AstraZeneca COVID-19 vaccine induces robust broadly cross-reactive antibody responses in Malawian adults previously infected with SARS-CoV-2

Using a prospective study design, recovered mild/moderate COVID-19 adult patients were recruited during the first two epidemic waves in Malawi (Blantyre City, southern region) and followed up every 30 days for a maximum of 270 days. We used a convenience sampling approach, whereby the study was advertised electronically and by word of mouth, with support from the Blantyre District Health Office and Malawi-Liverpool-Wellcome programme. Inclusion criteria for the study included being a Blantyre resident, aged between 18 and 65 years old, and having previous history of laboratory-confirmed diagnosis of COVID-19 not less than 28 days at the time of recruitment. The exclusion criteria included withholding consent and having symptoms suggestive of COVID-19 at time of recruitment. Peripheral blood samples were collected at recruitment and subsequent follow-ups. We used electronic case report forms (eCRFs) to collect clinical history and demographic data. The first wave in Malawi peaked in July 2020 and the second wave peaked in January 2021, driven by the original variant and Beta variant, respectively [19, 20].

The SARS-CoV-2 original (D614G) spike and RBD proteins were expressed in human embryonic kidney (HEK) 293F suspension cells by transfecting the cells with the spike plasmid. After incubating for 6 days at 37 °C, 70% humidity and 10% CO₂, proteins were purified using a nickel resin followed by size-exclusion chromatography. Relevant fractions were collected and flash-frozen until use. Spike or RBD protein (2 μg/ml) was used to coat 96-well, high-binding plates and incubated overnight at 4 °C. The plates were incubated in a blocking buffer consisting of 5% skimmed milk powder, 0.05% Tween 20, 1x PBS. Plasma samples were diluted to a 1:100 starting dilution in blocking buffer and added to the plates. Secondary antibody was diluted to 1:3000 in blocking buffer and added to the plates followed by TMB substrate (Thermofisher Scientific). Upon stopping the reaction with 1 M H₂SO₄, absorbance was measured at a 450-nm wavelength. In all instances, the CR3022 mAb was used as a positive control and palivizumab was used as a negative control.

The assay was performed as previously reported [21]. In brief, the expression plasmid encoding for SARS-CoV-2 RBD and full-length Spike were obtained from the Florian Krammer, Mount Sinai, USA. The recombinant trimeric Spike and RBD proteins were expressed as described previously [22] and were coupled to the magnetic microsphere beads (Bio-Rad, USA) using a two-step carbodiimide reaction [23]. An in-house references serum was developed by pooling convalescent serum from adult COVID-19 positive patients. This interim reference serum was calibrated against research reagent NIBSC 20/130 distributed by the National Institute for Standards and Biological Control (NIBSC) for the purpose of development and evaluation of serological assays for the detection of antibodies against SARS-CoV-2 (NIBSC, Potters Bar, UK). The binding antibody units (BAU) values assigned to in-house reference serum were 1242 BAU/mL and 2819 BAU/mL for RBD and full-length Spike IgG, respectively. Serum samples collected before 2020 were used for the analysis of assay specificity. Values of 26 BAU/mL and 32 BAU/mL were selected as the threshold indicative of SARS-CoV-2 antibodies, based on the highest value of RBD and full-length Spike IgG in samples from pre-COVID-19. Sensitivity of the assay in detecting past or current infection was assessed using serum samples obtained from randomly selected participants (n = 15) who tested SARS-CoV-2 PCR positive and who had serial sampling before and after post-symptom onset, including cases with mild-moderate illness and asymptomatic infections. The sensitivity of the IgG assay was 75% for samples taken 7–14 days and 100 % for samples taken above 14 days following the PCR positive for SARS-CoV-2 for both RBD and full-length Spike IgG. The assay was also evaluated against a COVID-19 convalescent plasma panel (NIBSC code 20/118) intended for the development and evaluation of serological assays for the detection of antibodies against SARS-CoV-2. The in-house characterisation of the plasma panel was in line with recommended criteria with 20/120 having the highest anti-RBD antibody titre followed by 20/122, which had a mid-antibody titre, 20/124 and 20/126 had the lowest titres, and 20/128 as negative. The optimal serum/plasma and secondary antibody dilutions for the assay were 1:100 and 1:200, respectively. The over-range samples were re-tested at higher dilutions (1:200-1:1000). Samples were analysed in duplicate, and each plate included two in-house control sera. Bead fluorescence was read with the Bio-Plex 200 instrument (Bio-Rad) using Bio-Plex manager 5.0 software (Bio-Rad).

