key: cord-1037504-iyz2ykdj
authors: Rodda, Lauren B.; Morawski, Peter A.; Pruner, Kurt B.; Fahning, Mitchell L.; Howard, Christian A.; Franko, Nicholas; Logue, Jennifer; Eggenberger, Julie; Stokes, Caleb; Golez, Inah; Hale, Malika; Gale, Michael; Chu, Helen Y.; Campbell, Daniel J.; Pepper, Marion
title: Imprinted SARS-CoV-2-specific memory lymphocytes define hybrid immunity
date: 2022-03-17
journal: Cell
DOI: 10.1016/j.cell.2022.03.018
sha: df17f34207807d74d2f873bca3381fe4d2a44157
doc_id: 1037504
cord_uid: iyz2ykdj

Immune memory is tailored by cues that lymphocytes perceive during priming. The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic created a situation in which nascent memory could be tracked through additional antigen exposures. Both SARS-CoV-2 infection and vaccination induce multifaceted, functional immune memory, but together they engender improved protection from disease, termed hybrid immunity. We therefore investigated how vaccine-induced memory is shaped by previous infection. We found that following vaccination, previously infected individuals generated more SARS-CoV-2 RBD-specific memory B cells and variant-neutralizing antibodies and a distinct population of IFN-γ and IL-10-expressing memory SARS-CoV-2 spike-specific CD4+ T cells than previously naive individuals. While additional vaccination could increase humoral memory in previously naive individuals, it did not recapitulate the distinct CD4+ T cell cytokine profile observed in previously infected subjects. Thus, imprinted features of SARS-CoV-2-specific memory lymphocytes define hybrid immunity.

The generation of immune memory is influenced by signals that B and T lymphocytes of the adaptive immune system perceive during a primary immune response. This system has evolved such that key cues from an invading pathogen or vaccine are relayed to lymphocytes so they can provide specific functional outputs to combat infection and protect from re-infection. The site of antigen encounter, specific inflammatory signals, and the number, timing, and frequency of antigen exposures all influence the resulting memory pool, yet the specific rules governing memory formation remain undefined.

Understanding how the confluence of these parameters influences immune memory function and maintenance is critical for optimizing protective vaccines.

In the context of the ongoing SARS-CoV-2 pandemic, so-called "hybrid-immunity" induced by a combination of prior SARS-CoV-2 infection and subsequent vaccination, provides greater protection against re-infection and severe COVID-19 disease than either infection or vaccination alone (Abu-Raddad et al., 2021; Crotty 2021; Gazit et al., 2021; Goldberg et al., 2021; Reynolds et al., 2021) . Studies investigating immune correlates of hybrid-immunity have revealed improved breadth and neutralizing ability of circulating antibodies from previously infected individuals (Cho et al., 2021; Schmidt et al., 2021b; Stamatatos et al., 2021; Wang et al., 2021) , but the specific changes in the cellular immune compartment associated with this immune state remain undefined. Additionally, it is not understood whether further activation of immune memory in vaccinated only individuals could achieve similar qualities, as previously infected individuals have had an additional antigen exposure.

J o u r n a l P r e -p r o o f 4 To answer these questions, we tracked circulating SARS-CoV-2-specific antibodies and memory lymphocytes in a cohort of Naive (N) or SARS-CoV-2-Previously Infected (PI) subjects over the course of three vaccinations. We focused on visualizing RBD-specific antibodies and B cells, and spike (S)-specific CD4 + T cells, as they are critical mediators of protection in infected individuals over a two-year time period and induced by SARS-CoV-2-directed vaccines (Corbett et al., 2021; Feng et al., 2021; Gilbert et al., 2021; Khoury et al., 2021) .

We find that the immune memory landscape elicited by vaccination of previously infected subjects is distinct from the immune memory of SARS-CoV-2-naive individuals. Three months after two doses of vaccination, PI individuals maintained a higher quantity of RBDspecific plasma antibodies with superior plasma neutralization of viral variants including omicron. PI individuals also retained increased numbers of RBD-specific B cells enriched for IgG + classical and activated memory B cells (MBCs) compared to N individuals.

However, these differences were normalized by administration of a third vaccine dose.

Examination of the T cell compartment revealed that although N and PI individuals generated equivalent numbers of S-specific CD4 + T cells after vaccination, there was a profound functional skewing towards a Th1 phenotype in PI subjects. CD4 + T cell functional differences between N and PI subjects persisted following the administration of a vaccine booster. Further, although the third dose of vaccine provided a boost in circulating antibodies, we observed no increase in memory T or B cells, indicating that the immune memory compartment is likely maximized after the two-dose regimen. Thus, J o u r n a l P r e -p r o o f 5 our data support a model in which the priming environment induced by SARS-CoV-2 infection imprints immune memory with multiple features of enhanced type-1 antiviral immunity. These likely contribute to the increased protection associated with hybrid immunity and are not fully recapitulated by repeat vaccination.

To investigate how prior SARS-CoV-2 infection impacts the quantity and quality of vaccine-induced SARS-CoV-2-specific immune memory, we collected plasma and peripheral blood mononuclear cells (PBMCs) from 24 SARS-CoV-2 N and 30 SARS-CoV-2 PI individuals before vaccination and 1 week, 3 months and 6 months after two vaccinations with BNT162b2 (Pfizer-BioNTech) or mRNA-1273 (Moderna) COVID-19 vaccines ( Figure 1A and Table 1 ) (Baden et al., 2021; Polack et al., 2020 ). An average of 8 months following initial vaccination, 10 participants from each group received a third vaccine dose and we collected samples 2 weeks (mean: 18 days, SD: 8 days) later. N participants had no detectable SARS-CoV-2 RBD-specific IgG plasma antibodies above a cutoff set by historical negative controls prior to vaccination ( Figure S1A) . PI participants reported a positive SARS-CoV-2 test and mildly symptomatic disease an average of 9 months before their pre-vaccination blood draw.

