Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) global transmission is still posing an unignorable threat to human health1,2. SARS-CoV-2 spike proteins target human angiotensin-converting enzyme 2 (hACE2) molecules on respiratory cells through its receptor binding domain (RBD), resulting in virus infection3,4,5. SARS-CoV-2 spikes, existing as homotrimers, are class I fusion proteins extending from the virus’ lipid envelopes. The S1 and S2 domains make up the ectodomain of each spike monomer, respectively. The S1 domain, encompassing the N-terminal domain (NTD), receptor binding domain (RBD) and CTD1/2 domains, is responsible for SARS-CoV-2 spikes’ targeting hACE2 molecules on respiratory cells6,7,8; S2 domain contains fusion peptide (FP), proteolytic cleavage site at S2′ and two heptad repeat regions which, upon spikes’ binding to ACE2, will undergo structural transitions from prefusion to postfusion state to catalyze the membrane fusion between virus and target host cell membranes9,10. Spike RBDs’ binding to hACE2 primes S2’ cleavage and initiates the membrane fusion process11,12. For prophylaxis purpose, SARS-CoV-2 spike RBDs’ binding affinity changes and immune evasion advantages gained through adaptive mutations will provide critical information facilitating vaccine’s development, detection of its host tropism range (hot spot mutations overcoming cross-species transmission barriers) and exploration of potential virus evolutionary origin13,14,15,16,17,18.

SARS-CoV-2 spikes’ affinity for hACE2 involves structural effectors not only limited to its receptor binding domains. In the prefusion state, spike trimers of SARS-CoV-2 and VOCs (variants of concern) exhibited different hACE2-accessible states by protomers’ RBDs adopting ‘up’ or ‘down’ configurations6,12,19,20. RBD ‘up’ state contributed to higher hACE2 binding21,22. Structural components distributed in S1/S2 domains were proposed as allosteric control elements, which include FPPR (fusion peptide proximal region), 630 loops, N2R linker connecting NTD and RBD in a protomer, and D614G mutation or other mutations such as A570D and S982A modulating spikes’ ‘up’ or ‘down’ states21,23,24,25. Even key RBD residues could play a role in spikes’ conformational shift such as S371L, S373P, and S375F detected in Omicron mutants and key hACE2-contact residues at 417 or 48421,22,26,27. Still spike RBDs play a determinant role in hACE2 binding, as mentioned above in virus entry11,12.

Extensive studies have disclosed SARS-CoV-2’s strong binding hot spots (residues at 417, 486, 493, 498, 501) that are well-defined along its receptor binding motif (RBM) in direct contact with hACE2 binding interface (residues at 30, 31, 35, 38, and 353)5,26,28,29,30,31,32. Structural studies showed that mutated hot spots at 498, 501 can be adaptive for SARS-CoV-2’s using rat or mink ACE2 as receptors16,30. These hot spots echo well with those identified in SARS-CoV-1(479 and 487), residue options of which are critical to cross-species infection with SARS-CoV-1 and determine its host tropism33,34,35,36. Residues at 493, 498, and 501 are also evolutionarily important, as single amino-acid mutation can enable phylogenetically diverse sarbecoviruses or SARS-CoV-2-like bat coronaviruses binding to hACE215,18,37. As SARS-CoV-2 evolves, these RBD hotspots are still adaptively mutating, with positive or negative effect on binding which is accompanied by the advantages of evading human immune surveillance24,26,38,39,40. How SARS-CoV-2 mutates to gain an immune evasion advantage while compensating for the binding-detrimental mutations is still worth exploring41,42.

SARS-CoV-2 spike RBD interfacial to hACE2 (RBM) encompasses two protruding ridges composed of loops and an intermediate concave part of anti-β sheets. Hotspots Q498 and N501 of wild type (WT) SARS-CoV-2 RBD formed a dense network of polar contact with hotspots Y41, D38, and K353 in hACE229,31; Hotspots K417 and Q493 were either located in the vicinity of or in the intermediate subregion, which presented a hydrophobic environment, and was electrostatically complementary to the hACE2 interfacial subregion harboring hotspots D30, K31, and E3529,31. Certain structural properties of interacting proteins’ interface can affect binding affinity via different effects on binding kinetics: favorable long range Coulombic electrostatic forces between interacting proteins in the vicinity of the binding surface would increase the rate of association (ka), while dissociation kinetics was dictated by the strength of short-range interactions43. We hypothesize that RBD subregions of distinct structural properties will affect RBDs’ binding kinetics differently and play distinct roles in affecting RBD affinity for hACE2.

This study will explore potential distinct contributions to hACE2 affinity by interfacial subregions of SARS-CoV-2 VOCs and phylogenetically close sarbecoviruses’ spike RBDs, and dissect specific subregion’s impact on the binding kinetics. Moreover, affinity and binding kinetics data of selected sarbecoviruses will help evaluate certain interfacial subregions’ role in SARS-CoV-2’s adaptive evolution and sarbecoviruses’ evolvable ACE2 binding. The interface of SARS-CoV-2 spike RBD to hACE2 was structurally partitioned into CR1, CR2, and CR3 subregions. Through chimeric swap mutants of respective subdomains from SARS-CoV-1, bat RaTG13, and selected SARS-CoV-2 VOCs, their impacts on binding affinity and kinetics were differentiated through a comprehensive SPR measurement of hACE2 binding and further compared with those of the included single mutations across RBM and around the RBD core. In contrast to single mutations’ impact, the CR3 subregion of dense polar contact networks poses an unexpectedly strong contribution to RBD/hACE2 affinity, which increases as SARS-CoV-2 adaptively mutates. Hot spot Y493 of CR2 from bat RaTG13’s RBD supports its critical role in ACE2 binding as discovered in ancestral and evolvable sarbecoviruses. CR2 involving key residue at 417 and CR3 subregion are the rate-limiting effectors of association kinetics and dissociation kinetics, respectively. The inherent binding mode offers direct clues to the valid epitopes in designing new antibodies that the coronavirus can’t elude.

Result

Structurally partitioned CR1, CR2, CR3 subdomains along intermolecular interface display distinct impacts on RBD/hACE2 interaction

SARS-CoV-2 spike RBD protein shares varied sequence identities with those of other phylogenetically close sarbecovirus subgenus species, such as bat sarbecovirus RaTG13 and SARS-CoV-131,44. Despite that the reported X-ray structures of such sarbecovirus RBD/hACE2 complexes shared a significant degree of the protein backbone’s structural conservation5,31,37, different binding affinities were detected among sarbecovirus RBDs’ interacting with hACE2 by up-to two to three orders magnitude of difference7,20,37,44. It was the specifically interfaced residues of the respective RBD/hACE2 complex and locally formed bonding networks that dictated such distinct binding affinities45. The key contact residues are distributed among different RBD subregions interfacial to hACE2, which differ across different clades of sarbecoviruses and are where adaptive mutations occur during SARS-CoV-2 evolution. Elucidation of these subregions’ potential distinct effect on RBD binding affinity & kinetics may offer unique perspective to evaluate SARS-CoV-2’s evolution and other sarbecoviruses’ evolvable ACE2 binding. The SARS-CoV-2 RBD’s hACE2-interacting interface is structurally partitioned into three consecutive CR1, CR3, and CR2 subdomains across RBM, referring to either protruding ridges or the middle concave domain (with either boundary spatially defined by F456/F490 and Y449/G496 as referred to PDB# 6M0J) of RBM, respectively (Fig. 1A) (detailed in Methods).