The haemagglutination test (HAT) has been developed in Prof Alain Townsend’s (AT) laboratory at the University of Oxford and uses the IH4-RBD. The IH4-RBD reagent is based on the camelid nanobody VHH-IH4, linked to the RBD of SARS-2 Spike protein. IH4 is specific for a conserved epitope on Glycophorin A, comprised of residues 52–55 (YPPE). The IH4-RBD fusion was designed by AT and has been produced by Absolute Antibody (Oxford, UK) in bulk (1g). One milligram is enough for 10,000 tests (100 ng/well). The RBD proteins were derived from the original D614G strain and the four variants of concern, namely alpha, beta, gamma, and delta. We conducted the assay as previously reported [24, 25]. In brief, O negative red blood cells obtained from the Malawi Blood Transfusion Services were diluted in phosphate-buffered saline (PBS) at 1:20 and plated in V-bottomed 96 well plates. Doubling dilution of serum samples starting from 1:20 were added to the plate until 1:640 dilution. CR3022 (a human monoclonal antibody isolated from a SARS recovered patient) and EY-6A (an antibody isolated from a COVID-19 recovered patient) reagents were used as positive controls, as they all bind to similar epitopes on RBD of all the variants, allowing cross-linking between red blood cells (RBCs) labelled with IH4-RBD. IH4-RBD was added to each well and RBCs were allowed to settle for an hour. The plate was then tilted for at least 30 seconds and photographed. The HAT titre was defined by the last well in which the teardrop fails to form. Partial teardrop was regarded as negative.

Samples that were seropositive for anti-Spike binding antibodies in the semi-quantitative ELISA were screened for neutralising activity as previously described [6, 18]. SARS-CoV-2-pseudotyped lentiviruses were prepared by co-transfecting the HEK 293T cell line with either the SARS-CoV-2 original spike (D614G) or the SARS-CoV-2 beta or delta spike plasmids in conjunction with a firefly luciferase encoding pNL4 lentivirus backbone plasmid. The parental plasmids were kindly provided by Drs. Elise Landais and Devin Sok (The International AIDS Vaccine Initiative (IAVI), USA). For the neutralisation assay, heat-inactivated seropositive serum samples were incubated with the SARS- CoV-2 pseudotyped virus for 1 h at 37 °C, 5% CO₂. Subsequently, 1 × 104 HEK 293 T cells that were engineered to over-express ACE-2, kindly provided by Dr. Michael Farzan (Scripps Research), were added and incubated at 37 °C, 5% CO₂ for 72 h upon which the luminescence of the luciferase gene was measured.

Data visualisation and statistical analyses were performed in GraphPad Prism software (version 9.1.2). The antibody binding and neutralisation activity data were log10 transformed. Ordinary or repeated measures one-way ANOVA was used to compare log10 transformed data, with p value adjusted for multiple comparisons using Šídák’s or Holm-Šídák’s multiple comparisons test. Effects were considered statistically significant when the p-value was less than 0.05.

Here, we studied a cohort of individuals that had recovered from mild/moderate COVID-19 (n = 52) (Table 1). The median time from laboratory-confirmed COVID-19 diagnosis to recruitment was 70 days (59–87) for the first wave and 63 days (39–71) for the second wave. Based on genomic surveillance data from Malawi [19], it is expected that the participants were infected by either ancestral variants in the first wave or the beta variant in the second wave. However, as genomic data were not collected as part of this study, we cannot completely rule out the possibility of infections from other variants of concern (VOCs) during the second wave, especially the alpha variant. Eight out of 52 individuals reported having a comorbidity, defined as diabetes, hypertension, or HIV. Twelve (23% [12/52]) of the study participants were vaccinated with one dose of the AstraZeneca COVID-19 vaccine during the follow-up period of the study as part of the routine vaccination programme. COVID-19 vaccination using the AstraZeneca COVID-19 vaccine was rolled out in Malawi from 11 March 2021 with 768,212 fully vaccinated as of 24 January 2022.

We tested the longevity of the neutralising activity in convalescent serum using a pseudovirus neutralising assay against the original D614G variant and the beta variant. Only 46 out of 52 individuals met the criteria for pseudovirus neutralisation assays (Fig. 1A), but one sample was not tested. Anti-Spike seropositive sera from individuals infected during the first wave were assessed for neutralisation activity against the D614G variant, while that from individuals infected in the second wave was tested against the beta variant. The neutralisation activity at 30–90 days was higher than that at 120 to 180 days (p = 0.0203) or 210 to 270 days (p = 0.0093), but there was no statistically significant difference between 120 to 180 days and 210 to 270 days (Fig. 1B). These data show significant decline in serum pseudovirus neutralisation activity within 6 months post laboratory-confirmed mild/moderate COVID-19.

