Characterizing SARS-CoV-2 neutralization profiles after bivalent boosting using antigenic cartography

We have previously described an antigenic map containing major pre-Omicron variants as well as BA.1, BA.2, and BA.5 Omicron variants⁷. Meanwhile, many more Omicron variants have emerged. To characterize antigenic relations of these newly emerged Omicron variants relative to older variants, we aimed to isolate a representative set of variants from BA.2.75, BA.5, recombinant XBB, and recombinant XBF linages (Supplementary Figs. 1–5). We generated virus stocks for three BA.2.75 variants (CB.1, BR.3, CH.1.1), six BA.5 variants (BA.5.2.1, BE.1.1, BF.7, BQ.1.3, BQ.1.1, BQ.1.18), two XBB recombinant variants (XBB.1, XBB.1.5.1), and one XBF recombinant variant (XBF.3) and confirmed their identity via third generation sequencing (Supplementary Table 2). Spike mutations for all variants used in this study relative to ancestral Wuhan-1 are shown in Supplementary Figs. 1–5.

In a second step, we characterized neutralization profiles for these new variants and determined their antigenic relation to early Omicron as well as pre-Omicron variants. Antigenic relations can be visualized in an antigenic map, which is generated by translating fold changes of neutralization titers between variants into antigenic map distances²³. To create antigenic maps that reflect the basic antigenic relationships among variants it is crucial to use single variant exposure sera, as multiple exposures increase cross-neutralization and therefore will potentially skew antigenic relations. We have earlier collected a number of first exposure sera for ancestral, alpha, beta, delta, BA.1 Omicron and BA.2 Omicron variants⁷. To increase resolution of an antigenic map in the area covered by Omicron variants, we included also one BA.5 first exposure serum and two CK.2.1.1 (BA.5.2 variant) first exposure sera in this study (Supplementary Fig. 21 for mutations in CK.2.1.1 spike compared to BA.5). We analyzed neutralizing antibodies for first infection sera and two dose BNT162b2 (BNT, Comirnaty, BioNTech/Pfizer) vaccinated individuals (BNT/BNT) against our panel of recently isolated Omicron variants as well as early BA.1, BA.2, and BA.5 Omicron variants and pre-Omicron variants (Fig. 1 and Supplementary Fig. 22). Three times BNT vaccinated individuals (BNT/BNT/BNT) were included as reference but the data from this cohort has not been used for calculation of the antigenic map as multi exposure sera likely underestimate antigenic relationships due to increased cross-reactivity and we previously showed that neutralization profiles of three dose vaccinated individuals were more similar to those of individuals after re-infection with an antigenically distinct variant than after single infection⁷. However, we constructed antibody landscapes for this group to visualize neutralization profiles (Supplementary Fig. 23).

Single variant exposure plasma samples were collected, either from unvaccinated individuals after first infection with ancestral virus (n = 5), alpha (n = 10), beta (n = 4), delta (n = 7), BA.1 (n = 16), BA.2 (n = 12), BA.5 (n = 1), or CK.2.1.1 (n = 2) variant or from two (BNT/BNT, n = 6) or three-dose (BNT/BNT/BNT, n = 6) vaccinated (ancestral) individuals. Titers of neutralizing antibodies against BA.2.75 variants (CB.1, BR.3, CH.1.1), BA.5 variants (BA.5.2.1, BE.1.1, BF.7, BQ.1.3, BQ.1.1, BQ.1.18), recombinant XBB variants (XBB.1, XBB.1.5.1), or recombinant XBF.3 for individual patients (circles) and geometric mean (red line) are shown. Titers below 16 were treated as negative (dotted line), and titers below 1 were set to 1. BNT = BNT162b2, IC₅₀ titer = 50% neutralization titer.

Titers against the BA.2.75 variants CB.1, BR.3, and CH.1.1 were low or undetectable for most single-exposure sera indicating the strong immune escape phenotype of these variants (Fig. 1, purple box). BA.5 variants could be divided into two groups regarding their neutralization profiles. BA.5.2.1, BE.1.1, and BF.7 were more similar to the initial BA.5 variant, while BQ variants showed a greater drop in neutralizing antibodies (Fig. 1, petrol box). The three BQ variants analyzed in this study differed in spike mainly by presence of the 144 deletion and the R346T mutation. BQ.1.18 contains both, BQ.1.1 only the R346T mutation and BQ.1.3 neither (Supplementary Fig. 4). However, BQ.1.3 has an additional E619Q mutation. All three variants showed similar neutralization profiles except for neutralization by the CK.2.1.1 convalescent sera (Supplementary Fig. 21). While BQ.1.3 and BQ.1.18 were neutralized by both CK.2.1.1 sera, no neutralizing antibodies against BQ.1.1 were detected. In contrast, the single BA.5 convalescent sample did not contain neutralizing antibodies against any of the three analyzed BQ variants. Titers for both XBB recombinant variants dropped relative to earlier variants with a similar pattern for XBB.1 and XBB.1.5 (Fig. 1, orange box). The XBF.3 recombinant variant showed a similar neutralization pattern as the BA.2.75 variants, which was not unexpected as XBF.3 contains a BA.2.75 variant spike (Fig. 1, green box). Overall, although many of the new variants were poorly neutralized by single infection sera, most individuals vaccinated with three doses of BNT were able to neutralize the whole panel of analyzed variants at least at a low level.

