Molecular Insights into Striking Antibody Evasion of SARS-CoV-2\\ Omicron Variant

Zeng-Shuai Yan(阎增帅)$^{1}$, Yao Xu(徐耀)$^{1}$, Hong-Ming Ding(丁泓铭)$^{2\hyperlink{s}{*}}$, and Yu-Qiang Ma(马余强)$^{1\hyperlink{s}{*}}$

doi:10.1088/0256-307X/39/10/108701

⇦Chinese Physics Letters, 2022, Vol. 39, No. 10, Article code 108701⇨ Molecular Insights into Striking Antibody Evasion of SARS-CoV-2 Omicron Variant Zeng-Shuai Yan (阎增帅)1, Yao Xu (徐耀)1, Hong-Ming Ding (丁泓铭)2*, and Yu-Qiang Ma (马余强)1* Affiliations 1National Laboratory of Solid State Microstructures and Department of Physics, Collaborative Innovation Center of Advanced Microstructures, Nanjing University, Nanjing 210093, China 2Center for Soft Condensed Matter Physics and Interdisciplinary Research, School of Physical Science and Technology, Soochow University, Suzhou 215006, China Received 19 July 2022; accepted manuscript online 8 September 2022; published online 16 September 2022 ^*Corresponding authors. Email: myqiang@nju.edu.cn, dinghm@suda.edu.cn Citation Text: Yan Z S, Xu Y, Ding H M et al. 2022 Chin. Phys. Lett. 39 108701 Abstract The SARS-CoV-2 Omicron variant has become the dominant variant in the world. Uncovering the structural basis of altered immune response and enhanced transmission of Omicron is particularly important. Here, taking twenty-five antibodies from four groups as examples, we comprehensively reveal the underlying mechanism of how mutations in Omicron induces the weak neutralization by using molecular simulations. Overall, the binding strength of 68% antibodies is weakened in Omicron, much larger than that in Delta (40%). Specifically, the percentage of the weakened antibodies vary largely in different groups. Moreover, the mutation-induced repulsion is mainly responsive for the weak neutralization in AB/CD groups but does not take effect in EF group. Significantly, we demonstrate that the disappearance of hydrophobic interaction and salt bridges due to residue deletions contributes to the decreased binding energy in NTD group. This work provides unprecedented atomistic details for the distinct neutralization of WT/Delta/Omicron, which informs prospective efforts to design antibodies/vaccines against Omicron.