Fig. 1: Scheme of the spatially distributed subdomains of the SARS-CoV-2 spike RBD and designment of the RBD chimeric mutants of specific swapped subdomains to disclose structural determinants governing SARS-CoV-2 RBD/hACE2 interaction.
figure 1

A WT SARS-CoV-2’s spike RBD, especially RBM can be spatially partitioned into three subdomains owing to their specific hACE2 contact resides and spatial arrangement. Similar RBD structural organizational patterns are detected in SARS-CoV-2 Omicron (B.1.1.529), SARS-CoV-1 and bat coronavirus RaTG13; The RBD structures as presented above are from spike RBD/hACE2 crystal structures with PDB# 6M0J, 7TN0, 2AJF, and 7DRV31,37,44,51. B SPR measurement of RBDs’ binding to hACE2(1-740)-FC immobilized on Series S Sensor Chip (Protein A), which cover those from above sarbecovirus species, but not that of RaTG13. The equilibrium dissociation constants (KD) from triplicate measurements were calculated with BIAcore@ 8 K Evaluation Software (GE Healthcare) by fitting to a 1:1 Langmuir binding model. C RBD protein sequence alignment between WT SARS-CoV-2, Omicron (B.1.1.529), SARS-CoV-1 and bat coronavirus RaTG13. The secondary structures as discovered in WT SARS-CoV-2 RBD is labeled on top of sequence alignment; Black filled triangles point to the direct contact resides in WT SARS-CoV-2 RBD/hACE2 complex. D Experimental design of chimeric mutants of swapped subdomains with WT SARS-CoV-2 RBD as the template. For each partitioned subdomain of wild type SARS-CoV-2 RBD, key residues of polar or non-polar contact within 4 Å cutoff between spike RBD and hACE2 interfacing chains are identified and those different residues are swapped with homologous counterparts identified from above selected crystal structures. For each swapped subdomain mutant, the mutated residues corresponding to each subdomain are marked in red color.

Full size image

To differentiate these subdomains’ specific impact on RBD/hACE2 interaction, respective CR1/2/3 subdomain-swapped chimeric RBD mutants were designed and generated by swapping key contact residues of particular subdomain of WT SARS-CoV-2 to those counterparts from other species’ RBDs (SARS-CoV-1, bat RaTG13, and Omicron (B.1.1.529 and BA.5.2)) while preserving WT spike RBDs’ remaining structural components. SARS-CoV-1 and bat RaTG13 are selected because they belong to phylogenetically close clades to SARS-CoV-2 (SARS-CoV-1 clade and SARS-CoV-2 clade respectively) but possess lower hACE2 affinity, while two Omicron VOCs are selected due to more clustered mutations detected around RBDs as compared to pre-Omicron VOCs. Omicron B.1.1.529 was the first Omicron subvariant identified in Africa escaping from protection conferred by then developed vaccines and therapeutic mAbs46. Omicron BA.5.2 belonged to BA.4/5 lineage, spike RBDs of which lineage possessed more substitutions conferring more immune evasion/growth efficacy and hACE2 binding strength47; It was the dominant circulating SARS-CoV-2 subvariant in south China when this study was conducted48, which retained the key hot mutations across RBM still in presently circulating Omicron VOCs. By sequence alignment and structural analysis of selected X-ray RBD/hACE2 complex structures, key polar or non-polar contact residues within 4 Å cutoff of specific RBD and hACE2 interfacing chains are identified and partitioned into CR1, CR2, and CR3 as described above (Fig. 1C, D). It should be noted that K417 contributes significantly to binding affinity and positions near the concave interface of RBD, which is the reason it is partitioned to CR2 outside of SARS-CoV-2 RBM.

First, RBDs of WT SARS-CoV-2, SARS-CoV-1, and Omicron B.1.1.529 were tested for their respective binding affinities to hACE2-FC (1-740) through triplicate SPR measurements. WT SARS-CoV-2 RBD has a ∼3 fold stronger binding affinity than that of SARS-CoV-1 and binds ∼3 fold weaker as compared to Omicron (B.1.1.529) RBD based on measured KD (equilibrium dissociation constant) values (Fig. 1B, Table 1). RaTG13 RBD’s hACE2 binding affinity was not measured, as it was already well tested to be weak (KD = 3.86 μM)37. Next, for respective subdomain’s swapping from selected sarbecovirus species, unexpected contrasting while pronounced impacts are observed: While the binding is nearly eliminated by the RaTG13-swapped CR3, the SARS-CoV-1/Omicron (B.1.1.529) CR3 shows a significant increase in binding over the WT SARS-CoV-2 RBD. Remarkably, a two-order magnitude strengthening is induced by swapped CR3 from SARS-CoV-1, in contrast to a significantly weaker binding from SARS-CoV-1 RBD (Figs. 2C, 1B) (Table 1). In RaTG13, a single Q493Y-mutated CR2 induces a binding increase of approximately five times over WT SARS-CoV-2 RBD, whereas those from SARS-CoV-1 and Omicron (B.1.1.529) weakens the RBD/hACE2 interaction by four to eight times (Fig. 2B). It should be noted that the SPR binding data analysis of SARS-CoV-1 RBD and CR2 chimera mutants from SARS-CoV-1 & Omicron (B.1.1.529) does not produce as satisfactory data fitting curve as that for other mutants, even after experiments were repeated (Figs. 1B and 2B). Considering these CR2s’ encompassing predominantly hydrophobic residues, their negative impact onto RBD protein stability cannot be excluded49. Moreover, their characteristic structural properties plus its central location in binding interface may cause interfacial instability as discussed in following text. Regarding CR1 switching, the swapped CR1 from RaTG13 reduces binding by a factor of seven, whereas that from Omicron (B.1.1.529) increases binding by a factor of more than three (Fig. 2A) (Table 1). The CR1-swapped RBD protein from SARS-CoV-1 was not obtained by in vitro cell culture, possibly due to solvent accessibility change of specific surface-exposed hydrophobic residues and resulting negative impact onto SARS-CoV-2 RBDs’ proper folding and stability (Fig. 1D)49. Considering that the targeted residues to be swapped are P462 and L472 only (corresponding to A475P and F486L mutations in WT RBD) and referring to CR1 swapped from RaTG13 harboring a single F486L mutation (Fig. 2A) and a single A475P mutant’s slightly negative effect (data shown later), we can only infer that CR1 from SARS-CoV-1 potentially deals with a similar negative impact.

