Prior SARS-CoV-2 infection rescues B and T cell responses to variants after first vaccine dose

During worldwide rollout of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccines, it is vital to understand how vaccination influences immune responses and protection among those who have had prior natural SARS-CoV-2 infection. This is a knowledge gap because a history of previous infection was an exclusion criterion in phase 3 vaccine trials (1). Countries have adopted diverse approaches—among them, the UK policy to maximize deployment of first doses to the largest possible number of people by extending the time interval to second dose. At the end of 2020, it became apparent that several virus variants had emerged (2, 3) and that these might affect vaccine rollout. The B.1.1.7 variant, possessing the spike Asn⁵⁰¹→Tyr (N501Y) mutation, first emerged in the UK in December 2020 and spread rapidly (4). Additional variants of concern (VOC) include the B.1.351 variant, which emerged at about the same time in South Africa, and the P.1 variant, which emerged in January 2021 in Brazil. In addition to the N501Y mutation, both of these variants have the E484K mutation, which is implicated in escape from neutralizing antibodies (nAbs) (5, 6).

The Pfizer/BioNTech mRNA vaccine BNT162b2 encodes a prefusion-stabilized, membrane-anchored SARS-CoV-2 full-length spike protein modified by two proline substitutions (1, 7, 8). A two-dose regimen of 30 μg BNT162b2, 21 days apart, confers 95% protection against Wuhan-Hu-1 SARS-CoV-2 (1), eliciting high nAb titers as well as CD4 and CD8 cell responses (8). When given as a single 60-μg dose, BNT162b1 induced virus Ab neutralization, but T cell responses were reduced compared with the standard prime-boost regime (8). A single 30-μg dose of BNT162b1 was not reported beyond day 21. However, the cumulative incidence of COVID-19 cases among 21,676 placebo and 21,699 vaccine recipients diverged 12 days after the first dose, indicating possible early-onset first-dose protection (1). For those who were previously infected, single-dose vaccination may act as a boost after natural infection. Therefore, we aimed to test the impact of prior SARS-CoV-2 infection on T and B cell responses to first-dose vaccination.

To do this, we analyzed T and B cell immunity after the first 30-μg dose of the Pfizer/BioNTech mRNA vaccine BNT162b2 in a cohort of UK hospital health care workers (HCW) (9–12). The COVIDsortium HCW cohort has been studied longitudinally since the end of March 2020, providing accurate infection and immune history in the context of genotyping, including human leukocyte antigen (HLA) imputation (10–12). Our aim was to compare T and B cell immunity after a first dose of vaccine in December 2020 in postinfection (after natural infection), vaccinated postinfection (vaccination in the context of prior SARS-CoV-2 infection), and vaccinated naïve (single-dose vaccination) individuals. We sought to explore whether there is evidence for altered T cell recognition of the B.1.1.7 and B.1.351 variants and, in particular, of the N501Y mutation shared by several VOC.

The UK has deployed a heterodox vaccination regimen to maximize immune protection and slow spread of the B.1.1.7 lineage, giving an initial 30-μg dose of BNT162b2 followed by boosting up to 12 weeks later (13). A cross-sectional substudy (n = 51 individuals) of the existing longitudinal HCW cohort (9–12) was recruited 22 (±2) days after the first dose. After the start of the study, the majority of acute infections had already occurred among this cohort (11). At the time of receiving their first vaccine dose in December 2020, prior to the emergence of VOC, 25 individuals were ~39 weeks removed from SARS-CoV-2 infection with the Wuhan-Hu-1 strain, and 26 were confirmed uninfected, having tested negative in longitudinal serology for spike and nucleocapsid (N) proteins (table S1 and fig. S1).

We first measured SARS-CoV-2 N antibody longitudinally up to 16 to 18 weeks, then at 28 to 30 weeks, and finally at 42 weeks after recruitment, to confirm that there was no laboratory evidence of new infection at the time of drawing blood for the vaccine study at 42 weeks; none of the previously uninfected HCW had become seropositive (Fig. 1A). T cell responses to spike protein and mapped epitope peptides (MEPs) in either postinfection, vaccinated postinfection, and vaccinated naïve individuals were compared (Fig. 1B). Ninety-six percent (22/23) of vaccinated postinfection individuals mounted a T cell response to spike protein compared with 70% (16/23) of vaccinated naïve individuals, with a fourfold increase in the magnitude of the T cell response. Furthermore, while the T cell response to spike protein in vaccinated naïve individuals increased (P = 0.0440), it was lower than that of vaccinated postinfection individuals (P = 0.0557) (Fig. 1C). As expected, there was no significant change in T cell response to N (a measure of immunity to natural infection) (fig. S2A).