DOI:10.1088/0256-307X/39/10/108701 © 2022 Chinese Physics Society Article Text The spread of the coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has become a global health crisis. Since it is a single-stranded RNA virus and prone to high mutation rates,[1] there have been a few variants of the SARS-CoV-2 (e.g., Alpha, Beta, Gamma and Delta) since December 2019. Noticeably, the Omicron variant is the latest variant of concern (VOC) by the World Health Organization in November 2021.[2] Up to date, the confirmed cases caused by Omicron have been reported in more than 149 countries and regions.[3] Some studies indicated that the doubling time of Omicron was shorter than those of the Delta and wild-type (WT).[4,5] Although the underlying mechanism of increased transmission efficiency is still debatable,[6-9] the large number of the mutations in Omicron is believed to be highly related to its high transmissibility: Omicron has 30 single-point mutations, 3 deletion mutations and 1 insertion mutation in the spike protein, with 15 mutations in the receptor binding domain (RBD) and 8 mutations in the N-terminal domain (NTD). Extensive studies have been performed by the experimental and clinical researchers on its high transmissibility since the discovery of the Omicron.[10-12] The escape of the Omicron from the immune system, especially the resistance to neutralization by antibodies is the most likely reason.[13-18] For example, Cameroni et al. found that 38 out of a panel of 44 NTD- or RBD-specific antibodies lost neutralizing activity against Omicron.[19] Xie and co-authors tested the neutralizing capacity of 9 RBD-specific antibodies and found Omicron exhibiting more potent antibodies escape capacity than Delta and WT.[20] The striking antibody evasion of Omicron was also reported in other recent works.[21-24] Moreover, some experimental studies reported that the binding affinity of Omicron RBD to ACE2 was strengthened, which may also be another reason for its high transmissibility.[19,25,26] However, some studies observed the similar binding affinity (to ACE2) between the WT and Omicron,[27,28] indicating an uncertain relationship between the binding affinity (to ACE2) and the high transmissibility. More recently, Chan and co-authors found that Omicron had faster and enhanced viral replication efficiency in the human bronchus,[6] and the Omicron spikes were predominantly in the open conformation,[29] suggesting that Omicron has other intrinsic capacities of the higher transmissibility compared to the precedent lineages. Apart from the great progress in the experimental and clinical studies, there have been some simulation works that investigated the possible mechanism of the high transmissibility of Omicron.[30,31] For example, Lan et al. identified that the binding affinity of Omicron RBD to ACE2 was enhanced than that of the WT by using all-atom molecular dynamics (MD) simulations, which was mainly caused by four key mutations (S477N, G496S, Q498R and N501Y).[32] The enhanced binding strength of RBD-ACE2 in the Omicron was also reported by Lupala et al. using the similar simulation methods.[33] However, Daniel and co-authors indicated that the RBD-ACE2 binding affinity was weakened in the Omicron,[34] again indicating an uncertain relationship between the RBD-ACE2 binding and the high transmissibility. Moreover, the neutralizing ability of antibodies against Omicron variant was also investigated in some MD works. Wu et al. found that the binding strength of Etesevimab and BD-368-2 became weaker in the Omicron, whereas that of Bebtelovimab remained the same in the Omicron (as in the WT).[28] In addition, some antibodies in different groups such as CB6, RENG10933, S309 and S2X259 lost or reduced neutralizing capacity against Omicron, as indicated by another simulation work.[35] Notably, there are many groups of antibodies denoted by the binding epitopes, and each group contains a lot of different antibodies.[20] Although the weak neutralizing ability of some antibodies in Omicron has been reported before by simulation works, the consideration of only several antibodies cannot give convincible evidence for the immune escape of the Omicron and cannot reveal the relationship between the specific mutation and the weakened neutralization (i.e., which mutation causes the reduction in neutralization efficacy). In this work, for the purpose of providing a general and statistical insight into the effect of mutations on the antibody evasion in Omicron, we systematically investigated the interactions of twenty-five antibodies from different groups with the SARS-CoV-2 spike protein (RBD and NTD) in WT and Omicron by using the all-atom molecular dynamics simulation and free energy calculation. For the sake of comparison, the interactions of these antibodies with the SARS-CoV-2 Delta variant are also considered. The key mutations (that induce the antibody resistance) and the possible mechanism at the atomic level are well revealed through the detailed structural analysis and energy decomposition. Modeling and Methods. Firstly, we describe the structural preparation. The initial structures of the antibodies with the RBD or NTD of SARS-CoV-2 (WT) were obtained from Protein Data Bank (https://www.rcsb.org). The details of all antibodies were listed in Table S1 in the Supplementary Information. Based on the WT RBD structure, the RBD