Vaccine-induced immunity provides more robust heterotypic immunity than natural infection to emerging SARS-CoV-2 variants of concern.

Both natural infection with SARS-CoV-2 and immunization with a number of vaccines induce protective immunity. However, the ability of such immune responses to recognize and therefore protect against emerging variants is a matter of increasing importance. Such variants of concern (VOC) include isolates of lineage B1.1.7, rst identied in the UK, and B1.351, rst identied in South Africa. Our data conrm that VOC, particularly those with substitutions at residues 484 and 417 escape neutralization by antibodies directed to the ACE2-binding Class 1 and the adjacent Class 2 epitopes but are susceptible to neutralization by the generally less potent antibodies directed to Class 3 and 4 epitopes on the anks RBD. To address this potential threat, we sampled a SARS-CoV-2 uninfected UK cohort recently vaccinated with BNT162b2 (Pzer-BioNTech, two doses delivered 18-28 days apart), alongside a cohort naturally infected in the rst wave of the epidemic in Spring 2020. We tested antibody and T cell responses against a reference isolate (VIC001) representing the original circulating lineage B and the impact of sequence variation in these two VOCs. We identied a reduction in antibody neutralization against the VOCs which was most evident in the B1.351 variant. However, the majority of the T cell response was directed against epitopes conserved across all three strains. The reduction in antibody neutralization was less marked in post-boost vaccine-induced than in naturally-induced immune responses and could be largely explained by the potency of the homotypic antibody response. However, after a single vaccination, which induced only modestly neutralizing homotypic antibody titres, neutralization against the VOCs was completely abrogated in the majority of vaccinees. These data indicate that VOCs may evade protective neutralising responses induced by prior infection, and to a lesser extent by immunization, particularly after a single vaccine, but the impact of the VOCs on T cell responses appears less marked. The results emphasize the need to generate high potency immune responses through vaccination in order to provide protection against these and other emergent variants. We observed that two doses of vaccine also induced a signicant increase in binding antibodies to spike of both SARS-CoV-1 & MERS, in addition to the four common coronaviruses currently circulating in the UK. The impact of antigenic imprinting on the potency of humoral and cellular heterotypic protection generated by the next generation of variant-directed vaccines remains to be determined. BBIP-CorV and AZD1222 nCoV-19), intermediate values both relative potency generating neutralising antibody responses 18,19 . mRNA and Adenovirus-vectored vaccines generate high magnitude SARS-CoV-2 multispecic CD4+ and CD8+ T cells responses. Reports of vaccines assessed in South Africa where B.1.351 dominates are currently emerging and include Ad26.COV2.S (single dose Ad26 vectored vaccine), Novovax (recombinant Spike/adjuvant) and CAZD1222. Each of the studies report reduced ecacy in South African populations. Vaccine correlates of protection, and the relative contribution of T cell and humoral immunity are yet to be precisely dened


