Greater IgG and FcγR-binding antibodies to ancestral SARS-CoV-2 spike and RBD in BNT162b2 recipients compared to mild-to-moderate convalescence

Sixteen individuals vaccinated with BNT162b2 had plasma samples drawn before vaccination (baseline) and two weeks (median: 13 days; IQR: 0.75 days) following the second dose (Table 1). We also studied plasma from fifteen convalescent individuals (median: 38 days post symptom onset; IQR: 10.5 days) with mild-to-moderate COVID-19 disease. Importantly, convalescent samples were collected during the first wave of the pandemic (between March 2020 and May 2020 in Victoria, Australia), providing an opportunity to study responses in individuals infected by viruses that harbor less phylogenic deviation from the ancestral reference strain used in the BNT162b2 mRNA vaccine.

First, plasma humoral profile to SARS-CoV-2 proteins (NP, RBD, S1, Trimer Spike) found in the ancestral wild-type virus was characterized using a SARS-CoV-2 multiplex bead assay. Robust antibody responses to NP were uniquely associated with convalescent COVID-19 individuals, indicating no prior infection in the BNT162b2-vaccinees (Supplementary Fig. 2). Convalescent and BNT162b2-vaccinees both elicited elevated levels of IgG to spike antigens; however as expected, BNT162b2-vaccinees demonstrated significantly greater responses in comparison to convalescent individuals which displayed more heterogenous responses (S1, 3.8-fold higher; trimer, 1,sixfold higher; RBD, 3.5-fold higher), (Fig. 1B–D; Supplementary Fig. 2). Next, we measured the capacity for antigen-specific antibodies to engage soluble FcγRIIa and FcγRIIIa-dimer constructs, which mimic the surface of innate immune cells and are a proxy readout for ADCP and ADCC activity, respectively. As before, plasma from BNT162b2-vaccinees showed significantly higher FcγRIIa and FcγRIIIa-dimer binding levels by anti-spike antibodies than mild-to-moderate convalescent patients (Supplementary Fig. 2; Supplementary Fig. 3). Notably IgG1 subclass levels strongly correlated with both FcγRs (FcγRIIa-H131: Pearson r = 0.94, p < 0.0001; FcγRIIa-H131: Pearson r = 0.90, p < 0.0001, Supplementary Fig. 3j). Unsupervised principal component analysis (PCA), used to broadly examine this repertoire of antibody responses, showed that baseline, BNT162b2 (2 weeks post dose two), and convalescent (mild-to-moderate) subjects were naturally separated into three distinct groups based on their antibody signatures (Supplementary Fig. 2E). This analysis confirms that BNT162b2 vaccination induces more robust anti-ancestral spike IgG responses with greater FcγR-binding capacity, in comparison to our mild-to-moderate convalescent donors.

Fig. 1

IgG binding and hACE2 inhibition to ancestral RBD and 38-RBD variants. A Pan-IgG binding levels (reciprocal ED50) and B human wild-type ACE2 (hACE2) inhibition levels (reciprocal ID50) by BNT162b2 (2 weeks following second dose; blue; n = 16) and mild/moderate convalescent COVID-19 patients (green; median 38 days post symptom onset; n = 15) as measured by multiplex to RBD variants. Reciprocal ED50/ID50 values, respectively, were calculated using the normalized binding MFI value for an 8-point serial plasma titration. C Fold change of pan-IgG binding levels of RBD variants in comparison to ancestral RBD and D Fold change of hACE2 inhibition of RBD variants in comparison to ancestral RBD by BNT162b2 plasma. Fold change was calculated as follows: (geometric mean of reciprocal ED50 or ID50 Ancestral RBD) divided by (geometric mean reciprocal ED50 or ID50 variant RBD). Friedman’s non-parametric test with Dunn’s multiple comparisons was used to assess statistical significance. E Pan-IgG binding levels (reciprocal ED50) and (F) hACE2 inhibition levels (reciprocal ID50) to RBD of Omicron BA.2 variant with fold change as indicated above in blue. Mann–Whitney U-test used for statistical evaluation. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***) and p < 0.0001 (****)

