Neutralizing antibodies correlate with protection from SARS-CoV-2 in humans during a fishery vessel outbreak with high attack rate

The development of vaccines against SARS-CoV-2 would be greatly facilitated by the identification of immunological correlates of protection in humans. However, to date, studies on protective immunity have only been performed in animal models and correlates of protection have not been established in humans. Here, we describe an outbreak of SARS-CoV-2 on a fishing vessel associated with a high attack rate. Predeparture serological and viral RT-PCR testing along with repeat testing after return to shore was available for 120 of the 122 persons on board over a median follow-up of 32.5 days (range 18.8 to 50.5 days). A total of 104 individuals had an RT-PCR positive viral test with Ct <35 or seroconverted during the follow-up period, yielding an attack rate on board of 85.2% (104/122 individuals). Metagenomic sequencing of 39 viral genomes suggested the outbreak originated largely from a single viral clade. Only three crewmembers tested seropositive prior to the boat’s departure in initial serological screening and also had neutralizing and spike-reactive antibodies in follow-up assays. None of these crewmembers with neutralizing antibody titers showed evidence of bona fide viral infection or experienced any symptoms during the viral outbreak. Therefore, the presence of neutralizing antibodies from prior infection was significantly associated with protection against re-infection (Fisher’s exact test, p=0.002).


Abstract 23
The development of vaccines against SARS-CoV-2 would be greatly facilitated by the 24 identification of immunological correlates of protection in humans. However, to date, 25 studies on protective immunity have only been performed in animal models and 26 correlates of protection have not been established in humans. Here, we describe an 27 outbreak of SARS-CoV-2 on a fishing vessel associated with a high attack rate. 28 Predeparture serological and viral RT-PCR testing along with repeat testing after return 29 to shore was available for 120 of the 122 persons on board over a median follow-up of 30 32.5 days (range 18.8 to 50.5 days). A total of 104 individuals had an RT-PCR positive 31 viral test with Ct <35 or seroconverted during the follow-up period, yielding an attack 32 rate on board of 85.2% (104/122 individuals). Metagenomic sequencing of 39 viral 33 genomes suggested the outbreak originated largely from a single viral clade. Only three 34 crewmembers tested seropositive prior to the boat's departure in initial serological 35 screening and also had neutralizing and spike-reactive antibodies in follow-up assays. 36 None of these crewmembers with neutralizing antibody titers showed evidence of bona 37 fide viral infection or experienced any symptoms during the viral outbreak. Therefore, 38 the presence of neutralizing antibodies from prior infection was significantly associated 39 with protection against re-infection (Fisher's exact test, p=0.002). 40 41 . CC-BY 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)

Introduction 42
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused 43 tens of millions of infections and hundreds of thousands of deaths worldwide since its 44 emergence in December 2019. Multiple vaccine candidates are currently in Phase III 45 trials (1)(2)(3). The success of these vaccines could be helped by further insights into the 46 protective nature of neutralizing antibodies in humans. 47

Neutralizing antibodies have been isolated from individuals previously infected 48
with SARS-CoV-2 (4, 5). These antibodies often target the receptor binding domain 49 (RBD) of the SARS-CoV-2 spike (S) protein and prevent the binding interaction between 50 the spike protein and the host's angiotensin-converting enzyme 2 (ACE2) (4, 5), 51 although neutralizing antibodies that do not inhibit spike's binding to ACE2 have also 52 been identified (6, 7). In animal models, neutralizing antibodies are protective against 53 9). 54

SARS
Vaccines currently in development against SARS-CoV-2 have been shown to 55 elicit levels of neutralizing antibodies comparable to those observed in naturally infected 56 persons (1)(2)(3). However, the protective nature of both vaccine-and infection-elicited 57 neutralizing antibodies in humans remains unproven, with animal models being used to 58 make inferences about protection (10, 11). Human challenge trials, which could provide 59 rapid information about the protection conferred by neutralizing antibodies (12, 13), are 60 controversial due to the severity and unknown long-term impacts of SARS-CoV-2 61 infection and concerns over ethical administration of such trials (14,15). 62 Given the high number of people exposed to SARS-CoV-2 every day, 63 retrospective analyses of outbreak events may provide insights into the protective 64 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101 the manufacturer's recommendations with 10µL serum diluted into 90µL dilution buffer 156 and read using the DS2 microplate reader (Dynex technologies). 157 Neutralization assays with spike-pseudotyped lentiviral particles were performed 158 as described previously (33), with a few modifications. Briefly, cells were seeded in 159 black-walled, clear bottom, poly-L-lysine coated 96-well plates (Greiner, 655936). About 160 14 hours later, serum samples were diluted in D10 media (DMEM with 10% heat-161 inactivated FBS, 2 mM l-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin) 162 starting with a 1:20 dilution followed by 6 serial 3-fold dilutions. An equal volume of full-163 length spike-pseudotyped lentiviral particles as diluted serum was added to the serum 164 dilutions and incubated at 37C for 1 hour. 100µL of the virus plus serum dilutions were 165 then added to the cells ~16 hours after the cells were seeded. 166 About 52 hours post-infection, luciferase activity was measured as described 167 previously (33) except luciferase activity was read out directly in the assay plates 168 without transferring to black, opaque bottom plates. Two "no serum" wells were included 169 in each row of the neutralization plate and fraction infectivity was calculated by dividing 170 the luciferase readings from the wells with serum by the average of the "no serum" wells 171 in the same row. After calculating the fraction infectivity, we used the neutcurve 172 Python package (https://jbloomlab.github.io/neutcurve/) to calculate the serum dilution 173 that inhibited infection by 50% (IC50) by fitting a Hill curve with the bottom fixed at 0 and 174 the top fixed at 1. All serum samples were measured in duplicate. To calibrate our 175 neutralization assays, we also ran them on the NIBSC reference serum sample (product 176 number 20/130) and measured an IC50 of 1:2395. 177 178 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101 Results 179

