Distinct evolution of infection-enhancing and neutralizing epitopes in the spike protein of SARS-CoV-2 variants (from alpha to omicron) : a structural and molecular epidemiology study


 Infection-enhancing antibodies may limit the efficiency of Covid-19 vaccines. We analyzed the evolution ofneutralizing and facilitating epitopes in 1,860,489 SARS-CoV-2 genomes stored in the Los Alamos databasefrom June to November 2021. The structural dynamics of these epitopes was determined by molecular modelingof the spike protein on a representative panel of SARS-CoV-2 variants. D614, which belongs to an antibody-dependent-enhancement (ADE) epitope common to SARS-CoV-1 and SARS-CoV-2, has mutated to D614G in2020, which could explain why ADE has not been detected following mass vaccination. A second epitopelocated in the N-terminal domain (NTD), specific of SARS-CoV-2, is highly conserved among most variants. Incontrast, the neutralizing epitope of the NTD showed extensive variations in SARS-CoV-2 variants. The balancebetween facilitating and neutralizing antibodies is in favor of neutralization for the Wuhan strain, alpha and betavariants, but not for gamma, delta, lambda, and mu. The recently emerging omicron variant is atypic as itsmutational profiles affects both neutralization and ADE epitopes. Overall, our data reveal that the evolution ofSARS-CoV-2 has dramatically affected the ADE/neutralization balance. Future vaccines should consider thesefindings to design new formulations adapted to SARS-CoV-2 variants and lacking ADE epitopes in the spikeprotein.


Introduction
Cytotoxic T-cells and neutralizing antibodies play a key role in the control of viral infections, especially in the case of respiratory viruses [1,2]. However, virus-specific antibodies can also promote pathology, a phenomenon referred to as antibody-dependent enhancement (ADE) [3]. ADE of virus infection is generally due to virus-specific antibodies that enhance the entry of virus into host cells, and in some cases, virus replication in monocytes, dendritic cells and macrophages through antibody binding to Fc receptors [4]. In addition, alternative mechanisms of ADE involving the complement component C1q have been reported [5]. ADE has been observed in two typical situations: i) reinfection with a virus variant after primary infection with a different strain [6] or a cross-reactive virus [7], and ii) as the result of viral infection in vaccinated people [8]. The ADE phenomenon was initially discovered in flaviviruses in the late 1960's [9] and experimentally demonstrated in the early 1970's [10]. It concerns a broad range of viruses including dengue [11], Ebola [12], Zika [13], HIV [14], influenza [15], and various animal and human coronaviruses [16].
As early as in June 2020, at a time when Covid-19 vaccines had just entered clinical evaluation, Akiko Iwasaki and Yexin Yang from Yale University School of Medicine alerted that "ADE should be given full consideration in the safety evaluation of emerging candidate vaccines for SARS-CoV-2" [17]. A similar warning on vaccine safety due to potential risks of ADE was independently published by Shibo Jiang [18]. In contrast, several authors considered the risk to be null or minimal in the case of SARS-CoV-2 [19] [20] [21] [22].
However, several pieces of evidence strongly argue in favor of an ADE issue for SARS-CoV-2. i) ADE has been reported for animal coronaviruses such as feline infectious peritonitis virus [23]. In the most dramatic cases, kittens previously vaccinated with a recombinant virus containing the spike protein gene succumbed of early death after a coronavirus challenge [24].
ii) ADE epitopes were characterized in the spike protein of this feline coronavirus [25]. iii) ADE epitopes have also been found in human coronaviruses related to SARS-CoV-2, i.e. [28]. The case of SARS-CoV-1 is particularly interesting since its spike protein displays a linear ADE epitope, 597-LYQDVNC-603 (recognized by the monoclonal antibody 43-3-14) [26] that is fully conserved in the SARS- [32]. In this context, we recently reported that facilitating anti-spike antibodies targeting the NTD have a higher affinity for the delta variant than for the initial Wuhan strain. We also reported that the main neutralizing epitope of the NTD is almost lost in δ variants [31]. This finding is of critical importance since ADE infection of coronaviruses is known to be induced by the presence of sub-neutralizing levels of anti-spike antibodies [33]. Overall, our data suggested that the balance between neutralizing and facilitating antibodies may greatly differ according to the virus strain.