Reduced IgG-binding antibodies to SARS-CoV-2 RBD variants, but not always concordant to loss in ACE2 inhibition

The RBD is a mutational hotspot for SARS-CoV-2. Because of its position at the ACE2 interface, mutations in the RBD can alter ACE2 affinity as well as result in the loss of epitopes for antibodies [2, 40, 41]. Thus, to characterize the impact of RBD mutations on plasma IgG binding in our convalescent and BNT162b2 donors, we generated a custom panel of 39 RBDs, including RBDs of 33 common-point mutations, 5 variants of concern/interest (Beta, Delta, Gamma, Kappa, and Omicron) and the ancestral RBD, and measured antibody responses using multiplex. Higher levels of IgG binding responses (larger reciprocal ED50 values) were observed in BNT162b2-vaccines compared to convalescent individuals across all RBDs assessed (Fig. 1A, E), consistent with our previous observation that BNT162b2 elicits more robust anti-spike IgG responses. Despite the overall greater levels of IgG in vaccinees, significantly reduced IgG recognition was observed against the VOCs Beta (7.1-fold reduction, p < 0.0001) and Gamma (15.5-fold reduction, p < 0.0001) by BNT162b2 plasma in comparison to the ancestral RBD (Fig. 1C). Drop in IgG binding was also observed with Omicron (BA.2) (2.9-fold reduction, p < 0.0001) (Fig. 1E). Furthermore, significant decreases in IgG recognition to various point mutations, including G446S (5.4-fold reduction, p < 0.0001) found in Omicron, E484K (5.9-fold reduction, p < 0.0001) found in Gamma and Beta, as well as F490S (5.4-fold reduction, p < 0.0001) and N501T (6.8-fold reduction, p < 0.0001) (Fig. 1C). These data highlight the influence of particular point mutations on recognition of the RBD by IgG antibodies.

Since the RBD is a dominant target for neutralizing antibodies, we next considered the impact of these RBD variants on blocking ACE2 binding, using a surrogate neutralization assay. Using a previously established in vitro ACE2 inhibition assay, we measured the capacity for BNT162b2 and convalescent plasma to block hACE2 binding in a competitive format. Overall, the capacity for BNT162b2 plasma to block ACE2 binding was greater (larger reciprocal ID50 values) in comparison to convalescent plasma across all 39 RBDs (Fig. 1B, F). Consistent with the loss of IgG binding, we observed decreased ACE2 inhibition to all assessed VOC, with a significant decrease observed for Beta (3.6-fold reduction) and Gamma (5.1-fold reduction), though this reduction was notably smaller than that for IgG binding (Fig. 1C). A substantial drop in ACE2 inhibition was observed with Omicron (BA.2) (26-fold reduction; p < 0.0001) highlighting the evasive nature of this variant (Fig. 1F). On the other hand, most of the RBD point mutations which displayed reduced IgG binding did not necessarily translate to a significant drop in ACE2 inhibition (Fig. 1C, D). While a decrease in IgG binding was observed for the RBD mutations G446S and N501T, the loss in ACE2 inhibition observed was much weaker (G446S = 1.6-fold reduction, p = ns; N501T = 2.0-fold reduction, p = 0.0013) (Fig. 1D; Supplementary Table 3). A 1.7-fold reduction (p = 0.04) in ACE2-blocking was also observed for L452R, despite having comparable IgG binding (p = ns) to the ancestral RBD. Furthermore, the RBD mutation N501Y, the prime mutation found in the RBD Alpha-VOC, demonstrated no significant decrease in IgG binding (1.6-fold reduction, p > 0.99) but showed a significant drop in capacity to block ACE2 from binding (5.4-fold reduction, p < 0.0001) (Fig. 1C, D; Supplementary Table 3). Together, these data highlight the influence of naturally occurring RBD mutations on the potential for polyclonal antibodies to recognize and block ACE2 binding.

