Insights on SARS-CoV-2’s Mutations for Evading Human Antibodies: Sacrifice and Survival

Recent mutations on the receptor binding domain (RBD) of the SARS-CoV-2’s spike protein have been manifested as the major cause of the wide and rapid spread of the virus. Especially, the variant B.1.351 in South Africa with the hallmark of triple mutations (N501Y, K417N and E484K) is worrisome. Quickly after the outbreak of this new variant, several studies showed that both N501Y and E484K can enhance the binding between RBD and the human ACE2 receptor. However, the mutation K417N seems to be unfavorable because it removes one interfacial salt-bridge. So far, it is still not well understood why the K417N mutation is selected in the viral evolution. Here, we show that despite the loss in the binding affinity (1.48 kcal/mol) between RBD and ACE2 the K417N mutation abolishes a buried interfacial salt-bridge between RBD and the neutralizing antibody CB6 and thus substantially reduces their binding energy by 9.59 kcal/mol, facilitating the variants to efficiently elude CB6 (as well as many other antibodies). Thus, when proliferating from person to person the virus might have adapted to the human immune system through evasive mutations. Taking into account limited and relevant experimental works in the field, we show that our theoretical predictions are consistent with existing experimental findings. By harnessing the revealed molecular mechanism for variants, it becomes feasible to redesign therapeutic antibodies accordingly to make them more efficacious.


COVID-19. 3
Experimental studies have demonstrated that the angiotensin converting enzyme 2 (ACE2), a protein expressed on the surface of human cells in various organs, plays a crucial role in the viral infection of SARS-CoV-2. [4][5][6] As a prelude of SARS-CoV-2's cell entry, the receptor binding domain (RBD) of the spike glycoprotein (S-protein) 7 on the virion surface binds ACE2 on a host cell. Therefore, the S-protein of SARS-CoV-2 becomes a primary target of the host immune system and is used as the leading antigen for vaccine development. Despite being detected from three different continents, the fact that the three aforementioned SARS-CoV-2 variants share some of the same mutations at the S-protein's RBD suggests that these mutations might have conferred an evolutionary advantage to the virus. While various mutations as well as deletions outside the RBD of S-protein are also important for the enhanced fitness (of variants) for entering host cells, here we focus on mutations in RBD that harbors the binding site of ACE2 and the epitopes for neutralizing monoclonal antibodies (mAbs).
For the U.K. variant (with the N501Y mutation on RBD), experimental studies have demonstrated that the reproductive number which measures its infectiousness is about 0.4 to 0.7 higher than other strains of the virus 8 and determined recently it is unlikely to escape the BNT162b2-vaccine-mediated protection. 9 Our previous work 10 showed that this mutation in the U.K. variant can increase the RBD's binding affinity with ACE2, but has no obvious effect on mAbs. Besides the N501Y mutation, the South Africa and Brazil variants also contain the K417N and E484K mutations. Recent studies showed that the E484K mutation can not only enhance the RBD-ACE2 binding 11 but also help virus escape the therapeutically relevant mAbs. 12 However, reflected by the paucity of existing work the significance of the K417N mutation so far is still elusive, which prevents us from fully understanding these new variants' infection mechanism. As shown in Fig. 1a, in the wild-type RBD, K417 forms a salt-bridge with D30 in ACE2. Thus, the K417N mutation resulting in the abolishment of this favorable interfacial interaction (i.e. reducing the RBD-ACE2 binding affinity as also being verified in experiment 11 ) is highly unintuitive. Here, we are motivated to investigate the molecular mechanism of the K417N mutation, in order to unveil its key benefits for the virus to evolve through this path.
Complementary to experimental efforts, the all-atom molecular dynamics (MD) simulations with sophisticated and well calibrated force fields have been widely used to image nanoscale events and investigate the molecular mechanism of proteins. [13][14][15][16] In this work, we conducted a computational analysis on the K417N mutation in the South Africa variant using MD simulations with explicit solvent, aiming to gain a better understanding of its underlying molecular mechanism. Besides the RBD-ACE2 interaction, we also investigated the RBD's interaction with mAbs. In particularly, we explored the mAb CB6 that recognizes an epitope site in the RBD overlapping the binding site of ACE2, and investigated its binding competition with ACE2 over RBD. Our results may help provide invaluable insights on why K417N has been selected in the viral evolution and inspire a better design of more efficacious mAbs for treating COVID-19 patients infected with the new SARS-CoV-2 variants. calculated the interfacial contact areas for the complexes using the solvent accessible surface area method. 17 On average, the contact area is about 8.6 nm 2 between RBD and ACE2 while the one for the RBD-CB6 contact is larger and is about 10.3 nm 2 . During the entire simulation time, values of contact areas fluctuated around the mean, indicating that these two complex-structures were stable and equilibrated in the electrolyte.
By examining the simulation trajectories, we found that K417 indeed played an important role in stabilizing the RBD-ACE2 and RBD-CB6 complexes. In Fig. 2b, we highlight the time-dependent number N of heavy (or non-hydrogen) atoms in ACE2 or CB6 that were within 5 Å of K417 in RBD. Notably, K417 in RBD interacts with many more atoms in CB6 than in ACE2. The average saturated number N for the RBD-CB6 complex is about 23 atoms (orange line in Fig. 2b), while the one for the RBD-ACE2 complex is only about 6 atoms (blue line in Fig. 2b). Additionally, the number N for the RBD-ACE2 complex fluctuates much more than the one for the RBD-CB6 complex, which indicates that the interaction in the latter complex is more stable.