Table 1 SPR data comparison of RBD binding to hACE2 for SARS-CoV-2 VOCs, interfacial subdomain-swap mutants and SARS-CoV-1
Full size table
Fig. 2: Affinity measurement of the impact on the SARS-CoV-2 RBD/hACE2 interaction by swapped RBD subdomains from RaTG13, SARS-CoV-1 and Omicron (B.1.1.529).
figure 2

A WT SARS-CoV-2’s CR1 subdomain was swapped and the resulting chimeric mutants’ binding to hACE2(1-740) was measured by SPR. Chimeric swapped CR1 mutant from SARS-CoV-1 could not be expressed, which was replaced by a SARS-CoV-2 RBD chimera containing SARS-CoV-1 RBM replacement, binding affinity measurement of which reflects inherent coordination among CR1, CR2 and CR3 subdomains. B WT SARS-CoV-2’s CR2 subdomain was swapped and the resulting chimeric mutants’ binding to hACE2(1-740) was measured by SPR. C WT SARS-CoV-2’s CR3 subdomain was swapped and the resulting chimeric mutants’ binding to hACE2(1-740) was measured by SPR.

Full size image

In summary, partitioned subdomains of the RBD/hACE2 interface make quite distinct contributions to the intermolecular interaction. An inherent coordination mechanism probably governs whole molecular behavior, affecting binding affinity. The resulting binding affinity from swapping RBM (covering CR1/2/3) of SARS-CoV-2 with that of SARS-CoV-1 (KD = 88.63 ± 1.82 nM) supports this in that the serious negative impact of concave CR2 and potential CR1 is compensated by CR3’s strong contribution (Fig. 2).

CR3 makes a dominant contribution in high affinity hACE2 interaction among representative world-wide circulating SARS-CoV-2 VOCs

The representative SARS-CoV-2 mutant strains are selected to measure their respective RBD/hACE2 binding affinities including VOCs such as Alpha(B.1.1.7), Gamma(P.1), Delta(B.1.617.2), Omicron (B.1.1.529) and BA.5.2. Although the hACE2 binding interface is usually divided into two patches (Fig. 3A, top left), the mutated residues are similarly distributed as structurally partitioned (Fig. 3A, B)37,50. We mainly focus on mutations’ effect on RBM in this section.

Fig. 3: Scheme of the natural mutations distributed on RBD, especially on the RBM of SARS-CoV-2 and RBD/hACE2 binding affinities’ comparison among representative world-wide circulating SARS-CoV-2 VOCs.
figure 3

A Top view of the respective SARS-CoV-2 VOCs’ RBM with key residues in contact with hACE2 of WT SARS-CoV-2(top left) or corresponding mutated residues from other VOCs colored accordingly. Also array of key mutated residue locations clustered from all selected VOCs are projected onto the same WT SARS-CoV-2 RBD structure (2nd image on top left). Protein structures of wild type, Alpha(B.1.17), Beta(B.1.351), Gamma(P.1), Delta(B.1.617.2), and Omicron (B.1.1.529, BA.5.2) RBDs are referred to PDB# 6M0J, 7MJN, 7V80, 7NXC, 7WBQ, 7TN0, and 7XWA respectively31,47,50,51,66,67. B Scheme of natural mutations distributed along the SARS-CoV-2 VOCs’ primary RBD protein sequence. For WT SARS-CoV-2 RBD, the RBM sequence coverage is labeled by a two-way arrow. For Omicron (BA.5.2), additional natural mutations to those of Omicron (B.1.1.529) are displayed in deep blue and retro-mutated residues of R493Q and S496G in RBM are colored in green. C RBD binding affinities to hACE2 are compared among selected SARS-CoV-2 VOCs by triplicate SPR measurements each. D Measurement of the impact on the SARS-CoV-2 RBD/hACE2 interaction by CR1, CR2 and CR3 swapped from Omicron (BA.5.2) respectively. Alongside is a SPR scan of Omicron (B.1.1.529) RBD chimera with CR3 swapped from SARS-CoV-1 to further corroborate CR3’s dominant role in high affinity RBD/hACE2 interaction.

Full size image

Compared to the WT SARS-CoV-2 RBD/hACE2 interaction (KD = 44 ± 2.06 nM), Alpha, Omicron (BA.5.2) VOC’s RBDs display the strongest binding enhancement by about five folds, whereas those from Gamma and Omicron (B.1.1.529) follow with a slightly lower affinity increase. Delta RBD binding to hACE2 ranks lowest (Fig. 3C) (Table 1). The sole N501Y mutation of CR3 stands out due to its conferring Alpha RBD the strongest hACE2 binding affinity except in Omicron strains. In contrast, Delta RBD lacking N501Y ranks the weakest, although L452R in CR2 strengthens interaction through a positive charge increase, as reported and tested later (Fig. 4B)40. However, CR3 subdomains from omicron (B.1.1.529) and (BA.5.2) confer on the SARS-CoV-2 chimeric mutant a further four & seven folds’ enhancement over Alpha RBD, respectively (Figs. 2C, 3D). Among selected representative circulating VOCs, as per CR3 subdomains’ contribution, Delta<Alpha=Gamma<Omicron (B.1.1.529) <Omicron (BA.5.2) reflects an ascending order of interaction strength with hACE2 (Table 1).

Fig. 4: SPR scanning to differentiate selected single residues’ impact onto the SARS-CoV-2 RBD/hACE2 interaction.
figure 4

A Cartoon shows naturally occurring key RBD mutations in SARS-CoV-2 Omicron (B.1.1.529 and BA.5.2), and distinct residues of SARS-CoV-1 and RaTG13 from those of WT SARS-CoV-2 spike RBD, with their respective locations projected onto the WT SARS-CoV-2 RBD structure (PDB# 6M0J). B SPR measurements of binding affinity of single SARS-CoV-2 RBD mutants of RBM region interacting with hACE2. For each single mutant, the equilibrium dissociation constants (KD) from triplicate measurements were analyzed with BIAcore@ 8 K Evaluation Software (GE Healthcare) by fitting to a 1:1 Langmuir binding model. Purple, red and yellow colors represent mutants distributed in CR1, CR2 and CR3 subdomains respectively as presented in cartoon A. C SPR measurement of binding affinities of single SARS-CoV-2 RBD mutants of non-RBM region for hACE2.

Full size image

Chimeric mutants of swapped CR2 and CR1 from Omicron (BA.5.2) were also tested to illustrate the mutual coordination mode of respective subdomains (Fig. 3A, D). Like B.1.1.529, CR2 swapping issues a similarly negative impact (Table 1), whereas CR1 swapping has little influence on interaction. To further corroborate the CR3’s dominant role, CR3 of Omicron B.1.1.529 RBD (F375S mutation for efficient protein expression) was swapped with that of SARS-CoV-1. Despite the negative influence of adjacent CR2, as confirmed above, the B.1.1.529 chimeric RBD mutant still displays a strong hACE2 binding affinity comparable to other strong CR3-swapped chimeric mutants identified (Figs. 2C, 3D). These data suggest the CR3 subdomain of attributed structural properties makes a dominant contribution to the RBD/hACE2 interaction, which keeps rising as SARS-CoV-2 evolves.