Introduction
The emergence of new lineages of SARS-CoV-2 on three continents towards the end of 2020, and their rapid expansion at the expense of the previously dominant lineages, poses signi cant challenges to public health (WHO | SARS-CoV-2 Variants) 1 . In order to address these challenges effectively, there is an urgent need to understand the biological consequences of the mutations found in these lineages, and the consequential impact on their susceptibility to current control measures, including vaccines, drugs and non-pharmaceutical interventions.
All three variants share the N501Y substitution in the receptor-binding domain (RBD) of spike glycoprotein (S), which increases binding a nity of S with the virus's cellular receptor, ACE2 2 (see Figure  1). Lineage B.1.1.7, rst identi ed in the UK in September 2020 ( [1]), is characterized by additional mutations in S, such as deletion of residues 69 & 70 and the P681H substitution for which plausible effects on the virus biology are proposed, as well as ve other mutations in S, a premature stop codon in ORF8, three substitutions and a deletion in ORF1 and two amino acid substitutions in nucleoprotein (N) of as-yet unknown signi cance. Lineage B.1.351 3 was rst identi ed in November 2020 in South Africa and is characterized by two additional substitutions of likely signi cance in RBD, namely, K417N and E484K. The former is predicted to disrupt a salt bridge with D30 of ACE2, a characteristic of SARS-CoV-2 in distinction to SARS-CoV-1, but may not impact on binding, whereas the latter, which might disrupt the interaction of RBD with K31 of hACE2, may enhance ACE2 binding 24 . Very recently (Jan 2021) E484K has been detected rst in lineage B1.1.7 in the UK 5 and subsequently in lineages A23.1, B.1 and B.1.177, as well as in imported cases of B.1.51 and P.2 [2]. Our data con rm that VOC, particularly those such as B1.351 with substitutions at residues 484 and 417, escape neutralization by antibodies directed to the ACE2-binding Class 1 and the adjacent Class 2 epitopes, but are susceptible to neutralization by the generally less potent antibodies directed to Class 3 and 4 epitopes on the anks of the RBD The immune correlates of protection against infection and disease caused by SARS-CoV-2 are imperfectly understood (recently reviewed by 6 ). Classically, neutralization by antibody, measured by reduction in plaque or infectious foci by authentic virus in vitro is considered a major component of protection, though indirect effects of antibody, such as complement activation and opsonization may also play a role in vivo. Recent studies have demonstrated that symptomatic re-infection within six months after the rst wave in the UK was very rare in the presence of anti-S or anti-N IgG antibodies 7,8 . Virus-speci c lymphocytes may play an important direct role in protection, in addition to their indirect effect mediated through help to antibody-producing cells. Robust T cell immune responses (with CD4+ T cells dominating) to S, M, N and some ORF antigens are readily detected after infection, correlate with disease severity and are durable for at least several months [9][10][11] . Furthermore, CD8 depletion studies in non-human primate (NHP) challenge studies suggest T cells also play a protective role especially when antibody levels are low 12 13 . Nevertheless, passive infusion of neutralizing antibody has been shown to be su cient to mediate effective protection against SARS-CoV-2 in these rhesus macaque challenge studies 13 , encouraging the evaluation of both convalescent sera and monoclonal antibody (mAb) cocktails as early therapeutic interventions. In common with endemic human coronaviruses, immunity following recovery from natural infection does not appear to be always su cient to prevent re-infection, although the increase in frequency of such cases towards the end of 2020 appears to be linked to the emergence of antigenically distinct lineages (see above).
At the time of writing (February 2021) multiple vaccines have been reported to have e cacy against COVID-19 disease in phase III clinical trials. Of these, three -BNT162b2, mRNA-1273 and Sputnik Vthat were reported to have e cacies in the mid-90% range, had also induced classical neutralising antibody titres substantially higher than those found on average in convalescent patients [14][15][16] . In contrast, one -CoronaVac -that showed approximately 50% e cacy, had been reported to induce neutralising titres several-fold lower than those found in convalescent patients 17  Ad26.COV2.S (single dose Ad26 vectored vaccine), Novovax (recombinant Spike/adjuvant) and CAZD1222. Each of the studies report reduced e cacy in South African populations. Vaccine correlates of protection, and the relative contribution of T cell and humoral immunity are yet to be precisely de ned since detailed immune analysis in people with vaccine breakthrough infections is lacking.
In pseudotype virus assays -in which S is not chaperoned by viral membrane proteins M and E, buds at the plasma membrane rather than internally, and has an immature quaternary structure that appears to render it variably more sensitive to neutralization by antibody -it appears that convalescent sera from patients exposed to prototype strains of SARS-CoV-2, in distinction to vaccine-elicited responses, may not be effective in neutralizing lineage B1.351 20,21 . As the lineage-de ning substitutions include changes in previously identi ed antibody epitopes and regions of S associated with its processing and rearrangement during cellular infection, this is a very plausible observation. In order to test whether convalescent sera and sera from vaccine recipients were similarly affected in their ability to neutralize authentic virions, we have undertaken classical neutralization assays against reference isolates of both B1.1.7 and B1.351 compared to the prototype, early pandemic, B VIC01 isolate. We nd that, while crossneutralization of B1.1.7 is only modestly reduced compared to that of the prototype strain, crossneutralization of B1.351 is markedly reduced. This effect is particularly marked in convalescent sera and in a very minor (<10%) subset of those that had mounted somewhat modest neutralizing responses to the vaccine.
Since viral mutations may also affect T cell recognition, we evaluate the contribution of T cells that target epitopes located at sites of amino acid substitution in the spike glycoproteins of VOCs. We evaluated the T cell response to peptides spanning the entire spike protein in an interferon-gamma (IFN-γ) ELISpot assay and found that the majority of T cell responses in recipients of two doses of the BNT162b2 vaccine are generated by epitopes that are invariant between the prototype Victoria strain and the three VOCs.
Our results suggest that reformulation of vaccines to address new variant lineages ought to be considered and indicates that seasonal re-vaccination might be required for this virus. We also show that neutralization of VOC is signi cantly enhanced by a second boost vaccination. There is legitimate concern that antigenic imprinting during vaccination to prototype S antigen blunts the response to variant sequences incorporated into next-generation vaccines. However, our results show that vaccination not only induces enhanced reactivity to S from endemic human coronaviruses, but also signi cant crossreactivity to both SARS-CoV-1 and MERS-CoV. Encouragingly, we nd evidence that the T cell epitopes that dominate responses in these same individuals, are not subject to major substitution in the three variants of concern.