To visualize the antigenic relation between virus variants, we next performed antigenic cartography²³ using data from the first infection and two dose vaccinated groups (see Supplementary Table 3 for sera included in the calculation of the antigenic map). In Fig. 2, we show our previously described map after addition of three BA.5/CK.2.1.1 sera and the panel of new variants⁷. In the antigenic map, colored circles indicate the position of the analyzed virus variants, and squares or triangles the position of the single variant exposure sera. Map proximity of viruses indicates a similar neutralization phenotype, and hence map distance reflects phenotypically distinct antigenic relationships. The positions of pre-Omicron and early Omicron variants BA.1, BA.2, and BA.5 did not change compared to our previous map⁷. Newly emerged Omicron variants clustered in the map area between and around BA.1 and BA.5, however also extended antigenic space substantially further from pre-Omicron variants. Most BA.5 variants are located close to our initial BA.5 isolate (BA.5.3.2), with BQ.1.1 being furthest away. The spike sequence identical BA.5 variants (BA.5.3.2, BA.5.2.1, and BE.1.1 with only an additional Q1208H mutation in S2) are within one antigenic unit, a distance which can be attributed to measurement noise (Supplementary Fig. 10). XBF.3 shows the greatest escape from pre-Omicron variants, and XBB.1 and XBB.1.5.1 occupy distinct positions from each other in the map due to slightly more escape of XBB.1 from BA.5 and CK.2.1.1 sera (Fig. 1). BA.2.75, XBB, and XBF variants, which all contain a BA.2 derived spike sequence, were positioned rather distant from BA.2 and further away from BA.2 than for example delta from BA.2. This can be explained by the low level of neutralization of these variants by sera from BA.2 convalescent individuals (Fig. 1). Consequently, exclusion of BA.2 convalescent sera moved BA.2.75 variants closer to BA.2 (Supplementary Fig. 12). Given the impact of sera in certain map areas to resolve this region of antigenic space, a limitation of the map is the low number of sera located in the area covered by the newer Omicron variants, increasing their position uncertainty compared to earlier variants (Supplementary Figs. 9–11 and 16). However, human first infection sera from these variants are extremely difficult to obtain after over three years of global SARS-CoV-2 circulation and vaccination campaigns.

The antigenic map shows virus variants in colored circles and human sera as open squares in the color of their root variant or gray for vaccine sera and light blue for CK.2.1.1 sera. Triangles point in the direction of sera outside of the shown area (Supplement Fig. 6 for a non-zoomed in version). Each grid in the map corresponds to one twofold dilution of titers in the neutralization assay, making map distance a measure of antigenic similarity. Objects in the map are located relative to each other, x– and y-axis orientation is relative. Variants are labeled by pango lineage and colloquial name. For recent variants, spike substitutions are listed in the upper right of the map. The number of sera per cohort that was used to construct the map is shown in Supplementary Table 3.

As updated bivalent booster immunization contains either ancestral and BA.1 variant (BA.1 biv.) or ancestral and BA.4/5 variant (BA.4/5 biv.), we next investigated the level of cross-neutralizing antibodies in individuals who received three doses of ancestral virus vaccine followed by a fourth dose of one of the two bivalent boosters. We therefore collected plasma samples from individuals after a bivalent booster with or without previous infection history (Supplementary Table 1). In a first step, we analyzed antibody titers against the viral nucleocapsid (N) to detect previous infections. All four study participants with BA.4/5 biv. booster with a known history of infection (1 likely with alpha and 3 with BA.2 Omicron variant) were positive for N antibodies. Additionally, 16 of the participants without known infection history (5 in the BA.1 biv. boosted and 11 in the BA.4/5 biv. boosted group) were positive for N antibodies indicating a previous undetected infection (Supplementary Fig. 24 and Supplementary Table 1). Consequently, for these individuals the infecting variant is unfortunately unknown. Samples were grouped according to N antibody results in individuals with (BA.1 biv./N⁺ and BA.4/5 biv./N⁺) or without (BA.1 biv./N^– and BA.4/5 biv./N^–) previous infection. The interval between the last vaccine dose and blood collection had been approximately 1 month longer for the BA.1 biv. boosted groups compared to the BA.4/5 biv. boosted groups (Supplementary Table 1), which could influence overall titers of neutralizing antibodies. Therefore, we first plotted titers against D614G, BA.1, and BA.5 across all four groups over time (Supplementary Fig. 25). Although neutralizing antibody titers tended to be lower for samples collected longer after immunization, no clear correlation was observed.

