Mass-standardised antibody concentration and affinity immunity endotypes to the spike protein of SARS-CoV-2 variants in a high-exposure cohort

School staff members are uniquely exposed to asymptomatic infections in children when in enclosed, poorly ventilated environments. Prevalence of SARS-CoV-2 amongst school children in the Omicron wave was high, leading to a constant, high exposure of teachers to the virus. A cohort of staff members (n = 28) was recruited from a UK primary school setting. The prevalent variants in the preceding six months were Omicron BA.1, BA.4/5 and BA.2: 61% of the cohort reported a lateral-flow-confirmed positive test for SARS-CoV-2 infection in late 2021 or 2022. A mass-standardised quantitative antibody spectrum was measured for both antibody concentration and a new measure of antibody quality; the proportion of antibodies still bound to the epitope at pH 3.2 compared with pH 7.4. The antibody spectrum was measured for IgG to the spike protein for variants Wuhan, Alpha, Beta, Gamma, Delta and Omicron BA.1, BA.2.75, BA.2.12.1, BA.4 and BA.5. The new antibody quality measure shows a median of 39% with (IQR 24% - 52%). The cohort showed a Universal positive immunity endotype, U(+), with antibodies binding to the spike protein across the variant spectrum having an incidence of 78% (95% CI 60% - 88%) with good antibody concentrations to all ten variants; the U(+) incidence drops to 39% (95% CI 22% - 59%) when the affinity spectrum is measured. The teachers' cohort shows a significant antibody maturation towards the variants prevalent wave variants with good quality antibodies to Wuhan, Delta and the Omicron variants but not the bet and gamma variants. Teachers and, consequently, parents are a critical perimeter of the pandemic and a high priority for vaccination along with healthcare workers.

School staff members are uniquely exposed to asymptomatic and symptomatic infections in children when in enclosed, poorly ventilated environments, with one study noting respiratory tract infections to be a leading cause of school absenteeism in over 70% of cases during the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic (1, 2). A retrospective test-and-trace study from Italy (3) has shown that, of the school students in the age range 6–10 years, 50% (95%CI = 39%–60%) tested positive and were asymptomatic carriers for SARS-CoV-2, whilst 21.4% (95%CI = 14%–31%) had symptomatic infections. Similarly, studies in the UK following the release of school teaching restrictions showed a surge in student infections during the period from August to October 2021 (4), with household clusters adding to transmissions in parents, particularly from children aged 5–15 years. Other studies counter with findings that schoolchildren in the pandemic (5) in the age range 5–11 years did not have a prevalence of infection greater than the local working population and attribute the greatest risk of infection to within-household transmission (6), although the prevalence of asymptomatic carriers was not reported in these studies.

A teachers’ cohort represents a high-exposure setting to circulating infections such as SARS-CoV-2, providing constant challenges to immunity imprinting and original antigenic sin (7–9) evolving the antibody epitope selection and the T-cell response. In a previous study (10), we observed antibody immunity endotypes to a spectrum of variants: Wuhan, Alpha, Beta, Gamma, Delta, and the Omicron variants—BA.1, BA.4, BA.5, BA.2.75, and BA.2.12.1. The immunity endotype is the presence of antibodies produced by the immune system to a set of epitopes on the spike protein surface. A universal endotype showed antibody concentrations to all variants and was present in 75% of the cohort tested. The spike protein has two main regions: S1, which contains the receptor binding domain (RBD) and is the target of neutralising antibodies, and the S2 region, which is located closer to the membrane of the virus and is typically conserved because mutation could reduce viral entry into host cells. The difference in antibody spectrum binding seen within the cohort reflects the difference in binding to the S1 region (including the RBD) and highlights inter-individual differences in antibody production and response across the population of immunity endotypes. The binding endotypes would be less likely to show differentiation between variants if raised to the more conserved S2 region.

