Dual strategies for durable antibodies against SARS-CoV-2: Immune trajectories and viral fitness constraints

The continuous evolution of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) presents ongoing challenges and risks to public health, as it renders most reported monoclonal antibodies (mAbs) ineffective due to immune escape [1,2]. Unlike conventional strategies that rely on conserved epitopes across known viral subtypes, the extraordinary pace and unpredictability of SARS-CoV-2 mutations have progressively eroded these epitopes, thereby destabilizing the foundations of traditional broadly neutralizing antibody (bnAb) development [3]. By February 2024, a total of 12,916 mAbs against the spike protein has been reported, including 7116 human mAbs targeting the receptor-binding domain (RBD) of SARS-CoV-2. However, the majority has already lost activity due to immune escape [4]. We selected a panel of eight structurally characterized broadly neutralizing mAbs (S2K146, BD55-1205, VIR-7229, SA55, S309, L4.65, BIOLS56, and S2H97), along with one bispecific antibody (bsAb), Dia-19, all of which have demonstrated sustained breadth of neutralization potency from pre-Omicron strains to XBB [[5], [6], [7], [8], [9]]. We evaluated their activity against emerging variants/subvariants in pseudovirus neutralization assays, including four early SARS-CoV-2 Omicron subvariants (BA.2, XBB, XBB.1.5, and HK.3) and twelve post-JN.1 SARS-CoV-2 subvariants (JN.1, JN.1.7, JN.1.16.1, JN.1.18, KP.1.1, KP.2, KP.3, KS.1, LB.1, KP.3.1.1, XEC, and BA.3.2) (detailed methods in Supplementary data). Based on antibody escape profiles and development history, we summarized two strategies for bnAb development that address the unprecedented challenge of SARS-CoV-2’s rapid and stochastic evolution.

We mapped the mutation sites of 16 Omicron subvariants across eight RBD antibody classes to evaluate their impacts [6,10]. Our analysis revealed that nearly all epitopes were disrupted by mutations, with the notable exception of RBD-8 (Figure 1A). However, the apparent conservation of RBD-8 results from its cryptic nature—low immunogenic exposure reduces immune recognition and selective pressure—thus limiting the neutralizing potency of antibodies targeting this region (Figure 1B) [6,8]. Taken together, these results demonstrate the progressive disappearance of conserved neutralizing epitopes that previously served as the cornerstone of bnAb development.

Figure 1.

Epitope mapping and neutralization activity of RBD-targeting antibodies against emerging SARS-CoV-2 variants/subvariants

(A) Binding sites of RBD-1 to RBD-8 class antibodies (three representative antibodies per class) on the SARS-CoV-2 RBD, indicated by distinct colors. Mutations identified in emerging SARS-CoV-2 variants/subvariants are shown above the RBD, with the first, second, and third substitutions colored in blue, pink, and navy blue, respectively; unmutated residues in a given variant are shown in light blue, while residues fully conserved across all analyzed variants/subvariants remain white. (B) The neutralizing activity (IC₅₀) of seven monoclonal antibodies (mAbs) and one bispecific antibody (bsAb, Dia-19) against 16 SARS-CoV-2 Omicron subvariants and SARS-CoV-1. Each data point represents one of the two technical replicates from duplicate experiments. (C–F) Radar charts showing pseudovirus neutralization of BD55-1205 (C), VIR-7229 (D), SA55 (E), and Dia-19 (F) against 16 SARS-CoV-2 Omicron subvariants, colored in violet, green, orange, and red, respectively. Four antibodies are shown to highlight those retaining neutralization activity and to further visualize their potency: BD55-1205, VIR-7229, SA55, and Dia-19. Radar plots illustrate the reciprocal IC₅₀ values of the indicated antibodies. Lower IC₅₀ values correspond to higher neutralization activity; taking the reciprocal allows the plotted area to scale directly with antibody potency, facilitating intuitive comparison across antibodies. A larger covered area thus corresponds to lower IC₅₀ values, indicating stronger neutralization potency. (G) Distribution of mutation counts across the variable regions (V regions) of 7116 SARS-CoV-2 RBD-targeting mAbs obtained from the CoV-AbDab database. Heavy-chain variable regions were aligned to their corresponding human germline immunoglobulin genes (IGHV) using the hAb-Convergent tool, and the number of mutated residues was quantified. The mutation counts of the eight representative mAbs investigated in this study were indicated on the plot. Abbreviations: RBD, receptor-binding domain; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; mAbs, monoclonal antibodies.

Pseudovirus neutralization consistently positioned S309, L4.65, and S2K146 within a mutation-accumulation escape trajectory, where successive epitope changes culminated in near-complete loss of activity against variants emerging after JN.1 (Figure 1B). Two RBD-8 mAbs, S2H97 and BIOLS56, retained intact epitopes but nevertheless showed very limited neutralizing activity (Figures 1B and S1A–S1B). In contrast, BD55-1205, VIR-7229, SA55, and Dia-19 exhibited notable resistance to escape despite targeting highly exposed and mutation-prone epitopes. Radar plots further illustrated their broad protection against recent variants/subvariants (Figure 1B–1F).