Volunteer samples
Volunteers were recruited at Oxford University Hospitals NHS Foundation Trust in ethically approved studies and included: 1) Hospitalised patients with severe COVID-19 de ned as SARS-CoV-2 PCR positive and requiring in patient oxygen support, recruited under study CMORE (research ethics committee (REC): North West -Preston, REC reference 20/NW/0235); 2) Healthcare Workers (HCWs) with asymptomatic or mild symptomatic COVID-19 disease de ned as SARS-CoV-2 PCR positive disease and not requiring O2 support/hospitalization and; 3) HCWs not known to be previously infected with SARS-CoV-2, sampled 7-17 days after receiving COVID-19 mRNA Vaccine BNT162b2, 30 micrograms, administered intramuscularly after dilution as a series of two doses (0.3 mL each) 18 Microneutralization Assay (MNA) The study was performed in the CL3 Facility of the University of Oxford operating under license from the HSE, on the basis of an agreed Code of Practice, Risk Assessments (under ACDP) and Standard Operating Procedures. The microneutralization assay determines the concentration of antibody that produces a 50% reduction in infectious focus-forming units of authentic SARS-CoV-2 in Vero CCL81 cells, as follows. Quadruplicate serial dilutions of serum or monoclonal antibody were preincubated with a xed dose of SARS-CoV-2 before incubation with Vero cells. A 1.5% carboxymethyl cellulose-containing overlay was used to prevent satellite focus formation. Twenty hours post-infection, the monolayers were xed with 4% paraformaldehyde, permeabilized with 2% Triton X-100 and stained for the nucleocapsid (N) antigen or spike (S) antigen using mAbs EY 2A and EY 6A, respectively 24  and resuspended in R10 and counted using the Guava® ViaCount TM assay on the Muse Cell Analyzer (Luminex Cooperation). PBMCs were frozen and stored in liquid nitrogen before use.

Expression and puri cation of monoclonal antibodies
Monoclonal antibodies used in this study were produced as previously described 25 . The variable heavy and light genes of these mAbs were cloned into the human IgG1, human kappa light chain or lambda light chain expression vectors (AbVec-hIgG1, AbVec-hIgKappa or AbVec-hIgLambda). The heavy and light antibody plasmids were transfected into the Expi293F (Gibco) cell line for expression of recombinant fulllength human IgG1 according to manufacturer's protocol. Harvested supernatant was clari ed at 3,000g for 10min and ltered (0.22 μm) before a nity puri cation using a HiTrap MabSelect SuRe column (Cytivia, Westborough MA, USA) according to manufacturer's instruction. Eluted IgG1 were concentrated, and buffer exchanged into PBS using an Amicon Ultra-15 lter unit with 30k molecular weight cut-off.

Results
Homotypic and Heterotypic neutralization of key SARS-CoV-2 lineages by antibody.
The primary structure of the Spike glycoprotein, and the characteristic sequence variants of the current three lineages of concern are illustrated in Figure 1. In this study, we analysed the homotypic neutralization of the prototype strain, Victoria/01/2020 (PANGO lineage A), by monoclonal antibodies (mAbs), sera from convalescent individuals following COVID-19, and recipients of the BNT162b2 (P zer) vaccine, which are each induced by prototypic S antigen, and heterotypic neutralization by these same antibodies of two new lineages of concern (B1.1.7 and B1.351). In Figure 1, we indicate the residues of S at which the respective lineage -as well as a third lineage of concern, P.1 -differ from prototype.
Neutralization by sera from vaccine recipients.
After a single dose of BNT162b2 vaccine, 10/11 sera showed homotypic neutralization potency equivalent to that of the more modest half of convalescent sera (NT50 ~ 1/100, see Figure 2b).
The data from convalescent individuals and those having received vaccine are summarized in Fig2D, showing the trend to lower neutralization potency across the three genotypes of virus, which is least marked in those receiving two doses of vaccine. The relationship of the neutralizing titre of each individual's serum to B VIC01 to their neutralizing titre against each VOC is displayed in Fig 2E and  Neutralization by monoclonal, neutralizing mAbs to the four epitopes of RBD and by reference serum In order to understand better the reasons for differences between the homotypic and heterotypic neutralization potency of polyclonal human sera seen above, we made use of a panel of six, epitopemapped neutralizing monoclonal antibodies (NmAbs, Figure 2F, and Fig S1) [25][26][27][28] . We have devised a "squirrel" diagram to help visualise the binding sites of the various mAbs on the RBD (Figure 2d). One NmAb, FI 3A, a Class 1 RBD mAb (binds to the left side of the head of the squirrel) whose homotypic NT50 is of the order of 1 nM, is largely unaffected by the changes in B1.  29 ). In contrast, NmAb FD 11A and S309, which are Class 3 RBD mAb, that bind to the right haunch of the squirrel, and EY 6A, Class 4 mAb, that binds to the left haunch of the squirrel, appears to be unaffected by the mutations in the VOCs.
Polyclonal responses generated by different individuals to natural infection or in response to vaccination may include a varying proportion of antibodies to these and other neutralization epitopes. We also noted signi cant deviations in heterotypic neutralization potency against a currently approved reference serum 20/130 (NIBSC, Fig 2G)