Key residues identified in related subdomains and structural analytical reference of CR3 versus CR1/2 by comparing SPR data of single mutants to those of swapped subregions

The single mutations detected in the RBDs of SARS-CoV-1, RaTG13, and Omicron (B.1.1.529)/(BA.5.2) are projected onto the WT SARS-CoV-2 RBD structure (Fig. 4A). Single mutants of S477N, T478K, and E484A for CR1 were generated, and only S477N has around 3 folds’ enhancement comparable to that of swapped CR1 (B.1.1.529) mutant whereas the other two are of no effect. (Fig. 4B) (Tables 1 and 2). Another functionally detected single residue in this subdomain is 486 F, which is mutated to 486 L in RaTG13 and 486 V in Omicron (BA.5.2), respectively. Either mutation attenuates the binding affinity significantly by around sixfold. Since S477N is the only residue that differs between these two Omicron subvariants according to this subdomain (Figs. 1D and 4B) (Tables 1 and 2), F486V’s negative effect cancels out S477N’s strengthening when comparing the binding of the corresponding swapped CR1 mutants from Omicron (BA.5.2) (KD = 13.59 ± 0.23 nM) to that of (B.1.1.529) (KD = 55.26 ± 0.34 nM). This can be ascribed to 486 F forming a hydrophobic core with M82/L79/Y83 of hACE2 and S477N hydrogen-bonded to S19 of hACE2 independently, which contributes to CR1’s binding energy change in an additive pattern31. To support this, swapped CR1 mutants’ binding energy change from Omicron (B.1.1.529 /BA.5.2), RaTG13, and double mutant S477N + F486V as compared to WT SARS-CoV-2 RBD were predicted by adding binding energy change from single mutations involved. The predicted energy changes are comparable to those of swapped subdomains, respectively (Table 3, Supplementary Fig. 1A).

Table 2 SPR data comparison of single RBD mutations’ impact onto RBD binding to hACE2 (for mutations in interfacial subdomains)
Full size table
Table 3 Binding energy change of RBD/hACE2 complex formation due to swapped CR1/2/3 subdomains and single residue mutations across RBM
Full size table

CR2 covers a hydrophobic concave region, which enwraps K417/L452/Q493 mutations and reduces the dielectric constant of this subdomain, strengthening the involved polar interactions inside29. K417 forms a critical salt bridge with D30 of hACE2 in WT SARS-CoV-2 RBD/hACE2 complex, while K417N mutation in Omicron (B.1.1.529) and (BA.5.2) leads to a five fold’s affinity decrease (Fig. 4B) (Supplementary Fig. 2)29,31. Similarly, Q493R and Q493N also impair binding affinities, although to a lesser extent. However, singly mutated Q493Y as detected in CR2 of RaTG13 has a significant enhancing effect by ~5 folds (KD = 8.8 ± 0.2 nM) as opposed to Q493R and Q493N (Fig. 4B) (Table 2). Referring to the CR2 interface of the RaTG13 RBD/hACE2 complex, Y493 not only establishes hydrogen bonds with K31/hACE2, but also fills the space left by the concave CR2 subregion, providing further interfacial stabilization. (Supplementary Fig. 2)37. Polar interactions of CR2 interface by key contact residues at 417/493/453 of RBD with D30/H34 and K31/E35 on either side of hACE2 N-terminal α1 helix also contribute to affinity additively, as evidenced by totaling binding energy change from single mutants in CR2 is comparable to those of swapped CR2s from Omicron (B.1.1.529 / BA.5.2) respectively (Table 3, Supplementary Figs. 2 and 1B)47,51.

Different from the above single contact residues’ additive effect on CR1/2 binding, most of the single mutations in the CR3 subdomain display quite an opposite role compared to their teamwork, as reflected in the swapped CR3 mutants’ impact. If working in an additive pattern, only the G496S mutation’s negative effect (KD = 256 ± 13 nM) can cancel the N501Y’s positive impact (KD = 7.39 ± 0.14 nM), not even counting Y505H’s further impairment (KD = 388.33 ± 13.65 nM) and other negative mutations such as G446S and Q498R (Fig. 4B). However, the overall swapped CR3 exhibit a very strong binding mode, as detected in swapped CR3 from SARS-CoV-1 and Omicron (BA.5.2 & B.1.1.529) (Fig. 2C) (Tables 1 and 2). These swapped subdomains make a dominant energy contribution to RBD’s binding to hACE2, which otherwise contradicts single key residues’ contribution except that of N501Y (Table 3, Supplementary Fig. 1C).

To illustrate the unexpectedly strong binding from swapped CR3 chimeric mutants, SARS-CoV-2 spike RBD/hACE2’s CR3 interfacial interaction is referenced to published crystal structures that can be spatially divided into two local polar interaction network groups which underpin hACE2 N-terminal α1-helix’s anchoring above CR3 ridge of RBD (Supplementary Fig. 3)31,37,44,47,51. The π-π interaction between Y41 of hACE2 and Y484 of RBD establishes the bridge, combining diagonal intermolecular tensions into a stronger pulling force in the SARS-CoV-1 RBD/hACE2 complex. A similar organization is observed in the structural model for the chimeric RBD mutant of SARS-CoV-1 CR3, in which an additional hydrogen bond detected between G496/RBD and K353/hACE2 adds another linkage (Supplementary Fig. 3). The G496-K353 residue pair can also be observed in the RBD/hACE2 complex of the WT SARS-CoV-2 and BA.2 subvariant, whose comparable effect can be evidenced by a single G496S mutation’s lowering the binding affinity by five folds and the DMS assay (Fig. 4B, Supplementary Fig. 3)17,28. In contrast, the π-π stacking between Y41/hACE2 and Y501/RBD belongs only to one local interaction network of two Omicron subvariants’ CR3, which explains their CR3s’ slightly lower binding than swapped CR3 from SARS-CoV-1(Supplementary Fig. 3) (Table 1). More drastically, Y449F mutation in RaTG13 RBD not only impairs one local bonding network’s binding strength due to the loss of hydrogen bonds formed with D38 & Q42, but breaks two local interaction networks’ balance even in presence of π-π interaction, rationalizing the abolished affinity of the chimeric mutant of swapped RaTG13 CR3 (Fig. 2C, Supplementary Fig. 3)37.

Mutations around SARS-CoV-2 RBD’s core affect SARS-CoV-2 spike RBD structural stability and intermolecular interaction

For Omicron B.1.1.529 & BA.5.2 and other Omicron subvariants, mutations also occur in non-RBM regions, for which the impacts of single mutants were quantitatively measured (Figs. 3B, 4C). Noticeably, D405N and R408S specific to Omicron BA5.2 attenuate binding by two folds, respectively (Fig. 4C) (Table 4). In the WT SARS-CoV-2 RBD/hACE2 structure, D405 acts as a central hub by forming hydrogen bonds with R403, G504, and R408 along the RBD’ core α helix, which was also conserved in complexes by Omicron B.1.1.529 and BA.5.2, even in the case of D405N/R408S detected in BA.5.2 RBD/hACE2 complex (Supplementary Fig. 4). D405N or R408S may disrupt the α-helix that connects the antiparallel β sheets of the RBD’s core, as these interlocked residues are part of it. It has been observed that R408 forms a hydrogen bond with the glycan that is preserved in both SARS-CoV-1 and SARS-CoV-2 and is attached to N90 of hACE229.

