Imprinted SARS-CoV-2 humoral immunity induces convergent Omicron RBD evolution

Recently, SARS-CoV-2 Omicron sublineages, such as BA.2.30, BA.2.75.2, BQ.1.1, XBB( a recombinant of BJ.1 and BM.1.1.1), have spread widely around the world with growth advantages over BA.5. Interestingly, despite their divergent evolutionary courses, these variants appear to harbor mutations on the same RBD sites, such as R346, K356, K444, V445, G446, N450, L452, N460, F486, F490, R493, and S494. We wonder to what extent these variants evade current herd immunity background and antibody-based drugs, the driving force behind these convergent mutations on RBD hotspots, and what the mutational convergence would lead to.

Firstly, it is observed that most therapeutic neutralizing antibody (NAbs) drugs have been heavily compromised by these variants (Fig. 1a). COV2-2196+COV2-2130 (Evusheld) is escaped or highly undermined by variants containing F486, R346, and K444-G446 mutations. LY-CoV1404 (Bebtelovimab) is vulnerable to K444N/T and the combination of K444M/G446S or V445P/G446S, evaded by some Omicron subvariants (Fig. 1a). SA55 is the only NAb with high potency against all tested Omicron subvariants. These convergent variants also escape the neutralization of plasma samples from individuals who had received 3 doses of CoronaVac with or without BA.1/BA.2/BA.5 post-vaccination breakthrough infection. BJ.1/BM.1.1.1 recombinant strain XBB and XBB.1 are the most humoral immune evasive among the strains tested (Fig. 1c-f). These convergent variants, while exhibiting high capability of immune evasion, also maintain sufficient ACE2-binding efficiency (Fig. 1b).

Secondly, we examined the antibody repertoires of individuals who had Omicron BA.2 and BA.5 breakthrough infection after receiving triple wildtype-based inactivated vaccines (CoronaVac). Similar to our previous reports on BA.1 breakthrough infection, immune imprinting, or so-called “original antigenic sin”, is also observed in BA.2 and BA.5 breakthrough infections. The majority of BA.2/BA.5 RBD-binding B cells also bind to WT RBD, while only a small proportion of the BA.2 RBD-binding B cells from BA.2 infection without previous vaccination are cross-reactive to WT RBD (Fig. 2a-c). Besides, cross-reactive mAbs demonstrate higher somatic hypermutation (SHM) rates (Fig. 2d), indicating that these BA.2/BA.5 RBD binding antibodies are most likely recalled from previous vaccination-induced immune memory.

In addition, we determined the escape mutation profiles of these antibodies by high-throughput deep mutational scanning (DMS) and obtained the DMS profiles of 3051 SARS-CoV-2 RBD-targeting mAbs. We co-embedded all mAbs using multidimensional scaling (MDS) based on their DMS results, followed by t-distributed stochastic neighbor embedding (t-SNE) for visualization and used KNN-based classification to determine the epitope groups of mAbs (Fig. 2e). Compared to WT infection or vaccination, BA.1/BA.2/BA.5 breakthrough infection elicited increased proportion of group E2.2, E3, and F1 mAbs (Fig. 2f), which exhibit weak neutralizing capacity and do not compete with ACE2 (Fig. 2g-h). Overall, the proportion and diversity of neutralizing antibody epitopes are reduced in Omicron breakthrough infection, especially in BA.5 breakthrough infection.

To examine the consequence of the immune imprinting and the reduction of effective NAb epitopes, we estimated the impact of mutations on the efficacy of humoral immunity by aggregating the DMS profiles of mAbs. In addition to the antibody escaping scores determined by DMS, ACE2 binding, RBD expression and neutralizing activity against the evolving strain are also weighted. For BA.2-elicited antibodies, various peaks of escape score were identified, while only R346T/S and K444E/Q/N/T/M peaks were observed when using BA.5 elicited antibodies (Fig. 3a). The concentrated immune pressure strikingly reflects the reduced diversity of NAbs elicited by BA.5 breakthrough infection due to immune imprinting, indicating that immune imprinting leads to concentrated immune pressure which in turn induces the convergent RBD evolution.