The universal endotype confers protection against all measured variants and suggests that the immune system has selected conserved epitopes on the S1 regions, such as the hinge region required for the spike protein to transition from the pre-fusion to the fusion state (11). Dropout immunity endotypes account for the remaining 25% of the cohort and show poor antibody binding to one or more variants, having made antibodies to epitopes that are not conserved between variants. Exposure to the dropout variant would have a risk of poor clearance and post-viral sequalae (12–14), such as long coronavirus disease (COVID). Interestingly, the prevalence RECOVER study identified the prevalence of long COVID to be lower in children, i.e., 4%, compared with 10%–26% in adults (15).

Immunity endotypes have been previously explored by comparing the concentrations of antibodies for different variant spike proteins against a threshold (10), such as a correlate of protection, with similar protective thresholds reported using normalised antibody measurements (16). However, the affinity of the antibody–antigen interactions is important, showing not just the binding to the epitope but also a measure of epitope–paratope complex stability, which is required if the antibody is to prevent binding and recruit other elements of the immune system, such as the complement cascade, to affect virus clearance. Affinity maturation is a critical part of the clone selection process (17). Measurement of antibody affinity requires the association and dissociation rates of the paratope–epitope complex to be accurately measured, leading to the definition of a complex half-life: t 1 / 2 = ln 2 k d, where t 1/2 is the half-life (in seconds) and k d is the complex dissociation rate (per second). For stable complexes and clone selection of high-affinity interactions, the half-life will be approximately several hours (18). A stable complex is required to recruit the other immune system components and effect clearance through processes such as antibody-dependent cellular cytotoxicity (19). Furthermore, affinity studies are typically performed at pH 7.4, which is a poor mimic of certain body environments such as the mucosa (20), where the pH may be more acidic, pH 5.5, but still requires the lifetime of the virus–antibody complex to be high.

In this paper, we explore a new method of assessing the stability of the paratope–epitope complex by measuring a ratio between the antibody–antigen binding responses at two different pH values: pH 3.2 and pH 7.4. Titration of the paratope–epitope complex shows that affinity is pH-dependent, becoming completely unstable at pH 1–2. The pH stability allows a new set of immunity endotypes linked to affinity to be explored. The variant dependence of the affinity endotypes will be sensitive to the prevalent waves (Figure 1) and, hence, affinity maturation. Furthermore, affinity endotypes may point to new correlates of protection for the mucosa, where neutralisation or clearance is required (22).

We report a prospective cohort study on 28 staff members in a primary school setting with samples taken on 13 July 2022. The local prevalent Omicron variants at the time of sampling were BA.4 and BA.5, whilst BA.1 and BA.2 were prevalent in the preceding 6 months (Figure 1). Demographic data were collected alongside vaccination histories and lateral flow confirmations of any recent SARS-CoV-2 infections. A COVID antibody spectrum test was performed to assess the concentration and pH-dependent affinity profiles of 10 spike protein variants: Wuhan, Alpha, Beta, Gamma, Delta, and the Omicron strains BA.1, BA.2.75, BA.2.12.1, BA.4, and BA.5. The full 30-dimensional profile of three endotypes was explored: antibody concentration (AC) and the new quality endotypes; high-quality antibody concentration (HQ); and the percentage of high-quality antibodies (HQ%). Statistical analysis was performed to compare all three endotype responses to the persistent viral strains and infection and vaccination history.

Materials

The materials used throughout the course of the experiments were used as supplied by the manufacturers, without further purification. Sigma-Aldrich supplied phosphate-buffered saline (PBS) in tablet form (P4417), phosphoric acid solution (85 ± 1 wt.% in water; 345245), and Tween-20 (P1379; Sigma, St. Louis, MO, USA). Glycine (analytical grade, G/0800/48) was provided by Fisher Scientific (Loughborough, UK). The assay running and dilution buffer was PBS with 0.005% (v / v) Tween-20, whilst the regeneration buffer was 0.1 M phosphoric acid with 0.02 M glycine solution in deionised water.