Table 4 SPR data comparison of single/multiple RBD mutations’ impact onto RBD binding to hACE2 (for mutations in non-RBM regions)
Full size table

Additionally, mutations to three sites are detected along a hairpin loop connecting the anti-β sheet core of RBDs, which are S375F, S373P, and S371L/F respectively. SPR data indicate that a single mutation of S375F attenuates binding by two folds; The binding impact is comparable to that by the triple mutant (S371F/S373P/S373F) (Table 4, Supplementary Fig. 4)47,51. Further, single S371L/P or S373P mutations’ impact is almost ignorable as compared to that by S375F(Table 4). Comparing the RBD/hACE2 complex of Omicron BA5.2 to that of wild-type complex, F375 formed a local hydrophobic core with I402, V407, and Y508 of adjacent anti-β sheets, which stretches the connected hairpin loop observably, destabilizing the adjoining α2 helix into a stretched loop (Supplementary Fig. 4)31,47. Similar impact on secondary structural transition is also observed in Omicron B.1.1.529 complex, although only shorter stretch of α helix adjoining the hairpin loop is detected (Supplementary Fig. 4)31,51. Consequently, a single S375F mutation has a significant impact on spike RBDs’ stability.

It was experimentally tested that Omicron RBD was not as thermally stable as that of other SARS-CoV-2 RBD structures52,53. These data not only give a structural explanation of the Omicron subvariants’ decreased RBD stability but corroborate the proposed non-RBM regions’ direct effect on the RBD/hACE2 interaction54.

CR2 and CR3 interfacial subregions play distinct roles in association and dissociation of RBD interacting with hACE2 respectively

SPR data inform interfacial RBD CR3, which is characterized by the densely interweaved polar bonding networks that distinguish it from those of key mutants in CR1 and CR2. We want to differentiate how these subregions may impact RBD/hACE2 association and dissociation kinetics (ka, kd) which determine the RBD affinity for hACE2 (KD = kd/ka). To quantify and compare each interfacial subregion’s effect on binding kinetics, the ratios of association/dissociation kinetics of the studied mutations relative to those of WT RBD interacting with hACE2 were calculated and plotted as shown in Fig. 5, which also included the equilibrium dissociation constants profile comparison.

Fig. 5: Binding kinetics analysis discloses distinct impact on RBD/hACE2 association and dissociation kinetics by CR2 and CR3 subdomains respectively.
figure 5

A Association kinetics profiles of RBD/hACE2 complex formation by SARS-CoV-2 chimeric mutants of swapped CR1/2/3 and single residue mutants reveal significant CR2’s impact on RBD/hACE2’s initial contact with hACE2, in which long range electrostatic force are involved. The mutations’ impact as compared to WT SARS-CoV-2 RBD was quantified as the ratio of ka(RBD mutant) to ka(wild type). For the swapped CR2 mutants from SARS-CoV-1 and Omicron (B.1.1.529 & BA.5.2), the mutated residues are marked and positively charged residues are labeled in orange. B Dissociation kinetics of RBD/hACE2 complex formation by SARS-CoV-2 chimeric mutants of swapped CR1/2/3 and single key residue mutants reveal CR3’s significant impact on RBD/ hACE2 complex stability. The mutations’ impact as compared to WT SARS-CoV-2 RBD was quantified as the ratio of kd(wild type) to kd(RBD mutant). C Relative binding affinity change due to swapped CR1/2/3 subdomains or single key residue mutations as compared to WT SARS-CoV-2 RBD interacting with hACE2. The mutations’ impact is quantitatively compared by ratio of KD(wild type) to KD(RBD mutant). D Based on the kinetics data in A, spike RBD surface potential map of swapped CR2 mutants from SARS-CoV-1 and Omicron (B.1.1.529 & BA.5.2) as compared to that of WT spike RBD suggests that the long range Coulombic electrostatic forces alter the spike RBD/hACE2 association kinetics. Below each subgraph, the range of surface potential distribution is labeled. The hACE2-contact surfaces of spike RBD’s CR1, CR2 and CR3 subregions are outlined with purple, red and green dash lines respectively. For all CR1/2/3 swapped mutants and single mutants, histograms reflecting binding kinetics are presented in purple, red and yellow respectively, while those of native RBDs of SARS-CoV-1/2 and subvariants are colored in dark grey. All error bars represent standard deviations resulting from ka, kd, and KD values obtained through three or more independent experiments.

Full size image

In contrast to swapped CR1 chimeric mutants, whose subdomain swapping has limited effect on association kinetics, swapped CR2 from SARS-CoV-1, Omicron subvariants demonstrate significant but varied negative impact, among which swapped CR2 (417 V/455Y/456 L/493 N) from SARS-CoV-1 is most impactive by attenuating association kinetics by around 16 folds (Fig. 5A). Noticeably, the K417V mutation changes the RBD’s surface electrostatic potential, specifically at the CR2 interface, resulting in an overall negatively charged local interface (Fig. 5D). As the hACE2 subregion interfacial to RBD CR2 shows negative surface electrostatic potential around key contact residues of 30D and 35E (Supplementary Fig. 5), K417V impairs the electrostatic complementarity of SARS-CoV-2 RBD/hACE2 CR2 interface. However, Q493R or L452R partly recovers positive surface potential adjacent to K417N in the case of swapped CR2s from Omicron (B.1.1.529) and (BA.5.2) respectively (Fig. 5D), which relieves the negative impact on the association by varied degrees (Fig. 5A). As the CR1 subarea exhibiting negative surface potential also complements the positively charged hACE2 contact interface (Fig. 5D, Supplementary Fig. 5), it is the electrostatic attraction between the RBD/hACE2 CR1 & CR2 interfaces that dictates association. RBD’s CR2 involving key residues at 417 is the rate-limiting effector.

The swapped CR3 from SARS-CoV-1, Omicron (B.1.1.529 & BA.5.2) exhibits a strong delay of the dissociation kinetics by up to 90 folds (Fig. 5B), but with an ignorable effect on the association, reflecting the key role of the strong short-range interactions characterized by densely interweaved bonding networks as assayed in the previous section. Since other mutations have much less effect on dissociation, including those from non-RBM (Fig. 5, Supplementary Fig. 6), CR3 acts as the rate limiting factor of dissociation, playing a pivotal role in stabilizing/strengthening the RBD/hACE2 complex.

Discussion

In this study, spike RBD protein instead of spike trimers is utilized to precisely measure RBD mutations’ impact by SPR because certain RBD residues that are adaptively mutated are potentially involved in adjacent RBDs’ coupling, affecting spikes’ ‘up’ or ‘down’ conformational shift21,22,26,27. For single mutations tested across RBM, the most impactive hotspots agree well with those already well-defined, including residues vital to SARS-CoV-1 and SARS-CoV-2’s binding and host tropism (498, 501, 449, 417, 493, 486)29,31,33,34,36,44. The new information comes from that interfacial CR2 and CR3 subregions can have unexpectedly strong impact onto binding as opposed to single mutations, even though these targeted subregions come from certain sarbecovirus RBDs of lower binding than WT SARS-CoV-2 RBD, implying their importance in sarbecoviruses’ evolvable hACE2 binding. The CR3 subdomain of attributed structural properties makes a dominant contribution to the RBD/hACE2 interaction, which keeps rising as SARS-CoV-2 evolves from pre-Omicron VOCs to Omicron subvariants. Further, CR2 involving key residues at 417/493 is the rate-limiting effector of association kinetics, and CR3 encompassing residues at 501/498/449 is the key binding energy contributor dictating dissociation kinetics. The new perspective and data offer new information on the differentiation of mutational impact during SARS-CoV-2’s adaptive evolution and transmission.

