Healthcare Workers

Serology

Detection 638 of IgM and IgG antibodies in patients with coronavirus disease 2019

Rapid Decay of Anti

Antibodies in Persons with Mild Covid-19

Evidence for sustained mucosal and systemic antibody responses to

SARS-CoV-2 antigens in COVID-19 patients

Dynamics and significance of the 649 antibody response to SARS-CoV-2 infection

ZBTB32 Restricts the Duration of Memory B Cell Recall Responses

murine cytomegalovirus infections, but not other repetitive challenges

Furin Cleavage Site Is Key to SARS

Neutralizing Antibody Production in Asymptomatic and Mild

Comparison with Pneumonic COVID-19 Patients

Findings from investigation and analysis of re-665 positive cases

Identification and characterization of the 668 constituent human serum antibodies elicited by vaccination

Structure, Function, and Evolution of Coronavirus Spike Proteins

Clinical features of COVID-19 in elderly patients: 672 A comparison with young and middle-aged patients

Clinical and immunological assessment of asymptomatic SARS-CoV-2 675 infections

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

Lifetime of plasma cells in the bone marrow

ENE-COVID): a nationwide, population-based seroepidemiological 684 study. The Lancet

The receptor-binding domain of the 687 viral spike protein is an immunodominant and highly specific target of antibodies in SARS-CoV-688 2 patients

Memory 690 B cells, but not long-lived plasma cells, possess antigen specificities for viral escape mutants

IgM Antibodies Against Severe Acute Respiratory Syndrome 694

Clinical Infectious Diseases

The behaviour of recent isolates of human respiratory coronavirus in vitro 696 and in volunteers: Evidence of heterogeneity among 229E-related strains

Cumulative incidence 700 and diagnosis of SARS-CoV-2 infection in New York

Measuring SARS-CoV-2 704 neutralizing antibody activity using pseudotyped and chimeric viruses

Robust T cell immunity in 707 convalescent individuals with asymptomatic or mild COVID-19

Longitudinal evaluation and decline of 710 antibody responses in SARS-CoV-2 infection

Humoral Immunity Due to Long-712

The extent of affinity 714 maturation differs between the memory and antibody-forming cell compartments in the primary 715 immune response

Cross-reactive serum and memory B cell responses to spike 718 protein in SARS-CoV-2 and endemic coronavirus infection

Structural Genomics of SARS-CoV-2 Indicates Evolutionary Conserved 721 Functional Regions of Viral Proteins

Seroconversion of a city: Longitudinal monitoring of

SARS-CoV-2 seroprevalence in New York City. MedRxiv 2020.06.28

Hospital-Wide SARS-CoV-2 Antibody Screening in 3056 Staff in a Tertiary Center

IgA dominates the early neutralizing antibody response to 730 SARS-CoV-2

SEROCoV-POP): a population-based study. The Lancet

Intrinsic Constraint on Plasmablast Growth and Extrinsic Limits of Plasma Cell Survival

A SARS-CoV-2 surrogate virus neutralization test based on 740 antibody-mediated blockage of ACE2-spike protein-protein interaction

COVID-19 re-infection by a phylogenetically distinct

SARS-coronavirus-2 strain confirmed by whole genome sequencing

Repeated In Vivo Stimulation of T and B Cell Responses in Old Mice Generates Protective 748 Immunity against Lethal West Nile Virus Encephalitis

SARS-CoV-2 infection induces robust, neutralizing 752 antibody responses that are stable for at least three months

Divergent transcriptional programming of class-specific B cell memory by

Escape from neutralizing 758 antibodies by SARS-CoV-2 spike protein variants

The emergence of SARS-CoV-2 in Europe

Cryo-EM structure of the 2019-nCoV spike in the prefusion 764 conformation

Viral RNA level, serum antibody responses, and transmission risk in 767 discharged COVID-19 patients with recurrent positive SARS-CoV-2 RNA test results: a 768 population-based observational cohort study

Protective 'immunity' by pre-existent neutralizing 770 antibody titers and preactivated T cells but not by so-called 'immunological memory

CD80 and PD-L2 define functionally 774 distinct memory B cell subsets that are independent of antibody isotype