A recombinant human antibody to the spike protein S2 subdomain, a chimeric monoclonal antibody (40590-D001, lot HA14AP2901; SinoBiological, Beijing, China), was used as a calibrant. The antibody was raised against the recombinant SARS-CoV-2/2019-nCoV spike S2 ECD protein (40590-V08B; SinoBiological). This was selected as it could bind to all 10 spike protein variants used herein.

NISTmAb, a recombinant humanised IgG1ĸ with a known sequence (23, 24) specific to the respiratory syncytial virus protein F (RSVF) (25), was obtained from the National Institute of Standards and Technology (RM8671). The detection mixture consisted of a 200-fold dilution of anti-IgG (IG8044 R2) from Randox (Crumlin, UK) in assay running buffer. The spike variant sensor chips were printed as detailed elsewhere (10) using proteins from SinoBiological and ACROBiosystems (Newark, DE, USA), as detailed in Supplementary Table 1. The sensor chips were also printed with a series of controls including recombinant human serum albumin from Sigma-Aldrich (A9731) and protein A/G from Thermo Fisher (21186; Waltham, MA, USA).

Methods

Biophotonic multiplexed immuno-kinetic assay

The multiplexed biophotonic array platform technology has been described in detail elsewhere (10) and has a CE mark. The array platform was used to perform the Attomarker COVID Antibody Spectrum Test (Exeter, UK), which has a UKCA mark. Briefly, the technology is a localised particle plasmon gold sensor that scatters light proportional to the mass of the material (formally refractive index) near to the gold nanoparticle surface in the plasmon field. The gold nanoparticles are printed into two arrays of 170 spots, which are then individually functionalised with the SARS-CoV-2 spike protein variants, as listed in Supplementary Table 1. Both sensor arrays contained one common spike protein variant, i.e., S (Wuhan), alongside control spots of recombinant human serum albumin for variation in illumination, temperature, and nonspecific binding. The integrity of the protein samples on the surface was assessed using an anti-S2 antibody (40590-D001), chosen for calibration as the corresponding epitope is present on all variants. Formally, S2 binding would not exclude the protein fragments on the surface. However, we have performed extensive S1 and S2 mapping on a previous study (10) with the antibodies described in Supplementary Table 2 in that study to show the spike protein integrity on the surface. The antibody ZHU1076 has binding kinetics measured for Supplementary Figures 3, 4 and has an epitope located in the RBD region. There was high correlation between the spike channel performance on the two array types (m, c > 0.92); therefore, the results were treated as aligned (Supplementary Table 2; Supplementary Figure 1).

The concentration of the anti-S2 antibody was calibrated against the NIST antibody RM8671, NISTmAb, a recombinant humanised IgG1ĸ with a known monomeric purity (23) to assume monomeric purity of the calibration antibody. This assay was performed by measuring the association curves of the two antibodies to the immobilised protein A/G on a sensor surface. High-control and low-control samples were made using the 40590-D001 antibody, and these controls were then used to quantify the results from human samples in units of milligrams per litre. Thus, the assay response is presented as a concentration of an antibody equivalent in its binding properties to the 40590-D001 control antibody.

The total AC endotype classification has been derived for the spike protein variant (10) using a threshold of 1.8 mg/L for the Wuhan variant at the beginning of the pandemic, an average of a diagnostic threshold of 200 PCR-positive patients compared with 200 pre-pandemic samples, determined as 1.34 ± 1.10 mg/L, and a diagnostic threshold derived from the anti-SARS-CoV-2 NISBC verification panel, which showed a separation of the seropositive and seronegative samples at 1.90 ± 1.10 mg/L. The mean mass-standardised value of the two thresholds (26) is 1.8 mg/L (95%CI = 0.2–3.4 mg/L). The limit of detection (LoD) of the technique is 0.2 mg/L.