Among those residues (in ACE2 or CB6) in contact with K417 in RBD, we discovered two key salt-bridges. One is formed by K417 in RBD and D30 in ACE2 (as shown in Fig. 1a).
We illustrate in Fig. 2c that this salt-bridge is on the surface of the RBD-ACE2 complex.
Thus, this salt-bridge is exposed to water and constantly disrupted by polar water molecules, accounting for the observed fluctuations in N (blue line in Fig. 2b). The other one is formed by K417 in RBD and D104 in CB6 as shown in Fig. 2d. Remarkably, this salt-bridge is buried among the heterotrimer composed of RBD and two variable domains of the heavy chain V H and light chain V L of CB6. This salt-bridge buried inside the protein complex was very stable during the simulation, consistent with the nearly constant number N in Fig. 2b (orange line).
To further demonstrate the stability of these two salt-bridges, we define the distance D between the NZ atom in the lysine (K) and the CG atom in the aspartate (D), as shown in the inset of Fig. 3a. The histograms in Fig. 3a show the probability distributions of the By definition, binding free energy changes for RBD with either ACE2 or CB6 (due to the K417N mutation) can be obtained as ∆∆G = ∆G 2 − ∆G 1 (see Fig. 3). In practice, it is not easy to directly calculate ∆G 1 and ∆G 2 , which can be circumvented by computing ∆G A and ∆G B instead using the thermodynamic cycle (see Fig. 3). Therefore, ∆∆G = ∆G A − ∆G B .
Through the ensemble average, 18 ∆G A and ∆G B can be calculated theoretically (see Method 3d and 3e) , the K417N mutation yielded a free-energy change ∆G B of -32.08 kcal/mol.
In the bound state for the RBD-CB6 (Figs. 3b and 3c), the free energy change ∆G A =-22.49 kcal/mol. Taking all together, the obtained value of ∆∆G is 9.59 kcal/mol, suggesting that the K417N mutation significantly reduced the binding affinity between RBD and CB6.
Similarly, for the RBD-ACE2 complex, we have ∆G A =-30.60 kcal/mol and consequently ∆∆G is 1.48 kcal/mol. Thus, the K417N mutation reduced the binding affinity between RBD and ACE2 as well, but the reduction is about 6.5 time less than that between RBD and CB6. Previously, it was demonstrated in experiment that the K417N mutation weakened the binding affinity between RBD and ACE2, 11 which is consistent with our simulation results.
The noticeably small errors listed in Tab. 1 manifest the accuracy and convergence afforded by the FEP methodology.  the complex of RBD (of the SARS-CoV-2's spike protein) and human antibodies, surprisingly we noticed that K417 in RBD can form a buried salt-bridge with either a glutamate or an aspartate in four other human antibodies as highlighted in IGHV-53 (Fig. 4a), C1AB3 (Fig.   4b), BD-236 (Fig. 4c) and CC12.1 (Fig. 4d). Thus, it is likely that all these antibodies may not be able to neutralize the 501Y.V2 variant. It is worth noting that while writing this paper, an experimental preprint posted on bioRxiv demonstrating that the South Africa variant can evade the human antibodies CB6 (Fig. 1c, also known as LyCoV016) and CC12.1 ( Fig. 4d)   mutations (Fig. S2a), we observed the same improved binding structure (see Fig. S2b and Movie S1 in the Supporting Information). For the E484K mutation, after about 190ns simulation, K484 moved toward ACE2 to form a salt-bridge exposed on the complex's surface with E75 (see Figs. S2b and S2c, and Movie S1 in the Supporting Information).
Therefore, the gain from the improved RBD-ACE2 binding resulted from both N501Y and E484K mutations might be enough to compensate the loss caused by the K417N mutation, and is likely to yield an even stronger interaction with ACE2 than the wild-type virus.
In summary, we have shown that K417 plays an important role in the binding between RBD and ACE2/CB6, by forming interfacial salt-bridges. The salt-bridge between K417 in RBD and D104 in CB6 is buried inside the complex and is therefore much more stable than the water exposed one formed between K417 in RBD and D30 in ACE2. Thus, the K417N mutation can weaken the RBD's binding with CB6 much more than with ACE2.
More dramatically, the K417N mutation allows the variant to escape from many human antibodies other than CB6 by removing a salt-bridge buried in the RBD-antibody interface.
Interestingly, the virus with the K417N mutation seems to scarify its binding affinity with ACE2 in order to survive the antibodies' attack. This strategy of eluding human antibodies might have been adopted by SARS-CoV as well, because after the sequence alignment the residue in SARS-CoV corresponding to K417 in SARS-CoV-2 is the valine (V).
On the other hand, a recent study has demonstrated that immune systems of human may also quickly respond to these new variants and produce corresponding neutralizing antibodies that contain more somatic hypermutation, increased potency and resistance to RBD mutations. 22 How the continued evolution of the humoral response to recent SARS-CoV-2 variants warrants further studies. Additionally, with the revealed mechanism underlying the K417N mutation, it is possible to design more efficacious antibody cocktails to treat COVID-19 patients infected with the variant 501Y.V2 as well as the recently discovered variant P.1 in Brazil.