Two main factors can be attributed to CR2’s being the association rate limit effector. The first is electrostatic complementarity between CR1/2 interfaces of WT SARS-CoV-2 RBDs and that of hACE2 (Fig. 5D, Supplementary Fig. 5A). Coulombic electrostatic forces will attract interacting proteins initially in the long range. Q493R or L452R mutations partially repair the impairing effect of swapped CR2 from SARS-CoV-1 (which lacks charged residues) on association kinetics (Fig. 5A); L452R strengthened interaction through a positive charge increase (Fig. 4B)40; Optimized D480 (equal to 494 of SARS-CoV-2, and no bonds formed) electrostatically complemented K31/hACE2 in the human-optimized SARS-CoV-1 RBD for hACE2 binding34. The second factor is CR2’s hydrophobic environment with charged residues (417 or 493) embedded. As illustrated by homologous residue 479 (equivalent to 493) of SARS-CoV-1, this structural property was energetically detrimental to spike RBD’s beginning contact with hACE2, but once polar bonds form, the interaction could be highly stabilized, benefiting from the low dielectric constant of hydrophobic environments29,33. The observed intermolecular conformational difference in structural comparisons (Supplementary Fig. 2) and molecular dynamics, which showed RBD and hACE2 undergoing symmetric twist about the axis of the N-terminal helix, may result from these attributed factors if concave CR2’s steric effect on its accessibility is also taken into account55.

SARS-CoV-2 RBD shares compatible sets of interfacial hot spots with SARS-CoV-1 in determining its host range: residues at 30, 31, 35, 38, 41, and 353 in ACE2 versus RBD residues at 486, 493, 498, and 501 (442, 472, 479, 480, 487 in SARS-CoV-1)33,34,36,56. Hot spots of residue 479 and 487 posed the major species barriers for SARS-CoV-1 infection of humans and palms civet33,34. SARS-CoV-2’s CR3 demonstrated substantial plasticity in adapting its RBD binding to other species’ ACE2s. Q498H and Q498R/N501Y enabled binding of RBD variants to rat ACE2, while the latter double mutations occurring in Omicron variants enabled binding to ACE2s from mouse and palm civets16,57; However, this doesn’t mean hot spots in CR2 or CR1 are not important. SARS-CoV-2 isolates from mink with 501 T, Y453F or F486L mutation displayed a higher binding capacity to mink ACE230. Residue 479 largely determined whether SARS-CoV-1 can infect humans33; Y493 in RaTG13’s CR2 enables its weak binding to hACE2 despite its CR3 losing binding strength due to 501D & 449 F and its CR1 affinity seriously impaired (Fig. 2). As to SARS-CoV-2’s origin and evolvable ACE2 binding, mutation K493Y in AncSarbecovirus RBD (ancestor of sarbecoviruses) and African sarbecovirus enabled binding to human ACE2, while the K493Y’s effect for latter was enhanced by T498W mutation15; These two key residues’ coordination in adaptive binding could also be observed in RBDs of SARS-CoV-2-like bat sarbecoviruses from Asia, illustrating the geographically and phylogenetically diverse sarbecoviruses’ similar evolutionary pathways to acquire hACE2 binding15,18. Our SPR data of swapped CR1 and CR3 from RaTG13 reminds us that while single mutation at 493 only confers sarbecoviruses’ RBD weak binding, further adaptive mutations in CR3 such as 501, 449, or both may substantially increase their human infection risks.

Based on SPR data, those swapped subdomains with strong affinities are referenced to structures already published (Fig. 6A). A phase-specific binding mode model is proposed to ascribe spike RBD/hACE2 affinity to distinct effectors in association and dissociation kinetics (Fig. 6B). As an extension, it will be interesting to inquire if there is still room for improvement in future SARS-CoV-2 spike RBD binding. As reported, Omicron BA.2.75 spike RBD affinity ranked highest (KD = 2.2 nM) among Omicron BA.1-5 subvariants58. While the key residues’ difference in RBD is only N460K(non-RBM) in BA.2.75 and L452R/F486V (CR1/2) in BA.4/5. Another reported super strong RBD mutant (RBD-62) of binding affinity over 1000 folds stronger than WT RBD encompassed typical S477N/E484K (CR1) and Q498R/N501Y (CR3) in the absence of negative Y505H/K417N mutational impact (Fig. 4B). Notably, the cryo-EM structure of CR3 subdomain of RBD-62 exhibited the similar favorable inter-bonding networks’ linkage by hydrogen bonds established between N501Y and Q498R as proposed (Fig. 6A)59. Based on our comprehensive SPR scanning across both RBM and non-RBM region, these data point to the critical negative impactors by Y505H (CR3), K417N (CR2) and F486V/L (CR1), whose impact on functional CR1/2 subdomains can potentially be compensated by adaptive mutations changing RBD surface charge such as N460K/E484K/L452R/R498Q and even allosteric regulations by R346K detected in Omicron BA.1.154,60,61. Considering the above converging mutations in developing Omicron subvariants, the future spike RBD mutants will only likely approach the subnanomolar affinity as detected for swapped Omicron’s CR3 mutants in the scenario of key K417N/F486/V/P mutations’ impact counteracted by other binding-favorite mutations across both CR1/2 subdomains and non-RBM regions, which, however, will be constrained by selective pressure driven by immune evasion.

Fig. 6: Structural reference of selected swapped CR1, CR2 and CR3 subdomains of strong affinities and a phase specific binding mode model attributing spike RBD/hACE2 affinity to distinct effectors in association and dissociation kinetics.
figure 6

A Structural reference of selected swapped CR1/CR2/CR3 subdomains of strong affinities based on SPR measurements in this study. For each subdomain contact interface’s structural display, key residues are marked and critical inter-residue bonding such as π-π interactions or hydrogen bonds were indicated by two-way arrows, with hydrogen-bonded atoms or pseudoatom centers of phenolic group of involved residues shown as colored spheres; The interaction networks can be spatially distinguished into separate polar interaction networks outlined by dashed ellipse. In the left top figure, the two bonding interaction networks of CR3 interface draw hACE2’s N terminal α1 helix towards RBD side by side (indicated by two parallel arrows), while in left bottom figure, two bonding networks linked by inter-group bonding generate a stronger combined pulling force to make RBD/hACE2 complex more stable. PDB# of the reference structure or modeling information for each structural presentation are marked. The referenced structures are PDB# 2AJF, 6M0J, 7XWA, 7DRV, 7TN031,37,44,47,51. B A phase-specific binding mode model of SARS-CoV-2 RBD interacting with hACE2. In phase 1, long range Coulombic electrostatic force from spike RBD CR1 and CR2 draw interfacial contact subregions from hACE2 N-terminal α1 helix close to initiate intermolecular contact, in which process the CR2 with positively charged K417 from WT SARS-CoV-2 spike RBD plays the rate-limiting effector in association kinetics. In phase 2, the dominant short-range interaction of RBD CR3 acts as the main binding energy contributor. These two spike RBD subdomains interacts with opposite hACE2 interfaces in a timely and spatially distinct manner owing to their specific structural properties. There may be interfacial rotation reflecting intermolecular instability as discussed, which will be reduced or stabilized by favorable polar interactions inherent to CR1/2 and CR3 as in A. Top figure is a structural cartoon proposing a phase-specific binding of hACE2’s N terminal α1 helix to the RBD interface, and bottom figure proposes high and low affinity interaction mode due to favorable (solid line) or unfavorable (dash line) long range Coulombic electrostatic forces or short range interactions as required in association and dissociation respectively. Curved double arrows represent potential interfacial rotation as RBD interacts with hACE2. The binding-favorable key contact residues are labelled in green while those unfavorable ones colored in red.