SARS-CoV-2 VOC RBDs have improved affinity to WT hACE2

Mutations within the RBD also have the potential to modulate affinity to hACE2 [2, 4, 41, 42]. Hence, we next characterized the binding affinity of wild type (WT)-ACE2 to VOC RBDs, as well as a subset of single-point mutations, using biolayer interferometry (Fig. 2; Supplementary Fig. 4). We found all VOC RBDs assessed demonstrated elevated binding affinity to WT ACE2 in comparison to ancestral RBD (Kd: 29.4 nM; t1/2: 83.6 s) (Fig. 2). These included Beta (Kd: 14.6 nM; t1/2: 164.83), Gamma (Kd: 15.0 nM; t1/2: 214.82 s) and Delta (Kd: 21.5 nM; t1/2: 143.77 s). Omicron (BA.2) demonstrated a 2.6-fold increase in binding (Kd: 11 nM; t1/2: 161.75 s), compared to ancestral RBD (Fig. 2). Together, these data reiterate how these mutations in the RBD, given its position at the ACE2 interface, can potentially improve viral fitness by enhancing affinity to ACE2. Furthermore, these data suggest that the RBD mutations observed in Beta and Gamma not only significantly reduced IgG binding, but also compounded with stronger ACE2 affinity to give the largest drops in ACE2 inhibition.

Fig. 2

Affinity of wild-type (WT) hACE2 to the RBD of variants. Sensograms of WT-human ACE2 (hACE2) binding affinity to A ancestral RBD, B Alpha (B.1.1.7), C Beta (B.1.351), D Delta (B.1.617.2) E Gamma (P1), F Omicron (BA.2) as measured by biolayer interferometry (BLI). Binding of twofold serial dilutions of each SARS-CoV-2 RBD variant (100 nM (red), 50 nM (yellow), 25 nM (green), 12 nM (blue), and 6 nM (pink)) was measured to immobilized hACE2 (3 μg/ml). Binding curves representative of 2 independent experiments for each variant are plotted (solid-colored lines), globally fitted to a 1:1 binding model (black line)

Previously, we and others have shown single-point mutations in RBD (such as N501Y found in Alpha, Beta, Gamma and Omicron) can enhance affinity to ACE2 (4.5-fold increase; Kd: 6.5 nM; t1.2: 329.1 s, Fig. 2B) [4, 33, 40]. With N501Y, we observed a reduced capacity for antibodies to neutralize, despite no loss in total IgG (Fig. 1). Similar to N501Y, here we observed using BLI that while L452R did not differ greatly from the ancestral RBD in total IgG binding, it exhibited a stronger affinity (Kd: 27.36 nM; t1.2: 126.4 s), which in turn could explain the drop in plasma-driven ACE2 inhibition (Supplementary Fig. 4; Supplementary Table 4). Likewise, using BLI, we observed that N501T had improved binding affinity (KD: 11.05 nM; t1.2: 217.8 s) as compared to the ancestral RBD (Supplementary Fig. 4; Supplementary Table 4). As such, the combination of reduced IgG binding due loss of epitope and stronger ACE2 affinity for N501T could further impeded ACE2 inhibition by plasma samples. Together, these results suggest that beyond antibody recognition, affinity interactions between RBD and ACE2 exert another layer of influence on the potential for antibodies to effectively block ACE2-binding and achieve neutralization.