SARS-CoV-1 [26] and MERS-CoV [27]
In the present study, we analyzed a panel of representative SARS-CoV-2 variants including alpha, beta, gamma, delta, lambda, mu as well as the most recent South-Africa strains C.1.2 (with no attributed Greek letter at the time of submission) and omicron. We used multiple amino acid sequence alignment methods combined with structural and molecular modeling approaches to determine the variability of ADE and neutralizing epitopes and the impact of this variability on antibody-spike protein interactions. Our main objectives were i) to decipher the evolution of neutralizing and facilitating epitopes since the beginning of the Covid-19 pandemic, and ii) to predict for each SARS-CoV-2 variant which way the balance between neutralization and facilitation is tipping.  [39].

Description of two distinct ADE epitopes in SARS-CoV-2 spike protein
The mutational patterns and geographic origins of the SARS-CoV-2 variants analyzed in this study are summarized in Table 1 Clearly, the NTD is key to understand how SARS-CoV-2 initially interacts with the plasma membrane of host cells.
The first ADE epitope studied is the 611-617 motif with the original amino acid sequence LYQDVNC recognized by the 43-3-14 antibody [26]. This ADE epitope is common to human coronaviruses SARS-CoV-1 and SARS CoV-2. Interestingly, this epitope is centered on position 614 which is an aspartic acid residue in the original Wuhan strain but has rapidly evolved to the ultra-dominant D614G during the first months of 2020 [40]. The localization of this epitope on the spike protein (Wuhan strain) is shown in Figure 1A (epitope colored in yellow, except for D614 highlighted in red). It is well exposed on the protein surface so that it can be recognized by facilitating antibodies generated during previous coronavirus infections in humans, especially in geographic areas previously exposed to SARS-CoV-1. The second ADE epitope targeted by facilitating antibodies is divided in two parts (both colored in blue in patients [30]. Although the two parts of this ADE epitope seem to be spatially distant, both are close to a flexible 20-amino acid residue loop (621-640) that is unresolved in PDB files but was added by molecular modeling in the structures shown in Figure 1. It is interesting to note that this loop (highlighted in green) is ideally located to connect the NTD and the RBD, but also to provide a conformational link between both ADE epitopes ( Figure 1B).
Once the NTD is bound to the cell membrane of the host cell, a conformational change unmasks the RBD which becomes available for a functional interaction with a viral receptor, chiefly ACE2 [37]. This spatial reorganization leads to the open, fusion-compatible conformation of the trimeric spike protein [41]. In the Wuhan strain, the closed conformation of the trimer [42] is stabilized by a hydrogen bond between D614 of one subunit and T859 of its neighbor (respectively chains B and C in Figure 2A). The global spreading of the pandemic during the first months of 2020 has been associated with the breakthrough of the first SARS-CoV-2 variant with a unique mutation in spike protein, D614G. As shown in Figure 2B, this mutation induces the loss of the hydrogen bond that stabilized the closed conformation. Thus, we analyzed the status of this hydrogen bond in the complex between the facilitating 1052 antibody and the spike protein trimer. As shown in Figure 2C, the antibody has a long range conformational effect on both D614 and T859, which renders impossible the formation of this hydrogen bond. It is likely that the 621-640 loop, which conformationally connects the 1052 and the 611-617 epitopes, mediates this distal effect. In this respect, it is interesting to note that this facilitation can be induced by two distinct mechanisms: i) the constitute the three-dimensional site recognized by the neutralizing 4A8 antibody [43]. The localization of the neutralization epitope of the NTD at the virus/host cell interface is consistent with this high variability as it is submitted to a strong pressure of selection for SARS-CoV-2 variants. Conversely, the ADE epitope, which is on the lateral side of the NTD, is not facing the plasma membrane of the host cell and for this reason is not subjected to such a high selective pressure.
The frequency of amino acid sequence variations of the ADE and neutralizing epitopes was analyzed by specific queries of the Los Alamos database over the last six-month period (2021-06-01 to 2021-11-27) ( Table 2). All the epitopes listed in Figure 3 were analyzed in 1,860,489 genomes. The ADE epitope of the NTD is highly conserved (>98% for all segments) except for the 64-69 motif at position H69 (variation of 5.46% with 1 mutation), mostly reflecting the alpha variant [44]. The ADE epitope 611-617 displays 1 mutation in 98.70% of cases, consistent with the worldwide dominance of the D614G mutant [45]. The

