Immune memory from SARS-CoV-2 infection in hamsters provides variant-independent protection but still allows virus transmission

Description SARS-CoV-2 infection results in cellular memory that protects against variants of concern but fails to block transmission. Modeling SARS-CoV-2 infection with hamsters Animal models to study SARS-CoV-2 are crucial in developing therapies and vaccines to effectively control infection and end the pandemic. Here, Horiuchi et al. used golden hamsters as a model to study the immune responses to SARS-CoV-2 infection. They found that SARS-CoV-2 infection induced immune responses different from influenza infection but could effectively lead to memory B and T cell responses. The generated memory T cells were able to protect against SARS-CoV-2 reinfection of animals with the same strain and a variant of concern. Despite immunity and protection to reinfection, hamsters with SARS-CoV-2 immune memory could still transmit the virus to naïve cage mates. Thus, golden hamsters represent a robust model for studying the immune responses to SARS-CoV-2 infection. SARS-CoV-2 has caused morbidity and mortality across the globe. As the virus spreads, new variants are arising that show enhanced capacity to bypass preexisting immunity. To understand the memory response to SARS-CoV-2, here, we monitored SARS-CoV-2–specific T and B cells in a longitudinal study of infected and recovered golden hamsters (Mesocricetus auratus). We demonstrated that engagement of the innate immune system after SARS-CoV-2 infection was delayed but was followed by a pronounced adaptive response. Moreover, T cell adoptive transfer conferred a reduction in virus levels and rapid induction of SARS-CoV-2–specific B cells, demonstrating that both lymphocyte populations contributed to the overall response. Reinfection of recovered animals with a SARS-CoV-2 variant of concern showed that SARS-CoV-2–specific T and B cells could effectively control the infection that associated with the rapid induction of neutralizing antibodies but failed to block transmission to both naïve and seroconverted animals. These data suggest that the adaptive immune response to SARS-CoV-2 is sufficient to provide protection to the host, independent of the emergence of variants.


INTRODUCTION
The host response to virus infection begins at the level of infected cells in mammals (1,2). The cellular intrinsic response to virus is mediated by the production of virus-specific replication intermediates referred to as pathogen-associated molecular patterns (PAMPs) (3,4). These conserved viral structures are recognized by protein sentinels in the cell, which induce the activation of the interferon regulatory factor (IRF) and nuclear factor B (NF-B) pathways to coordinate a very sophisticated host immune response (1,5). The host response to virus includes both the induction of type I interferon (IFN-I) to induce the fortification of cellular defenses in nearby cells alongside cytokines that recruit and activate cells of the adaptive immune response (6). The IFN-I response signals to neighboring cells and elicits a second signal transduction event that culminates in the up-regulation of hundreds of IFN-stimulated genes (ISGs) that collectively inhibit virus replication (7). Examples of ISGs include ISG15, which regulates IFN-I signaling (8) or IRF7, serving to expand the transcriptional footprint of the cell after virus infection (9). Concomitant with the induction of IFN-I biology, infected cells also recruit the adaptive immune response through the induction of chemoattractant cytokines or chemokines, such as CXCL10 and CCL5 (10)(11)(12). Together, these two complementary arms of the defense system limit the immediate spread of virus in vivo while rapidly developing a pathogen-specific immune response that ensures lifelong immunity.
Despite the powerful nature of the host antiviral response, virus disease still occurs. Without exception, virus-mediated morbidity and mortality are the product of antagonizing some aspect of the host defense biology. Influenza A virus (IAV) infection results in disease because the virus can successfully mask any aberrant RNA generated during infection through high levels of the nonstructural protein 1 (NS1) protein, resulting in a minimal transcriptional response in the infected cell (13). In contrast, the transcriptional footprint of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection also indicates significant interference imposed on host biology as it relates to the response to virus infection. Although SARS-CoV-2-infected cells do demonstrate a diminished IFN-I response as a result of a number of virus-specific products reported to antagonize PAMP detection, the infection seemingly demonstrates high levels of NF-B activation for reasons that remain unclear (14,15). This process of SARS-CoV-2 partially blocking the host innate response results in a lack of local replication control alongside a high level of proinflammatory cytokines and chemokines, which may constitute the molecular basis of coronavirus disease 2019 (COVID- 19), ultimately resulting in COVID-19. These data have also been corroborated using both clinical data and the golden hamster model (16)(17)(18).
The development of immune memory is initiated by antigenpresenting cells, which present peptide fragments of the virus to induce activation and differentiation of CD4 + and CD8 + T cells that express a T cell receptor capable of engaging the material presented. Effector CD4 + T cells further induce virus-specific B cell receptorexpressing B cells, which contribute to antibody production. Effector CD8 + T cells function as cytotoxic T cells and play a key role in virus clearance. After virus clearance, a small subset of virus-specific T and B cells survive indefinitely as memory cells. These memory T and B cells have the ability to respond rapidly after reinfection to clear the virus efficiently (19)(20)(21)(22). Here, we focus on how SARS-CoV-2 modulation of host antiviral biology contributes to the immune memory response.
Vaccination against SARS-CoV-2 Spike protein (here noted as S) is a highly effective means of preventing the development of COVID-19 (23). For SARS-CoV-2 infection, virus-specific T and B cells are detectable in the blood for up to 8 months after vaccination in humans, suggesting the presence of memory cells generated by the vaccine (24). However, the effectiveness of this memory response in blocking transmission or reinfection from SARS-CoV-2, or one of the variants that have arisen as a result of positive selection, remains unknown because immunological data concerning COVID-19 have relied heavily on clinical data, which lack the specific controls and experimental designs one can achieve with a small animal model (25)(26)(27).
Because the golden hamster model has proven to largely phenocopy COVID-19 biology, here, we aimed to determine the memory response to SARS-CoV-2 (18,(28)(29)(30)(31). To this end, we developed the necessary tools to analyze the adaptive immune response in the golden hamster. We found that SARS-CoV-2 infection induced a strong adaptive immune response compared with IAV and generated virus-specific T and B cells that were preserved as memory cells after viral clearance. Moreover, establishment of the memory response also protected animals from rechallenge, with the SARS-CoV-2 beta variant infection showing a significant decrease in viral titer and a corresponding increase in cross-reactive antibody production as compared with the control cohort. These data could be further resolved through the use of adoptive transfers of memory T cells, which, we found, conferred reduction in virus levels after the rapid induction of SARS-CoV-2-specific B cells, although transmission potential remained unchanged. These results shed light on the importance of memory T and B cells for protection against reinfection with SARS-CoV-2 and its emerged variants.