To determine the distinct humoral features of hybrid immunity, we first compared the circulating SARS-CoV-2 RBD-specific B cells in N and PI participants pre-and post-J o u r n a l P r e -p r o o f 6 vaccination. Tetramer enrichment allowed us to identify rare SARS-CoV-2 RBD-specific B cells from PBMCs to track the memory response ( Figure S1B ). Pre-vaccination, PI individuals had significantly more RBD-specific antigen-experienced B cells (CD21 + CD27 + and CD21 -CD27 +/-) (Cancro and Tamayko, 2021) (Figures 1B and 1C , S1C) and plasma RBD-specific IgG and IgA than N individuals (Figures 1D and 1E) , reflecting the sustained memory response to their previous SARS-CoV-2 infection (Rodda et al., 2021 . Two dose vaccination induced a robust humoral response in both N and PI individuals including increased numbers of RBD-specific antigenexperienced B cells (Figure 1C) , an acute burst of RBD-specific plasmablasts ( Figure   S1D ) and increased titers of RBD-specific IgG and IgA antibody (Figures 1D and 1E) .

The numbers of RBD-specific antigen-experienced B cells and RBD-specific antibodies in both groups contracted to levels higher than those at pre-vaccination. However, PI individuals formed and retained higher numbers of RBD-specific antigen-experienced B cells and elevated RBD-specific IgG and IgA compared with N individuals at 3 and 6 months post-vaccination (Figures 1C-E) . These findings corroborate recent work and support the likely functional role of the elevated RBD-specific antibody and MBCs in hybrid immunity .

To determine if a third vaccine dose could overcome these differences by engaging the MBCs in N individuals in additional expansion, we measured the humoral response in both groups 2 weeks after a third dose. While RBD-specific B cells in N individuals responded by increasing RBD-specific IgG and IgA antibody to the levels in PI individuals, the antibody response in N individuals to a third SARS-CoV-2 antigen exposure did not J o u r n a l P r e -p r o o f 7 match the acute response in PI individuals to their third exposure (vaccination 2) ( Figures   1D and 1E ). This muted response to a third exposure in N and PI individuals may reflect efficient clearance of antigen by the high quantity of antigen-specific memory now established in both groups, and the increase in RBD-specific antibody in both groups is likely driven largely by short-lived plasmablasts at two weeks post-third vaccination.

Comparison of N and PI subjects several months post-third vaccination is required to determine if a the third vaccination dose induces sustained expansion of RBD-specific MBCs and circulating antibody titers in N individuals that more closely match those in PI subjects.

CD4 + T cell responses to vaccination were also assessed in N and PI individuals using an activation induced marker (AIM) assay based on expression of CD69 and CD137 to detect SARS-CoV-2 specificity (Tarke et al., 2021) (Figures S2A-S2F) . PBMCs from N or PI participants isolated before and after vaccination were stimulated with SARS-CoV-2 spike (S) peptide pools optimized for the induction of MHC class II-dependent CD4 + T cell responses (15-mers), and containing T cell epitopes across a range of common HLA haplotypes against the viral membrane and nucleocapsid (M/N) or S proteins (Data S1).

Pre-vaccination, CD4 + T cells from PI participants showed responses to both M/N and S, consistent with prior SARS-CoV-2 infection and the presence of memory cells (Figures   2A and 2B) , whereas responses in N participants were undetectable above individual J o u r n a l P r e -p r o o f 8 donor background (Figures S2G and S2H) . Throughout the course of vaccination, responses to M/N remained negative in N participants, but endured, undiminished in PI participants at all time points, indicating that SARS-CoV-2-specific CD4 + T cells primed during infection, persist at least 18 months following initial antigen exposure ( Figure 2B ).

Following initial two-dose vaccination, S-reactive AIM + CD4 + T cells were significantly elevated above pre-vaccination levels in both N and PI individuals, with no significant differences in peak CD4 + T cell response between groups (Figures 2B, S2G, and S2H) .

Therefore COVID-19 vaccination induces numerically equivalent populations of S-specific cells in all subjects, regardless of prior SARS-CoV-2 exposure status, that persist for at least three months.

We next examined the relative composition of effector and memory populations, including circulating T follicular helper (Tfh)-like cells (Ueno, 2016) , within the SARS-CoV-2reactive T cell compartment in N and PI subjects. Prior to vaccination, there was no difference in the frequency of S-reactive central memory (CM) T cells between N and PI groups, however PI individuals maintained significantly elevated frequencies of effector memory (EM and EMRA) and circulating Tfh cells (Figures 2C and 2D) . Two-dose vaccination induced a significant increase in nearly all AIM + effector and memory populations in both groups, with EM and Tfh populations dominating the response ( Figure   2D ). In addition, two-dose vaccination was sufficient to normalize numerical differences identified in pre-vaccination N and PI participants, while a third dose was not associated with any additional increase in these populations in either N or PI participants ( Figure   2D ). Therefore, vaccination induces numerically equivalent S-specific memory Tfh and J o u r n a l P r e -p r o o f 9 EM CD4 + T cell responses in N and PI individuals that do not further increase in response to an additional vaccine dose.

Memory B cells function by activating, proliferating and forming antibody-secreting plasmablasts more rapidly than naive B cells (Cancro and Tamayko, 2021) . Prior to vaccination, RBD-specific B cells in N individuals were predominantly Naive (CD21 + CD27 -) while RBD-specific B cells in PI individuals were predominantly non-Naive antigen-experienced B cells (CD21 + CD27 + and CD21 -CD27 +/-) (Figures S3A and S3B) .