Full size image

High-throughput DMS assays concerning escape mutation profiles of antibodies against the evolving Omicron subvariants have identified converged mutational hot spots such as K417N/T, K444-G446, N450, L452, and especially E484K62. As pointed out, mutations of F486S/V, F490S, and L452R as partitioned into CR1 strongly attenuated the neutralizations by classified group B/C Nabs, which interface directly with hACE2. Strikingly, BA.2.75/BA.5 with extra four or five RBD mutations, including L452R, K444M/N, R346T, N460K, and K356T would enable them to completely evade plasma samples of most vaccinated and convalescent individuals62. As mutations around the RBD core conferred critical immune escape properties (S371F, D405N, R408S), they could also lead to a loss of thermal stability harmful to RBD/hACE2 interactions, as both tested here and reported (Table 4)17,52,53. This will make it logical for SARS-CoV-2 to enhance the RBD binding strength to offset the affinity loss occurred to both RBM and non-RBM regions exchanging for immune escape from developing vaccines and convalescent plasma.

Super strong CR3 binding itself should make the NAbs’ attack on this subdomain more inaccessible, as previously proposed when intermolecular affinity rises38,63. Accordingly, targeting the preliminary intermolecular contact stage will be an ideal strategy for designing new antibodies that the coronavirus can’t elude. Indeed, a mAb (C68.3) was identified exhibiting broad and potent neutralization to more evolved SARS-CoV-2 VOCs, including Omicron BQ.1.1 and XBB.1.5, with a very focused epitope centering residues A475 and G476 conserved in major circulating SARS-CoV-2 VOCs, and targeting exactly the CR1 interface to block hACE2 binding64,65.

Methods

Structural partitions of SARS-CoV-2 spike RBD interfaced with hACE2 across RBM (receptor binding motif)

Based on the characteristics of structural organization and bonding networks of SARS-CoV-2 spike RBD/hACE2 complex crystal structure (pdb#6M0J), three spatially contiguous subdomains are structurally partitioned across the SARS-CoV-2 RBM interfaced with hACE2 and designated as CR1, CR2 and CR3 respectively. Specifically, the CR1 subdomain is comprised of protein sequence spanning I472∼490 F, which assumes a protruding ridge configuration of RBD in contact with hACE2; On another RBD/hACE2 interface end, the CR3 subdomain is comprised of two protein sequence segments spanning G496∼505Y and K444∼Y449 respectively, which are sequentially discontinuous but spatially organized into another protruding ridge configuration of RBD. Between CR1 and CR3 is a concave intermediate region (Y449∼F456, F490∼G496) of overall hydrophobic nature with polar residues embedded, in which key K417 outside RBM is included as per its reported critical contribution to RBD/hACE2 affinity and relative spatial assignment. The boundary between the three contiguous subdomains is spatially demarcated by F456/F490 and Y449/G496 residue pairs as presented in RBD/hACE2 crystal structure (PDB# 6M0J).

Design of protein construct for spike RBDs of selected sarbecovirus species, CR1, CR2, CR3 subdomain swapped mutants and selected single key mutants of each subdomain

cDNA sequences encoding specific spike RBDs were optimized for mammalian cells’ protein expression for wild type SARS-CoV-2, Gamma subvariant (P.1), Delta subvariant (B.1.617.2), Omicron subvariants (B.1.1.529 and BA.5.2) and SARS-CoV-1. The specific RBD’s cDNA sequence was subcloned into the pcDNA3.1 plasmid via restriction endonuclease sites via EcoR V and Xba I, with protein encoding sequence’s N-terminal addition of an interleukin 2 (IL-2) signal peptide and carboxy-terminal fused with a hexa-histidine tag (6×His) facilitating nickel affinity protein purification. For swapped CR1, CR2 and CR3 subdomain mutants with the WT SARS-CoV-2 RBD as the main structural template, the necessary multisite mutants matching targeted swapped subdomains of specific sarbecovirus species or SARS-CoV-2 VOCs were generally generated through multiple points’ mutagenesis through circular plasmid PCR with Phusion™ High–Fidelity DNA Polymerase(Thermo Fisher Scientific)；To obtain the swapped RBM mutant from SARS-CoV-1, overlapping extension PCR was performed to obtain the swapped cDNA fragment encoding SARS-CoV-1 RBM which was then integrated to WT SARS-CoV-2 RBD scaffold to replace its prototype RBM by using seamless cloning kit. For spike RBD’s binding partner, optimized cDNA sequence encoding hACE2(1-740) was synthesized and subcloned into pcDNA3.1 plasmid with the carboxyl-terminal tagged with a human Fc tag.

Protein expression and purification

The protein construct was transiently transfected into Expi293F (Thermofisher) cells in SMM 293-TII Expression medium (SinoBiological), which was cultured in suspension (shaking at 130 r.p.m) in a humidified cell incubator with the setting of 37 °C, 8% CO2. The cells’ density upon transfection was two million per mL and the transfection agent applied to suspension-cultured Expi293F cells was PEI 40 K Transfection Reagent (Servicebio) mixed with SMM 293-TII Expression Medium. The culture was usually supplemented with SMS 293-SUPI cell culture supplement (SinoBiological) 24 and 72 h after transfection. The supernatants were collected after four or five days’ cell culture depending on the satisfactory coomassie blue staining check of secreted proteins’ level. After centrifugation of the harvested supernatants at 1000 r.p.m for 5 mins and 4500 r.p.m for 10 mins, the expressed protein samples either went immediately to the purification step or were stored at −80 °C.

The supernatants containing expressed spike RBD proteins were centrifuged at 4 °C for 15 min at 6000 g, which was then filtered through PVDF films of 0.22 μm pore size. His-tagged RBD protein was purified through two-purification steps of affinity chromatography and gel filtration respectively. First, the supernatants were applied to the prepackaged HisTrap HP column (Cytiva) which was preequilibrated with buffer A (20 mM Tris, 200 mM NaCl, 20 mM Imidazole, pH 7.5). After washing with the same buffer, the protein was collected through a gradient elution with varied mixture of buffer A and buffer B (20 mM Tris, 200 mM NaCl, 500 mM Imidazole, pH 7.5). The eluted peak solution containing the target proteins was further concentrated with 10kD MWCO Amicon Ultra 15 mL centrifugation filter (Milipore) before applied to Superdex 200 Increase 10/300GL column (Cytiva), which was pre-equilibrated in buffer C (10 mM HEPES、140 mM NaCl、pH 7.4). The protein was gel-filtered through the column in buffer C. The eluted protein samples from gel filtration purification step were aliquoted and stored at −80 °C. For hACE2(1-740)-Fc protein expressed, the first purification step was carried out with a prepackaged HiTrap Protein A HP column (Cytiva), which was preequilibrated with equilibration buffer (20 mM PB, 140 mM NaCl, pH 7.4) and the target protein was eluted through a two-step elution with elution buffers (100 mM Citric acid, pH4.5 and pH3.5) respectively. The protein sample was neutralized and concentrated with 30kD Amicon Ultra 15 mL centrifugation filter (Milipore) before applied to a Superdex 200 Increase 10/300GL column (Cytiva). The protein was gel-filtered through the column in buffer C (10 mM HEPES, 140 mM NaCl, pH 7.4). The eluted protein sample from gel filtration purification step was stored at −80 °C. All purified protein samples’ concentrations were measured by BCA methods and their satisfactory purities (at least above 90%) were verified through SDS-PAGE and coomassie blue staining check.