structures in Delta (K417N, L452R, T478K) and Omicron (G339D, S371L, S373P, S375F, K417N, N440K, G446S, S477N, T478K, E484A, Q493R, G496S, Q498R, N501Y, Y505H) were modeled. Similarly, the NTD structures in Delta (T19R, G142D, E156G, del157-158) and Omicron (A67V, del69-70, T95I, G142D, del143-145, N211I, del212, Ins214EPE) were also constructed based on the WT NTD structure. The substitution and deletion mutation of residues were performed using Pymol and MODELLER,[36] respectively. Secondly, we state the all-atom molecular dynamics (MD) simulation. The RBD- or NTD-antibody complex was initially placed in the center of a cubic box and the minimum distance from the surface of the box to the complex atoms was set as 1.5 nm. Then, each complex was solvated in TIP3P water[37] and NaCl was added to neutralize the system at 0.15 m. The system was energy-minimized by the steepest descent method until the convergence was reached. After the energy minimization, the system was gradually heated from 0 to 310 K in the NVT ensemble over a period of 500 ps and then performed in the NPT ensemble for 500 ps, where the heavy atoms of protein were restrained using 1000 kJ$\cdot$mol$^{-1}\cdot$nm$^{-2}$ harmonic constraints. The temperature was maintained at 310 K by the Velocity–Rescale thermostat with a time constant of 0.2 ps, and the pressure was kept at 1 bar by the berendsen barostat with a time constant of 2.0 ps.[38] Finally, 10 ns NPT with weak restrains (100 kJ$\cdot$mol$^{-1}\cdot$nm$^{-2}$) on the backbone atoms of the proteins were performed for each system. This treatment can prevent the structural drift to the incorrect protein structures and lead to the quick convergence in the interaction energy, which is commonly used in previous works.[39-43] All the MD simulations were performed using the GROMACS software package (version 2020.6)[44] with the Amber ff14sb force field.[45] It should be noted that the Amber ff14sb showed the best performance in predicting the protein-protein binding free energies among the four popular force fields (i.e., Amber ff14sb, Amber ff03, Charmm36, and OPLS) in a recent study.[46] The Lennard–Jones interaction was computed using a cutoff of 1.2 nm, and the particle-mesh Ewald (PME) method was used to treat the long-range electrostatic interactions.[47] The bonds involving hydrogen atoms were constrained by the LINCS algorithm.[48] Periodic boundary conditions were applied in all three directions. Thirdly, we report the free energy calculation. The binding free energy ($\Delta G_{\rm bind}$) between the RBD or NTD of SARS-COV-2 and the antibodies was calculated by \begin{align} \Delta G_{\rm bind}=\,&\Delta H-T\Delta S\notag\\ =\,&\Delta E_{\rm ele}+\Delta E_{\rm vdw}+\Delta G_{\rm PB}+\Delta G_{\rm SA}, \tag {1} \end{align} where $\Delta H$ was the enthalpy change (including the electrostatic energy ($\Delta E_{\rm ele}$), the van der Waals energy ($\Delta E_{\rm vdw}$), the polar term ($\Delta G_{\rm PB}$) and nonpolar term ($\Delta G_{\rm SA}$) of solvation energy), $-T\Delta S$ was the entropy change. There were 100 frames (with an interval of 10 ps in the last 1 ns) used for calculating the binding energy in each system with the screening molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) method.[40,46,49] All the MM/PBSA calculations were performed using the modified gmx_mmpbsa script (https://github.com/Jerkwin/gmxtools/). This script can call the APBS software (https://github.com/Electrostatics /apbs/) to solve the PB equation and obtain the polar term ($\Delta G_{\rm PB}$) and nonpolar term ($\Delta G_{\rm SA}$) of solvation energy. Also, there were 10000 frames (with an interval of 0.1 ps in the last 1 ns) used for calculating the entropy term with the interaction energy (IE) method,[50,51] and the code was written by ourselves. The experimental binding energy was estimated as $\Delta G=k_{\rm B}T\ln K_{\rm d}$, where $k_{\rm B}$ is the Boltzmann constant, $T$ is the temperature, and $K_{\rm d}$ is the equilibrium dissociation constant. Results. Firstly, we present general comparisons of the antibody neutralization among the WT, Delta and Omicron. In this work, twenty-five RBD-targeted and NTD-targeted antibodies were considered. According to the previous work by Xie and co-authors, the RBD-targeted antibodies can be further classified into six epitope groups (A–F).[20] Here, for the sake of simplicity, we divided the RBD-targeted antibodies into three groups (AB, CD, EF) since the neutralizing mechanisms between A/B, C/D, E/F are similar (Fig. 1). In total, there are four groups (i.e., AB, CD, EF, NTD) of antibodies in this work. Since the number of antibodies in group AB is much larger than other groups in the experiments, ten antibodies were chosen in group AB and five antibodies was chosen in other groups. We first calculated the binding free energy of the RBD- and NTD-antibodies in the cases of SARS-CoV-2 WT with the screening molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) method. As shown in Fig. S1 and Table S2 in the Supplementary Information, most of the absolute errors between the calculated free energies and the experimental ones were below 10 kcal$\cdot$mol$^{-1}$, with the mean absolute error (MAE) of about 5 kcal$\cdot$mol$^{-1}$. These results again indicate the good performance of the screening MM/PBSA on the binding free energy in SARS-CoV-2 systems.[40,52]