We observed two laboratory-confirmed SARS-CoV-2 reinfections during the study period. The first re-infected participant was a 46-year-old male who had mild COVID-19 during the first wave and was recruited into the study at 90 days post diagnosis (Fig. 2A). His last visit before PCR-confirmed reinfection was day 240, at which he had undetectable pseudovirus neutralisation activity (D614G, ID₅₀ < 20, beta, ID₅₀ < 20) and anti-Spike and anti-RBD IgG antibody levels of 5.9 BAU/ml and 0.8 BAU/ml, respectively. At his scheduled monthly visit (day 270), a noticeable increase in anti-Spike (58.8 BAU/ml) and anti-RBD (39.5 BAU/ml) was observed confirming reinfection. His reinfection episode was mild as the primary infection. Consistent with the time of reinfection (second wave), he had higher pseudovirus neutralisation titres against the beta variant than D614G (ID₅₀ 91 vs 37). The second re-infected participant was a 49-year-old female who had moderate COVID-19 during the first wave and was recruited at 60 days post diagnosis (Fig. 2B). Her last visit before reinfection was day 180, at which she had low pseudovirus neutralisation activity (D614G, ID₅₀ 46, beta, ID₅₀ < 20) and anti-Spike and anti-RBD IgG antibody levels of 669.8 BAU/ml and 778.7 BAU/ml, respectively. At her scheduled monthly visit of day 210, a noticeable increase in anti-Spike (1453.4 BAU/ml) and anti-RBD (1844.5 BAU/ml) was also observed confirming reinfection. Her reinfection episode was moderate like the primary infection. Again, consistent with a second wave infection, she had higher pseudovirus neutralisation titres against the beta variant than the D614G variant (ID₅₀ 414 vs 219). These data demonstrate that loss or absence of variant-specific pseudovirus neutralising antibodies was associated with reinfection in these two participants.

To obtain further insights on the association between binding antibody levels and neutralisation activity, we quantified the concentrations of anti-Spike and anti-RBD IgG antibodies in the convalescent sera from the first wave using a single-plex bead-based immunoassay [21]. We grouped the data into three tertiles based on the antibody concentrations, low (≤ 25th centile), medium (> 25th < 75th centile) and high (≥ 75th centile), and then assessed the level of pseudovirus neutralisation activity against D614G in these three groups. The level of pseudovirus neutralisation activity was higher in those with concentrations of anti-Spike or anti-RBD antibodies in the medium centile (anti-Spike, p = 0.0077; anti-RBD, p = 0.0119) and high centile (anti-Spike, p < 0.0001; anti-RBD, p < 0.0001), compared to those in low centile group (Fig. 3A, B). The level of pseudovirus neutralisation activity was higher in those with levels of anti-Spike or anti-RBD antibodies in the high centile compared to medium centile (anti-Spike, p < 0.0001; anti-RBD, p < 0.0001) (Fig. 3A, B). These data show that high concentrations of anti-Spike and anti-RBD binding antibodies in convalescent serum from mild/moderate COVID-19 was indicative of potent pseudovirus neutralisation activity against the infecting variant.

We next assessed whether vaccination with a single dose of the AstraZeneca COVID-19 vaccine in individuals who previously had mild/moderate COVID-19 could impact humoral responses against SARS-CoV-2, as has been observed with other COVID-19 vaccines [9‐11, 18]. A total of 12 participants in our study were vaccinated, as part of the routine national COVID-19 vaccine roll-out, at a median of 60 days (IQR 60–75) post laboratory-confirmed diagnosis of mild/moderate COVID-19. All the individuals were recruited in the second wave and hence were likely to have been infected with the beta variant. The median duration from vaccination to post-vaccination sample collection was 25 days (IQR 21–28). Antibody responses post-vaccination were assessed at a single time point. We observed a 2-fold (p = 0.0031) and 4-fold (p < 0.0001) increase in the levels of anti-Spike and RBD IgG antibodies against the original D614G variant following vaccination, respectively (Fig. 4A). Most of the individuals (92% [11/12]) had antibody levels beyond the suggested protective level of 154 BAU/ml [26] following vaccination (Fig. 3A). In a subset of the vaccinated participants with paired samples (n = 7), we assessed whether vaccination enhanced the breadth of the antibody response, defined as the presence of binding antibodies in serum against multiple VOCs. We used a recently published SARS-CoV-2 RBD haemagglutination test (HAT) that correlates well with neutralisation activity [24, 25]. We tested reactivity against D614G, alpha, beta, gamma, and delta variants in pre-vaccination and post-vaccination serum. Reactivity to all VOCs was observed in post-vaccination serum but not in pre-vaccination serum (Fig. 4B, C). These data show improved antibody recognition of RBD variations following vaccination in previously SARS-CoV-2 infected individuals.

To ascertain enhanced neutralisation activity after vaccination, sera from individuals who were infected during the second wave and subsequently vaccinated was tested for pseudovirus neutralisation activity against both the D614G, beta, and delta variants. Pseudovirus neutralisation activity was increased against D614G (p = 0.0156), beta (p = 0.0078), and delta (p = 0.0156) variants (Fig. 4D). Moreover, even in those that had no pre-vaccination pseudovirus neutralisation activity against the D614G (pre-vac 29% vs. post-vac 100%) or beta (pre-vac 63% vs. post-vac 100%) or delta variants (pre-vac 71% vs. post-vac 100%), vaccination rescued pseudovirus neutralisation activity against the variants (Fig. 4E). These data showed that vaccination improved pseudovirus neutralisation activity not only against the infecting variant (beta variant) but also the original D614G and delta variants.