Our antibody binding footprints analysis reveals that the S309 (developed as sotrovimab) accumulated mutations at 4 of 18 contact residues, and lost activity against JN.1, although it retains weak efficacy against other variants/subvariants (Figures 1B and S1C). L4.65, a component of the bsAb Dia-19, escaped after JN.1, with mutations in 6 of 18 residues (Figure S1D). The angiotensin-converting enzyme 2 (ACE2)-mimicking mAb S2K146 exhibited escape after JN.1, with mutations in 11 of 23 residues (Figure S1E). By contrast, BD55-1205 (RBD-1) exhibited balanced activity despite mutations in 9 of 24 contact residues (Figure S1F and S1I). VIR-7229 (RBD-2) and SA55 (RBD-3) displayed similar features, with 9 of 23 and 8 of 21 mutated residues, respectively (Figure S1G–S1I).

BD55-1205 and SA55 preserved nanomolar potency against all emerging SARS-CoV-2 variants/subvariants. VIR-7229 also demonstrated strong neutralization overall, with remarkable reduction observed against KP.3.1.1 and XEC (137-fold and 108-fold relative to JN.1), yet maintaining functional activity (Figures 1B–1E and S2). The finding showed that these mAbs that targeted highly mutated epitopes could still preserve primary binding interactions. Importantly, these resilient mAbs showed heavy-chain variable region somatic hypermutation (11 to 13 mutations), which was higher than the average mutations (7 to 8) across 7116 reported SARS-CoV-2 RBD-targeting mAbs. This strongly suggests that a robust evolutionary process during immune adaptation underlies their durability (Figure 1G, Tables S1–S2) [4,11], and highlights the necessity of leveraging antibody evolution to enhance the success rate of discovering bnAbs, leading to our first strategy—immune trajectory strategy.

This immune trajectory strategy seeks to increase bnAbs discovery rates by leveraging immune repertoires shaped by divergent exposures. For example, convalescent SARS-CoV-1 donors or individuals with breakthrough infections after ancestral SARS-CoV-2 infection often generate broadly neutralizing antibodies. Indeed, S309 was isolated from SARS-CoV-1 convalescent patients, SA55 was isolated from SARS-CoV-1 convalescent patients vaccinated with SARS-CoV-2, and both BD55-1205 and VIR-7229 (further engineered by affinity maturation) were isolated from a breakthrough infection [5,7,9]. This highlights the necessity of leveraging the antibody evolution process to enhance bnAb discovery.

The second strategy is derived from the bsAb Dia-19, which exhibited exceptional neutralizing activity across all emerging SARS-CoV-2 variants/subvariants [6,8]. As a single-chain diabody-Fc constructed from L4.65 (RBD-5) and BIOLS56 (RBD-8) (Figure S1I), Dia-19 elucidated a well-characterized synergistic rescue mechanism [8]. Most intriguingly, when either or both parental mAbs were evaded by viral mutations, Dia-19 demonstrated synergistic rescue of neutralizing activity, restoring its potency to nanomolar level (Figure 1B and 1F). Remarkably, epitope footprint analysis showed mutations in 6 out of 18 contact residues for the L4.65 component (Figure S1D). Since the L4.65 lost activity against JN.1, JN.1.7, JN.1.18, KP.1.1, LB.1, KP.3.1.1, and XEC, this implies that unlike the SA55, BD55-1205, and VIR-7229, the resistance of L4.65 in Dia-19 derives from synergy (Figure 1B). Based on this result, we propose the second rational design strategy targeting viral characteristics—viral fitness-constraint strategy.

This viral fitness-constraint strategy exploits highly constrained aspects of viral biology. Structurally, spike-mediated entry requires conformational rearrangements of a homotrimeric fusion protein, creating exploitable constraints [12]. Dia-19 leverages this by mechanically cross-linking spike trimers, forcing exposure of the cryptic RBD-8 epitope, thereby transforming structural conservation into therapeutic opportunity [8]. Functionally, the cost of altering receptor-binding is substantially higher than immune escape, making receptor-mimicking antibodies a logical approach. The key challenge is enabling receptor-mimicking antibodies to reproduce the six-loop binding interface precisely as even minor deviations accelerate immune escape. For example, the ACE2-mimicking antibody S2K146 targets the essential receptor-binding interface but has already been compromised by recent mutations (Figure 1B).

Overall, the emergence and continuous evolution of SARS-CoV-2 challenge the conventional paradigm of antibody development, where therapeutic design consistently lags behind viral diversification. Built on the large body of RBD-targeting mAbs characterized worldwide, we examined the maturation trajectory of SARS-CoV-2 RBD-directed antibodies and observed that high-maturity bnAbs tolerate mutations better across the variants tested. Guided by this result, we propose two practical strategies: a discovery strategy to identify high-value, higher-maturity candidates and an engineering strategy to upgrade low-maturity antibodies. As predictive modeling of mutations becomes more reliable, these approaches substantially increase the likelihood of identifying durable bnAbs. Finally, because bnAb breadth and durability depend on many factors, these strategies are guidelines rather than final answers. They are especially useful while prediction models of viral evolution are still being developed.

Author Contributions

Pengyue Gao: investigation, validation, data curation, visualization, writing – review & editing. Linh Nguyen: writing – review & editing. André Costa Lobato: data curation. George Fu Gao: conceptualization, project administration, supervision, funding acquisition, writing – review & editing. Zhou Tong: supervision, methodology, investigation, writing – original draft, writing – review & editing.

Peer-review Statement

George Fu Gao is one of the Editors-in-Chief of hLife. Given the role of the Editors-in-Chief, he was not involved in the peer review of this article and did not have access to information about the peer review. Full responsibility for the editorial process for this article was delegated to the Section Editor.

Ethics Approval

There are no ethical concerns associated with this study.

Data Availability

All data supporting the findings of this study are available within this research paper and are available from the corresponding author upon request.

Declaration of Competing Interests

The authors declare no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article.