Fig. 1. Schematic illustration of the antibodies in different groups. The locations (a) and the binding interfaces (b) of the RBD-targeted antibodies in A–F groups with the spike protein at the open conformation. Medium purple surface on the RBD depicts the RBM, and the binding epitopes are shown as colored outlines. (c) The location of NTD-targeted antibodies binding with the spike protein at the closed conformation. The NTD-targeted antibody and NTD are colored in blue and gray, respectively.

Fig. 2. The comparison of the neutralizing ability of antibodies against SARS-CoV-2 WT, Delta and Omicron. (a) The calculated binding free energies of antibodies in AB, CD, EF and NTD groups in the cases of WT, Delta and Omicron. (b) The evasion percentage of antibodies in AB, CD, EF and NTD groups, and that of total antibodies.

We then calculated the binding free energy in Delta and Omicron as shown in Fig. 2(a). To make a general comparison between the WT, Delta and Omicron, here a criterion of 5 kcal$\cdot$mol$^{-1}$ in binding free energy difference was used to distinguish whether the antibody was escaped in the Omicron and Delta.[53-55] The use of 5 kcal$\cdot$mol$^{-1}$ as the criterion is due to the following two reasons: (1) As indicated by previous studies, the accuracy of the MM/PBSA method is about 5 kcal/mol.[56,57] (2) The 5 kcal/mol difference corresponds to the dissociation constant ($K_{\rm d}$) change with over 3 orders of magnitude in experiment, which is sufficient to distinguish the difference in binding affinity. As summarized in Fig. 2(b), 68% of the antibodies were escaped in Omicron. This trend is consistent with recent experimental results, in which 38 out of a panel of 44 antibodies lost neutralizing activity against Omicron.[19] On the contrary, 40% of the antibodies were escaped in Delta. In comparison, although the Delta can lead to the antibody evasion (compared to SARS-CoV-2 WT), its evasion ability was weaker than that of the Omicron, which also agrees with recent experiments.[20-22] Notably, the antibodies can bind to the different regions of the spike protein of SARS-CoV-2. Since the mutations of the Omicron vary largely among these regions, the antibody-evasion behavior of these different antibodies could be different. As shown in Fig. 2(b), the majority of the antibodies in AB (80%), CD (80%), NTD (80%) groups were escaped in Omicron, but few antibodies in EF group (20%) were found. Such similar results were also observed in the case of Delta, where the percentages of the evasion antibodies in AB, CD, EF, and NTD groups were 50%, 20%, 0%, and 80%, respectively. These results indicate that the mutations may have distinct impacts on the neutralization of antibodies with different epitopes. Therefore, next we made a detailed investigation on the effect of mutations in different groups. Secondly, we try to realize different molecular origins for the striking escape ability of Omicron from the antibodies in AB group. The energy decomposition of the mutations for the ten antibodies in AB group was first analyzed. It was found that the T478K, E484A, Q493R and Y505H mutations were mainly responsible for weakening the binding strength (Fig. 3), three of them (E484A, Q493R and Y505H) were unique in the Omicron variant and one mutation (T478K) was shared by the Omicron and Delta variant.