We further analyzed neutralizing antibody titers against a broader panel of variants, i.e., D614G, beta, delta, BA.1, BA.2, CB.1, BR.3, CH.1.1, BA.5 (BA.5.3.2), BF.7, BQ.1.3, BQ.1.1, BQ.1.18, XBB.1, XBB.1.5.1, and XBF.3 variants (Fig. 3 and Supplementary Fig. 26). For all bivalently boosted groups titers of neutralizing antibodies were in general higher compared to single exposure or three dose vaccinated cohorts analyzed in Fig. 1. Titers were especially high for pre-Omicron and early Omicron variants, but dropped against BA.2.75, BQ, XBB, and XBF variants. Interestingly, single individuals exhibited high neutralization titers against these variants as well. However, mean titers against these variants dropped ~4-fold or more against the reference variants D614G, BA.1, and BA.5 (Supplementary Fig. 26). For hybrid immune individuals, this drop was less pronounced and most individuals neutralized all analyzed variants. We next constructed antibody landscapes to better compare neutralization profiles between individuals with and without N antibodies (see Fig. 4 GMT landscapes and Supplementary Fig. 27 for individual landscapes). The different intervals between booster dose and blood collection again limited direct comparison between BA.1 biv. and BA.4/5 biv. boosted individuals. However, a comparison between N⁺ and N⁻ shows that for both boosters the hybrid immunity landscapes were higher and flatter than the landscapes from N negative individuals indicating broader neutralization. The difference between hybrid and vaccine-only immunity was more pronounced in the BA.1 biv. groups. Timing between last vaccination and blood collection, which has been slightly longer for the BA.1 biv. groups compared to the BA.4/5 biv. groups could contribute to this difference.

Plasma was collected from individuals that received three doses of ancestral variant vaccine followed by a forth dose of either bivalent ancestral + BA.1 or ancestral + BA.4/5 vaccine. Antibodies against SARS-CoV-2 nucleocapsid (N) were determined using ELISA and samples were grouped accordingly: ancestral + BA.1 boost without detectable N antibodies (BA.1 biv./N⁻), n = 12; ancestral + BA.1 boost with positive N ELISA (BA.1 biv./N⁺), n = 5; ancestral + BA.4/5 boost without detectable N antibodies (BA.4/5 biv./N⁻), n = 16; ancestral + BA.4/5 boost with positive N ELISA (BA.4/5 biv./N⁺), n = 15. Titers of neutralizing antibodies against indicated variants are shown for individual patients as symbols connected by lines. Mean titers are shown as bars. Titers below 16 were treated as negative (dotted line), and titers below 1 were set to 1. IC₅₀ titer = 50% neutralization titer.

The map shown in Fig. 2 was used as a base map to construct antibody landscapes. Geometric mean titers (GMT) against each variant are shown on the z-axis above the corresponding position in the map. To construct landscapes, a continuous surface is fitted through these titers per individual. The geometric mean of these individual landscape fits is shown for a ancestral + BA.1 boost without detectable nucleocapsid (N) antibodies (BA.1 biv/N⁻, light red) and ancestral + BA.1 boost with positive N ELISA (BA.1 biv./N⁺, dark red) and b ancestral + BA.4/5 boost without detectable N antibodies (BA.4/5 biv./N⁻, light blue) and ancestral + BA.4/5 boost with positive N ELISA (BA.4/5 biv./N⁺, dark blue). The dots above each variant correspond to the GMT, which was calculated using the titertools R package²⁵.

Considering the almost complete escape from single exposure sera of the BQ, XBF, and BA.2.75 variants, their position in antigenic space could be even further away from pre-Omicron and early Omicron variants than in the current map^8,24. As many titers against these variants were below the LOD (limit of detection), the actual titer difference is censored by the LOD. Consequently, a variant may be positioned at a distance corresponding to the fold drop from maximum titer to the LOD, however, it cannot accurately be estimated how much further away the variant actually is (Supplementary Figs. 10C and 16). This could only be resolved by increasing the resolution in this area of antigenic space through addition of first exposure sera from these variants, which are challenging to obtain given the current situation of population immunity. To test the impact of the underestimation of antigenic escape of the recent Omicron variants on the antibody landscapes, we constructed landscapes where the new variants were not included in the fitting procedure (Supplementary Fig. 28). All landscapes with fitting only pre-Omicron and early BA.1, BA.2 and BA.5 Omicron variants were flatter than when fitting all variants, and neutralization breadth differed only based on N antibody status but not on received booster. This observation suggests that some of the newer Omicron variants are indeed further out in antigenic space compared to their position in our current antigenic map.