Discussion
Vaccine strategies against viral diseases are confronted to the risk of antibody facilitation (ADE), especially when the strain used for the immunization protocol is distinct from circulating viruses [46]. In the past, ADE has been evidenced for a broad range of human RNA viruses including HIV, influenza, filoviruses, and coronaviruses [4]. Although ADE antibodies have been consistently characterized in the serum from Covid-19 convalescent patients [29] [30], the risk of ADE linked to vaccination with spike protein-based vectors (either mRNA or adenovirus) has not been considered as critical. As a matter of fact, it has been generally assumed that ADE antibodies exhibited SARS-CoV-2 infection enhancement in vitro but not in vivo [30]. However, a potential caveat of these studies is that SARS-CoV-2 variants have not been specifically assessed. Moreover, surprising higher incidence rates in vaccinated vs. unvaccinated individuals in the 0-14 days after the first dose were recently reported in long-term care facility residents and health-care workers, which resulted in significant negative vaccine efficiency estimates of -37% and -113%, respectively [47]. To which extent this apparent enhancement of SARS-CoV-2 infection is due (or not due) to an imbalance between vaccine-induced (and/or pre-existing) neutralizing and facilitating antibodies warrants further investigation. Moreover, a recent report revealed that there is no clear relationship between the percentage of fully vaccinated individuals and new Covid-19 cases in 68 countries including Israel, a pioneer in mass vaccination against SARS-CoV-2 [48]. Taken together, these observations suggested that ADE, or more specifically the ADE/neutralization balance, could pose a problem for Covid-19 vaccine strategies, especially during the outbreak of SARS-CoV-2 variants. Finally, it is worth noting that ADE has been suspected to increase the severity of Covid-19 symptoms in selected geographic areas [49].
The objective of the present study was thus to assess the potential risk of ADE in vaccinated individuals challenged with SARS-CoV-2 variants. To this end, we studied the amino acid sequence variability of ADE and neutralizing epitopes in the NTD and rod-like regions of the spike protein. Then we used our target-based molecular modeling strategy to interpret these data at the level of the three-dimensional structure of the spike proteins.
We focused our attention on two distinct ADE epitopes: one linear epitope common to SARS- Both epitopes are present on the spike protein generated by mRNA vaccines as the original formulas are based on the Wuhan strain [50]. Therefore, it is of high importance to determine whether these epitopes are still expressed and accessible on SARS-CoV-2 variants. The 611-617 epitope has probably escaped facilitating antibodies because the D614G variant has rapidly replaced the original strain [45]. Although in the initial study of the D614G mutation the authors mentioned the presence of D614 in a conserved ADE epitope, they did not comment further this important issue [40]. Our modeling approaches revealed a common molecular mechanism leading to enhanced infectivity for the D614G variant and for ADE antibodies with the Wuhan strain ( Figure 2). In both cases, the loss of a stabilizing hydrogen bond between amino acid residues 614 and 859 of two vicinal spike protein chains relaxes the trimer and facilitates the conformational change that unmasks the RBD. A major outcome of our study is the identification of the 621-640 loop, which is missing in PDB files, as the conformational transmitter that allows the 1052 antibody to induce distant effects on amino acid residue 614. In this respect, the enhancement of infection provided by this ADE antibody involves two distinct Fc-independent mechanisms: a long range conformational effect and a stabilization of the NTD bound to a lipid raft [31]. infected with alpha or beta strains from the ADE risk ( Figure 3). Nevertheless, these variants also showed significant variability of the neutralizing epitope, which could have decreased vaccine efficiency [54]. The situation is more dramatic for the delta variant. Indeed, several studies converged to alert on the potential risk of ADE when a delta SARS-CoV-2 variant infects a vaccinated individual [31] [32]. Our study confirms this possibility and further extends it to other circulating variants, including lambda and mu, for which the neutralization/facilitation balance is unfavorable. A useful approach to anticipate such ADE risk in face of any variant is to analyze both the ADE and neutralizing epitopes of the NTD, as developed in Figure 3. At first glance, one can determine the balance between neutralization and facilitation and assess the risks of virus escape, ADE and/or both. Our molecular modeling approaches confirmed that hot mutational spots in ADE and neutralizing epitopes of the NTD give reliable information on antibody recognition of the spike protein, allowing us to determine which way the balance between neutralization and facilitation is tipping.
We recently hypothesized that the delta variant is dominating because its electrostatic surface potential of the NTD region that faces the host cell membrane has evolved to a large        Mutations patterns in the NTD and rod-like regions of the SARS-CoV-2 spike protein were obtained from the GISAID database (https://www.gisaid.org/hcov19-variants). Deletions (∆) and insertions (+) are underscored. The frequency of mutations of each epitope sequence is calculated as the percentage of identity with the reference amino acid sequence of the SARS-CoV-2 spike protein (Wuhan strain). The most variable amino acid residues of each epitope are underlined. 1,860,489 sequences were analyzed from 2021-06-01 to 2021-11-27. The raw data were obtained from the Los Alamos website (https://cov.lanl.gov/content/sequence/ANALYZEALIGN/analyze_align.html).