Differential escape of neutralizing antibodies by SARS-CoV-2 Omicron and pre-emergent sarbecoviruses

The SARS-CoV-2 B.1.1.529 lineage, Omicron variant, was first detected in November 2021 and carries 32 amino acid mutations in the spike protein (15 in RBD) and exhibits significant escape of neutralizing antibodies targeting the parental SARS-CoV-2 virus. Here, we performed a high-resolution multiplex (16-plex) surrogate virus neutralization assay covering all major SARS-CoV-2 variants and pre-emergent ACE2-binding sarbecoviruses against 20 different human serum panels from infected, vaccinated and hybrid immune individuals which had vaccine-breakthrough infections or infection followed by vaccination. Among all sarbecoviruses tested, we observed 1.1 to 4.7-, 2.3 to 10.3- and 0.7 to 33.3-fold reduction in neutralization activities to SARS-CoV-2 Beta, Omicron and SARS-CoV-1, respectively. Among the SARS-CoV-2 related sarbecoviruses, it is found that the genetically more distant bat RaTG13 and pangolin GX-P5L sarbecoviruses had less neutralization escape than Omicron. Our data suggest that the SARS-CoV-2 variants emerged from the changed immune landscape of human populations are more potent in escaping neutralizing antibodies, from infection or vaccination, than pre-emergent sarbecoviruses naturally evolved in animal populations with no or less immune selection pressure.


Introduction
The coronavirus disease 2019 (COVID-19) pandemic started in December 2019 and has caused 364 million cases and claimed 5.6 million lives, as of 27 January 2020. The causative agent, SARS-CoV-2 1 , is a member of the subgenus Sarbecovirus, same as for the SARS-CoV-1 that caused the SARS outbreak 19 years ago 2 . Bats are known reservoirs for SARS-related coronaviruses 3,4 . Multiple sarbecoviruses have been detected in bats [5][6][7][8] and more recently in pangolins 9 . SARS-CoV-2 variants of concern (VOC) emerged since late 2020 after 11 months of evolution via massive human-to-human transmission, probably in response to the changing immune dynamics of the human population 10 , with ve major VOCs recognized by WHO hitherto, namely Alpha, Beta, Gamma, Delta and Omicron. SARS-CoV-2 VOCs have either developed resistance/escape to neutralizing antibodies [11][12][13][14][15][16] or acquired mutations that enhanced transmission or pathogenicity 17,18 . The SARS-CoV-2 Omicron variant was rst detected in South Africa and Botswana and has rapidly spread to many countries as the most rapidly transmitting VOC to date.
There are two major features making Omicron different from previous VOCs: it contains the most mutations in the spike protein (32 aa mutations) and the receptor binding domain (15 aa mutations) (Fig.   1a) and it is much more capable of escaping neutralizing antibodies (NAbs) than any prior VOC [19][20][21][22] (Fig   1a). The enhanced transmission phenotype of Omicron is mostly linked to its ability of NAb escaping.
There have been several published studies comparing effectiveness of NAbs on Omicron versus the parental strain or other VOCs using conventional live virus neutralization test (cVNT) or spike pseudovirus VNT (pVNT) using either infected or vaccinated serum panels. Due to the inter-assay variability and labor-intensive nature of cVNT and pVNT, it is di cult, if not impossible, to conduct a pansarbecovirus NAb-escape study using a large number of human immune sera. Our current study aims to ll this gap. Here we employed a 16-plex surrogate virus neutralization test (sVNT) platform to provide a high-resolution examination of NAb-escape characteristics of all known human SARS-CoV-2 VOCs as well as pre-emergent sarbecoviruses of bat and pangolin origin. The sVNT developed in early 2020 23,24 has been validated and applied in multiple studies examining NAbs from infected, vaccinated and vaccinebreakthrough cohorts as well as virus origin and human-to-animal transmission studies 5,[25][26][27][28] . The sVNT measures RBD-targeting NAbs, which accounts for more than 90% of total NAbs 29,30 .
Combining this powerful NAb detection platform with a large collection of 20 different human serum panels, we found that Omicron's ability to escape NAbs is more potent than one would expect from a natural virus evolution path in a non-infected or unvaccinated population, pointing to the possibility of Omicron emergence as a result of immune selection during human transmission in a changed immune landscape after almost two years of SARS-CoV-2 circulation and vaccination program around the globe.
As expected, all RBDs bind to human ACE2 in a dose-dependent manner in the multiplex Luminex system.