Predeparture PCR and serology testing 180
There were a total of 122 people (113 men and 9 women) on the manifest of the ship. 181 Prior to the ship's departure, crewmembers were screened for active SARS-CoV-2 182 infection by RT-PCR, or for serological evidence of prior or ongoing infection using the 183 Abbott Architect assay which detects antibodies against the viral nucleoprotein (N). 184 Predeparture RT-PCR and serology test data were available for 120 crewmembers.

191
After becoming aware of the subsequent SARS-CoV-2 outbreak on the ship (see 192 next section), we tested residual predeparture serum samples from the six individuals 193 who were seropositive in the Abbott Architect assay to characterize the neutralizing and 194 spike-binding activity of their sera. The sera of three of these six individuals had potent 195 neutralizing activity against SARS-CoV-2 spike pseudotyped lentiviral particles (Table 1, 196 Figure 1B). The neutralizing titers (1:174, 1:161, 1:3082) are in the typical range of titers 197 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint observed in humans who have been infected with SARS-CoV-2 within the previous few 198 months (29,34,35). The sera of the three individuals with neutralizing titers also had 199 high activity in an assay that measure the ability of antibodies to block RBD binding to 200 ACE2, as well as in IgG ELISAs against spike and RBD (Table 1, Figure 1C

212
Notably, the sera of the other three individuals who were seropositive in the 213 Abbott Architect assay but did not have neutralizing activity had lower quantitative 214 readings in the Abbott assay (including two that were close to the cutoff of 1.40; Figure  215 1A) and readings comparable to those from negative controls in the RBD and spike 216 ELISA assays ( Figure 1C). Therefore, we speculate that the three individuals without 217 neutralizing activity were false positives in the initial serological screening. However, 218 they could have been in the early stages of active infection, since the Abbott Architect 219 detects antibodies against N while all the other assays we used detect antibodies 220 against spike, and anti-N antibodies appear earlier after infection than anti-spike 221

A B C
. CC-BY 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.

