Neutralising reactivity against SARS-CoV-2 B.1.617.2 (Delta) variant by vaccination status and pre-exposure


 In February and March 2020, one of the first Italian clusters of SARS-CoV-2 infection was detected in the municipality of Vo’. Positive subjects were followed up at 2 and 9 months post-infection with different immuno-assays and a micro-neutralisation test. Here we report on the results of the third serosurvey conducted in the same population in June 2021, 15 months post-infection, when we tested 61% of the infected individuals (n=76). Antibodies against the spike (S) antigen significantly decreased (P<0.006, Kruskal-Wallis test) among unvaccinated subjects (n=35) and increased (P<0.0001) in vaccinated individuals (n=41), whereas those against the nucleocapsid (N) decreased in the whole cohort. From the comparison with two control groups (naïve Vo’ inhabitants (n=20) and healthcare workers (HCW, n=61)), subjects vaccinated post exposure (hybrid immunity) had higher antibody levels (P<0.0001) than subjects vaccinated when naïve. Two doses of vaccine elicited stronger anti-S antibody response than natural infection (P<0.0001). Finally, the neutralising reactivity of sera against the B.1.617.2 (Delta) was lower than compared to the B.1 strain (median 1:320 versus 1:1280 1/dil, P<0.0001, and 1:640 versus 1:2560 1/dil, P=0.0014, after one or two vaccine doses, respectively), although subjects with hybrid immunity maintained neutralising titres above 1:40 1/dil.


48
Introduction 49 50 Understanding the extent and duration of protection developed upon natural SARS-CoV-2 infections and 51 vaccination is a current research priority. Evidence suggests that more than 90% of COVID-19 patients 52 seroconvert after natural infection and develop variable levels of neutralising antibodies 1-3 , and 53 demonstrates that the currently EMA and FDA approved vaccines induce humoral and cellular immunity in 54 most individuals 4-7 . However, antibody titres have been reported to wane over time 8 . Although memory B 55 cells and cellular immunity can offer a quick and potent response in case of re-exposure to the virus 9 , 56 preventing re-infections 10-15 and offering long-term protection regardless of the presence of antibody-57 escaping mutations [16][17][18] , the interplay between antibody and cellular immunity, and the variation of 58 naturally-and vaccine-acquired protection, remain to be fully characterised and understood. From an 59 immunological perspective, there can be significant differences in the immune response generated by 60 vaccines in individuals who were not exposed to SARS-CoV-2 before vaccination, and in subjects who 61 recovered from a naturally acquired infection (so called 'hybrid immunity'). Recent studies have reported of 62 increased potency of 'hybrid immunity', with viral antigen persistence in some tissues being hypothesised 63 as a potential mechanism driving the process of memory B and T cell maturation, resulting in an increased 64 affinity against viral antigens. escaping the immune protection elicited from infection and vaccination, thus posing further challenges for 68 epidemic control. The most widespread variants currently circulating are called Delta plus strains 19,20 and 69 belong to a group of sub-lineages of the B.1.617.2 variant of concern (VOC), which emerged in India in 70 October 2020 21  February 2020. This study follows on from the previous serosurveys conducted in the same population at 79 two and nine months after the initial SARS-CoV-2 outbreak 8,28 , and provides unique longitudinal data on the 80 magnitude, neutralizing ability, and persistence of the antibody response against the spike (S) and 81 nucleocapsid (N) antigens in unvaccinated pre-exposed subjects as well as vaccinated pre-exposed and 82 naïve subjects, against both a B.1 SARS-CoV-2 strain circulating at the start of the pandemic and the In June 2021, 76 subjects infected by SARS-CoV-2 in February/March 2020 (as defined by the ground truth 90 definition, see Methods) were tested with the same methods applied in the previous surveys (Methods) 91 ( Fig. 1). Overall, all 76 (100%, 95% Confidence Interval (CI) 95.3-100%) individuals tested positive to at least 92 one assay, with 9 (11.8%, 95% CI 5.6-21.3%) being positive to all three of them. As observed in our previous 93 surveys, in June 2021 we observed strong differences in the proportion of positive subjects depending on 94 the assay used, with 11.8% (9 out of 76, 95% CI 5. 6 CoV-2 according to the ground truth definition had received at least one dose of vaccine at least seven days 113 before testing. As expected, vaccination had a strong impact on S-targeting antibody levels ( Fig. 2) but not 114 on those directed against the N antigen (Fig. 3). All vaccinated subjects showed reactivity against the S 115 antigen, had a neutralising titre greater than 1:40 (41 out of 41 for both DiaSorin and neutralisation, 95% CI 116 91.4-100%) (Fig. 2b, 2d), and they still showed reactivity against the N antigen either when using Abbott 117 (22.0%, 9 out of 41, 95% CI 10.6-37.7%) or Roche (92.7%, 38 out of 41, 95% CI 80.1-98.4%) assays 118 (Supplementary Fig. 1b and 1d). In the unvaccinated group the serum reactivity against the S antigen was 119 significantly lower compared to the vaccinated subjects, with positivity rates of 57.1% (20 out of 35, 95% CI 120 39.3-73.7%) and 5.7% (2 out of 35, 95% CI 0.7-19.2%) for DiaSorin and microneutralisation assays, 121 respectively ( Fig. 2a- antibodies directed against the S antigen decreased significantly, as measured by DiaSorin and 173 neutralisation (Wilcoxon matched-pairs signed rank test p < 0.0001 for both cases) (Fig. 2). 174 The serum reactivity against the N antigen progressively decreased with time irrespectively of the utilised 175 assay in the whole cohort (Fig. 3, Wilcoxon matched-pairs signed rank test p < 0.001 for both Abbott and 176 Roche assays) and among vaccinated and unvaccinated individuals separately (Wilcoxon matched-pairs 177 signed rank tests p < 0.001) ( Supplementary Fig. 1). Nonetheless, we observed a significant difference 178 between anti-N antibody titres detected in June 2021 between vaccinated and unvaccinated individuals 179 (Mann Whitney test P = 0.0003 and P = 0.0005 for Roche and Abbott, respectively) ( Supplementary Fig. 2).