(CD21 -CD27 +/-CD11c + ) (Cancro and Tomayko, 2021) in both groups at 1 week postvaccination with PI individuals displaying higher numbers than N individuals (Figures 3A   and 3B) . Although CD21 -CD27 -CD11c + 'atypical MBCs' have been associated with chronic infection and autoimmunity, the expression pattern of these markers on MBCs is also associated with recently activated B cells following acute antigen re-exposure. While the proportion of CD21 -CD11c + activated MBCs decreased in both groups at 3 months post-infection, PI individuals retained higher numbers and a higher proportion than N individuals ( Figure 3B ). This may reflect prolonged antigen retention and a longer or larger ongoing germinal center (GC) response, which would contribute to the increased quantity of RBD-specific antigen-experienced cells including classical MBCs and long lived plasma cells producing RBD-specific antibodies in PI individuals with hybrid immunity (Knox et al., 2017; Kim et al., 2019) . J o u r n a l P r e -p r o o f 10 PI individuals also displayed higher proportions and numbers of IgG + classical MBCs (cMBCs, CD21 + CD27 + ) among RBD-specific antigen experienced cells compared to N donors prior to vaccination, indicative of their previous SARS-CoV-2 infection ( Figure   3C ). IgG + cMBCs rapidly form IgG-secreting plasmablasts upon antigen exposure and are likely important to protection (Laidlaw and Ellebedy, 2021) . Due to the expansion of CD21 -CD11c + activated MBCs, IgG + cMBCs made up a reduced proportion of the RBDspecific antigen experienced cells in both groups at 1 week post-vaccination. In response to vaccination, the numbers of IgG + cMBCs increased in both groups at 1 week postvaccination. The numbers of IgG + cMBCs in N individuals further increased at 3 months post-vaccination, but did not quite reach the numbers of IgG + cMBCs in PI subjects at this time point. N and PI individuals presented similar frequencies of IgG + cMBCs at 3 months post-vaccination. This quantitative gap between N and PI individuals was subsequently normalized by a third dose of vaccine ( Figure 3C) . Thus, the higher number of IgG + cMBCs in PI individuals likely contributes to immune protection associated with hybrid immunity and this benefit can also be achieved by three vaccinations. While mucosal infections such as SARS-CoV-2 are associated with IgA + MBCs particularly at the site of infection, PI individuals had only a small proportion of circulating IgA + RBD-specific antigen-experienced B cells at 9 months post-infection and they were not further induced in N or PI individuals by vaccination ( Figure S3C ).

To assess if infection prior to vaccination impacts the breadth or ability of SARS-CoV-2 specific MBCs and antibodies to recognize and combat SARS-CoV-2 variants of concern J o u r n a l P r e -p r o o f 11 (VOCs), we measured the ability of RBD-specific MBCs and antibody to recognize the B.1.351 (Beta, β) variant and plasma antibody to neutralize the B.1.617.2 (Delta, Δ) and B.1.1.529 (Omicron, o) variants. SARS-CoV-2(β) and SARS-CoV-2(o) have been found to significantly reduce neutralization by vaccine-induced antibodies as compared to the original SARS-CoV-2(Wu-1) strain which was also used to make the mRNA vaccines relevant to this study (Cameroni et al., 2021 , Collier et al., 2021 . Using SARS-CoV-2 RBD(Wu-1) and SARS-CoV-2 RBD(β) tetramers simultaneously (Figure S1B ), we found that the majority of RBD-specific antigen-experienced B cells in all N and PI participants could recognize both variants (Figures S3D and S3E ). This technique does not allow us to distinguish cross-reactive cells that bind epitopes shared by the variant RBDs from cells that bind epitopes with the 3 amino acid differences between the variants. While vaccination induced increased RBD(β)-specific IgG antibody titers in both groups, PI individuals had significantly higher RBD(β)-specific IgG titers than N individuals pre-and 6 months post-vaccination ( Figure S3F ). This is corroborated by similar recent findings (Stamatatos et al. 2021) . Additionally, we found that plasma from PI individuals 6 months post-vaccination was significantly better at neutralizing SARS-CoV-2(WA-1) virus ( Figure   3D and S3G) as well as pseudotyped SARS-CoV-2(Δ)-spike and SARS-CoV-2(o)-spike lentiviruses than plasma from N individuals (Figures 3E and S3G) . SARS-CoV-2(WA-1) neutralization was highly correlated with RBD(Wu-1)-specific IgG antibody titers as we and others have shown previously, and the higher titers in PI individuals post-vaccination likely contribute to the improved variant neutralization over N individuals ( Figure   S3H ) (Rodda et al., 2021; Goel et al., 2021) . The increased antibody response to a third vaccine dose in N individuals (Figures 1D, 1E and S3F) correlated with sharply increased J o u r n a l P r e -p r o o f 12 neutralization capacity of the plasma antibody against the variants tested reaching the levels in PI individuals pre-and post-third vaccination (Figures 3D and S3F) . Thus, vaccination induces robust humoral memory in N and PI participants, but a third dose is needed for N individuals to achieve the breadth and VOC neutralizing activity of the plasma antibody formed in hybrid immunity.

During T cell priming, the context of the initial antigen exposure directs CD4 + T cells towards different helper and memory fates with distinct functional capacities. To determine how a primary infection versus vaccination or total number of antigen exposures differentially impacted functional outcomes, we compared the CD4 + T helper lineage composition of SARS-CoV-2 AIM + T cells based on differential chemokine receptor expression in N and PI subjects. Prior to vaccination, we found that S-reactive T cells in PI subjects were primarily Th1 and Th1/17 memory cells (Figure 4A and 4B) . In response to vaccination, S-reactive cells displayed features of activation, such as the downregulation of CD127 (Figure S4A) , and the frequency of S-specific cells increased in nearly all T helper subsets examined in both N and PI individuals, consistent with previous reports Guerrera et al., 2021; Painter et al., 2021) . However, PI subjects maintained an elevated frequency of both CXCR3 + CCR6 -Th1 and CXCR3 + CCR6 + Th1/17 cells, whereas N participants showed increased numbers of CXCR3 -CCR6 -CCR4 + Th2 cells by 3 months post-vaccine two ( Figure 4B ). Th1 and Th1/17 AIM + cells continued to trend higher in PI individuals following a third vaccine J o u r n a l P r e -p r o o f 13 dose, but this difference did not reach statistical significance. In addition, at this acute timepoint post-third vaccine, the frequency of Th2 populations were normalized between groups ( Figure 4B) . These data demonstrate that T helper polarization in response to initial vaccination differs according to prior SARS-CoV-2 infection status, and that additional vaccine doses may have a normalizing effect on some aspects of this response.