Fig. 3. The weakened neutralizing ability of SARS-CoV-2 Delta and Omicron by antibodies in AB group. [(a), (d), (g), (j)] The energy decomposition of key residues in WT, Delta and Omicron. [(b), (c)] The key interaction of residue 478 at the RBD-CT-P59 binding interface in WT and Omicron, respectively. [(e), (f)] The key interaction of residue 493 at the RBD-C1A-B3 binding interface in WT and Omicron, respectively. [(h), (i)] The key interaction of residue 484 at the RBD-AZD8895 binding interface in WT and Omicron, respectively. [(k), (l)] The key interaction of residue 505 at the RBD-S2H14 binding interface in WT and Omicron, respectively. Black dashed line, red solid line, green arrow and light blue shading indicate hydrogen bond, salt bridge, residues repulsion and hydrophobic interaction, respectively.

Figure 3(a) shows that the T478K mutation disfavored RBD-antibodies binding due to a positive energy contribution in both Delta and Omicron (while it was nearly zero in WT). For example, in CT-P59, there was a lysine (K54) in the antibody near the T478 in WT [Fig. 3(b)], but T478 was a charge-neutral residue so that its interaction with charged K54 was weak. When mutated to K478 in Omicron, there was an electrostatic repulsion between K478 and K54 [Fig. 3(c)], which weakened the binding strength. Likewise, this repulsion was also found in other RBD-antibody systems both in Delta and Omicron [Figs. S2(a)-S2(d)]. Moreover, such mutation-induced repulsion was also observed in Q493R mutation in Omicron and L452R in Delta [Figs. 3(d)–3(f), Figs. S2(e)–S2(h) and S3], which contributed to a positive binding energy due to the electrostatic repulsion between the residues carrying the same charge. Interestingly, we note that previous experiments also reported that T478K and Q493R participated in reducing the binding of the Omicron RBD to CT-P59, AZD8895 (COV2-2196) and other antibodies of AB group.[24] Figure 3(g) shows that the E484A contributed to a decreased binding energy in Omicron (it is negative in WT but becomes nearly zero in Omicron). As illustrated in Fig. 3(h), there was a salt bridge between E484 and R24 of antibody AZD8895, which enhanced RBD-AZD8895 binding in WT. While in Omicron, the critical salt bridge interaction was lost after E484A mutation [Fig. 3(i)]. The salt-bridge loss due to the E484A mutation was also observed in other antibodies such as S2H14 and C1A-B3 [Figs. S2(i)–S2(l)]. Moreover, the Y505H mutation also contributed to a decreased binding energy in Omicron [Fig. 3(j)]. However, the molecular origin was different. As illustrated in Fig. 3(k), there were some hydrophobic residues at the RBD-antibody interface such as Y33, Y105 and Y107 in S2H14, the hydrophobic interaction between these residues and Y505 of WT RBD enhanced binding of RBD to S2H14. The hydrophobic pocket could also be found in other antibodies such as DXP-604 and BDII-196 in WT [Figs. S2(m)–S2(o)]. Moreover, the hydrogen bonds such as Y505-E51 of S2H14 and Y505-Q90 and L91 of DXP-604 also enhanced the WT RBD-antibodies binding [Fig. 3(k) and Fig. S2(m)]. However, due to Y505H mutation, H505 in Omicron was a polar residue, and significantly decreased the hydrophobic interactions [Fig. 3(l) and Figs. S2(n)–S2(p)]. Meanwhile, the hydrogen bonds between RBD-S2H14 and DXP-604 also disappeared, which further weakened the RBD-antibodies binding [Fig. 3(l) and Fig. S2(n)]. Noticeably, Dejnirattisai et al. found that Y505H participated in reducing the neutralization of an antibody mAb222 in the experiment.[21] Apart from the above mutations, most of the other mutations did not cause the change of key interactions at the interface and even did not participate in the binding. Hence, these mutations (i.e., S373P, S477N, G496S) had no significant effect on the binding strength [Figs. S4(a)–S4(c)]. Notably, the N501Y could even enhance the binding of some antibodies to Omicron by contributing the hydrophobic interaction [Figs. S4(d)–S4(f)].

Fig. 4. The weakened neutralizing ability of SARS-CoV-2 Delta and Omicron by antibodies in CD group. [(a), (d), (g)] The energy decomposition of key residues in WT, Delta and Omicron. [(b), (c)] The key interaction of residue 478 at the RBD-LY-CoV555 binding interface in WT and Omicron, respectively. [(e), (f)] The key interaction of residue 493 at the RBD-COVOX-316 binding interface in WT and Omicron, respectively. [(h), (i)] The key interaction of residue 498 at the RBD-LY-CoV555 binding interface in WT and Omicron, respectively. Black dashed line and green arrow indicate hydrogen bond and residues repulsion, respectively.