Innate immune engagement correlates with the timing of viral clearance
In an effort to better understand the immune response to SARS-CoV-2, we compared this infection to challenge with an unrelated RNA respiratory virus responsible for a past pandemic (Influenza A/California/04/2009) (18,32). The idea to compare these two viral platforms was to provide a benchmark for hamster immunity by including a virus model that has been extensively studied to better understand how the response to SARS-CoV-2 may differ. In this regard, intranasal inoculation of IAV resulted in robust replication in the respiratory tract, showing peak titers of ~10 8 plaque-forming units (PFU) at 3 days postinfection (dpi). Despite the high viral load, IAV was rapidly cleared from the lungs, showing no infectious particles by 7 dpi (Fig. 1A). This replication trajectory was similar when comparing the titers of SARS-CoV-2 Washington strain (SARS-CoV-2WA), with the highest virus levels being achieved at 3 dpi with virus levels reaching ~10 8 PFU (Fig. 1B). In contrast, whereas IAV clearance was already evident starting at 5 dpi, SARS-CoV-2WA dynamics showed continual virus production with little change in titers until 7 dpi (Fig. 1, A and B). The inability to control SARS-CoV-2WA also correlated with body weight because infected animals showed delayed growth as compared with both mock-and IAVinfected animals (fig. S1A).
To assess the innate immune responses in these model systems, we performed mRNA sequencing at 3-and 5-dpi lungs to determine differential gene expression (DEG) in the respiratory tract. Reflecting viral titers, IAV-infected hamsters showed a transcriptional footprint indicative of NF-B activation, including high levels of chemokines at 3 dpi (Fig. 1, C and D). In contrast, at 5 dpi, these signatures largely returned to baseline, correlating with the decrease in IAV titers (Fig. 1, A, C, and D). These results suggest a diminished IFN-I response that enabled IAV titers to reach ~10 8 PFU but a subsequent strong induction of an NF-B-mediated response, which successfully cleared the infection. In contrast to IAV, SARS-CoV-2WA-infected hamsters showed a delayed induction of NF-B-related genes, requiring two additional days to generate a comparable response with that observed for IAV despite similar viral loads (Fig. 1, A, C, and D). Together, these data demonstrated a delayed induction of innate immune response upon SARS-CoV-2WA infection (Fig. 1E). The delay in the host antiviral response after SARS-CoV-2WA infection as compared with IAV was further corroborated by quantitative reverse transcription polymerase chain reaction (qRT-PCR) for key genes involved in the IFN-I system (Isg15 and Irf7) and chemokines (Cxcl10 and Ccl5) (Fig. 1F). These results reflect the interactions between host and virus, suggesting that SARS-CoV-2 can antagonize aspects of the host response in the hamster model more potently than IAV.