Correlation between two DiaSorin assays and neutralisation 186
We assessed in parallel the performance of two DiaSorin tests, the first version targeting antibodies against 187 the S1/S2 antigen and the updated version containing a full trimeric spike antigen. The two assays showed 188 a strong correlation (Spearman's r = 0.820, 95% CI 0.670-0.906) ( Supplementary Fig. 3) and concordance 189 (Supplementary Table 1). We estimated a conversion factor between the two assays of 3.190 (95% CI We investigated the impact of past SARS-CoV-2 exposure to the humoral immune response induced by subjects from the Vo' cohort (n = 20) vaccinated when naïve to vaccinated individuals post exposure (n = 200 41) we observed significantly higher titres in previously exposed individuals (Mann Whitney test, P < 201 0.0001). Two vaccine doses reduced the observed difference in antibody titres between subjects vaccinated 202 when naïve and subjects vaccinated post exposure (Fig. 4a, Kruskal-Wallis test, P = 0.01). Neutralisation and 203 anti-S titres observed in pre-exposed vaccinated subjects after one and two vaccine doses were statistically 204 comparable (Kruskal-Wallis test, P = 1 for both DiaSorin and neutralisation). Similar trends were observed 205 when comparing the group of vaccinated subjects previously exposed to SARS-CoV-2 with an independent 206 cohort of healthcare workers (HCW, n = 61) vaccinated when naïve from the complex operational unit 207 (U.O.C.) of Microbiology and Virology of Padua University Hospital (Kruskal-Wallis test, P < 0.0001 for both 208 DiaSorin and neutralisation) (Fig. 4b, 4d). 209 210 Fig. 4. Antibody levels in vaccinated naïve and vaccinated pre-exposed individuals according to DiaSorin 211 and micro-neutralisation assays. a-b) Observed antibody levels measured by DiaSorin assays in vaccinated 212 naïve and pre-exposed individuals with at least one dose of vaccine (Mann Whitney test, P < 0.0001) and 213 with one or two doses of vaccine (Kruskal-Wallis test, vaccinated naïve versus pre-exposed subjects after 214 one vaccine dose, P < 0.0001; after two vaccine doses, P = 0.01; vaccinated naïve HCW versus pre-exposed 215 subjects after two vaccine doses, P < 0.0001). c-d) Observed neutralising antibody titres measured by a 216 micro-neutralisation assay in vaccinated naïve and pre-exposed individuals with at least one dose of vaccine versus pre-exposed subjects after one or two vaccine doses, P < 0.0001; vaccinated naïve HCW versus pre-219 exposed subjects after two vaccine doses, P < 0.0001). Asterisks indicate *p < 0.05, **p < 0.01, 220 ***p < 0.001, ****p < 0.0001. GT: ground truth, infected Vo' population; HCW: healthcare workers. 221 222 Two vaccine doses in naïve individuals trigger higher anti-S antibodies and neutralising titres than natural 223 infection 224 Using the conversion factor calculated to convert the results of the old DiaSorin S1/S2 assay into the new 225 DiaSorin trimericS assay, we compared the antibody response after natural infection with the response to 226 vaccination, roughly two months after the immune stimulus. Vo' subjects exposed to SARS-CoV-2 in 227 February/March 2020 and tested in May 2020 showed lower anti-S antibody levels with respect to both 228 naïve subjects from Vo' (Kruskal-Wallis test, P < 0.0001) and HCW (Mann-Whitney test, P < 0.0001) after 229 two doses of vaccine ( Fig. 5a and 5b). A similar trend was observed for neutralising antibody titres, 230 although the difference is significant only between exposed subjects and vaccinated HCW (Mann-Whitney 231 test, P = 0.0002) ( Fig. 5c and 5d). 232 233