Paired analysis of T cell immunity to spike protein in previously uninfected individuals, analyzed at the 16- to 18-week time point and 3 weeks after vaccination, showed a significantly increased response (P = 0.0089) (Fig. 1D). Three individuals who previously showed a response, despite lack of laboratory evidence for infection (therefore presumably a cross-reactive response to an endemic human coronavirus), showed an unchanged or decreased response to spike after vaccination.

The size of the SARS-CoV-2 spike subunit 1 (S1)–specific memory B cell (MBC) pool was investigated by B cell enzyme-linked immunosorbent spot (ELISpot) assay (Fig. 1E and fig. S2B). As for T cell responses, the number of S1-specific immunoglobulin G (IgG⁺) antibody-secreting cells (ASCs) was far greater in vaccinated postinfection individuals than in vaccinated naïve individuals (P < 0.0001). Prior infection generated a 63-fold increase in S1-specific ASCs. There were no preexisting S1-specific ASCs in uninfected HCW before vaccination. Twenty of 22 vaccinated naïve individuals had detectable S1-specific ASCs composing 0.02 to 1.54% of the MBC pool. By comparison, all vaccinated postinfection individuals had detectable S1-specific ASCs (1.90 to 50% of the MBC pool). We previously reported (14) spike receptor binding domain (RBD) enhanced Ab responses in the vaccinated postinfection group. In this work, the vaccinated naïve group attained antibody titers similar to those of the postinfection group at 16 to 18 weeks and 28 to 30 weeks (Fig. 1F). Vaccinated naïve individuals demonstrated a lower nAb response to wild-type virus than was seen after natural infection at 16 to 18 weeks, although this did not achieve statistical significance. In line with the findings for MBC and RBD binding, there was a significantly enhanced nAb response in vaccinated postinfection individuals compared with the vaccinated naïve group (Fig. 1G), with a mean value of 25,273 compared with 420, that is, a 60-fold increase. To put this in context, these values are 43-fold higher than the values recorded after two vaccine doses in the phase 1 trial (7). There was no correlation between the magnitude of the spike protein T cell response and the percentage of S1-specific ASCs (Fig. 1H). As expected, there was a positive correlation between the percentage of S1-specific ASCs and the serum titer of RBD antibody in the vaccinated postinfection individuals [correlation coefficient (r) = 0.6502; P = 0.0008] (Fig. 1I). After vaccination, two previously infected individuals showed lower percentages of S1-specific memory B cells and reduced serum RBD-specific antibody levels than the rest of the group; prior infection involving case-definition symptoms tended to be associated with a higher specific B cell frequency than milder disease (Fig. 1F and fig. S2C). These individuals who, despite infection, had also not shown a detectable T cell response (one never seroconverted, and the other rapidly became seronegative during longitudinal follow-up) had a poor or absent response to infection that was only minimally overcome by vaccination.

The data in Fig. 1 indicate that there is a strong prime-boosting effect of prior infection on single-dose vaccination. Augmentation is seen more strongly in MBC frequency, anti-RBD, and nAb responses than for T cell response frequency. Furthermore, there was no correlation between S1 ASC frequency and T cell response frequency (Fig. 1H). There is, however, a correlation between S1 ASC and RBD antibody titers, indicating that individuals with higher numbers of MBCs mount stronger antibody responses, and individuals who had experienced infection clustered at the higher end of this response (Fig. 1I).