hACE2 polymorphisms alter affinity to SARS-CoV-2 RBD variants

Several ACE2 polymorphisms are present at low frequency within the human population, with certain alleles more prevalent in different populations (Supplementary Table 6). In particular, K26R has been observed at low frequencies across most populations, whereas low frequencies of S19P are predominantly detected within African/African American populations (0.3%) [22, 24]. Previous work has shown these polymorphisms confer altered affinity to ancestral RBD; however, the binding kinetics to emerged VOC RBD, especially Delta and Omicron, has yet to be described. Here, we assessed the binding profiles of three low-frequency ACE2 polymorphisms (E35K, K26R and S19P) to 5 VOC RBDs (Alpha, Beta, Delta, Gamma, Omicron (BA.2)) and ancestral RBD using biolayer interferometry (BLI). We found that the polymorphism S19P (KD: 14.38 nM) and K26R (KD: 20.0) demonstrated a 2.04-fold and 1.47-fold increase in binding affinity, respectively, to ancestral RBD when compared to WT ACE2 (KD: 28.99 nM) (Fig. 3, Supplementary Table 5). Furthermore, this enhanced affinity of the S19P and K26R polymorphisms remains overall elevated against the RBD’s of Beta (KD: S19P = 10.51 nM, K26R = 17.46 nM), Omicron (BA.2) (KD: S19P = 4.97 nM, K26R = 5.18 nM) and Delta (KD: S19P = 8.56, K26R = 13.47) (Fig. 3, Supplementary Table 5). These data suggest that individuals who harbor S19P or K26R ACE2 polymorphisms may have greater susceptibility to infection by SARS-CoV-2 due to increased ACE2 affinity. Contrastingly, ACE2 polymorphism E35K demonstrated weaker binding when compared to ancestral RBD (KD: 36.63 nM), Omicron (BA.2) (Kd: 18.50 nM), and Delta (Kd: 43.34 nM), suggesting that individuals harboring the E35K ACE2 polymorphism may demonstrate a more protective phenotype due to reduced ACE2 affinity for RBD (Fig. 3; Supplementary Table 5). Together, these data suggest that ACE2 polymorphisms, though low in frequency, can modulate binding to ACE2, and by extension may change an individual’s susceptibility to infection.

Fig. 3

Affinity of hACE2 polymorphisms to VOC RBD’s. Binding profiles of WT hACE2 (purple) and three polymorphisms (E35K, green; K26R, red; S19P, blue) to A ancestral RBD and RBD of variants of concern B Alpha (B.1.1.7), C Beta (B.1.351), D Delta (B.1.617.2), E Omicron (BA.2). Sensograms show 100 nM of immobilized RBD of each variant run against 100 nm (3 μg/ml) ACE2 of each polymorphism to measure binding kinetics

FcγR-binding antibodies to RBD variants

Several previous studies have highlighted the importance for Fc-effector functions during SARS-CoV-2 infection, yet to date, little work has widely characterized the impact of RBD mutations on FcγR-binding [14, 15, 17, 18, 20]. Given the observed loss of IgG to certain RBD variants, we, thus, next investigated the impact that these mutations conferred on the capacity to for anti-RBD antibodies to engage our FcγRIIa and FcγRIIIa-dimer constructs, a surrogate measure for ADCP and ADCC activity, respectively. We used our multiplex panel of 39-RBD to address the impact of these mutations on FcR-binding antibodies as compared to the ancestral RBD. Consistent with our previous observation to total IgG levels, BNT162b2 vaccinee plasma had a significantly greater capacity (higher reciprocal ED50 vales) to engage both our FcγRIIa and FcγRIIIa-dimer constructs across all variants, in comparison to convalescent plasma (Fig. 4A, B). Indeed, antigen-specific IgG levels highly correlated with FcγRIIa and FcγRIIIa-binding of their respective variants (Supplementary Fig. 5). Compared to ancestral RBD, VOC Beta and Gamma demonstrated a decrease in anti-RBD FcγRIIa (9.8- and 14.9-fold reduction respectively; p < 0.0001) and FcγRIIIa-dimer-binding antibodies (8.4-, and 7.2-fold reduction respectively; p < 0.0001) (Fig. 4C, D). Furthermore, we observed that single-point mutations with poor IgG binding, such as G446S, displayed a reduction in FcγR-binding as well (FcγRIIa: 6.2-, 5.0-fold reduction, respectively; FcγRIIIa 5.5-, 2.1-fold reduction, respectively) (p < 0.0001) (Fig. 4C, D; Supplementary Table 3). Likewise, a drop was observed to Omicron BA.2 with FcγRIIa binding, with a marginally larger decrease to FcγRIIIa observed (3.1- and 3.7-fold reduction, respectively; p < 0.0001) (Fig. 4E, F). Although interestingly, we noted this drop against Omicron was less pronounce than the decrease in ACE2 inhibition (26-fold reduction; p < 0.0001), suggesting FcγR-binding antibodies may be more resistant to mutations found in Omicron than ACE2 inhibition (Fig. 1F).