To complement the phenotypic analysis of CD4 + T cells, we employed a similar assay to directly interrogate cytokine production from SARS-CoV-2-specific CD4 + T cells following stimulation with M/N or S peptide pools. As demonstrated previously, CD154 and CD69

can be used to identify activated, antigen-reactive T cells (Reiss et al., 2017 ) that represent the primary cytokine producers in response to stimulation (Figures S4B and   S4C ). Importantly, the response measured in CD154 + CD69 + SARS-CoV-2 reactive T cells to peptide stimulation mirrors that of AIM + CD69 + CD137 + cells shown in Figure 2 , including an M/N response exclusive to PI subjects and robust induction of S-specific cells following two-dose immunization that was maintained after a third vaccination dose in both groups (Figures S4D-S4F) .

In PI subjects, we measured a strong cytokine response to infection in M/N-(Figures 4C-4F , S6A, and S6B) that was dominated by the production of IFN- and IL-2, but also included modest levels of IL-10. Cytokine production in response to M/N peptides in PI individuals was not enhanced by vaccination ( Figures S5A-E) . Neither N nor PI donor S-reactive T cells produced any detectable IL-J o u r n a l P r e -p r o o f 14 4 ( Figures 4G and 4H) , CD107a, (Figures S6C-S6H ) prior to vaccination. The response in N individuals to two-doses of vaccination generated a functionally diverse CD4 + T cell response, exhibiting strong induction of IFN-, IL-10, IL-4 ( Figures 4C-4H) , IL-2, and IL-21 ( Figures S6A-D) , while we observed no increased expression of IL-17A, CD107a, or IL-22 over pre-vaccination levels ( Figures S6F-H) . IL-4 production, a hallmark of both Th2 and Tfh CD4 + T cell responses (Ruterbusch et al., 2020) , was enriched specifically at one week post-vaccination in N individuals ( Figures   4G and 4H) . However, few spike-reactive T cells co-produced IL-4 and IL-13 ( Figure   S6E ), suggesting the absence of bona fide Th2 polarization. Consistent with our phenotypic analysis of T helper polarization based on differential chemokine receptor expression, PI individuals not only exhibited a strong induction of IFN--producing Sreactive CD4 + T cells following vaccination, but this response remained significantly elevated compared with the N group following both second and third vaccine doses ( Figures 4C and 4D) . PI donor CD4 + T cells also produced higher levels of IL-10 after two-dose vaccination compared with N participants, which also trended higher after a third vaccine dose despite failing to reach statistical significance (Figures 4E and 4F) .

Production of IL-21, although not detected following infection, was observed at comparable levels in both N and PI individuals following two-and three-dose vaccination ( Figures S6C and S6D) .

The S-reactive CD4 + T cell compartment was further distilled into subpopulations using a dimensionality reduction analysis on pooled CD154 + CD69 + cells from all participants.

UMAP followed by FlowSOM analysis identified 15 distinct clusters, of which 13 were J o u r n a l P r e -p r o o f 15 expanded following infection or vaccination, and 6 included cells which produced cytokine(s) (Figures 5A-P) . Two of these clusters, CL4 and CL13, were specifically enriched in PI donor T cells following vaccination. CL4 was defined largely by the production of IFN- and low CD127 expression, indicative of terminally differentiated Th1 cells ( Figure 5E ). CL13 was defined by co-production of IFN- and IL-10, while also displaying low CD127 expression, resembling FOXP3type-1 regulatory (Tr1) cells ( Figure 5N ) (Jankovic et al., 2007; Sun et al., 2009 ). In contrast, increased vaccineinduced T cells producing IL-4 were specifically enriched in CL15 only in N participants.

These cells additionally co-produced IL-2, IFN-, and small amounts of IL-21 ( Figure 5P ), suggesting a Tfh-like phenotype (Olatunde et al., 2021) . Collectively, this unbiased analysis of the S-reactive CD4 + T cell compartment identifies the emergence of functionally distinct subsets which differ following vaccination in PI and N participants according to prior SARS-CoV-2 infection history.

To determine how the number of antigen exposures may alter distinct cytokine production profiles seen in N and PI individuals, we compared S-reactive T cells from N individuals after three doses of vaccination (3 antigen exposures) to those from PI individuals after two doses of vaccination (3 antigen exposures). A comparison of AIM + S-specific CD4 + T cells in triple vaccinated N subjects to double vaccinated PI individuals normalized differences between Th1 and Th1/17 frequencies as assessed by CXCR3 and CCR6 expression (Figures 4A and 4B) . We also found similar frequencies of CM, EM, EMRA and and Tfh cells when normalizing for the number of antigen exposures (Figures 2B   and 2D) . However, when cytokine production was assessed between triple vaccinated N J o u r n a l P r e -p r o o f 16 participants and double vaccinated PI participants, the PI group maintained a dramatically elevated frequency of IFN--and IL-10-producing S-specific CD4 + T cells compared with the N group (Figures 4C and 4D) . These data indicate that T cell priming by infection or vaccination promotes distinct effector states in which phenotypic outcomes can be normalized by additional antigen exposure, while functional differences persist through multiple vaccine doses.

Protective immunity generated in response to infection or vaccination is characterized by an expanded population of antigen-specific memory cells that rapidly expresses effector molecules upon antigen re-exposure. In order to prevent or control reinfection, memory cells are functionally tailored by the initial priming event, however it is unclear how subsequent antigen exposures alter the quantity and quality of the response. The emergence of a novel coronavirus and expeditious generation of vaccines provided an unprecedented opportunity to examine the formation of a nascent antigen-specific memory pool in human volunteers and the impact of previous infection on the quantity, quality and durability of vaccine-induced immune memory. This is particularly important as SARS-CoV-2 infection prior to vaccination, referred to as "hybrid immunity", provides a protective advantage over vaccination alone (Abu-Raddad et al., 2021; Crotty et al. 2021; Gazit et al., 2021; Goldberg et al., 2021) .