Adaptive immunity to SARS-CoV-2 infection exceeds the response to IAV
To compare the adaptive immune response to SARS-CoV-2 versus IAV, we analyzed T and B cell frequency in the lung after viral clearance by developing an assay to characterize immune cells by fluorescence-activated cell sorting (FACS). Given the limitations of available reagents for the golden hamster, we first tested antibodies that cross-reacted with divergent species and therefore were likely to recognize the ortholog in hamsters (33)(34)(35). In addition to reported cross-reactive antibodies, we also combined commercialized anti-golden hamster antibodies and adjusted the titration to create the greatest disparity between different cellular populations to aid in our immune cell profiling ( fig. S2, A to C). On the basis of these markers [major histocompatibility complex class II (MHCII), CD3, CD4, Bcl6, FoxP3, CXCR3, CXCR5, immunoglobulin G2 (IgG2), and IgG], we profiled total lymphocytes, T cells, and B cells in the lungs, which conformed to the general frequency observed in other model vertebrates (36). After gating total lymphoid populations and screening out dead cells, total T cells were defined as CD3 + MHCll − , whereas B cells were denoted as CD3 − MHCll + . A subset of T cells could be further delineated as CD4 + T cells by gating on CD3 + CD4 + , whereas CD8 + T cells were defined by CD3 + CD4 − . CD4 + T cell subsets were further separated by the expression of CXCR5, CXCR3, FoxP3, and Bcl6 [CXCR3 + FoxP3 − as T helper 1 (T H 1), CXCR3 − FoxP3 + as T regulatory (T reg ), and CXCR5 + Bcl6 + as T follicular helper (T FH ) cells]. CD8 + T cells were subdivided as CXCR3 + versus CXCR3 − CD8 + T cells. Last, B cell isotypes were identified as either IgG2 + or IgG2 − IgG + to denote class switch in addition to being Bcl6 + versus Bcl6 − , indicating germinal versus nongerminal B cells, respectively ( fig. S2A).
To profile the host adaptive immune response to SARS-CoV-2 and IAV, hamsters were infected intranasally or treated with phosphatebuffered saline (PBS) as a noninfected control ( Fig. 2A). T and B cell populations in the lungs at 7 and 14 dpi were analyzed by FACS using the gating strategies outlined above ( fig. S2A). The Ficoll-separated cell number after Ficoll separation was significantly elevated in response to SARS-CoV-2WA at 7 dpi, suggesting increased lymphoid cell migration in agreement with the enhanced chemokine expression profile observed (Figs. 1C and 2B). Profiling total T cells (CD3 + ) demonstrated a significant increase in frequency after SARS-CoV-2WA infection as compared with IAV at 7 dpi, which returned to baseline in all cohorts by 14 dpi (Fig. 2C). To further delineate the nature of this increased frequency, we again profiled total T cells for CD4 expression to reveal that SARS-CoV-2WA induced a marked increase specific to CD8 + cells (CD4 − CD3 + ) (Fig. 2D).
Next, we determined the frequency of T H 1 and FoxP3 + T reg cells because these have been implicated in the respiratory response to virus infection (22,37,38). In agreement with published studies, both IAV and SARS-CoV-2WA infection induced CXCR3 + T H 1 cells (Fig. 2E). In contrast, we observed differences in CXCR3 − /FoxP3 + T reg cells in response to SARS-CoV-2WA that were significantly lower when compared with IAV, which had a corresponding increase in CXCR3 + / FoxP3 + cells ( Fig. 2E and fig. S2C) (39). Profiling the frequency of CXCR5 + Bcl6 + T FH cells showed a significantly more robust response to SARS-CoV-2WA as compared with IAV, which was sustained for the duration of the experiment (Fig. 2F). Given recent evidence for the role of CD8 + cells in combating a respiratory viral infection (40), we measured CD3 + /CXCR3 + CD4 − populations and found that SARS-CoV-2WA infection induced stronger CD8 + T cells in the lung (Fig. 2E), as compared with IAV, suggesting robust infiltration consistent with COVID-19. Together, these data demonstrate a marked change in CD4 + T cell subsets in response to SARS-CoV-2WA. Despite showing no increase in total cell number over mock-or IAV-infected animals, SARS-CoV-2WA infection induced a marked loss of T reg cells with a contrasting increase in T FH and CD8 + T cells (Fig. 2, F and G). Last, we profiled B cells in these model systems but found no significant difference between the response to SARS-CoV-2WA and IAV ( fig. S2, D and E). Collectively, we concluded that SARS-CoV-2 infection induces a stronger effector T cell response in the lung compared with IAV infection.