Hence, the universal positive AC endotype [AC U(+)] has the AC for all variants above the 1.8-mg/L threshold. A single dropout has an AC for one or more variants below the threshold [e.g., β(−)], a double dropout, two variants [e.g., α(−)β(−)], and so on, for all variants, and these are all labelled U(±), i.e., with one or more variant ACs<1.8 mg/L. A full correlation analysis for all combinations of responses was also performed from which the correlation coefficients and the lines of best fit were derived.

The new antibody quality endotypes, i.e., HQ and HQ%, were derived by comparing the antibody binding on the surface of the target antigen after a 30-s wash step at either physiological serum pH 7.4 or pH 3.2. Full dissociation occurs at pH 1.9, the pH used during regeneration of the sensor array surfaces. The titration plots are shown in Supplementary Figure 2 for four COVID-related proteins of nucleocapsid, S (Wuhan), S (Omicron), and RBD. Full details are in Supplementary Table 1. The HQ affinity measure may be compared with an avidity measure derived for HIV antibodies (27), where the epitope–paratope complex stability was challenged by the chaotrope guanidine. The affinity endotypes were generated against a threshold derived from the lower interquartile range (IQR) and derived for all variants. The percentage was then used to calculate the high-quality antibody level. HQ%, U(+) is then defined as all variants above the threshold; similarly, HQ U(+) denotes all variants above the HQ threshold. Any one dropout is classified as U(±).

The full patient antibody profile is parameterised by the total AC, the HQ, and the HQ% (calculated as HQ/AC × 100). The cohort was used to determine reference ranges for these two new immunological measures. The HQ% from the cohort was used to estimate the HQ threshold of a sterilising serum, scaling the AC threshold derived elsewhere (26). The quality endotype is defined as having an HQ% below the lower quartile of the distribution. The incidence of each endotype is generated as a proportion of the cohort with Wilson’s 95% confidence limits.

Patient cohort

The staff members were recruited from a primary school setting, and the UKCA-marked Attomarker COVID Antibody Spectrum Test was performed after giving informed consent with a registered healthcare professional for the use of their residual blood samples to aid understanding of the pandemic. The vacutainer blood samples were centrifuged at 1,300 rpm for 10 min to separate the blood constituents. The serum was then collected and used for the variant profile screening using the Attomarker biophotonic array platform. These results were subsequently provided to the participants.

Ethical approval

The use of the Attomarker clinical samples was approved by the Bioscience Research Ethics Committee, University of Exeter, and individuals provided informed consent prior to their venous blood draw.

The cohort recruited from the school comprised 86% women and 14% men: 68% had an initial AstraZeneca (AZ) two-vaccination dose, 25% Pfizer, 4% Moderna, and 4% undeclared (Table 1 and Supplementary Table 7). There were seven teachers in the cohort, or 25% (95%CI = 13%–43%), who were un-boosted. Each teacher had a full concentration and pH-dependent affinity profile assessment for each of the 10 spike proteins. A pH-dependent affinity titration for a single sample is shown in Supplementary Figure 2 for nucleocapsid, spike protein, the Omicron spike protein, and RBD. Each of the paratope–epitope titrations showed a partial fall in binding at pH 3.2, which was selected as the differentiating pH for the HQ and HQ% to provide a good dynamic range. The mean and the IQR for HQ% were derived from all variants measured to give HQ% as 39% (IQR = 24%–52%) (Supplementary Table 8). The full dynamic range of HQ% was 2%–100%.

Example patient endotype spectra are shown in Figure 2, set against the protective AC threshold of 1.8 mg/L and the HQ% lower IQR marker at 24%. The teacher cohort had a confirmed infection incidence in the preceding 6 months of 61% (95%CI = 42%–76%) and an AC U(+) endotype incidence of 78% (95%CI = 60%–88%) consistent with the incidence observed in our previous study (28), shown in Figure 2A and Table 1. Of the remaining six teachers, five showed a Delta dropout U(±) endotype either alone or in combination with the Beta variant: the two Beta variant dropout endotypes were associated with a Pfizer vaccination history similar to previous observations (28). The AC U(+) incidence is not reflected in the incidence of HQ U(+) and HQ% U(+), as can be seen for patient 3 (Figures 2G–I). The cohort incidence of HQ U(+) was 43% (95%CI = 25%–63%) and that of HQ% U(+) was 21% (95%CI = 9%–41%). The dropout variant classification is shown for each patient independently (Table 1), the majority of which are from early variants, α,β,γ,δ(−), and never solely from Omicron variants. The universal immunity endotype [U(+)] for all three measures was 14% (95%CI = 5%–34%) (Table 1).