In this study, we examined the distinguishing features of hybrid immunity to SARS-CoV-2 in comparison to vaccination alone over time and with subsequent antigen exposures.

We found that infection prior to vaccination induced higher numbers of RBD-specific MBC, more sustained IgG and IgA circulating antibodies, and increased plasma neutralization of multiple VOCs including SARS-CoV-2 (omicron). RBD-binding and neutralizing antibody has been proposed to be essential to the enhanced protection described in PI individuals after vaccination (PV2) (Cho et al., 2021; Schmidt et al., 2021a; Schmidt et al., 2021b; Stamatatos et al., 2021; Wang et al., 2021) . These features were normalized following booster vaccination, and therefore could be a function of the number of antigen exposures. However, while SARS-CoV-2 infection and COVID-19 immunization both induce robust Th1 immune memory, as found here and by other groups Guerrera et al., 2021; Painter et al., 2021) , we found that the spike-specific CD4 + T cells induced in previously infected and then vaccinated individuals exhibited a unique IFN- and IL-10 cytokine-producing profile not recapitulated in N individuals by repeat vaccination. This represents the first immune correlate of hybrid immunity that is stably imprinted on responding cells during initial infection and could contribute to protection from symptomatic infection by providing an antiviral program.

There are many parameters that could contribute to the distinct vaccine-induced immune responses seen in N or PI individuals including the timing, route or inflammatory conditions associated with a given SARS-CoV-2 antigen exposure. In this study, we were able to control for the number of antigen exposures by comparing double and triple vaccinated N and PI individuals. However, the timing of exposures varies substantially between the groups. In PI individuals, the first and second exposures to antigen were separated by ~10 months, whereas in N individuals, the time between first and second J o u r n a l P r e -p r o o f 18 exposures was only ~3.5 weeks. Longer intervals between vaccinations in N individuals results in more neutralizing antibody and more SARS-CoV-2-specific B cells which could then drive a larger response given a subsequent antigen exposure (Payne et al. 2021 ).

The benefit of time was seen again in N individuals receiving their third vaccine 8 months after their second, which likely helped to normalize quantitative differences in SARS-CoV-2-specific antibody and MBCs seen in N and PI individuals post-vaccination 3.

We found that RBD-specific MBCs in PI individuals sustained a higher proportion and number of CD21 -CD11c + activated MBCs at 3 months post-vaccination 2 (after three exposures). Work by Pape et al. also found that after two exposures, PI individuals had higher numbers of CD21 -CD27 -CD11c + MBCs than N individuals (Pape et al. 2021) further suggesting this could be an imprinted quality of prior infection or a difference in the quality of MBCs responding to vaccination as discussed above. CD21 -CD11c + memory cells have been described in the varied contexts of chronic viral infection, autoimmunity and vaccination and found to have similarly varied features including T-bet expression, potent antigen presentation, recent activation and resistance to activation (Knox et al., 2017; Glass et al., 2020) . The COVID-19 mRNA vaccines induce GC responses that last for months in N individuals, but whether they are larger or last longer in PI individuals is yet to be determined (Turner et al. 2021a; Cho et al., 2021; Kim et al., 2021; Turner et al., 2021b) . We hypothesize that CD21 -CD11c + MBCs are likely recently activated, and may be the result of ongoing GC responses and the sustained antigen driving them. This population could be larger in PI subjects because they are further supported by the elevated frequency of IFN--expressing CD4 + T cells which can drive T-J o u r n a l P r e -p r o o f 19 bet expression and support B cell activation. Larger and longer GCs support the generation of high affinity and broadly neutralizing antibodies which are critical for combating variants and are likely contributing to the protective advantage of hybrid immunity (Turner et al., 2021b; Goel et al., 2021) . Additional data from several months after a third vaccination are needed to determine whether this enhanced activation can be matched in N individuals by repeat vaccination.

Differences in route or inflammation associated with antigen exposure between N and PI individuals are likely key factors contributing to the differences in T cell phenotype and functionality associated with hybrid immunity. Pulmonary Infection with SARS-CoV-2 virus and intramuscular immunization with full-length spike mRNA involve distinct inflammatory environments for immune activation to spike protein. Prolonged viral replication in pulmonary tissue has been associated with production of various inflammatory cytokines in COVID-19 patients (Galani et al., 2021; Lucas et al., 2020) .

Specifically, the inflammatory cytokines IL-12, IFN-, and IFN-, produced by innate lymphocytes in response to viral PAMPs, directly drive T cells to upregulate the transcription factor T-bet, the master transcription factor for Th1 cells (Brinkmann et al., 1993; Lazarevic et al., 2013) , and may induce the CD4 + memory T cell IFN- and IL-10 signature we describe in PI individuals that is maintained through vaccination. Further, the cytokines IL-6, IL-12, or IL-27 drive the expression of IL-10 in FOXP3 -CD4 + helper T cells through the adapter c-maf (Gabryšová et al., 2018; Pot et al., 2009; Yang et al., 2005) . Functional differences, including the production of IL-10, have previously been described in SARS-CoV-2-specific CD4 + T cells from individuals that experienced J o u r n a l P r e -p r o o f 20 different levels of COVID-19 severity, raising questions about how these cytokines may be involved in immune protection (Le Bert et al., 2021) .

Additionally, infection with SARS-CoV-2 recruits effector T cells to the lung tissue, generating tissue-resident memory (Trm) cells in the human lung (Poon et al., 2021) . T cells that enter the lung tissue or interact with lung-derived innate cells in the lung-draining lymph node, may encounter a distinct inflammatory environment not present during intramuscular mRNA vaccination. Indeed, tissue-experienced CD4 + Trm cells produce the highest levels of IFN- and IL-10 in a mouse-adapted SARS-CoV infection (Zhao et al., 2016) . Therefore, it may be the case that the increase in cytokine production we see in PI participants is due to the presence of circulating S-reactive T cells which previously received tissue-specific cues. Although primarily studied in the mouse, Trm cells have been shown to be able to re-enter the circulating memory T cell compartment from the intestine in a phenomenon termed "retrograde migration" following systemic LCMV or VSV infections, as well as from the lung following influenza infection (Wijeyesinghe et al., 2021; Fonseca et al., 2020; Stolley et al., 2020) . A similar phenomenon has also been described in humans as skin-resident CD4 + Trm cells can re-enter the blood (Klicznik et al., 2019) . This raises the possibility that lung-transiting CoV-2-specific T cells may be captured in AIM assays of PBMCs, and contribute to the increased cytokine production we observe in PI donors.