Thirdly, we investigate the mutation-induced repulsion causing the striking escape ability of Omicron from the antibodies in CD group. The energy decomposition of the mutations for the five antibodies in CD group was then analyzed, where three residue mutations T478K, Q493R and Q498R were highlighted. The T478K contributed a positive energy in both Delta and Omicron in three antibodies [Fig. 4(a)]. Similar to group AB, the T478K also caused an electrostatic repulsion between the K478 in Omicron and R50 & R96 in LY-CoV555 [Figs. 4(b) and 4(c)]. Moreover, the repulsion between the K478 and positively charged residue of antibodies could also be found in other antibodies in both Delta and Omicron [Figs. S5(a)–S5(d)]. Similarly, the Q493R and Q498R mutations also weakened the RBD-antibody binding by contributing a positive energy due to the electrostatic repulsion [Figs. 4(d)–4(i) and Figs. S5(e)–S5(h)]. Specifically, there also existed the loss of a hydrogen bond between the Q493 and S103 of COVOX-316 due to the mutation, which further weakened the binding [Figs. 4(e)–4(f)]. Notably, the Q493R mutation preventing the neutralization through steric hindrance was also found in the experiments using LY-CoV555 or other antibody like REGN10933 (belong to CD group).[7,23,24] However, we found that, although T478K, Q493R and Q498R almost did not affect RBD-REGN10987 binding, its binding affinity was weakened significantly, which resulted from the N440K mutation. Interestingly, such mutation also caused an electrostatic repulsion [Figs. S6(a)–S6(c)], in agreement with a recent experiment report.[23] Apart from the above mutations, the L452R that is a unique mutation in Delta can weaken the binding strength of some antibodies due to the electrostatic repulsion [Figs. S6(d)–S6(f)]. The other mutations (i.e., S477N, G496S) did not affect the binding strength obviously [Figs. S6(g) and S6(h)]. Fourthly, we state the retained ability of the antibodies in EF group to neutralize the Omicron. Similar to the above two groups, we also analyzed the energy decomposition of the mutations for the five antibodies in group EF, where five mutations (i.e., G339D, S317L, S373P, S375F and N440K) may occur in the RBD-antibodies interface. However, as shown in Figs. 5(a)–5(e), there was no significant difference of the binding energy contributed by the five mutations between WT and Omicron. This result was also revealed by the structural analysis, namely these mutations did not involve in any key interaction like hydrogen bond, salt bridge and hydrophobic interaction both in WT and Omicron [Figs. 5(f)–5(i)]. Moreover, no mutations took place at the RBD-antibodies interface in Delta (Fig. S7). Therefore, the antibodies in the EF group may still neutralize the Omicron and Delta, which deserves to be further considered in the clinic use.

Fig. 5. The weakened neutralizing ability of SARS-CoV-2 Omicron by antibodies in group EF. (a)–(e) The energy decomposition of residues 339, 371, 373, 375 and 440 in WT and Omicron, respectively. [(f), (g)] The location of residues 371, 371 and 375 at the RBD-CR3022 binding interface in WT and Omicron, respectively. [(h), (i)] The location of residues 339 and 440 at the RBD-S309 binding interface in WT and Omicron, respectively.

Fifthly, we describe the deletion mutations causing the striking escape ability of Omicron from the antibodies in NTD group. Apart from the mutations, there are some deletions of the residues in the NTD of Omicron and Delta, which is slightly different from that in the RBD. By analyzing the energy decomposition of the mutated or deleted residues, we found that the deletion of the residues (i.e., 143–145 in Omicron and 157–158 in Delta) was mainly responsible for weakening the NTD-antibodies binding [Figs. 6(a) and 6(b)]. As illustrated in Fig. 6(c), the hydrophobic pocket composed of residues 143–145 in the NTD (of WT) enhanced the hydrophobic interactions with the hydrophobic and/or aromatic residues in the antibodies. However, the deletion of the residues in Omicron caused the loss of key hydrophobic interaction between the NTD and antibodies [Fig. 6(d) and Figs. S8(a)–S8(d)]. Moreover, the deletion mutation may also induce the local structural change in Omicron, which further caused the loss of the salt bridges (e.g., K147-E73, K150-D56) that formed in the WT [Figs. 6(c) and 6(d)]. Similarly, as shown in Fig. 6(e), the residues 157–158 in WT facilitate the binding through the hydrophobic interaction (e.g., F157-F53 in FC05) and the salt bridges (e.g., R158-D101/D102 in FC05). The deletion of the two residues caused the loss of the above key interactions [Fig. 6(f) and Figs. S8(e)–S8(h)] and thus weakened the binding strength. Apart from the above mutations, the other mutations (i.e., T19R, G142D in Delta and A67V, T95I in Omicron) did not involve in the change of key interactions and were even not in the epitopes, thus they had no impact on the antibody binding (Fig. S9). In general, due to the residue deletions, the neutralization ability of the antibodies in NTD groups was greatly decreased in both Omicron and Delta, which should be taken cautions in the clinic use.

Fig. 6. The weakened neutralizing ability of SARS-CoV-2 Delta and Omicron by antibodies in NTD group. [(a), (b)] The energy decomposition of residue 143–145 and residue 157–158 in WT. [(c), (d)] The key interactions at the NTD-FC05 binding interface contributed by residue 143–145 in WT and deletion of that in Omicron, respectively. [(e), (f)] The key interactions at the NTD-C12C9 binding interface contributed by residue 157–158 in WT and deletion of that in Delta, respectively. Black dashed line, red solid line and light blue shading indicate hydrogen bond, salt bridge and hydrophobic interaction, respectively.