SARS-CoV-2 variants in individuals vaccinated 267
naïve or pre-exposed and unvaccinated pre-exposed subjects. a-b) Neutralising antibody titres against B.1 268 and B.1.617.2 SARS-CoV-2 variants among pre-exposed a) vaccinated and b) unvaccinated individuals 269 (Kruskal-Wallis test, pre-exposed subjects after one (P < 0.0001) or two (P = 0.0014) doses of vaccine; Mann 270 Whitney test, pre-exposed unvaccinated P = 0.00002). c) Neutralising antibody titres against B.1 and Delta 271 SARS-CoV-2 variants among vaccinated naive individuals (Kruskal-Wallis test, vaccinated naïve after one (P 272 = 1) or two (P = 0.15) doses of vaccine). Asterisks indicate *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 273 0.0001. GT: ground truth, infected Vo' population. The Vo' cohort is a highly characterised population including a core of individuals identified as exposed to 282 SARS-CoV-2 back in February/March 2020, which has been followed-up through time in sequential swab 283 and serological surveys until June 2021, roughly 15 months after viral exposure, thus offering unique 284 insights into the long term antibody dynamics. The results presented in this study confirm the trends 285 observed in our previous follow-up, performed at 9 months since the first wave in Vo' 8 , with strong 286 variability observed among serological tests, especially for the two assays targeting the N viral antigen. 287 Of the identified SARS-CoV-2 cases who acquired the infection in February/March 2020, only 11.8% (95% CI 288 5.6-21.3%) tested positive by Abbott while 93.4% (95% CI 85.3-97.8%) tested positive by Roche after 15 289 months. This discrepancy could be due to differences in the employed antigens and to the fact that the N 290 epitopes recognised by antibodies might change with time. In perspective, given that these two tests could allow to discern recent from past infections, they could be employed in future seroprevalence studies to 292 assess the attack rate in vaccinated subjects and thus provide new data on the frequency of breakthrough 293 infections as well as re-infection. 294 We found that all individuals infected at the start of the pandemic and tested 15 months later are positive 295 to at least one serological assay, although the decreasing trend of antibody levels against both S and N 296 antigens is confirmed, independently from the type of test used. In the absence of vaccination, the 297 neutralising titres of the infected subjects drop almost completely below the 1:80 (1/dil) threshold. 298 While the observed decrease in antibody titres is in line with other recent reports, it does not necessarily 299 translate into an impaired immunity in these subjects, since humoral response is one arm of the adaptive 300 immune response, which also includes cellular immunity and reactivation upon stimulation of memory B 301 and T cells 14,30,31 . 302 Unexpectedly, we found a significant difference in the amount of circulating N targeting antibodies 303 between vaccinated and unvaccinated subjects pre-exposed to SARS-CoV-2. To investigate this pattern we 304 retrospectively analysed the differences in N targeting antibodies present in the two groups in the 305 November serosurvey, before the beginning of the vaccination campaign. We found that the two groups 306 are significantly different in terms of age, with vaccinated subjects being older than unvaccinated subjects 307 (Suppl. fig. 2e). The age difference can be explained by the vaccination strategy and agrees with our 308 previous observation that antibody levels were higher with increasing age in this cohort 8 . 309 We found that the response to vaccination is different among subjects vaccinated after pre-exposure and 310 when naïve: while a marked increase in S-targeting antibodies is observed in all individuals, antibodies 311 induced by vaccination are higher in pre-exposed subjects. In vaccinated pre-exposed subjects, a single 312 dose of vaccine saturates the dynamic range of DiaSorin assay and is shown to boost a strong neutralisation 313 response, as confirmed in other studies 32,33 . This suggests that a single dose of vaccine in pre-exposed 314 patients induces a robust immune response in support of the vaccination strategy implemented in 315 Germany, France, Italy, and Israel among other countries. It has been shown that B cell maturation due to 316 somatic hypermutation, possibly stimulated by long-term persistence of viral antigens in specific body 317 niches 9,34-36 , can produce stronger and more specific antibodies 37 . 318 By comparing the antibody levels in vaccinated naïve subjects (in June 2021) with those of patients who 319 recovered from natural infection in May 2020, we demonstrate that a complete vaccination course confers 320 stronger immunity than natural infection alone, at least in terms of serum antibodies as detected by both 321 DiaSorin and neutralisation. 322 We tested the ability of antibodies developed against SARS-CoV-2 strains circulating early in the pandemic 323 to neutralise the Delta variant of concern (VOC B.1.617.2), which is characterised by several mutations in 324 the spike protein and an increased transmissibility that allowed this variant to become prevalent exposed and unvaccinated, vaccinated pre-exposed and vaccinated naïve). The reduction in neutralising 327 reactivity is more evident in the vaccinated pre-exposed subjects (Fig. 6a) Tubes (SST) and centrifuged for 10 min at 1000-1300 RCF (g). Serological tests were performed by trained 356 laboratory staff using the same commercial kits employed in previous serosurveys 8 and produced by 357 Abbott 41 , DiaSorin 42 , and Roche 43 , applying the detection thresholds provided by the manufacturer (Table  358 1). For DiaSorin, both the new TrimericS and the previous S1/S2 kits were used for comparison. 359 Two independent assays were set up in parallel to assess the neutralisation ability of patients' seric 362 antibodies against two viral isolates, a third passage B.1 strain isolated in March 2020 (GenBank accession 363 MW468415) and a third passage B.1.617.2 strain from August 2021 (GenBank accession …(waiting for the 364 release of the accession number)). Heat-inactivated serum samples (30 min at 56 °C) were diluted 1:10 with 365 DMEM FBS Free medium and filtered (0.22 μm pore size). 50 μl of viral isolate, diluted in DMEM FBS Free to 366 the final concentration of 100 median tissue culture infective dose (TCID50), were mixed with an equal 367 volume of two-fold serial dilutions of sera in 96-wells microplates and incubated for 1 h at 37 °C in a 368 humidified atmosphere with 5% CO2. Following incubation, 100 µL of VERO E6 cells suspended in DMEM 369 6% FBS were added to each well and incubated at 37°C. After 72h, cytopathic effect was assessed; the 370 supernatant was removed and 120 μl of 5% formaldehyde Gram's crystal violet 40% m/v were added to 371 each well, followed by 30 min of incubation. After a washing step with water, plates were allowed to dry 372 and the absorbance was read at 595 nm. The neutralisation titre was determined as the highest serum 373 dilution showing an optical density (OD) of 90% or more with respect to the control sera. 374 375