Antigen-specific T and B cells generated after SARS-CoV-2 infection are maintained as memory cells
Given the unique adaptive immune response to SARS-CoV-2 in the initial phase of infection, we next chose to evaluate the generation Hamsters were intranasally inoculated with 10 3 PFU of SARS-CoV-2WA or 10 5 PFU of IAV and harvested at 1, 3, 5, or 7 dpi. (C and D) Heatmap depicting the expression levels of NF-B-related genes containing chemokines differentially expressed in lung samples of hamsters at 3 and 5 dpi. (E) GSEA analysis using NF-B (top) and chemokine (bottom) signaling pathway gene sets was conducted on differential expression data generated from a direct comparison of SARS-CoV-2WA-and IAV-infected lung tissues at 3 dpi in red and 5 dpi in blue. (F) mRNA expression level of ISG15, IRF7, Cxcl10, and CCL5 represented by qRT-PCR. RNA samples of lungs from IAV-or SARS-CoV-2WAinfected hamsters at 3 and 5 dpi were assessed. All data were generated, with error bars denoting SD across samples (n = 4).
of memory B and T cells. To assess the memory response, it is imperative to perform longitudinal studies to identify times of peak infiltration followed by a return to baseline. To this end, we infected golden hamsters with SARS-CoV-2WA and obtained lung, lung-draining mediastinal lymph node (mLN), blood, and/or spleen at 7, 14, and 42 dpi (Fig. 3A). On the basis of cell frequency, we defined the acute phase of the immune response to SARS-CoV-2WA to encompass the first 14 dpi ( fig. S3, A and B). In contrast, at 42 dpi, we observed the memory phase as defined by a contraction of the frequency of immune cell population ( fig. S3, A and B).
To identify antigen-specific T and B cells in golden hamsters, we combined several methods widely used in human studies (41)(42)(43)(44). For SARS-CoV-2-specific CD4 + T cells, Ficoll-separated cells collected from either lungs, peripheral blood mononuclear cells (PBMCs), mLNs, or spleens were cultured for 24 hours in the presence of a titrated SARS-CoV-2 Spike (S), nucleocapsid (N), and matrix (M) peptide mix after overnight resting. These data identified a CD4 + T cell population with high levels of Ki67 and IRF4, indicating specific recognition of the aforementioned peptides and a productive adaptive response (fig. S3, C and E) (41,43).
In an effort to evaluate the B cell response to infection, biotin-labeled SARS-CoV-2 S protein was added to individual immune cells from either lungs, PBMCs, mLNs, or spleens in the presence of two fluorescently labeled streptavidins (BV786 and BV650). In this assay, B cells capable of recognizing antigens of S will bind the biotinlabeled substrate, which then further associates with both streptavidin fluorophores. To ensure specificity, we performed this assay using splenocytes from either SARS-CoV-2WAor IAV-infected hamsters before the experiment ( fig. S3D).
Upon establishing a methodology for antigen-specific B and T cells in the hamster model, we next applied it to our acute and memory phase after SARS-CoV-2WA infection. SARS-CoV-2-specific CD4 + T cells and S-specific B cells were detected at 7 and 14 dpi (Fig. 3, B and C). The frequencies of antigen-specific cells were variable depending on both time after infection and cell population tested. These data demonstrated a robust antigen-specific CD4 + T cell response in the lung as early as 7 dpi that diminished thereafter but was still present at 14 dpi (Fig. 3B). In PBMCs, SARS-CoV-2specific CD4 + T cells were undetectable at 7 dpi but significantly increased in frequency thereafter (Fig. 3B). Similar trends for SARS-CoV-2-specific B cells could be observed in the lung and PBMCs as compared with CD4 + T cells, with the lung showing a high frequency of antigen-specific cells at 7 dpi, which decreased over time in contrast to blood where this pattern was inversed (Fig. 3C). For B cells, we also examined the mLN and found SARS-CoV-2-specific cells showing the same trend in frequency as PBMCs (Fig. 3C).
After acute infection, a subset of the antigen-specific cells are preserved for a long time in the body as memory cells (45). In clinical samples of SARS-CoV-2, memory T and B cells were detected in peripheral blood for up to 8 months (24). To determine whether SARS-CoV-2 infection in golden hamsters could induce memory T and B cells, we next examined the immune cell population beyond 40 dpi (Fig. 3, D and E). We found SARS-CoV-2-specific CD4 + T cells in the lungs of hamsters 5 weeks after viral clearance (Figs. 1A and 3D). SARS-CoV-2-specific-CD4 + T cells could also be observed in the spleen beyond 40 dpi but were below the level of detection in PBMCs (Fig. 3D). Parallel trends observed between T and B cells during the acute phase of infection were also mirrored beyond 40 dpi, with SARS-CoV-2-specific B cells displaying high frequencies in the lung, spleen, and PBMCs (Fig. 3E). As expected, antibodies recognizing the receptor binding domain (RBD) of S as well as general neutralization antibody titers peaked within 2 weeks and decreased thereafter, plateauing at levels that maintained significance over mock (Fig. 3, F and G). Together, these data suggest that SARS-CoV-2 infection in hamsters induces antigen-specific T and B cells that are preserved as memory cells.

Adaptive immunity protects the host but does not prevent transmission after secondary exposure to SARS-CoV-2
Clinical studies from recovered SARS-CoV-2-infected individuals have demonstrated a decrease in antibody titer and antigen-specific T and B cells (46,47). Despite being a relatively normal immune phenomenon, these findings gave way to concerns about the possibility of reinfection. Given that these trends were also observed in our golden hamster model, we next subjected a cohort of SARS-CoV-2WA-recovered hamsters (4 months after primary infection) to rechallenge (Fig. 4A). Rechallenged animals were monitored for infectious virus from both nasal washes and lung tissue, demonstrating near-complete protection among seroconverted animals (Fig. 4B). These data suggest that hamsters that have recovered from SARS-CoV-2 infection are protected from reinfection for at least 4 months after clearance.
To determine whether this level of individual protection also prevented the ability of seroconverted animals to transmit virus, we next cohoused rechallenged animals at 1 dpi (and 4 months before primary infection) with naïve animals (Fig. 4A). To mitigate the possibility of direct viral carryover from the rechallenged animals, we waited 1 day after virus administration and moved both the rechallenged and naïve animals to a new cage. These data found that, despite our inability to detect a single infectious particle from nasal washes, transmission was observed in all eight transmission experiments, where virus levels were robust in both the nasal washes and lungs of the naïve exposed animals (Fig. 4C). Examining the rechallenged animals demonstrated that, despite the detection of SARS-CoV-2-specific B cells, the amount of antibody was similar to that detected at 72 dpi ( Fig. 3F and fig. S4, A and B). The frequency of SARS-CoV-2 Spike-specific B cells and anti-RBD-specific antibody titers did not correlate after SARS-CoV-2WA reinfection (fig. S4C). These results suggest that the antibody titer and the presence of SARS-CoV-2 Spike-specific B cells are not the result of a memory response. Hence, protection of these animals after second infection may have been the result of residual antibodies from the primary infection ( fig. S4, A to D). Together, these data suggest that recovered individuals are protected from reinfection of the same SARS-CoV-2-specific strain but maintain the capacity to transmit virus.