The distinct CD4 + T cell cytokine signature we describe likely contributes to the increased protection observed in PI individuals with hybrid-immunity. IFN- is produced rapidly by memory T cells in vivo, driving an antiviral response program which is characterized by J o u r n a l P r e -p r o o f 21 the upregulation of interferon-stimulated genes. Our data showing elevated frequencies of IFN--producing memory CD4 + T cells in PI individuals and the observed localization of CD4 + Trm in the lung (Poon et al., 2021 ) together support a model in which early antiviral programs are bolstered rapidly by the inclusion of memory T cells locally at the site of infection. Indeed, pre-treatment of naive mice with recombinant IFN- completely inhibits SARS-CoV infection of the lung (Zhao et al., 2016) . This study also reported that concomitant treatment with recombinant TNF- exacerbates SARS-CoV-mediated disease in mice. IL-10 can inhibit TNF- and other pro-inflammatory cytokines in order to regulate inflammatory responses to infection (De Waal Malefyt et al., 1991; Gazzinelli et al., 1996; Hunter et al., 1997) . Therefore, increased production of IL-10 by S-specific Th1 cells potentiated through viral infection could be a critical element of hybrid immunity that limits future symptomatic SARS-CoV-2 infection. Additional work is needed to explore this crucial role of memory T cell-derived IL-10 in secondary challenge. 

M.P. is a member of the Scientific Advisory Board of VaxArt and NeoLeukin Inc.

We worked to ensure gender balance in the recruitment of human subjects. We worked to ensure that the study questionnaires were prepared in an inclusive way. One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in science. Statistics determined by Wilcoxon matched-paired signed rank test for longitudinal analyses and multiple unpaired Mann-Whitney test for group analyses: not significant (ns), *P < 0.05, **P < 0.005, ***P < 0.0005, and ****P < 0.0001. Error bars represent mean and SD. See also: Figure S1 .

Analysis of non-naive CD4 + CD69 + CD137 + T cells (AIM CD4 + ) in SARS-CoV-2 Naive 

Infected ( 

Requests for further information and reagents should be directed to and will be fulfilled by Marion Pepper (mpepper@uw.edu).

Unique tetramer reagents generated in this study are available upon reasonable request.

• All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Participants were enrolled in the Hospitalized or Ambulatory Adults with Respiratory Viral 

Participants were enrolled in one of three prospective cohort studies for longitudinal tracking of immune responses to COVID-19 vaccination in WA, USA. Cohort details are reported in Table 1 

Unless otherwise noted, cell lines were cultured in DMEM with 10% heat-inactivated FBS, 2 mM L-glutamine, 10mM HEPES, 100 U/mL penicillin, and 100 µg/mL streptomycin in a humidified atmosphere with 5% CO2 at 37 o C. 

Venous blood from study volunteers was collected in EDTA tubes and spun at 700 x g for 10 min. Plasma was collected, heat-inactivated at 56 o C for 30 min and stored at -80 o C.

Cellular fraction was resuspended in phosphate buffered saline (PBS) and PBMCs were separated from red blood cells using Sepmate PBMC Isolation Tubes (STEMCELL Technologies) according to manufacturer's instruction or cells were resuspended in HBSS, washed, overlaid with ficoll, spun at 400 x g for 30 min with no brake and PBMCs collected. Cells were washed twice in HBSS and frozen at -80 o C before being stored in liquid nitrogen.

Recombinant SARS-CoV-2 RBD (Wuhan-1, Wu-1) was generated by standard transient transfection followed by IMAC purification as described previously .

Recombinant SARS-CoV-2 RBD (B.1.1.351, β) was generated by transient transfection from the SARS-CoV-2-Beta-RBD-Avi (K417N-E484K-N501Y) construct synthesized by GenScript into CMVR with an N-terminal mu-phosphatase signal peptide and a C-terminal octa-histidine tag, flexible linker, and avi tag (GHHHHHHHHGGSSGLNDIFEAQKIEWHE) as described previously (Tortorici et al., 2021) . For tetramer generation, RBD proteins were biotinylated with the BirA500 kit (Avidity), tetramerized with streptavidinphycoerythrin (SA-PE) or streptavidin-allophycocyanin (SA-APC) (Agilent) and stored in 50% glycerol at -20 o C as previously described (Krishnamurty et al., 2016) . Decoy J o u r n a l P r e -p r o o f 38 reagents were generated by tetramerizing an irrelevant biotinylated protein with SA-PE previously conjugated to Dylight594 NHS Ester (ThermoFisher) and Dylight650 NHS Ester (ThermoFisher) or SA-APC previously conjugated to Dylight755 NHS Ester (ThermoFisher).

96-well plates (Corning) were coated with 2 ug/mL of recombinant SARS-CoV-2 RBD protein diluted in PBS and incubated at 4 o C overnight. Plates were washed with PBS-T (PBS containing 0.05% Tween-20) and incubated with blocking buffer (PBS-T and 3% milk) for 1h at room temperature (RT). Plasma, culture supernatants or monoclonal antibodies were serially diluted in dilution buffer (PBS-T and 1% milk) in triplicate, added to plates, and incubated at RT for 2 h. Secondary antibodies were diluted in dilution buffer as follows: anti-human IgG-HRP (Jackson ImmunoResearch) at 1:3000 or anti-human IgA-HRP (Southern Biotech) at 1:1500. Plates were incubated with secondary antibodies for 1h at RT, then detected with 1X 3,30,5,50-Tetramethylbenzidine (TMB) (Invitrogen) and quenched with 1M HCl. Sample optical density (OD) was measured by a spectrophotometer at 450 nm and 570 nm.