Surface plasmon resonance detection of spike RBDs’ dynamic interaction with hACE2

The binding assay of dynamic interaction with hACE2 was carried out on BIAcore 8 K (GE Healthcare), for which the multiple cycle capture method was applied and performed at 25 °C. Specifically, hACE2(1-740)-FC was captured as the ligand onto the Series S Sensor Chip Protein A (Cytiva) with the capture response values generally ranging from 400 to 600 RU (Response Unit). The specific RBDs as the solute analyte in flow phase was flown across the Protein A sensor chip at a flow speed of 30μL/min with series of selected concentrations set respectively for each test (0, 3.7, 11.1, 33.3, 100, 300, 900 nM, respectively). The association time was set as 240 sec and the dissociation period was set as 600 sec for full representation of the collective molecular interaction behavior of the test system. Upon each association/dissociation cycle finished, the regeneration time was set at 30 sec. The running buffer for SPR measurement is buffer H (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.005% Tween-20). The buffer for regeneration of the Protein A chip was buffer I (10 mmol/L Glycine-HCl (pH 1.5) flowing across the chip for 30 sec or 10 mmol/L Glycine-HCl (pH 2.0) flowing for 50 sec. For each tested RBD species or mutant, the binding assay was performed generally in triplicate and the experimental data were processed through BIAcore@ 8 K Evaluation Software (GE Healthcare) to obtain the association rate constants(ka), dissociation rate constants(kd), which are used to calculate equilibrium dissociation constants (KD = kd/ka) by fitting to a 1:1 Langmuir binding model. The setup basics of the SPR experimental measurement were listed in supplemental information (Table 1) and the measured binding kinetic parameters of targeted swapped RBD mutants, RBD species and single RBD mutants were listed in Tables 1, 2, and 4 respectively.

Homology Modeling of chimeric RBD mutant of swapped CR3 from SARS-CoV-1 interacting with hACE2

Structural interaction of the RBD mutant of swapped CR3 subdomain from SARS-CoV-1 with hACE2 was modelled on reported WT RBD/hACE2 crystal structure (PDB# 6M0J). The targeted CR3 subdomain of SARS-CoV-1 was of only three critical residues different (Fig. 1D) whose distribution surrounds the triangular core of key local CR3 interaction networks generally identified among SARS-CoV-2 VOCs’ RBDs (Y41/ACE2, Y/N/T501 and R/Y/498). Modeling was carried out through Schrödinger suite. The structural complex template was prepared in the protein preparation wizard to check and correct the potential missing side chains and loops; The protonation states of amino acid residues were assigned and the placement of hydrogen atoms was optimized. The swapped RBD sequence was aligned with template sequence in Prime module with conserved residues identified and predicted secondary structures assigned. The structural model building was subsequently executed through the energy-based method, after which model refinement was applied. According to above limited mutations of swapped SARS CR3 subregion as compared to the WT RBD/hACE2 complex, main model refinement steps applied included (i) optimization of side chains; (ii) minimization of non-template resides; (iii) checking any missing side chains and final minimization using the all-atom minimization technique. Validation of the structural model was based on Ramachandran Plot statistics with most favored regions occupying 93.2%, additional allowed regions 6.7% and generously allowed regions 0.1%. Overall average of G-factors of the structural modeling was −0.01(a measure of ‘how unusual, or out-of-the-ordinary’ structure properties with values above −0.5 determined as usual and below −1.0 as highly unusual).

Data analysis and graph preparation

Binding energy change of RBD/hACE2 complex formation due to swapped CR1/2/3 subdomain and key residue mutations of both RBM and non-RBM region as compared to WT SARS-CoV-2 RBD was calculated according to the formula ΔΔG = ΔGWT-ΔGMutant. For each Binding energy change of specific RBD/hACE2 binding measurement, it was calculated according to formula ΔG = R ∗ T ∗ lnKD, where R = 1.987 cal mol–1 K–1 = 8.314 J mol–1 K–1, T = 298.15 K, and KD as the equilibrium dissociation constants from experimental SPR measurements (in units M). The binding energy change profile was graphed as the histograms using GraphPad Prism 8 software. All error bar represents standard deviations resulting from KD values obtained through three or more independent experiments.

To quantitively compare the relative mutations’ impact onto the RBD/hACE2 intermolecular binding kinetics, the experimentally measured binding kinetics parameters were compared to those of WT RBD/hACE2 interactions respectively. Specifically, the relative RBD mutants’ impact on RBD/hACE2 association kinetics as compared to WT SARS-CoV-2 RBD was quantified as the ratio of ka(RBD mutant) to ka(wild type), and corresponding comparison of the dissociation kinetics was quantified as the ratio of kd(wild type) to kd(RBD mutant). Relative binding affinity change due to swapped CR1/2/3 subdomains or single key residue mutations as compared to WT SARS-CoV-2 RBD interacting with hACE2 was quantified by ratio of KD(wild type) to KD(RBD mutant). The ratio value of larger than one represents a positive enhancement while that lower than one means the opposite, which both reflect the degree of the impact and can be converted to fold’s change of either increase or decrease. The association and dissociation kinetics profiles’ comparison was prepared and visualized in histograms using GraphPad Prism 8 software. All error bars represented standard deviations resulting from relevant kinetics parameters obtained through three or more independent experiments.

Pymol software was generally applied for the general protein structural analysis and the molecular structures’ graph generation. To generate the surface electrostatic potential of spike RBDs, the WT spike RBD’s structure is obtained from the WT SARS-CoV-2 spike RBD/hACE2 complex (PDB#6M0J), and used as the structural template to generate swapped RBD mutations for surface potential mapping by Pymol. Both Pymol and Adobe Illustrator were applied to generate cartoons showing selected structural presentations of CR1/2/3 subregions which contribute to detected strong RBD/hACE2 binding and a phase-specific binding mode model illustrating distinct effectors in association and dissociation kinetics that determine spike RBD/hACE2 binding affinity.

Statistics and reproducibility

For SPR measurement of each selected spike RBD/mutant’s binding to hACE2, three or more independent experiments were performed. The association/disassociation kinetics values (ka, kd) and binding affinity values (equilibrium dissociation constants, KD) were collected as described above in SPR method section. The mean values with standard deviations for each selected RBD/mutant’s hACE2-binding properties were calculated by GraphPad Prism 8 software. The ratios of association/dissociation kinetics of the studied mutations relative to those of WT RBD were graphed as histograms with error bars defining the standard deviations.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.