SARS-CoV-2-specific memory T cells contribute to viral clearance
To assess the specific contribution of the memory T cell response, we first needed to control for any interference from residual antibodies derived from the primary infection ( fig. S4, A to C). To this end, we established an adoptive transfer system for the golden hamster because previous studies have determined that these animals are sufficiently inbred to prevent graft-versus-host disease (48). We isolated and treated total lymphoid cells with the CellTrace Violet (CTV) reagent before transferring them to naïve animals. We confirmed that lymphoid cells collected from the lungs and spleens of naïve donors remained in the recipient hamsters in all organs tested for up to 4 days after intravenous administration by the presence of CTV ( fig.  S5, A and B). Having demonstrated the feasibility of this technique, we next collected lymphoid cells from lungs, PBMCs, and spleens from SARS-CoV-2WA-recovered animals (40 dpi) and transferred them to naïve recipients ( fig. S5C). One day after adoptive transfer, treated animals were infected with SARS-CoV-2WA, and blood was collected at 3 and 8 dpi. At 3 dpi, we observed a significant boost in anti-RBD antibodies as compared with animals transfused with control cells and challenged ( fig. S5D). These data indicate that our transferred B cells produced functional antibodies in the recipient animals. At 8 dpi, we observed that the primary response to virus outcompetes any boost provided by the adoptive transfer ( fig. S5D).
To parse out the contribution of T versus B cells in this memory response, we next repeated the adoptive transfer after B cell depletion by using anti-MHCll, anti-IgG, and anti-IgM with the combination of magnetic beads from immune cells derived from lungs, PBMCs, and spleens ( fig. S5, E and F). Adoptive transfer of these individual enriched T cell populations was administered to hamsters 1 day before SARS-CoV-2WA infection (Fig. 5A). Virus titers from treated animals were determined from nasal wash at 1, 3, and 5 dpi (Fig. 5B). These data demonstrated that memory T cells from recovered hamsters contribute to early virus clearance in the upper respiratory tract because titers from SARS-CoV-2WA-recovered animals all demonstrated significantly lower viral loads (Fig. 5B). Of note, we observed some baseline differences in viral load between different cohorts, suggesting that the administration of different cell lineages may result in distinct host responses that might influence virus replication. Because the results derived from these data were only compared within treatment cohorts, we are still able to conclude that lung cells, PBMCs, and splenic cells from previously infected animals all provide a level of protection. The observed decrease in viral titers from SARS-CoV-2WA-recovered T cell donor recipients was also reflected in total lung where virus levels trended downward from all T cell populations tested (Fig. 5C). Last, because CD4 + T cells contribute to antibody production during virus infection in the lung (36), we analyzed the frequency of S-specific B cells from SARS-CoV-2WA-recovered CD3 + T cell-transferred donor hamster organs after infection. S-specific B cells were readily detectable in all donors that received T cells from recovered recipients (Fig. 5D). These data suggest that the presence of SARS-CoV-2-specific memory CD3 + T cells contributes to early induction of virus-specific antibodies without the presence of memory B cells.

Memory B and T cells generated after SARS-CoV-2 infection contribute to SARS-CoV-2 variant clearance
The recent emergence of multiple variants displaying different S mutations has caused significant concern worldwide because their potential to evade preexisting immunity could result in a second pandemic (49)(50)(51)(52)(53). This concern has been partially justified because variants such as the B.1.351 beta variant (SARS-CoV-2) demonstrate some level of antibody evasion when tested in vitro (49)(50)(51)(52)(53)(54). To evaluate the possibility of variants evading preexisting immunity in the context of a physiological model, we next challenged hamsters that had recovered from SARS-CoV-2WA with the same variant (SARS-CoV-2) in vivo.
As previously reported, we could confirm using our hamster model that the antibodies elicited by challenge with SARS-CoV-2WA were less neutralizing for SARS-CoV-2 when tested in vitro (Fig. 6A). However, because the host response to virus infection requires both B and T cells, we next evaluated this same dynamic in vivo (Fig. 6B). Nasal washes from either naïve or previously exposed animals rechallenged with the SARS-CoV-2 variant showed comparable titers in the upper respiratory tract at 1 and 2 dpi (Fig. 6C). However, at 4 dpi, animals that had been previously exposed to SARS-CoV-2WA showed complete clearance in contrast to the control group where infectious material could be detected until 5 dpi (Fig. 6C). These results were further corroborated in lung titers (Fig. 6C). To ascertain how the memory response may contribute to these findings, anti-SARS-CoV-2 S-specific B cells and SARS-CoV-2 RBD antibody titers were analyzed (Fig. 6, D and E). These analyses demonstrated that SARS-CoV-2 S-specific B cells and RBD-specific antibodies were robustly induced (Fig. 6, D and E). Correlations between RBD antibody titer and WA S-specific B cell frequencies in these animals suggest the successful establishment of an immune memory response (Fig. 6F and fig. S6). In addition, SARS-CoV-2 reinfected hamsters showed strong increase in neutralization titer against both SARS-CoV-2WA and SARS-CoV-2 (Fig. 6G). Moreover, these results suggest that SARS-CoV-2 S-specific memory B cells reactivate in the lungs and/or spleens of recovered individuals and may induce cross-reactive antibodies to both SARS-CoV-2WA and the SARS-CoV-2 variant. In all, these data demonstrate that B and T cell memory generated after A B C