The cohort bee swarm box plots (Figure 3) showed the median concentrations, i.e., AC, for all variants to be equivalent (Supplementary Table 6), but with significantly lower median pH-dependent affinity values (Wilcoxon: Wuhan vs. Beta: p = 6.1 × 10 −8; Wuhan vs. Gamma: p = 5.3 × 10 −6). The median HQ% quantities for the β and γ variants were significantly lower than for the vast majority of other variants (Supplementary Table 6).

The teacher cohort was uniquely exposed in the primary school setting to children, 5–11 years old, who were either asymptomatic carriers or symptomatic cases (3, 4) during the COVID pandemic. The repeated, evolving exposure is the ‘new-normal’, where repeated waves of infection lead to a high virus prevalence and infection (Figure 1). Schools have been identified (29) as foci of the pandemic, although the children themselves may not be vulnerable to infection, leading to difficult discussions about child vaccination (30). Countries have differed in their approaches despite the World Health Organisation (WHO) recommendations (31). The effect of reopening schools in the UK was a significant increase in infection rates amongst schoolchildren (29). The teachers and parents of schoolchildren represent one of the critical immunity perimeters of the pandemic with consequent implications for workforce sickness and impact on economic activity.

The constant, high-exposure teachers’ cohort, whether unvaccinated, vaccinated, or boosted, are exposed repeatedly to the prevalent circulating virus to which there is a significantly refined immunity response (Figure 1). Immunity imprinting has been reported for SARS-CoV-2 vaccination (32), and we have reported the immunity endotypes for antibody concentration against a spectrum of variant spike proteins (28), suggesting that the initial exposure to a virus or a vaccine may imprint an immune memory response. Antibody quality or antibody affinity maturation has been reported for SARS-CoV-2 immunity (33, 34), with mechanisms of maturation proposed in the germinal centre B-cell response (35, 36) critical to the adaptive immunity response (37). Sustained antigen availability has also indicated non-standard vaccination strategies to optimise the adaptive response (38, 39). The teacher cohort is therefore an ideal setting for the observation of antibody immunity imprinting and consequent maturation: the forgiveness of original antigenic sin.

There is a spectrum of immunity in the teacher cohort derived from vaccination and natural infection (four staff members reported sickness and had time off work) (Table 1). The incidence of lateral flow test-confirmed infection in the preceding 6 months was 63%, suggesting exposure to BA.1.159, BA.4/5, and BA.2 (Figure 1). The antibody concentration profile across the spectrum of spike proteins showed that the cohort incidence of U(+) was 75% and reflects previous measurements indicating that many had moderate–high antibody levels to all variants, potentially from targeting a universal epitope (40) on the S1 region of the protein. The AC endotype classification is threshold-dependent. We have derived a 1.8-mg/L threshold from a set of recovery cohorts and reference materials, which may also be compared to the threshold derived by Plotkin (41) of 5 mg/L and a study determining the absolute protective antibody levels (16) for COVID determined at 7.5 mg/L. The convergence of these threshold data begins to point to a fundamental AC endotype property of COVID infections.