Binding of antibodies to Alpha and Beta Coronavirus proteins
We probed the antibody-binding properties of sera from vaccinated, convalescent and control sera using a customised MSD coronavirus antigen array ELISA (Figure 3). We observed that sera from individuals receiving two doses of the P zer vaccine showed signi cantly higher binding to both SARS-CoV-2 spike and RBD compared to those receiving single dose and to convalescent individuals one month after infection (Fig 3 A, and B, respectively). The absence of antibody binding to N (Fig 3C) con rms that all vaccinated individuals are naïve for SARS-CoV-2 infection.
There was signi cant antibody binding to both SARS-CoV-1 and MERS spike protein in vaccinated and COVID-19 convalescent individuals compared to the negative control sera (Fig 3D & E, respectively). This was particularly marked for SARS-CoV-1 reactivity in fully vaccinated individuals, suggesting that vaccine can induce a broad response to widely shared epitopes, such as those exempli ed by EY 6A (see above) and CR3022 30 .
We also screened for antibody binding to the spike antigen of the four common coronaviruses circulating in the UK (Fig 3 F-I). There is a signi cant increase in binding to all four, particularly to the Betacoronavirus clade A isolates, HKU1 and OC43, in vaccinated and COVID-19 convalescent sera (p<0.0001). Binding to the Alphacoronavirus isolates, 229E and, to a lesser extent, NL63S was also greater in the post-boost vaccinees, but not in convalescent sera.

Prediction of heterotypic neutralization by immunoassay
Authentic virus neutralization assays require specialist staff and facilities that are not widely available, and access to reference isolates of virus that are laborious to distribute. Accordingly, we asked whether high throughput ELISA-style immunoassays could provide a degree of predictive value for heterotypic neutralization. Using the Mesoscale discovery (MSD) assay, we investigated the relationship between the binding activity of serum to homotypic SARS-CoV-2 S and its heterotypic neutralization potency against both SARS-CoV-2 VOCs, and its ability to bind to other pandemic and endemic human coronaviruses. The correlation coe cients are summarized and the individual data are shown in Figure 5. Unsurprisingly, binding to SARS-CoV-2 S correlates very strongly with binding to SARS-CoV-2 S in convalescent sera, though surprisingly less well in post-boost vaccinees, who, reassuringly, showed no reactivity to SARS-CoV-2 N. S binding in vaccinees predicted homotypic and heterotypic neutralization moderately (0.5 >Spearman r > 0.4, *). S binding offered no predictive value for either homotypic or heterotypic neutralization in convalescent sera. Interestingly, binding activity in post-vaccine sera to SARS-CoV-2 R predicted binding to MERS-CoV S very well (r = 0.69, ***) but not to SARS-CoV-1 S. While SARS-CoV-2 S binding activity in vaccinee sera predicted binding to the S of one endemic beta coronavirus (OC43 S r = 0.57, **; HKU1 S, r = 0.44, *) it did not predict binding to the other two endemic coronaviruses.