Fig. 4. Adaptive immunity protects the host but does not prevent transmission after secondary exposure to SARS-CoV-2. (A) Experimental protocol for hamster rechallenge and transmission. Syrian golden hamsters infected intranasally with either mock or 10 3 PFU of SARS-CoV-2WA
were rechallenged with the same SARS-CoV-2WA strain after more than 120 days after primary infection. For transmission, naïve hamsters were cohoused with the rechallenged hamsters, one to one, 1 day after rechallenge in a new cage. Nasal wash was collected at 2 and 5 days after rechallenge, and lungs were collected at 5 days after rechallenge to analyze PFU. (B) PFU from nasal wash at 2 and 5 dpi (left) and the lungs at 5 dpi (right) in rechallenged hamsters. (C) PFU from nasal wash at 2 and 5 dpi (left) and the lungs at 5 dpi (right) in cohoused hamsters with the rechallenged hamsters. A representative dataset of eight individuals is shown. ***P < 0.001.

of 14
SARS-CoV-2 infection can readjust and contribute to the clearance of related variants.

DISCUSSION
Viruses are not inherent drivers of morbidity and mortality. Despite the global burden imposed by viruses, these represent only a small fraction of the vertebrate virome (55). The vast majority of viruses do not cause disease and can be readily controlled by our immune system. What makes pathogens noteworthy is their capacity to antagonize our antiviral biology. Characterization of IAV and SARS-CoV-2 infections has demonstrated that both viruses dampen production of IFN-I, albeit with different kinetics and potency. We found that the innate host response to SARS-CoV-2 was delayed when compared with IAV, a dynamic that provides an explanation for how COVID-19 develops. However, this virus-host interaction is only a partial contributor to the biology of SARS-CoV-2. Here, we further focused on the development of the memory response to SARS-CoV-2. Recent studies have suggested the possibility of memory T cells participating in the protection against SARS-CoV-2 and variant of concerns in humans (24,56). However, there are very few studies that demonstrate T cell protection after secondary infection (28,57,58). Because the golden hamster model has proven to largely phenocopy COVID-19 biology, we aimed to determine the functionality of the memory response to SARS-CoV-2.
To define the breadth of immune memory to SARS-CoV-2, we first needed to identify reagents that could monitor key immune cells in a model system with little to no commercial reagents available. To analyze hamster markers by FACS, we introduced crossreactive antibodies commercialized for human and murine targets. Although some of the antibodies used for this assay were previously reported (33)(34)(35), this study demonstrated the use of cross-reactive antibodies, CXCR3, CXCR5, and Bcl6 for hamster research. Of note, the cellular expression patterns from these antibodies are consistent with our present understanding of the frequency of different immune cell populations. Upon developing the methodology to monitor the adaptive immune response to SARS-CoV-2 in the hamster, we focused on the behavior of T and B cells in various immune compartments. These data revealed an expansion of immune cells in response to SARS-CoV-2 including the establishment of a memory response. SARS-CoV-2 included a significant population of resident memory T cells in the lung, which may serve to protect virus infection in the airway, similar to what has been described for IAV (59,60).
To better define the individual contributions of T and B cells, we next performed an adoptive transfer of only T cells from previously infected animals. These data demonstrated that virus titers could be lowered, even in the near absence of B cell support, suggesting that T cells can directly contribute to the inhibition of virus and illustrating the formation of immune memory to SARS-CoV-2. Given these results, we next focused our model on addressing two important and unanswered questions as it pertains to memory. First, we sought to determine whether transmission was interrupted after the establishment of memory. Second, we chose to examine whether memory to the founder strain of SARS-CoV-2 was sufficient to protect the host from the emerging variants. On the basis of previous studies, we know that vaccination reduces the risk of transmission but does not prevent it outright, a result that is compatible in the golden hamster model (61,62).
Here, we demonstrated that, despite the delay in the initial cellular response to infection, SARS-CoV-2 induced a canonical adaptive host response that culminates in the establishment of immune memory. However, despite this, we found that generation of SARS-CoV-2specific B and T cells failed to block transmission despite showing detectable infectious particles in those animals that had seroconverted. In the case of the hamster model, very low amounts of infectious particles are still sufficient to be transmitted (18,28). We believe that the amount of virus necessary for transmission was below our level of detection and thus allows for the transmission events observed. These data have very significant implications for this pandemic because it would suggest that individuals that are vaccine hesitant will remain at risk indefinitely as SARS-CoV-2 may continue to circulate globally for years to come without the knowledge of those who have recovered from or been vaccinated for the virus.
Because the golden hamster is not a commonly used small animal model, these studies required some level of unconventional methods to monitor host immunology in the absence of certain commercial reagents. For example, identification of T and B cell populations was only possible with a limited number of commercialized anti-hamster antibodies and cross-reactive antibodies, which were performed using flow cytometry. For this reason, we were also not able to identify B cell-specific or CD8 + T cell populations. Moreover, because CD3 staining was only possible after fixation, we were not able to sort live cells for this population. Last, because the lymphocyte culture condition was not adjusted to hamsters and also the antibodies we used are detecting the target by cross-reactivity, chemokine markers were not detectable after stimulation and culture. Also, because we were not able to stain CD8 + T cells by anti-CD8 antibodies, the background of Ki67 and IRF4 was too high to identify the stimulation-specific increase of the population. Therefore, we were not able to provide high resolution of antigen-specific CD8 + T cells or subsets of CD4 + T cells in this assay.
Last, we addressed the growing concern that the emergence of the SARS-CoV-2 variants may to lead to another pandemic wave even among vaccinated and/or recovered individuals. This concern is mainly based on the position of many of the mutations that have emerged in the S open reading frame located near a site commonly recognized after seroconversion. Antibodies generated to the parental strain have been found to be neutralizing to many of the widely circulating variants (25)(26)(27). However, because the virus neutralization assay was performed by incubating serum and virus in vitro (63), it does not account for the contribution of T cells and other aspects of the immune response that can only be achieved under physiological conditions. The SARS-CoV-2 rechallenge study has been conducted in other reports (63,64). To this end, we challenged animals that had previously recovered to the parental strain of SARS-CoV-2 and compared these animals with an established memory response to a de novo challenge experiment with the beta variant of SARS-CoV-2. The results of this rechallenge experiment showed no initial difference in the host response, confirming an absence of residual antibodies or inflammation. However, despite the common trajectory of virus replication before day 4, at this time point, the cohort phenotype bifurcated markedly with preexposed animals, showing a complete loss of infectious virus. These data demonstrate that memory response to even the founder strain of SARS-CoV-2 was sufficient to neutralize infection of a variant that deviates from the founder S attachment protein. Together, this work demonstrates that, although the establishment of immune memory does not prevent SARS-CoV-2 transmission, it is sufficient at protecting us against both the founder and the variant of concern, suggesting that, with aggressive vaccination campaigns, this pandemic can be resolved.