The SARS-CoV-2(o)-spike pseudotyping plasmid was assembled by introducing gene 

PVNT assays were performed as previously described (Crawford et al., 2020) . Briefly, heat inactivated plasma was diluted 1:5 followed by four 3-fold serial dilutions all in duplicate and mixed 1:1 with 10 6 relative luciferase units of SARS-CoV-2(Δ)-spike or SARS-CoV-2(o)-spike pseudotyped lentivirus in DMEM with 10% heat-inactivated FBS, 2 mM L-glutamine, 100 U/mL penicillin, and 100 µg/mL streptomycin. After 1 hour 

PRNT assays were performed as previously described (Erasmus et al., 2020) . Briefly, heat inactivated plasma was diluted 1:5 followed by four 4-fold serial dilutions and mixed 1:1 with 600 PFU/mL SARS-CoV-2 WA-1 (BEI resources; NR-52281) in PBS+0.3% cold water fish skin gelatin (Sigma). After 30 min of incubation at 37 o C, the plasma/virus mixtures were added to 12 well plates of Vero cells and incubated for 1 h at 37 o C, rocking every 15 min. All dilutions were done in duplicate, along with virus only and no virus controls. Plates were then washed with PBS and overlaid with a 1:1 mixture of 2.4% Avicel RC-591 (FMC) and 2X MEM (ThermoFisher) supplemented with 4% heat-inactivated FBS and Penicillin/Streptomycin (Fisher Scientific.) After a 48h incubation, the overlay was removed, plates were washed with PBS, fixed with 10% formaldehyde (Sigma-Aldrich) in PBS for 30 min at room temp and stained with 1% crystal violet (Sigma-Aldrich) in 20% EtOH. Percent neutralization was calculated as (1 -# sample plaques/# positive control plaques) x 100. Data was analyzed in Prism (GraphPad).

J o u r n a l P r e -p r o o f 41 SARS-CoV-2 15-mer peptides, 1mg each (BEI Resources), were provided lyophilized and stored at -80C. Peptides were selected for reactivity against a broad range of class I and class II HLA sub-types for targeted coverage of CD4 + and CD8 + T cell epitopes identified previously (Tarke et al., 2021 ) (Data S1). Before use, peptides were warmed to room temperature for 1 hour then reconstituted in DMSO to a concentration of 10 mg/mL. Individual peptides were combined in equal ratios to create Membrane/Nucleocapsid (182 ug/mL each, 55 peptides) or Spike (200ug/mL each, 49 peptides) megapools, maintaining a total concentration of 10mg/mL.

Approximately 20x10 6 PBMC from SARS-CoV-2 naive or previously infected individuals were divided in two for full phenotypic or cytokine analysis. For broad surface phenotyping 10x10 6 PBMC per sample were further divided into four 5mL polystyrene tubes and cells were pelleted at 250 x g for 5 minutes. Pellets were resuspended at 5x10 6 /mL in one of the following treatment conditions: DMSO (Sigma-Aldrich, >99.5% cell culture grade), 1ug/mL CEFX Ultra SuperStim Pool (JPT, PM-CEFX-2), or 5ug/mL SARS-CoV-2

Membrane and Nucleocapsid or Spike peptide megapools. Stimulation was performed for 18 hours in ImmunoCult-XF T cell Expansion Medium (StemCell Technologies). For intracellular cytokine assessment 2x10 6 /mL PBMC were stimulated using 6.6ug/mL SARS-CoV-2 peptide megapools, or an equivalent volume of DMSO (Sigma-Aldrich, >99.5% cell culture grade) for 12 hours in RPMI complete T cell medium containing anti- 

PBMCs were thawed at 37 o C and washed twice before first staining with decoy tetramer and then with RBD tetramer prior to incubation with anti-PE magnetic beads and magnetic bead enrichment (Miltenyi Biotec) as previously described (Krishnamurty et al., 2016) .

Cells in the positive fraction were stained with surface antibodies for B cell phenotypes.

Subsequent analysis of T cells was performed on unlabeled PBMC flow-through resulting from B cell RBD-tetramer pulldown.

PBMC flow-through were stimulated with peptide megapools as described above. For surface phenotyping, cells were washed and barcoded using four different fluorescently labeled CD45 antibodies to create eight unique barcodes as previously described (Becht et al., 2021) . Using this method to limit technical variability, all four stimulation conditions for both pre-and post-vaccination samples were combined and fully stained simultaneously for 20 minutes at 37 o C. Cells were then fixed for 10 minutes at room temperature using 1% paraformaldehyde (Sigma-Aldrich). For intracellular cytokine staining, cells were first incubated with anti-CXCR5 antibody at room temperature for 40 minutes. Cells were then stained with surface antibodies for 25 minutes at 4 o C and then fixed and permeabilized with Fixation/Permeabilization kit (Becton Dickinson) for 15 minutes at room temperature before staining for intracellular cytokines for 20 minutes at room temperature. 

AIM-positive (CD154 + CD69 + ) cells from all data files were concatenated with keywords and subjected to Phenograph clustering algorithm using k=40 nearest neighbors (Levine et al., 2015) and UMAP dimensionality reduction plugins using parameters IL-2, IFN-, IL-10, IL-4, IL-21, CD127, CD25, and CXCR5 in FlowJo 10 (Becton Dickinson). Clusters were then enumerated per 10 6 non-naive T cells for each sample by multiplying the percentage "cluster X parameter" by percentage "AIM + " and then by 1 million.