The Omicron RBD showed the weakest binding in this assay system (Extended Data Fig. 1a). We rst tested the 16-plex sVNT using the WHO international standard 20/136 31 and the neutralization activities were reduced from SARS-CoV-2 in the following descending order for the human SARS-CoV-2 variants: Delta, Alpha, Delta plus, Lambda, Gamma, Beta, Mu and Omicron. Only limited cross-reactivity to SARS-CoV-1 and clade-1 sarbecoviruses was observed (Extended Data Fig 1b). The calibration of WHO international standard using SARS-CoV-2 ancestral multiplex sVNT was modelled (Extended Data Fig 1c) and the IU/ml at 1:80 dilution of all the tested serum panels in this study are listed in Supplementary Data Table 1. To ensure that our RBD-targeted sVNT assay is well correlated with NAb assays based on whole spike protein, we conducted a comparative analysis using the multiplex sVNT and pseudovirus-based VNT (pVNT) against the ancestral and Omicron viruses, respectively. From a well-de ned panel of 120 sera with varying levels of neutralizing antibodies (NAbs), the data showed a good correlation between sVNT and pVNT with R 2 of 0.83 and 0.73 for the ancestral and Omicron viruses, respectively (Extended Data Fig 1d, 1e).
Reduction of neutralization activities of different human serum panels to SARS-CoV-2 Omicron After con rming the functional performance of the 16-plex sVNT platform, we examined the ability of different human serum panels to neutralize Omicron. The serum panels were in three different groups: 1) convalescent sera from natural infection (4 panels), 2) serum from vaccinees, with or without the booster (12 panels); and 3) the hybrid immune group which had a mixture of infection and vaccination or vice versa (4 panels). See Supplementary Data Table 2 for details of the 20 different serum panels used in this study.
For group 1: we had access to convalescent sera from patients infected with SARS-CoV-2 ancestral, Beta and Delta variants and SARS-CoV-1. Sera from all three SARS-CoV-2 related viruses had reduced level of neutralization against Omicron, with more signi cant reduction from the Ancestral-and Beta-infected patients, with geometric mean NT50 of 10.3-fold and 6.7-fold reduction from the homologous RBD,  Fig 2d). It is interesting to note that the NAb level to Omicron is similar between the Omicron-and Delta-breakthrough groups. None of the vaccine-breakthrough and COVID-19-vaccinated individuals had neutralization activity to clade-1 sarbecoviruses. In other words, the SARS-vaccinated group is the only one who had genuine pan-sarbecovirus NAbs despite the fact that they also suffered some level of reduction in neutralization against Omicron.

Correlation between degree of neutralizing antibody escape and number of RBD mutations
Bat CoV RaTG13 and pangolin CoV GX-P5L are both phylogenetically placed in the SARS-CoV-2 clade, which are more distantly related to SARS-CoV-2 than Omicron with 90.1% and 86.6% amino acid sequence identity, respectively, of the RBD sequence in comparison to Omicron at 93.3% (Fig 1b, c). When the level of NAb titers were examined side by side, it was found that Omicron had a more signi cant NAb escape than either RaTG13 or GX-P5L (Fig 2a-d, Fig 3a, Supplementary Data Fig 1a-d), indicating that Omicron is antigenically more distant than the two pre-emergent sarbecoviruses despite the fact that there are 15 aa differences in Omicron RBD in comparison to 22 and 30 aa differences in the RBD of RaTG13 and GX-P5L, respectively. Our data further indicated that the degree of NAb escape in VOCs was more signi cant than that exhibited by pre-emergent sarbecoviruses of animal origin using serum panels from 2-dose mRNA vaccinated (Fig 3b), 3-dose mRNA vaccinated (Fig 3c) and vaccine-breakthrough (Fig  3d) groups. In other words, to achieve the same level of NAb escape, the number of aa mutations in the RBD region needs to be greater than those in the VOCs. Importantly, it was shown that the degree of NAb escape was less for the BNT162b2-vaccinated SARS survivor group with the exception of VOC Omicron (Fig 3f), again highlighting the greater-than-expected NAb escape displayed by Omicron in relation to the number of aa mutations in the RBD sequence.
Apart from phylogenic distance based on genetic differences, antigenic cartography was used to generate antigenic maps that could re ect the antigenic properties of closely related pathogens 32,33 .
Generated from neutralization titer 50% of the examined 187 sera on all 16 sarbecoviruses (Supplementary Data Table 1), the antigenic map showed the clustering of SARS-CoV-2 clade and the antigenic distance between the SARS-CoV-2 and other clade 2 sarbecoviruses in the following descending order: Delta plus, Delta, Lambda, Alpha, Gamma, Mu, RaTG13, Beta, GX-P5L and Omicron (Extended Data  Fig 2a). Among all clade-2 sarbecoviruses, only Omicron showed an extremely large distance (>2 Antigenic Unit) between SARS-CoV-2, revealing the unusual antigenic drift of Omicron. Similar observation is observed with vaccinated serum panels (Extended Data Fig 2b). On the antigenic map generated from SARS-vaccinated serum panel (Extended Data Fig 2c), the antigenic distance between SARS-CoV-2 clade and SARS-CoV-1 clade coronaviruses was closer, yet the Omicron is antigenic distinct. Consistent with our earlier observation, Omicron is more antigenic distant to SARS-CoV-2 than GX-P5L or RaTG13.