Fig. 4

Loss of anti-RBD FcγR-binding antibodies and ADCP responses to VOC RBDs in BNT162b2 and convalescent plasma. A FcγRIIa-H131 and B FcγRIIIa-V158 binding levels of RBD-specific antibodies using plasma of BNT162b2 (2 weeks following second dose; blue; n = 16) and mild/moderate convalescent COVID-19 patients (green; median 38 days post symptom onset; n = 15) as measured via multiplexing. C Fold change of anti-RBD FcγRIIa-H131 binding levels to RBD variants in comparison to ancestral RBD of BNT162b2 plasma. D Fold change of FcγRIIIa-V158 levels of RBD variants in comparison to ancestral of BNT162b2 plasma. Reciprocal ED50 values were calculated using the normalized FcγR binding MFI value for an 8-point serial plasma titration. Fold change was calculated as follows: (geometric mean of reciprocal ED50 or ID50, respectively ancestral RBD) divided by (geometric mean reciprocal ED50 or ID50, respectively variant RBD). Friedman’s non-parametric test with Dunn’s multiple comparisons was used to assess statistical significance. E FcγRIIa-H131binding (Reciprocal ED50) and F FcγRIIIa-V158 inhibition levels to RBD of Omicron variant with fold change of BNT162b2 plasma indicated above in blue. Mann–Whitney U-test used to assess statistical significance. G, H Spearman correlations of phagocytosis activity (reciprocal ED50) using THP-1 monocytes and multiplex FcRIIa-H131 data (reciprocal ED50) against the ancestral and Beta RBD. I ADCP by THP-1 monocytes induced by the plasma of BNT162b2 against the ancestral and Beta RBD shown as phagocytic score ED50. Wilcoxon matched pairs signed rank test was used to assess statistical significance. p < 0.05 (*), p < 0.01 (**), p < 0.001 (***) and p < 0.0001 (****)

To validate our findings using the surrogate FcγR dimers, we next measured the capacity for BNT162b2 plasma to induced ADCP using a duplex bead-based assay with a THP-1 monocyte cell line. Phagocytic score reciprocal ED50 were calculated, with highly concordant values obtained when non-specific background responses was adjusted for using either a no plasma control or corresponding matched baseline samples (Pearson r = 0.98; p < 0.0001) (Supplementary Fig. 1D–F). FcγRIIa reciprocal ED50 values correlated well wit phagocytosis ED50 levels of ancestral RBD by THP-1 monocytes (spearman r = 0.59, p = 0.017), as well as the RBD of Beta (spearman r = 0.52, p = 0.04) (Fig. 4G, H). For both ancestral and beta RBD, IgG binding (reciprocal ED50) correlated strongly with FcγRIIa-dimer (Spearman r = 0.77 and 0.79, respectively; p = 0.0006 and 0.0004, respectively), and phagocytosis levels (Spearman r = 0.81 and 0.72, respectively; p = 0.0002 and 0.002, respectively) (Supplementary Fig. 6A–D). We observed a significant decrease in ADCP activity against ancestral vs Beta RBD (p = 0.0026), supporting the mutations found in VOCs are detrimental to FcγR-binding antibodies (Fig. 4I). Altogether, our data suggest that a reduced capacity for antibodies to induce Fc-effector functions against the RBD variants compared to RBD of wild type. However, in the context of Omicron, FcγR-binding antibodies be more readily maintained than ACE2 inhibiting antibodies.