Discussion. Taking twenty-five antibodies from different groups as examples, we showed here 68% and 40% of the binding free energies are weakened in the Omicron and the Delta, respectively, which are in agreement with the previous experimental results on their escape ability. Importantly, the escape behavior of different antibodies is largely varied; the percentages of the weakened antibodies to Omicron are 80%, 80%, 20%, and 80% of in AB, CD, EF, NTD groups. Moreover, combining the effort of the energy decomposition of mutations and structural analysis of the binding interfaces, we found that the mutations can cause the electrostatic repulsion, the loss of the salt bridge and hydrogen bond, and weakening the hydrophobic interaction in AB group; while the electrostatic repulsion is mainly responsible for weakening binding strength in CD group. By contrast, the neutralization ability of antibodies in EF group are less affected since most of the mutations are not involved in crucial interactions at the interfaces. Further, the deletion of the residues is the unique mutation in the NTD, which significantly decreases the hydrophobic interactions and salt bridges between the antibodies and NTD, and results in the weak binding both in Omicron and Delta. In general, this study provides unprecedented in silico data to support the striking antibody evasion of the Omicron, and also gives clear molecular origins for the mutation-induced-weak binding, which may shed some light on the clinic implications of specific antibodies against the Omicron. Acknowledgments. This work was supported by the National Natural Science Foundation of China (Grant Nos. 11874045, 12222506, and 12174184), and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (Grant No. KYCX22_0086). We are grateful to the High Performance Computing Center (HPCC) of Nanjing University for the numerical calculations. References Genome-wide mapping of SARS-CoV-2 RNA structures identifies therapeutically-relevant elementsThe Spread of SARS-CoV-2 Variant Omicron with a Doubling Time of 2.0–3.3 Days Can Be Explained by Immune EvasionOmicron SARS-CoV-2 variant: a new chapter in the COVID-19 pandemicSARS-CoV-2 Omicron variant replication in human bronchus and lung ex vivoAltered TMPRSS2 usage by SARS-CoV-2 Omicron impacts infectivity and fusogenicityDelta spike P681R mutation enhances SARS-CoV-2 fitness over Alpha variantSpread of Gamma (P.1) Sub-Lineages Carrying Spike Mutations Close to the Furin Cleavage Site and Deletions in the N-Terminal Domain Drives Ongoing Transmission of SARS-CoV-2 in Amazonas, BrazilSerial Intervals and Household Transmission of SARS-CoV-2 Omicron Variant, South Korea, 2021Omicron variant (B.1.1.529) of SARS-CoV-2: Mutation, infectivity, transmission, and vaccine resistanceOmicron variant of SARS‐CoV‐2: Genomics, transmissibility, and responses to current COVID‐19 vaccinesThe significant immune escape of pseudotyped SARS-CoV-2 variant OmicronIncreased immune escape of the new SARS-CoV-2 variant of concern OmicronParallel profiling of antigenicity alteration and immune escape of SARS-CoV-2 Omicron and other variantsSARS-CoV-2 Omicron strain exhibits potent capabilities for immune evasion and viral entranceActivity of convalescent and vaccine serum against SARS-CoV-2 OmicronReduced neutralisation of SARS-CoV-2 omicron B.1.1.529 variant by post-immunisation serumBroadly neutralizing antibodies overcome SARS-CoV-2 Omicron antigenic shiftOmicron escapes the majority of existing SARS-CoV-2 neutralizing antibodiesSARS-CoV-2 Omicron-B.1.1.529 leads to widespread escape from neutralizing antibody responsesConsiderable escape of SARS-CoV-2 Omicron to antibody neutralizationStriking antibody evasion manifested by the Omicron variant of SARS-CoV-2Structural basis of SARS-CoV-2 Omicron immune evasion and receptor engagementStructures of the Omicron spike trimer with ACE2 and an anti-Omicron antibodyCharacterization of SARS-CoV-2 Omicron spike RBD reveals significantly decreased stability, severe evasion of neutralizing-antibody recognition but unaffected engagement by decoy ACE2 modified for enhanced RBD bindingReceptor binding and complex structures of human ACE2 to spike RBD from omicron and delta SARS-CoV-2SARS-CoV-2 Omicron RBD shows weaker binding affinity than the currently dominant Delta variant to human ACE2Cryo-EM structure of a SARS-CoV-2 omicron spike protein ectodomainSARS-CoV-2 Omicron spike glycoprotein receptor binding domain exhibits super-binder ability with ACE2 but not convalescent monoclonal antibodyScanning the RBD-ACE2 molecular interactions in Omicron variantStructural insights into the SARS-CoV-2 Omicron RBD-ACE2 interactionMutations on RBD of SARS-CoV-2 Omicron variant result in stronger binding to human ACE2 receptorSimulation of the omicron variant of SARS-CoV-2 shows broad antibody escape, weakened ACE2 binding, and modest increase in furin bindingStructural insight into antibody evasion of SARS-CoV-2 omicron variantComparative Protein Structure Modeling Using MODELLERQuantum and statistical mechanical studies of liquids. 25. Solvation and conformation of methanol in waterMolecular dynamics with coupling to an external bathAn optimized MM/PBSA virtual screening approach applied to an HIV-1 gp41 fusion peptide inhibitorAccurate Evaluation on the Interactions of SARS-CoV-2 with Its Receptor ACE2 and Antibodies CR3022/CB6Molecular dynamics of ion transport through the open conformation of a bacterial voltage-gated sodium channelRefinement of protein structure homology models via long, all-atom molecular dynamics simulationsProtein Structure Refinement through Structure Selection and Averaging from Molecular Dynamics EnsemblesGROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputersff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SBImproving the Performance of MM/PBSA in Protein–Protein Interactions via the Screening Electrostatic EnergyA smooth particle mesh Ewald methodLINCS: A linear constraint solver for molecular simulationsEnd-Point Binding Free Energy Calculation with MM/PBSA and MM/GBSA: Strategies and Applications in Drug DesignInteraction entropy for protein-protein bindingInteraction Entropy: A New Paradigm for Highly Efficient and Reliable Computation of Protein–Ligand Binding Free EnergyInteraction of serum proteins with SARS-CoV-2 RBDThe structural and energetic basis for high selectivity in a high-affinity protein-protein interactionNeEMO: a method using residue interaction networks to improve prediction of protein stability upon mutationIn silico screening of potential anti–COVID-19 bioactive natural constituents from food sources by molecular dockingMM/GBSA and LIE estimates of host–guest affinities: dependence on charges and solvation modelLigand Affinities Predicted with the MM/PBSA Method: Dependence on the Simulation Method and the Force Field