To further dissect the differential level of NAb escape in relation to RBD aa mutations, we employed a panel of hyperimmune rabbit sera raised against recombinant RBD proteins of different clade-2 sarbecoviruses, including human SARS-CoV-2, bat CoV RaTG13, pangolin CoV GX-P5L, and two non-ACE2 binding RBDs from bat CoV RmYN02 and bat CoV SL-ZC45 5 . Consistent with the previous data obtained from human sera, the SARS-CoV-2 RBD rabbit hyperimmune sera exhibited a 9.1-fold reduction of NAb titers against Omicron. Sera raised against RaTG13 and GX-P5L RBDs had highest NAb titer to their homologous viruses in the 16-plex sVNT and displayed a 6.8-fold and 4.1-fold NAb titer reduction, respectively, to SARS-CoV-2 (Extended Data Fig 3a, b). In contrast, Omicron showed an almost complete NAb escape against these sera, with a 64.8-fold and 33.2-fold reduction, respectively, against the RaTG13 and GX-P5L hyper immune sera (Extended Data Fig 3a, b), again demonstrating the greater NAb escape ability unique to Omicron in the context of RBD mutations or sequence relatedness. Sera derived from SARS-CoV-1 immunized rabbits exhibited 28.4-fold reduction to Omicron. For the hyperimmune sera raised against the non-ACE2 binding RBDs of RmYN02 and SL-ZC45, it was interesting to note that they had the highest NAb titers against VOC Delta Plus, with GMT of 86 and 289, respectively (Extended Data  Fig 3c). As a negative control, no NAb was detected for the rabbit hyperimmune sera raised against a non-sarbecovirus bat CoV HKU1 RBD (Extended Data Fig 3c).

Discussion
In less than a year after COVID-19 pandemic, vaccine candidates based on the original Wuhan-hu-1 (ancestral) strain spike protein were developed, including mRNA, inactivated, protein subunit and virusvectored vaccines. As of 27 January 2022, more than 9.8 billion doses of vaccine have been administered worldwide. Of the vaccine induced immunity, NAb levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection 34 . With the great ability of Omicron in NAb escape, it has been shown that an mRNA-based vaccine booster is essential to increase NAb to protect against Omicron infection [19][20][21][22]35 and our data in this study support previous ndings (Fig 2c, e). Our study further demonstrates that three doses of inactivated virus vaccine failed to elicit su cient NAbs against Omicron (Fig 2c). However, for those who have received two doses of inactivated virus vaccines, heterologous boosting with either mRNA vaccine BNT162b2 or viral-vectored vaccine AZD1222 could induce signi cantly higher amount of NAbs against Omicron (Fig 2c). Breakthrough infection signi cantly increased overall NAb titers against all SARS-CoV-2 variants, including Omicron (Fig 2d). In general, Delta-breakthrough infection seems to elicit higher levels of NAbs against different clade-2 sarbecoviruses than Omicron-breakthrough infection (Fig 2d), probably due to the less pathogenic nature of the Omicron virus 36,37 . Despite showing reduced NAb activities against Omicron, BNT162b2-vaccinated SARS survivors remains the only group demonstrating true pan-sarbecovirus NAbs 23 (Fig 2d-e).
It should be noted that our sVNT is based on detection of RBD-targeted NAbs by blocking RBD-ACE2 interaction as a biochemical simulation of a genuine live virus-based VNT. For that reason, the sVNT will be less sensitive as it is unable to detect the total NAbs against the whole Spike protein as in PRNT or pVNT assays. The fold of reduction in our study is generally less than those published in previous studies using pVNT or PRNT. However, the comparative fold of reduction across different sarbecoviruses or different serum panels is consistent regardless of the VNT platform used.