Shortly before the vaccination program was initiated, several VOC emerged, including B.1.1.7. This variant has nine mutations in the spike protein. Several studies have reported weaker nAb responses to B.1.1.7 relative to the previously circulating Wuhan-Hu-1 strain (2–6, 15–18). The majority of SARS-CoV-2 immune naïve individuals made no nAb response to the B.1.1.7 (18/20) and B.1.351 (17/20) variants after single-dose vaccination. In contrast, almost all vaccinated postinfection individuals made a strong nAb response to the B.1.1.7 (24/24) and B.1.351 (23/24) variants after a single-dose vaccination, with a 46-fold (B.1.1.7) and 63-fold (B.1.351) increase in mean nAb half-maximal inhibitory concentration (IC₅₀) in vaccinated postinfection individuals compared with vaccinated naïve individuals. In a paired analysis, we observed in vitro significantly reduced nAb potency to authentic B.1.1.7 variant (mean: 35) with a 96% fall compared to that of Wuhan-Hu-1(mean: 866; P < 0.0001) in sera from individuals with a past medical history of natural infection (Fig. 2B). Worryingly, after single-dose vaccination, 90% (18/20) of vaccinated naïve individuals showed no detectable nAbs (IC₅₀ < 50) against B.1.1.7 (mean IC₅₀: 37; range: 0 to 184; P = 0.2090), but they did show demonstrable nAb responses to Wuhan-Hu-1 SARS-CoV-2 virus (mean IC₅₀: 420; range: 80 to 2004; P = 0.0046). In contrast, all vaccinated postinfection individuals responded to single-dose vaccination with substantially enhanced nAb responses, neutralizing not just Wuhan-Hu-1 SARS-CoV-2 (mean IC₅₀: 25,273; range: 581 to 76,369) but also the B.1.1.7 (mean IC₅₀: 1717; range: 52 to 4919) and B.1.351 (mean IC₅₀: 5451; range: 41 to 20,411) variants (Fig. 2, A and B, and fig. S3). We show a 93% reduction in neutralization (IC₅₀) responses to the SARS-CoV-2 B.1.1.7 variant (mean: 1717) compared with the Wuhan-Hu-1 (mean: 25,273) virus in vaccinated postinfection individuals. However, despite this fall, the majority (22/24) remain within a “protective threshold.” This was not the case for vaccinated naïve individuals. There was a 91% reduction in neutralization (IC₅₀) responses against the SARS-CoV-2 B.1.1.7 variant (mean: 37) compared with the Wuhan-Hu-1 virus (mean: 420), resulting in the majority of individuals (19/20) falling below the “protective threshold.” This result was mirrored in the SARS-CoV-2 S1-specific MBC pool, where reduced numbers of S1-specific IgG⁺ ASC are seen (in vaccinated naïve individuals compared with vaccinated postinfection individuals) responding to S1 antigen containing the N501Y, K417N, and E484K mutations. Prior infection substantially enhances the specific MBC pool after single-dose vaccination (Fig. 2C). We looked at correlations between RBD binding antibodies, B cell responses, T cell responses, and IC₅₀, comparing neutralization of Wuhan-Hu-1, B.1.1.7, and B.1.351 live virus (Fig. 2D). Despite the lower neutralization of B.1.1.7 and B.1.351 variants, the pattern was retained of strong correlation between RBD antibody titer and S1-specific B cell frequency and neutralization and somewhat weaker correlation between T cell response and neutralization.

A lack of Ab-mediated protection in single-dose vaccinees could be mitigated by a broader repertoire of T cell responses (18). To investigate differences in T cell recognition, we designed peptide pools covering the affected regions of Wuhan-Hu-1, B.1.1.7, and B.1.351 variant sequence (table S2). We compared T cell responses to these peptide pools in peripheral blood mononuclear cells (PBMCs) from vaccinated postinfection and vaccinated naïve individuals (Fig. 2E). Responses in postinfection vaccinees were in general higher than in the vaccinated naïve individuals (note an enhanced response to the B.1.1.7 peptide pool). T cell responses were heterogeneous; responses to variant pools could be either higher or lower than to Wuhan-Hu-1 pools. Alterations in affinity for the T cell receptor can lead to altered peptide ligand effects and differential polarization of cytokine effector programs, as we have previously observed in Zika virus infection (19). We wondered whether this was also occurring for SARS-CoV-2; however, we found no evidence for immune deviation to interleukin (IL)–4, IL-5, IL-10, IL-13, IL-17A, or IL-23 (fig. S3).

For B.1.1.7 and B.1.351, attention has centered on the N501Y mutation, as this is implicated in altered angiotensin-converting enzyme 2 (ACE2) binding and enhanced infectivity and transmission but is also a target for B and T cell recognition. We initially looked at T cell responses after natural infection and found that at 16 to 18 weeks postinfection, the N501Y mutation appeared to have no substantial differential impact on the T cell response (Fig. 2F), unlike nAb recognition (5).

The specific impact of any T cell epitope changes on the immune response against VOC depends on changes in peptide binding to the peptide-presenting HLA molecules. Because the HLA complex is the most polymorphic part of the human genome, any alteration to core HLA binding motifs will differentially affect people with certain HLA alleles over others. We performed in silico analysis (using NetMHCIIpan) to predict which of the B.1.1.7 and B.1.351 mutations were found in HLA core binding motifs and how this might affect binding to common HLAII alleles (DRB1*0101, DRB1*0301, DRB1*0401, DRB1*0701, DRB1*1101, DRB1*1301, and DRB1*1501) (tables S3 and S4). Some of the mutations did not fall in a region predicted to bind the HLAII alleles tested (D3L, T716I, T1001I, A1708D, and 3675-7 SGF del). Although several mutations were not predicted to significantly change affinity for the HLAII alleles, others did show predicted differential affinities depending on host HLAII type (tables S3 and S4). Analyzing altered responses to the D1118H mutation, we noted that individuals who carried DRB1*0301 and DRB1*0401 showed enhanced T cell responses to the Wuhan-Hu-1 peptide compared with those who did not (P = 0.0072) (Fig. 2G). T cell responses to the variant peptide appeared to be reduced in individuals carrying DRB1*0301 and DRB1*0401 (Fig. 2H). There is a basis for this in terms of differential HLAII binding as the D-to-H mutation is predicted to lose the T cell epitope for people carrying DRB1*0301 and DRB1*0401 but not, for example, in those who carry DRB1*0701 or DRB1*1501, who would be predicted to show an enhanced response (table S3). People carrying DRB1*1301 are predicted to gain a response as a consequence of this mutation. Analyzing responses to the E484K mutation seen in B.1.351 and P.1 variants, we noted that it did not fall in a region predicted to bind the HLAII alleles tested (table S4). The mutation appeared to have no substantial or differential impact on T cell responses (Fig. 2I).