Discussion
Our results show that overall, both binding and neutralization by antibodies induced by prototypic S protein is diminished to S from recent variants of concern (VOC); B1.351 to a greater extent than B1.1.7.
This broad trend masks both qualitative and quantitative differences in antibody responses by individuals, whose serum may contain differing proportions of antibodies to neutralizing epitopes that we show here are sometimes conserved between strains and always reduced in VOCs, particularly to B1.351.
Although, in principle, neutralization results using pseudotype viruses and in vitro binding assays may not reliably predict the potency of heterotypic immune protection afforded by natural infection or immunization by prototypic SARS-CoV-2, our results using authentic virus largely substantiate provisional conclusions recently made using such data. Encouragingly, we nd that the majority of T cell responses in recipients of two doses of the BNT162b2 vaccine are generated by epitopes that are invariant between the prototype and two of the current variants of concern (B1.1.7 and B1.351). T cell responses to SARS-CoV-2 are known to target a wide range of regions in spike 31 , 8 . Moreover, in over 90% of these recipients, heterotypic neutralizing titres (NT50) remain comfortably above the level associated with immune protection in recent vaccine trials. However, in a majority of individuals whose homotypic neutralization titres were more modest -including over 50% of convalescent COVID-19 individuals and recipients of a single dose of vaccine -heterotypic neutralization dropped to negligible levels. This loss of crossneutralization was particularly notable against B1.351 with potential implications for vaccine e cacy in populations where this VOC dominates and when only moderate levels of S antibodies are generated after vaccination.
It should be noted that neutralization escape, observed in a well of a micro-titre plate, is not direct evidence of vaccine failure [32][33][34] . Non-neutralising antigen-speci c antibodies, T cells and innate lymphocytes clearly have the potential to contribute to vaccine e cacy 35 . The acceptance that prior infection with in uenza virus results in reduced disease against subsequent infection with heterosubtypic strains, in both human and animal challenge studies, provides further evidence that cellular components and non-neutralising antibodies make an important contribution to protection 33 . We also note that the recent South African and UK Novavax vaccine clinical trials showed 50-80% protective e cacy against infection for the B1.351 and B1.1.7 VOC respectively. Furthermore, cases of vaccinated individuals requiring hospitalization due to severe disease were extremely rare for these VOCs. Ongoing analysis of real-world vaccine roll out will illuminate the extent of vaccine breakthrough with VOCs.
Nevertheless, our results re-emphasize the urgent need to deploy the most effective vaccine strategies as widely and rapidly as possible in order to provide population protection against the emerging lineages of concern of SARS-CoV-2. Our ndings show clearly that the weaker responses generated for example by proteolytically removed. Following folding, trimer assembly and glycosylation in the ER and Golgi, the trans-Golgi localized protease, furin, cleaves the boundary between the S1 and S2 polypeptides.
Following binding of the receptor-binding domain (RBD, cyan) to ACE2 on host cells, cell-surface TMPRSS2 proteolytically cleaves the S2' site, facilitating conformational changes to spike that result in fusion of the virus envelope with the plasma membrane. Variant residue positions are indicated below, and their approximate location on the S polypeptide is indicated. Residue identities are shown at each of these positions for a prototype isolate, Victoria/01/2020 (VIC01, PANGO lineage B), and at each position in the three lineages of interest (B1.1.7, B1.351, and P.1) at which the respective lineage differs from prototype. Δ indicates deletion of one or more residues. Note, there are lineage-de ning substitutions outside RBD, in the N-terminal domain (NTD) and C-terminal domain (CTD) of S1 (dark blue), and in S2 (tan). These may include changes that directly or indirectly affect antibody-mediated neutralization by loss or altered dynamics of epitope, respectively.   Variants shown as a percentage of total spike response to Victoria strain as determined by ex vivo IFN-γ ELISpot C. T cell responses to 22 individual peptides in Victoria strain that have corresponding to mutations in B1.1.7, B1.351 and P1 variants. Each bar represents one volunteer with a positive response (de ned as a response to the peptide minus the background that was greater than a twice the background). N=24, SFC/106 PBMC = spot forming cells per million peripheral blood mononuclear cells, with background subtracted.

Figure 5
Cross-correlation of immune parameters. Pairwise Spearman correlation analyses were undertaken between the value of binding of post-boost vaccinee or convalescent serum antibody to SARS-CoV-2 S, as determined by the MSD immunoasssay platform (see Figure 3), and the homotypic and heterotypic neutralizing titre of the same sera (see Figure 2) and the MSD-determined binding to SARS-CoV-2 RBD and N, and to the S protein of other coronaviruses. Spearman's r parameter, in the associated two-tailed P value and its interpretation are given for each pairwise comparison. In the panels, below, the MSD values (left axis) and NT50 values (right axis) for each individual are plotted against the corresponding MSD value for SARS-CoV-2 S.