Bats are known natural reservoir for SARS-related coronaviruses 3,4 . Rich diversity of SARS-related coronaviruses detected in bats and pangolins highlight the risk of future spillover event 8,9,38,39 . The emergence of viral variants is within expectation with massive transmission events happening since the start of pandemic. Drivers for variant emergence include change of population immunity landscape 10 , selective pressure from convalescent plasma or monoclonal antibody therapy on immunocompromised or chronically infected individuals [40][41][42] and reverse zoonosis 25,43,44 . The ndings presented in this study would suggest that the emergence of Omicron is most likely due to host's immune selective pressure 45 . This is based on the fact that pre-emergent clade-2 sarbecovirus GX-P5L and RaTG13, which are evolutionarily more distantly related to SARS-CoV-2 than Omicron (Fig 1b), showed less degree of NAb escape than Omicron (Fig 3a-e). This is true in general for VOCs in comparison to pre-emergent sarbecoviruses of animal origin which are unlikely to be exposed to SARS-CoV-2 related immune selective pressure.
Our high-resolution sVNT analysis indicated the presence of four major antigenic clades based on cross-NAb activities: SARS-CoV-2 and clade-2 sarbecoviruses (minus Omicron), VOC Omicron, bat CoV RsSHC014, and SARS-CoV-1 and its related bat viruses. It is important to note that bat RaTG13 and pangolin GX-P5L, but not Omicron, are antigenically grouped with SARS-CoV-2. This was rst shown from the human sera studies, subsequently supported by the cross-NAb activities of the rabbit hyperimmune sera raised against the RBDs of these two viruses, i.e., these rabbit sera neutralized SARS-CoV-2 and most variants of concern, but not Omicron (Extended Data Fig 3a-c).
In conclusion, our study demonstrated the following important observations and ndings. First, a high resolution NAb assessment platform is highly powerful in dissecting differential NAb escape abilities in the context of its correlation with RBD mutations. Second, the large number of different human serum panels provided extra con dence in con rming the greater-than-expected NAb escape ability of Omicron.
Third, it is clear that Omicron is exceptional in its NAb escape capability when considering that GX-P5L, with twice as many RBD mutations, displayed less degree of NAb escape, suggesting that the emergence of Omicron is likely due to immune selective pressure. Fourth, Omicron breakthrough infections did not induce higher or broader NAbs against clade-2 sarbecoviruses in comparison to Delta breakthrough infections. Fifth, among all the human serum panels, the BNT162b2-vaccinated SARS survivors is the only group that displayed true pan-sarbecovirus NAbs.
The emergence of Omicron highlights the urgent needs to develop more broadly protective vaccines which can act in a VOC-agnostic manner. Despite 15 aa mutations in its RBD, Omicron already shows a distinctive antigenic grouping different from all other clade-2 sarbecoviruses. It is highly possible that such variants may emerge if VOC-speci c vaccines are mass deployed as we have experienced with the two major VOCs, Delta and Omicron, in 2021 since mass vaccination started for the ancestral viral strain. Data from the Omicron breakthrough infection group further suggest that an Omicron-speci c vaccine is unlikely to provide the pan-sarbecovirus NAbs required to prevent future VOC infection. Data from the vaccinated SARS survivors would argue that cross-clade immunization approach may be the most effective in inducing genuinely pan-sarbecovirus NAbs.

Pseudovirus neutralization test (pVNT)
SARS-CoV-2 Wuhan-hu-1 (ancestral), delta, Omicron and GX-P5L full-length spike pseudotyped viruses were produced and packaged as previously described 23,24 . Brie y, 5 million of HEK293T cells were transfected with 20 µg of pCAGGS spike plasmid using FuGENE6 (Promega). At 24h post transfection, cells were incubated with VSV∆G luc seed virus (at MOI of 5) for 2h. Following two PBS washes, infected cells were replenished with complete growth media supplemented with 1:5000 diluted anti-VSV-G mAb (Clone 8GF11, Kerafast). At 24h post infection, pseudoviruses were harvested by centrifugation at 2,000 x g for 5 min. For pVNT assay, 3 x 10 6 RLU of pseudoviruses were pre-incubated with four-fold serial diluted test serum in a nal volume of 50 µl for 1h at 37°C, followed by infection of ACE2-stablely-expressing A549 cells. At 20-24h post-infection, an equal volume of ONE-Glo luciferase substrate (Promega) was added and the luminescence signal was measured using the Cytation 5 microplate reader (BioTek) with Gen5 software version 3.10.
Phylogenetic analysis ACE2-binding sarbecovirus RBD sequences were either directly retrieved from NCBI or translated from nucleotide sequences retrieved from GISAID. Further analysis was performed in the Geneious Prime (version 2021.1). RBD protein sequences were aligned with MAFFT to construct the phylogenetic tree by the maximum-likelihood method with the blosum62 model using 1,000 bootstrap replicates in the PHYML 3.0 software.

Antigenic cartography
Antigenic map was generated using the R package "Racmacs" (version 1.1.18) in R (version 4.1.2) based on the matrix of neutralization titer 50% of serum on the 16 sarbecoviruses on multiplex sVNT. The number of optimizations was set to 1000.