Study design
This study characterizes the innate and adaptive immune response in golden hamsters after SARS-CoV-2 infection. Golden hamsters were intranasally challenged by SARS-CoV-2 (10 3 PFU, 100 l) or IAV (10 5 PFU, 100 l) to analyze the host response. At indicated time points, lungs were collected and used to define virus titer, assess gene expression changes, and monitor immune cell profiles by flow cytometry. To this end, we established a flow cytometry-based assay to identify both total and SARS-CoV-2-specific T and B cell populations. All animals were purchased from Charles River Laboratories, and the studies were conducted in accordance with animal welfare and Institutional Animal Care and Use Committee (IACUC) guidelines. All experiments involving virus infections were carried out in a U.S. Centers for Disease Control and Prevention (CDC)/ U.S. Department of Agriculture (USDA)-approved high-containment biosafety level three (BSL-3) facility.

In vivo infections and harvest
Before intranasal infection, hamsters were anesthetized by intraperitoneal injection with 200 l of a ketamine/xylazine solution (3:1) (100 mg/kg). SARS-CoV-2 (10 3 PFU) or 10 5 PFU of IAV stock was resuspended in 100 l of PBS. One hundred microliters of PBS or each virus was administered intranasally for each animal. Hamsters were euthanized by intraperitoneal injection of pentobarbital. Blood was collected from the aorta during necropsy at the end of each indicated time point.

Plaque assay
Virus titers of IAV and each SARS-CoV-2 strain were determined by means of plaque assay in MDCK cells or Vero-E6 cells, respectively. At 48 hours after infection, plaques were fixed in 5% formaldehyde and stained with crystal violet.

Plaque reduction assay
Each SARS-CoV-2 strain (100 l, 200 PFU) was incubated with twofold serial dilutions of sera (100 l) for 1 hour at room temperature. One hundred microliters of antibody-virus mixtures containing 100 PFU of SARS-CoV-2 was inoculated into Vero-E6 cells to perform the plaque assay. At 48 hours after infection, plaques were visualized and counted.