Definition of COVID-19 recovered patients (ground truth, GT) 376
Multiple rounds of mass testing, that included oropharyngeal swabs and serological assays, allowed for the 377 identification of all the residents in the municipality of Vo' who were infected and recovered from SARS-378 CoV-2 infection during the first wave, between February and March 2020. To be included among COVID-19 379 recovered individuals, one of the following criteria had to be satisfied: i) a positive swab, ii) a viral 380 neutralization titre greater than 1:40, or iii) serum reactivity against two serological tests with different 381 antigen targets. We refer to this group as baseline ground truth (GT). It included 125 subjects, a size that 382 perfectly fitted the seroprevalence estimated through a multinomial likelihood model 8 . These subjects were 383 followed up at several time points to monitor the presence and persistence of antibodies against both the 384 spike (S) and the nucleocapsid (N) antigens (Figure 1), as well as to investigate the presence of virus 385 neutralising antibodies (Tables 1 and 2). We previously reported that all subjects belonging to the GT were 386 positive to at least one serological assay in May 2020, about two months after the time of their infection 387 (Fig. 1). On occasion of a second serological survey conducted in November 98.8% of GT subjects were still 388 positive nearly 9 months after the infection 8 , although with strong differences depending on the test. 389 390

Statistical methods 397
Estimates of antibody decay rate and association analysis 398 The antibody decay rate was estimated at the individual level as the logarithmic change in antibody values 399 observed between May 2020 and June 2021 (within the same subject) divided by the number of days 400 between the two serosurveys (400 days). The antibody half-life was estimated as the natural logarithm of 401 0.5 divided by the antibody decay rate, and was calculated on all subjects testing positive in May 2020 402 serosurvey and without doubling antibody levels in November 2020 and June 2021 (Abbott n=65, DiaSorin The associations between antibody levels and symptom occurrence, hospitalisation, sex, age-group and 405 BMI and between vaccination and pre-exposure, symptom occurrence and sex were assessed using the 406 Kruskal-Wallis test. We used Fisher's exact test to assess the association between vaccination and previous 407 hospitalisation. 408 409

Ethical approval statement 412
The third serosurvey of the Vo' population was approved by the Ethics Committee for Clinical Research of 413 the province of Padova. Study participation was by consent. For participants under 18 years of age, consent 414 was provided by a parent or legal guardian. 415