Statistics used are described in figure legends and were determined using Prism J o u r n a l P r e -p r o o f

Protection afforded by the BNT162b2 and mRNA-1273 COVID-19 vaccines in fully vaccinated cohorts with and without prior infection

Efficacy and Safety of the mRNA-1273 SARS-CoV-2 Vaccine

High-throughput single-cell quantification of hundreds of proteins using conventional flow cytometry and machine learning

Interferon alpha increases the frequency of interferon gamma-producing human CD4+ T cells

Broadly neutralizing antibodies overcome SARS-CoV-2 Omicron antigenic shift

Memory B cells and plasma cells: The differentiative continuum of humoral immunity

Sensitivity of SARS-CoV-2 B.1.1.7 to mRNA vaccine-elicited antibodies

Immune correlates of protection by mRNA-1273 vaccine against SARS-CoV-2 in nonhuman primates

Protocol and Reagents for Pseudotyping Lentiviral Particles with SARS-CoV-2 Spike Protein for Neutralization Assays

Hybrid immunity

Immunological memory to SARS-CoV-2 assessed for up to 8 months after infection

Interleukin 10 (IL-10) Inhibits Cytokine Synthesis by Human Monocytes: An Autoregulatory Role of IL-10 Produced by Monocytes

An Alphavirus-RNA vaccine induces SARS-CoV-2 neutralizing antibody and T cell responses in mice and nonhuman primates

Correlates of protection against symptomatic and asymptomatic SARS-CoV-2 infection

Developmental plasticity allows outside-in immune responses by resident memory T cells

c-Maf controls immune responses by regulating disease-specific gene networks and repressing IL-2 in CD4+ T cells

Untuned antiviral immunity in COVID-19 revealed by temporal type I/III interferon patterns and flu comparison

Comparing SARS-CoV-2 natural immunity to vaccine-induced immunity: reinfections versus breakthrough infections

In the absence of endogenous ILon CD4+ T cells and accompanied by overproduction of IL-12, IFN-gamma and TNF-alpha

Immune correlates analysis of the mRNA-1273 COVID-19 vaccine efficacy clinical trial

An Integrated Multi-omic Single-Cell Atlas of Human B Cell Identity

mRNA vaccines induce durable immune memory to SARS-CoV-2 and variants of concern

Protection and waning of natural and hybrid COVID-19 immunity

BNT162b2 vaccination induces durable SARS-CoV-2 specific T cells with a stem cell memory phenotype

1L-10 Is Required to Prevent Immune Hyperactivity During Infection with Trypanosoma cruzi

Conventional T-bet+Foxp3− Th1 cells are the major source of host-protective regulatory IL-10 during intracellular protozoan infection

Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection

Germinal centre-driven maturation of B cell response to SARS-CoV-2 vaccination

Human CD4+CD103+ cutaneous resident memory T cells are found in the circulation of healthy individuals

T-bet-expressing B cells during HIV and HCV infections

Somatically Hypermutated Plasmodium-Specific IgM+ Memory B Cells Are Rapid, Plastic, Early Responders upon Malaria Rechallenge

The germinal centre B cell response to SARS-CoV-2

T-bet: a bridge between innate and adaptive immunity

Highly functional virus-specific cellular immune response in asymptomatic SARS-CoV-2 infection

Data-Driven Phenotypic Dissection of AML Reveals Progenitor-like Cells that Correlate with Prognosis

Longitudinal analyses reveal immunological misfiring in severe COVID-19

TH1 cells control themselves by producing interleukin-10

Cytokine-skewed Tfh cells: functional consequences for B cell help

Rapid induction of antigen-specific CD4+ T cells is associated with coordinated humoral and cellular immunity to SARS-CoV-2 mRNA vaccination

High-affinity memory B cells induced by SARS-CoV-2 infection produce more plasmablasts and atypical memory B cells than those primed by mRNA vaccines

Immunogenicity of standard and extended dosing intervals of BNT162b2 mRNA vaccine

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

SARS-CoV-2 infection generates tissue-localized immunological memory in humans

Cutting Edge: IL-27 Induces the Transcription Factor c-Maf

Coordinately Act Together to Promote Differentiation of IL-10-Producing Tr1 Cells

Comparative analysis of activation induced marker (AIM) assays for sensitive identification of antigen-specific CD4 T cells

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

Functional SARS-CoV-2-Specific Immune Memory Persists after Mild COVID-19

Vivo CD4+ T Cell Differentiation and Function: Revisiting the Th1/Th2 Paradigm

Plasma Neutralization of the SARS-CoV-2 Omicron Variant

High genetic barrier to SARS-CoV-2 polyclonal neutralizing antibody escape

mRNA vaccination boosts cross-variant neutralizing antibodies elicited by SARS-CoV-2 infection

Retrograde migration supplies resident memory T cells to lung-draining LN after influenza infection

Effector T cells control lung inflammation during acute influenza virus infection by producing IL-10

Comprehensive analysis of T cell immunodominance and immunoprevalence of SARS-CoV-2 epitopes in COVID-19 cases

Broad sarbecovirus neutralization by a human monoclonal antibody

SARS-CoV-2 infection induces long-lived bone marrow plasma cells in humans

SARS-CoV-2 mRNA vaccines induce persistent human germinal centre responses

Human Circulating T Follicular Helper Cell Subsets in Health and Disease

Elicitation of Potent Neutralizing Antibody Responses by Designed Protein Nanoparticle Vaccines for SARS-CoV-2

Naturally enhanced neutralizing breadth against SARS-CoV-2 one year after infection

Expansible residence decentralizes immune homeostasis

IL-6 Plays a Unique Role in Initiating c-Maf Expression during Early Stage of CD4 T Cell Activation

Airway Memory CD4+ T Cells Mediate Protective Immunity against Emerging Respiratory Coronaviruses

Anti-human CCR4-PE/Dazzle594, CLone L291H4 Biolegend CAT#359420 Anti-human CCR6-BV650, Clone G043G3 Biolegend CAT#353426 Anti-human CCR7-BV605, Clone G043H7 Biolegend CAT#353224 Anti-human CD107a BV510, Clone H4A3 Biolegend CAT#328632 Anti-human CD11c-PEDazzle594, Clone 3.9BioLegend CAT#301642 Anti-human CD127 AlexaFlour 488, Clone AO19D5 Biolegend CAT#351314 Anti-human CD127-PE/Cy7, Clone hIL7Rm21Biolegend CAT#560822 Anti-human CD134-PerCP/cy5.5, Clone BerACT35Biolegend CAT#350010 Anti-human CD137-BV750, Clone 4B4-