Transcriptome analysis
One microgram of total RNA was enriched for polyadenylated RNA species and prepared for short-read next-generation sequencing using the TruSeq Stranded mRNA Library Prep Kit (Illumina) according to the manufacturer's instructions. Sequencing libraries were sequenced on an Illumina NextSeq 500 platform. Fastq files were generated using bcl2fastq (Illumina) and aligned to the Syrian golden hamster genome (MesAur 1.0, ensembl) using the RNA-Seq Alignment application (Basespace, Illumina). DEG was determined using the DESeq2 pipeline (65). All genes with a P-adjusted value (P adj ) of <0.1 were classified as differentially expressed genes. Determining the number of reads mapping to the viral genome (GenBank: MN985325.1) was performed using bowtie2 (66). Gene set enrichment analysis (GSEA) was performed with GSEA Java application GSEA_4.1.0 (made available by the Broad Institute and University of California San Diego) (https://www.nature.com/articles/ ng1180). GSEA was conducted as a preranked analysis using a ranking statistic of −log 10 (P value)/sign(log 2 fold change). Gene sets used for analysis were part of Hallmark and curated human phenotype and gene ontology gene sets maintained as part of MSigDB (v7.4). All visualizations of RNA-sequencing differential expression data were created in R using ggplot2 and gplots packages. Gene set enrichment plots were adapted from VisualizeRNAseq (https://github. com/GryderArt/VisualizeRNAseq).
Quantitative reverse transcription polymerase chain reaction RNA was reverse-transcribed into complementary DNA using oligo(dT) primers with SuperScript II Reverse Transcriptase (Thermo Fisher Scientific). qPCR was performed using primers specific for -actin, ISG15, IRF7, Cxcl10, or CCL5 (18) with KAPA SYBR Fast qPCR Master Mix (KAPA Biosystems) on LightCycler 480 Instrument II (Roche). Delta-delta-cycle threshold was determined relative to noninfected samples.

Anti-RBD hamster IgG ELISA
The enzyme-linked immunosorbent assay (ELISA) protocol was adapted from a previously established protocol (67). Ninety-six-well plates (Immulon 4 HBX, Thermo Fisher Scientific) were coated with 50 ml of recombinant RBD (2 mg/ml) in PBS at 4°C overnight. After overnight incubation, the coating solutions were removed, and the plates were blocked with 100 ml of 3% nonfat milk (AmericanBio) prepared in 0.1% Tween 20 containing PBS (PBS-T) for 1 hour at room temperature. To reduce the risk of containing live virus, serum samples were heated for 1 hour at 56°C before use and were handled in a BSL-3 facility. Serial dilution of serum samples and dilution of secondary antibody were done in 1% nonfat milk prepared in PBS-T. After blocking, solutions were removed and 100 ml of diluted serum was added after 2 hours of incubation at room temperature. After the solutions were removed, the wells were washed with 250 ml of PBS-T three times. One hundred milliliters of 1:7500 diluted anti-Syrian hamster IgG horseradish peroxidase (Thermo Fisher Scientific) secondary antibody was added and incubated for 1 hour at room temperature. After the solutions were removed, the wells were washed with 250 ml of PBS-T three times. Once the wells were completely dry, 100 ml of SIGMAFAST o-phenylenediamine dihydrochloride (Millipore Sigma) solution was added and the reaction was stopped with 30 ml of 3 M HCl. The optical density at 490 nm was measured, and the concentrations of the antibody were analyzed by area under the curve (AUC) using Prism 8 (GraphPad).

Antigen-specific CD4 + T cell and B cell detection assay
To identify SARS-CoV-2-specific CD4 + T cells, Ficoll-separated cells collected from either lungs, PBMCs, mLNs, or spleens were rested overnight at 37°C in nonserum medium (X-VIVO 15; Lonza) with the concentration at 1 × 10 6 cells per well in a 96-well U-bottom plate (TPP; 92097). Cells were further cultured for 24 hours at 37°C in the presence of SARS-CoV-2 Spike (S), nucleocapsid (N), and matrix (M) peptide mix (1 g/ml each; GenScript). Staphylococcal enterotoxin B (1 g/ml) stimulation was used as a positive control, and dimethyl sulfoxide was used as a negative control. After stimulation, cells were collected and further processed for FACS analysis.

Adoptive transfer assay
To identify transferred cells, lymphoid cells from the lungs and spleens were labeled with CTV (CellTrace Violet Cell Proliferation Kit, Thermo Fisher Scientific). The labeled cells were further transferred intravenously at 2 × 10 6 cells per hamster. After 4 days, lungs, PBMCs, and spleens were collected to detect CTV by FACS. For total cell transfer from SARS-CoV-2WA-recovered hamsters, lymphoid cells from the SARS-CoV-2WA-recovered hamster lungs, PBMCs, and spleens were collected. The cells were further transferred intravenously at 5 × 10 7 cells per hamster. After 1 day of cell transfer, hamsters were infected by SARS-CoV-2WA accordingly. For T cellenriched cell transfer experiment from SARS-CoV-2WA-recovered hamsters, lymphoid cells from the SARS-CoV-2WA-recovered hamster lungs, PBMCs, and spleens were collected. MHCll + cells were depleted with the combination of biotin-anti-MHCll antibody (1.2 g/ml; Invitrogen, MA5-17761), biotin-anti-hamster IgG (1.2 g/ml; BioLegend, 405601), and in-laboratory biotin-labeled anti-hamster IgM (1.2 g/ml; Rockland, 107-4107) after adjusting the cells to 1 × 10 8 cells/ml. Cells were further processed with Biotin Positive Selection Kit ll (STEMCELL Technologies), and the remaining cells were used as CD3 + T cell-enriched cells. For the cell transfer, 2 × 10 6 cells per hamster for the lung and 2 × 10 7 cells per hamster for PBMCs and spleens were transferred intravenously. After 1 day of cell transfer, hamsters were infected by SARS-CoV-2WA accordingly.

Statistics
The significance of the difference between groups in the experiments was evaluated by two-tailed paired t test by Prism (GraphPad) software. A value of P < 0.05 was considered significant.