When we primed transgenic mice expressing human HLA-DRB1*0401 with the Wuhan-Hu-1 peptide pool, T cell responses to the B.1.1.7 variant peptide pool were significantly reduced (P = 0.0286) (Fig. 2J). Furthermore, the T cell response to the spike N501Y mutation common to all three of the current VOC was ablated.

In this HCW cohort, vaccinated naïve individuals made an anti-S1 RBD Ab response with a mean titer of ~100 U/ml at 22 (±2) days after vaccination, roughly equivalent to the mean peak Ab response after natural infection (14). However, the spike T cell response after one dose was lower than after natural infection, and for 30% of vaccinees, no response could be measured. However, T cell responses are enhanced fourfold in those vaccinated postinfection. This T cell enhancement is small relative to the 63-fold change in ASCs and the corresponding 140-fold change in Roche anti-S (RBD) Ab levels we observed after one vaccine dose in HCW vaccinated postinfection (14). While much has been written about the impact of rapidly waning serum antibodies, our findings confirm that MBCs are nevertheless primed and able to contribute a rapid, large response to repeat exposure. The rather large effect on B cell priming and restimulation, relative to T cells, in previously infected single dose–vaccinated individuals may reflect the fact that, among the nuanced differences between the licensed SARS-CoV-2 vaccines, aspects of the mRNA adjuvant effect appear to skew immunity to high nAb titers, which may underpin its high efficacy. Our evidence for enhanced vaccine responses after infection supports the case that only one vaccine dose is necessary to maximize immune protection for SARS-CoV-2–experienced individuals (14, 20).

It is notable that the high IC₅₀ titers in those vaccinated after infection provide such a large protective margin that responses to authentic B.1.1.7 and B.1.351 variants are also high. In contrast, nAb responses in individuals several months on from mild infection show much lower IC₅₀ titers against B.1.1.7 and B.1.351, often less than 100. Similarly, the majority of responses in naïve individuals after one dose show weak recognition of B.1.1.7 and B.1.351. This finding indicates potentially poor protection against B.1.1.7 and B.1.351 in individuals who have experienced natural infection or who have only had one vaccine dose.

It is important to map the effect of VOC mutations on any evasion of T cell immunity. The case has been made that reductions in antibody neutralization of mutant spike may be mitigated by protective T cells (8). A case has been made for the role of T cells as correlates of protection (21). Our evidence from this analysis 22 (±2) days after one dose is that T cell immunity is mostly variably low but also relatively unperturbed by the N501Y mutation. The other mutations we considered that overlay CD4 epitopes were, as might be predicted, distributed across the range of HLAII polymorphisms. Those alleles associated with loss of CD4 response to the variant pool tended to be those with a lysine in pocket 4 of the groove (HLA-DR residue 71β), whereas those with an increased response to the variant pool tended to be those with a smaller amino acid, alanine. In HLA-DRB1*0401 transgenics, we confirmed that in the context of a given HLAII heterodimer, the N501Y mutation can result in ablation of this part of the T cell response, demonstrating that HLA polymorphisms are likely to be significant determinants of responder and nonresponder status with respect to vaccine escape.

SARS-CoV-2 immunity now encompasses postinfection plus either zero, one, or two vaccine doses and first and second dose naïve vaccinated. Single-dose vaccination after infection achieves similar levels of S1 RBD binding antibodies to two doses in naïve vaccinated individuals and second-dose vaccination in one-dose vaccinated postinfection individuals offers no additional enhancement (22). Moving forward, it will be important to resolve the quantitative and qualitative differences between these groups in terms of neutralizing antibody repertoire as well as phenotype and durability of memory B and T cell responses. Durability of immunity to natural infection and after vaccination as well as sustained vaccine efficacy and vaccine escape need to be monitored over time.