The study explored antibody affinity as an additional property of the host antibody response using binding at pH 3.2 to produce HQ%. The dynamic range of the new metric proved excellent, varying from 2% to 100% and showing significant variation in the cohort spectra (Figures 2, 3). The AC wanes with time, typically with a half-life of 60–200 days (42), suggesting that AC and HQ classifications change with time since acute infection or exposure to variants in the school setting. However, the HQ% endotype, the percentage of high-quality antibodies, may be conserved over time, reflecting clone selection and the choice memory B cells. The Wuhan and BA.5 antibodies generated by the boosted vaccination national strategy and high exposure to the variants in the school setting showed similar HQ and HQ% endotypes in the cohort (Figure 3). On the other hand, there were significantly worse HQ and HQ% endotype responses to the earlier variants, which did not form waves in the UK: Alpha, Beta, and Gamma (Kolmogorov–Smirnov test p -values: α = 0.016, β = 4.1 × 10 −8, γ = 1.84 × 10 −5, δ = 0.28, BA.2.12.1 = 0.016, BA.2.75 = 0.0023, BA.4 = 0.47, BA.5 = 0.91, and BA.1 = 0.91 vs. Wuhan). The HQ% endotype threshold was set at the lower quartile of percentage distribution of 24%, and U(+) had an incidence of 14% (95%CI = 5%–34%). However, there was no sole dropout variant for the post-Omicron variants, indicating that the performance of the vaccine to β and γ was poor and not improved by variant wave exposure. The poor pH 3.2 stability may also provide insights into why the vaccines did not prevent infection as there is poorer stability in acidic conditions such as the mucosa at pH 5.5.

The cohort distributions showed a significant antibody maturation and consequent vaccine efficacy towards the Omicron sub-variants (high-affinity concentration data Kolmogorov–Smirnov test p -values: α = 0.47, β = 0.0023, γ = 0.15, δ = 0.47, BA.2.12.1 = 0.0062, BA.2.75 = 0.016, BA.4/5 = 0.15, and BA.1 = 0.91 vs. Wuhan), with maximum efficiency for the variants BA.2.12.1 and BA.2.75, both descending from the widely circulated BA.2 lineage. These results indicate that the vaccine response is frangible in the presence of the circulating wave with ongoing antibody maturation and clone selection.

The new measure of pH-dependent affinity, i.e., HQ%, had a mean of 39% (IQR = 24%–52%) and represents a simple way to assess the dissociation constant of the epitope–paratope complex. The HQ threshold at 24% or 0.43 mg/L is comparable to the AC threshold of 1.8 mg/L at the time of exposure in the estimation cohorts. The HQ threshold may reflect a better correlate of protection than just neutralising antibody concentration. The new HQ% measure of complex half-life, corresponding to 0.5–2 h, and a new threshold in the quality of the immune response, points to a correlate of protection with a supporting mechanism. A linear correlation has been shown with the complex half-life and affinity constant, K D (in picomolar), calculated as K D = k d k a, where k a is the association rate constant for the antibody to the epitope (per molar per second) and k a is the dissociation rate constant (per second). Good quality commercial antibodies typically have values from 10 −9 to 10 −11 M for COVID neutralising antibodies (43). Furthermore, the physiological importance of a lower pH is also relevant to the nasal mucosa (20), which, at pH 5.5–6.5, is significantly lower than the pH of the serum and the buffers used for conventional affinity measurements.

The immunity profile for staff in the constant, high-exposure cohort has been characterised in 30 dimensions—10 variant spike proteins, total AC, HQ, and HQ%—which are now all measures of a sterilising serum for the clearance of different variants of SARS-CoV-2. Technique-independent thresholds and reference ranges have been set for these serum characteristics. The patient antibody spectra showed significant antibody maturation to all SARS-CoV-2 variant spike proteins, notably to the prevalent Omicron variants. Approximately 39% of the cohort had full, pan-variant protection, whereas the remainder had higher-quality antibodies focussed on the preceding wave variants. A smart-boosting programme focussed on high-transmission groups, such as teachers and parents/guardians of school-aged children, could significantly reduce the risk of high infection levels within the population during each wave, as the pandemic transitions to endemic. Furthermore, extensive and readily measured immunity profiles may highlight specific endotype vulnerability to infection and provide insights into conditions such as long COVID (44).