224
Overall, assuming that only individuals who were positive in the initial Abbott 225 Architect assay have neutralizing anti-spike antibodies, then just three of the 120 226 individuals with pre-departure screening data had neutralizing antibodies prior to 227 boarding the ship. We consider this assumption to be well supported by several lines of is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Serological responses among these individuals as measured by Abbott SARS-CoV-2 246
IgG index value increased for the majority of these individuals (Figure 2A).  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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint serology after returning to shore, though two of the three crewmembers tested negative 269 3 and 4 times, respectively, by RT-PCR over three weeks after returning. contrast, among the other 117 of 120 individuals with pre-departure serological data 289 who were seronegative or lacked spike-reactive antibodies prior to departure, 103 of 290 117 were infected using the same case definition (of the 2 individuals without pre-291 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101 departure serological screening, one tested positive and one tested negative by RT-292 PCR on return). Therefore, the overall rate of infection was 0 of 3 among individuals 293 with neutralizing antibodies, and 103 of 117 among individuals without such antibodies. 294 This difference is statistically significant ( The three crewmembers who were seropositive for anti-N antibodies by Abbott but did 301 not have neutralizing antibodies were all infected during follow-up, with minimum Cts of 302 17.6, 22.8, and 22.9 and increases in Abbott index values (Table 1). Sex did not differ 303 between uninfected and infected, with females composing 5.6% (1 of 18) and 7.7% (8 of 304 104) of these two groups, respectively (Fisher's exact test, p=1). 305 We also looked in detail at the viral testing results of the three crewmembers who 306 were positive for neutralizing antibodies to assess the strength of the evidence that they 307 were not re-infected during this ship outbreak. Two tested fully negative by RT-PCR on 308 3+ occasions, with negative tests on Days 18,25,35,and 36 and Days 18,35,and 36. 309 The third individual tested negative on the Roche cobas on Day 21 and Day 28, and 310 positive only by the E-gene primers/probe set (Ct 37.4) and negative by the orf1ab 311 . CC-BY 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 August 14, 2020. Here, we report an outbreak of SARS-CoV-2 on a fishing vessel with an attack 324 rate greater than 85%. Screening with the Abbott Architect anti-nucleocapsid IgG 325 antibody test followed by confirmation of positives with multiple anti-spike protein 326 antibody tests including neutralization assays demonstrated the protective nature of 327 neutralizing antibodies. In particular, none of the three individuals with pre-existing 328 neutralizing antibodies were infected, whereas the vast majority of other individuals 329 were infected. These findings are consistent with data from animal models, in which the 330 elicitation of high titers of neutralizing antibodies was protective against re-challenge 331 with SARS- 10,41). 332 An assumption of our analysis is that the only individuals who had pre-existing 333 neutralizing and anti-spike antibodies were those who tested seropositive in the initial 334 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101 pre-departure Abbot Architect anti-N serological screening, since only individuals 335 positive in that screening were subjected to additional serological assays for anti-spike 336 and neutralizing antibodies. However, this assumption is well supported by the validated 337 high sensitivity of the Abbott Architect assay (20), plus the well-established fact that 338 anti-N antibodies appear earlier that anti-spike antibodies (36,37). Additionally, our four 339 anti-spike antibody tests showed a high level of consistency among seropositive 340 samples, and prior work using the exact same assays has found neutralizing antibodies 341 only among individuals who were positive in the Abbott Architect assay (32). As shown 342 by others, the RBD ELISA and neutralizing antibody assays were highly consistent (42, 343 43). The ACE2 blockade of binding functional ELISA assay showed excellent 344 consistency with the more laborious pseudovirus neutralizing antibody assay (44). 345 It is intriguing that one individual who had predeparture neutralizing antibodies 346 and was classified as uninfected by our case definition nonetheless had a sporadic very 347 weak signal in viral testing on two different RT-PCR platforms. It is well-established that 348 SARS-CoV-2 can be detected for multiple weeks in the nasopharyngeal tract, well after 349 the resolution of symptoms and elicitation of an antiviral immune response (45,46). 350 However, it is unclear at this time whether immunity to SARS-CoV-2 will be sterilizing 351 (10, 47), and it is possible that the sporadic weak signal in viral testing for this individual 352 was the result of re-exposure to virus on the boat. 353 In prior studies, the Abbott SARS-CoV-2 IgG assay has shown excellent 354 performance characteristics with high specificity (99.1-99.9%) for prior infection with 355 48,49). Curiously, the positive predictive value for the Abbott SARS-356

SARS
CoV-2 IgG assay for neutralizing antibodies or protection in our population was only 357 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint 50% (3/6 crewmembers). It is difficult to conclusively determine whether these 358 represented false positives or just anti-N/anti-spike discrepants, particularly given that 359 anti-N antibodies tend to appear before anti-spike antibodies (36, 37). All three of the 360 individuals who were Abbott IgG positive prior to departure but lacked neutralizing and 361 anti-spike antibodies and were RT-PCR positive upon return showed strong increases 362 in index value. In addition, two of these three individuals had pre-departure Abbott index 363 values that were close to the positivity cut-off. Unfortunately, we did not have sufficient 364 residual pre-departure serum to run on a separate anti-N platform such as the Roche 365 Elecsys anti-SARS-CoV-2 (50). 366 This study is limited by lack of information on clinical symptoms for the majority of 367 crewmembers on the vessel and direct knowledge of contacts on the boat. We cannot 368 also necessarily know that the three individuals with neutralizing antibodies prior to 369 departure were exposed directly to SARS-CoV-2 on the vessel. The study is also limited 370 by the low seroprevalence in the predeparture cohort---which is consistent with the 371 approximate seroprevalence in May 2020 in the Seattle area, but means that there were 372 only three individuals with pre-existing neutralizing antibodies. Nonetheless, with an 373 overall attack rate of >85%, the lack of infection in the three individuals with neutralizing 374 antibodies was statistically significant compared to the rest of the boat's crew. Overall, 375 our results provide the first direct evidence anti-SARS-CoV-2 neutralizing antibodies are 376 protective against SARS-CoV-2 infection in humans. 377 378 Acknowledgements 379 . CC-BY 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 August 14, 2020. . https://doi.org/10.1101 The authors thank Nicole Lieberman for helpful comments and Nathan Breit for data 380 pulls. We also thank Brooke Fiala, Samuel Wrenn, Deleah Pettie, and Neil P. King at 381 the Institute for Protein Design for sharing reagents for ELISA assays. Work performed 382 in the clinical laboratory was supported by the Department of Laboratory Medicine and 383 Pathology. This research was supported by the following grants from the NIAID of the CC-BY 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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint . CC-BY 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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint . CC-BY 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 August 14, 2020. . https://doi.org/10.1101/2020.08.13.20173161 doi: medRxiv preprint