Furthermore, we combined DMS of antibodies from WT/BA.1/BA.2/BA.5 immune background to rationalize and predict the evolutionary trends of BA.2.75 and BA.5. It was observed that R346T/S, K356T, N417Y/H/I/T, K444E/Q/N/T/M, V445D/G/A, N450T/D/K/S, L452R, I468N, A484P, F486S/V, and F490S/Y are the most significant sites for BA.2.75, which are in line with recent mutation hotspots of BA.2.75. As for BA.5, the most significant sites are more concentrated on R346, K444-G446, and N450, followed by K356, N417, L455, N460, and A484, consistent with emerging BA.5 variants (Fig. 3b).

Based on the observed and predicted convergent hotspots on BA.2.75 and BA.5 RBD, we move on to predict the destination of convergent evolution. We constructed pseudoviruses with the most common convergent mutations from variants with growth advantage and their serial combinations. The neutralization of BA.2 neutralizing antibodies is generally in line with DMS profiles. R493Q and N417T are not major contributors to antibody evasion, but R493Q significantly benefits ACE2 binding; V445A and K444N caused slightly, and F486S/V caused significantly reduced ACE2-binding capability, consistent with the measurement of emerging subvariants (Fig 3c).  Notably, BA.5 with six additional mutations could evade the vast majority of RBD NAbs, while maintaining high hACE2-binding capability, despite the reduction caused by K444N/T and F486V. BQ.1.1, XBB, and CH.1.1 could also escape the majority of RBD-targeting NAbs.

We also evaluated the neutralization potency and breadth of Omicron-specific mAbs and NTD-targeting antibodies. These specific antibodies exhibit strong neutralizing potency against corresponding strains but poor neutralizing breadth against other strains. Importantly, most of these antibodies are escaped by BQ.1.1 and XBB (Fig. 3d). NTD-targeting NAbs exhibit moderate to low neutralization activity. Group α NTD NAbs are vulnerable to K147E and W152R carried by BA.2.75*, and Y144del harbored by BJ.1/XBB. Group δ is sensitive to V83A (XBB) and R237T. F157L, I210V and G257S of BA.2.75 did not heavily affect the tested mAbs. Y144del and V83A carried by BJ.1 or XBB would escape NTD antibodies, according to the above results. XBB.1 escaped all NTD-targeting NAbs tested (Fig. 3e).

Moreover, we evaluated the escaping activity of pseudoviruses constructed based on BA.2.75 and BA.5 harboring selected RBD and NTD escaping convergent mutations (Fig. 4a). These mutants escaped the majority of tested NAb drugs, except SA55 (Fig. 4b), while maintaining considerable ACE2 binding ability (Fig. 4c). The neutralization of plasma from vaccinees and convalescents after BA.1, BA.2, and even BA.5 breakthrough infection are highly impaired or evaded by these designed variants. The addition of L452R, K444M, R346T, F486V, and K356T RBD mutations to BA.2.75 would decrease the plasma neutralization titers of most vaccinees and the convalescents to the detection limit. Also, extra K444N, R346T, N460K, and K356T mutations adding to BA.5 would evade plasma samples from vaccinees and BA.1 convalescents. The plasma from BA.2/BA.5 convalescents can tolerate more mutations based on BA.5, and extra NTD mutations such as K147E and W152R are needed to completely eliminate their neutralization. Some variants similar to our predicted and constructed variants have already emerged while we performed the experiments, which again validates our prediction model.

To sum up, by characterizing the escape mutation profiles and neutralization activity of 3051 mAbs stimulated from different immune histories, we demonstrated that the diversity of antibody repertoire after BA.1/BA.2/BA.5 breakthrough infection shrank due to immune imprinting, thus concentrating the immune pressure and promoting convergent evolution, which will in turn further narrow down the NAb repertoire.