[1]	Manfredonia I, Nithin C, Ponce-Salvatierra A, Ghosh P, Wirecki T K, Marinus T, Ogando N S, Snijder E J, van Hemert M J, and Bujnicki J M 2020 Nucl. Acids Res. 48 12436
[2]	https://www.who.int/news/item/26-11-2021-classification-of-omicron-(b.1.1.529)-sars-cov-2-variant-of-concern
[3]	https://www.who.int/publications/m/item/enhancing-readiness-for-omicron-(b.1.1.529)-technical-brief-and-priority-actions-for-member-states
[4]	Grabowski F, Kochańczyk M, and Lipniacki T 2022 Viruses 14 294
[5]	Karim S S A and Karim Q A 2021 Lancet 398 2126
[6]	Hui K P, Ho J C, Cheung M C, Ng K C, Ching R H, Lai K L, Kam T T, Gu H G, Sit K Y, and Hsin M K 2022 Nature 603 715
[7]	Meng B, Abdullahi A, Ferreira I A, Goonawardane N, Saito A, Kimura I, Yamasoba D, Gerber P P, Fatihi S, and Rathore S 2022 Nature 603 706
[8]	Liu Y, Liu J Y, Johnson B A, Xia H J, Ku Z Q, Schindewolf C, Widen S G, An Z Q, Weaver S C, and Menachery V D 2022 Cell Rep. 39 110829
[9]	Naveca F G, Nascimento V, Souza V, Corado A D L, Nascimento F, Silva G, Mejía M C, Brandão M J, Á C, and Duarte D 2022 Microbiol. Spectrum 10 e02366-21
[10]	Song J S, Lee J, Kim M, Jeong H S, Kim M S, Kim S G, Yoo H N, Lee J J, Lee H Y, and Lee S E 2022 Emerging Infect. Dis. 28 756
[11]	Ren S Y, Wang W B, Gao R D, and Zhou A M 2022 World J. Clin. Cases 10 1
[12]	Araf Y, Akter F, Tang Y D, Fatemi R, Parvez S A, Zheng C, and Hossain G 2022 J. Med. Virol. 94 1825
[13]	Zhang L, Li Q Q, Liang Z T, Li T, Liu S, Cui Q Q, Nie J H, Wu Q, Qu X W, and Huang W J 2022 Emerging Microbes Infect. 11 1
[14]	Hu J, Peng P, Cao X X, Wu K, Chen J, Wang K, Tang N, and Huang A L 2022 Cell. & Mol. Immunol. 19 293
[15]	Sun C, Kang Y F, Liu Y T, Kong X W, Xu H Q, Xiong D, Xie C, Liu Y H, Peng S, and Feng G K 2022 Signal Transduct. Target. Ther. 7 42
[16]	Zhang X T, Wu S J, Wu B L, Yang Q R, Chen A, Li Y Z, Zhang Y W, Pan T, Zhang H, and He X 2021 Signal Transduct. Target. Ther. 6 430
[17]	Carreño J M, Alshammary H, Tcheou J, Singh G, Raskin A J, Kawabata H, Sominsky L A, Clark J J, Adelsberg D C, and Bielak D A 2022 Nature 602 682
[18]	Dejnirattisai W, Shaw R H, Supasa P, Liu C, Stuart A S, Pollard A J, Liu X X, Lambe T, Crook D, and Stuart D I 2022 Lancet 399 234
[19]	Cameroni E, Bowen J E, Rosen L E, Saliba C, Zepeda S K, Culap K, Pinto D, VanBlargan L A, De Marco A, and di Iulio J 2022 Nature 602 664
[20]	Cao Y L, Wang J, Jian F C, Xiao T H, Song W L, Yisimayi A, Huang W J, Li Q G, Wang P, and An R 2022 Nature 602 657
[21]	Dejnirattisai W, Huo J D, Zhou D, Zahradník J, Supasa P, Liu C, Duyvesteyn H M, Ginn H M, Mentzer A J, and Tuekprakhon A 2022 Cell 185 467
[22]	Planas D, Saunders N, Maes P, Guivel-Benhassine F, Planchais C, Buchrieser J, Bolland W H, Porrot F, Staropoli I, and Lemoine F 2022 Nature 602 671
[23]	Liu L, Iketani S, Guo Y C, Chan J F W, Wang M, Liu L Y, Luo Y, Chu H, Huang Y M, and Nair M S 2022 Nature 602 676
[24]	McCallum M, Czudnochowski N, Rosen L E, Zepeda S K, Bowen J E, Walls A C, Hauser K, Joshi A, Stewart C, and Dillen J R 2022 Science 375 864
[25]	Yin W C, Xu Y W, Xu P Y, Cao X D, Wu C R, Gu C Y, He X H, Wang X X, Huang S J, and Yuan Q N 2022 Science 375 1048
[26]	Lin S, Chen Z M, Zhang X D, Wen A, Yuan X, Yu C Z, Yang J, He B, Cao Y, and Lu G 2022 Signal Transduct. Target. Ther. 7 56
[27]	Han P C, Li L J, Liu S, Wang Q S, Zhang D, Xu Z P, Han P, Li X M, Peng Q, and Su C 2022 Cell 185 630
[28]	Wu L Y, Zhou L P, Mo M X, Liu T T, Wu C K, Gong C Y, Lu K, Gong L K, Zhu W L, and Xu Z J 2022 Signal Transduct. Target. Ther. 7 8
[29]	Ye G, Liu B, and Li F 2022 Nat. Commun. 13 1214
[30]	Omotuyi I, Afolabi E, Oyinloye B, Fatumo S, Femi-Oyewo M, and Bogoro S 2022 Comput. Biol. Med. 142 105226
[31]	Rath S L, Padhi A K, and Mandal N 2022 Biochem. Biophys. Res. Commun. 592 18
[32]	Lan J, He X H, Ren Y F, Wang Z Y, Zhou H, Fan S L, Zhu C Y, Liu D S, Shao B, and Liu T Y 2022 Cell Res. 32 593
[33]	Lupala C S, Ye Y J, Chen H, Su X D, and Liu H G 2022 Biochem. Biophys. Res. Commun. 590 34
[34]	Jawaid M Z, Baidya A, Mahboubi-Ardakani R, Davis R L, and Cox D L 2021 bioRxiv:10.1101/2021.12.14.472704
[35]	Verma J and Subbarao N 2022 bioRxiv:2022.01.25.477671
[36]	Webb B and Sali A 2016 Curr. Protoc. Bioinf. 54 5.6.1
[37]	Jorgensen W L and Madura J D 1983 J. Am. Chem. Soc. 105 1407
[38]	Berendsen H J, Postma J V, Van Gunsteren W F, DiNola A, and Haak J R 1984 J. Chem. Phys. 81 3684
[39]	Venken T, Krnavek D, Münch J, Kirchhoff F, Henklein P, De Maeyer M, and Voet A 2011 Proteins: Struct. Funct. Bioinf. 79 3221
[40]	Ding H M, Yin Y W, Sheng Y J, and Ma Y Q 2021 Chin. Phys. Lett. 38 018701
[41]	Ulmschneider M B, Bagnéris C, McCusker E C, DeCaen P G, Delling M, Clapham D E, Ulmschneider J P, and Wallace B A 2013 Proc. Natl. Acad. Sci. USA 110 6364
[42]	Raval A, Piana S, Eastwood M P, Dror R O, and Shaw D E 2012 Proteins: Struct. Funct. Bioinf. 80 2071
[43]	Mirjalili V and Feig M 2013 J. Chem. Theory Comput. 9 1294
[44]	Abraham M J, Murtola T, Schulz R, Páll S, Smith J C, Hess B, and Lindahl E 2015 SoftwareX 1–2 19
[45]	Maier J A, Martinez C, Kasavajhala K, Wickstrom L, Hauser K E, and Simmerling C 2015 J. Chem. Theory Comput. 11 3696
[46]	Sheng Y J, Yin Y W, Ma Y Q, and Ding H M 2021 J. Chem. Inf. Model. 61 2454
[47]	Essmann U, Perera L, Berkowitz M L, Darden T, Lee H, and Pedersen L G 1995 J. Chem. Phys. 103 8577
[48]	Hess B, Bekker H, Berendsen H J, and Fraaije J G 1997 J. Comput. Chem. 18 1463
[49]	Wang E C, Sun H Y, Wang J M, Wang Z, Liu H, Zhang J Z, and Hou T J 2019 Chem. Rev. 119 9478
[50]	Sun Z X, Yan Y N, Yang M Y, and Zhang J Z 2017 J. Chem. Phys. 146 124124
[51]	Duan L L, Liu X, and Zhang J Z 2016 J. Am. Chem. Soc. 138 5722
[52]	Yin Y W, Sheng Y J, Wang M, Ma Y Q, and Ding H M 2021 Nanoscale 13 12865
[53]	Meenan N A, Sharma A, Fleishman S J, MacDonald C J, Morel B, Boetzel R, Moore G R, Baker D, and Kleanthous C 2010 Proc. Natl. Acad. Sci. USA 107 10080
[54]	Giollo M, Martin A J, Walsh I, Ferrari C, and Tosatto S C 2014 BMC Genomics 15 S7
[55]	Xu J, Gao L, Liang H, and Chen S D 2021 Nutrition 82 111049
[56]	Genheden S 2011 J. Comput.-Aided Mol. Des. 25 1085
[57]	Weis A, Katebzadeh K, Söderhjelm P, Nilsson I, and Ryde U 2006 J. Med. Chem. 49 6596