Immune memory from SARS-CoV-2 infection in hamsters provides variant-independent protection but still allows virus transmission
Longitudinal tracking of antigen-specific lymphocytes in golden hamsters shows that memory from the founder strain of SARS-CoV-2 clears a beta variant rechallenge and that transferred memory T cells alone lower viral load, yet protected animals still infected every cohoused naive partner.
- Senior authors
- Benjamin R. tenOever
- Correspondence
- Benjamin R. tenOever
Research areas & themes
Citation
Horiuchi S, Oishi K, Carrau L, Frere J, Møller R, Panis M, tenOever BR. Immune memory from SARS-CoV-2 infection in hamsters provides variant-independent protection but still allows virus transmission. Science Immunology, 2021, volume 6, issue 66, article eabm3131.
DOI 10.1126/sciimmunol.abm3131. PMID 34699266.
One-sentence contribution
Longitudinal tracking of antigen-specific lymphocytes in golden hamsters shows that memory from the founder strain of SARS-CoV-2 clears a beta variant rechallenge and that transferred memory T cells alone lower viral load, yet protected animals still infected every cohoused naive partner.
Executive summary
Clinical immunology of COVID-19 was constrained by the absence of controlled rechallenge and transmission experiments, and the golden hamster had been shown to reproduce much of COVID-19 biology without the reagent base needed for detailed immune phenotyping. This study built that reagent base, using cross-reactive and commercial anti-hamster antibodies to resolve T cell and B cell populations by flow cytometry, and then used it to follow the response to SARS-CoV-2 longitudinally against influenza A virus as a benchmark. Both viruses reached comparable peak lung titers, but influenza was cleared by 7 days while SARS-CoV-2 persisted, and transcriptional profiling showed that NF-kB-associated and chemokine gene induction was delayed by about two days after SARS-CoV-2 infection. The delayed innate response was followed by an adaptive response that exceeded the one to influenza, with more CD8 T cells and T follicular helper cells and fewer regulatory T cells. Antigen-specific CD4 T cells and spike-specific B cells were detectable at 7 and 14 days and persisted beyond 40 days in lung, spleen and blood. Recovered animals rechallenged four months later with the same strain showed no detectable infectious virus, yet every cohoused naive animal became infected. Adoptive transfer of B cell-depleted lymphocytes from recovered animals reduced nasal wash titers in naive recipients and was followed by rapid appearance of spike-specific B cells. Rechallenge with the B.1.351 beta variant gave equal early titers but complete clearance by 4 days in previously exposed animals, with induction of antibodies neutralizing both the founder strain and the variant.
Scientific context
The host response to virus infection begins with recognition of pathogen-associated molecular patterns, activating the interferon regulatory factor and NF-kB pathways, inducing type I interferon and the interferon-stimulated genes, and recruiting the adaptive response through chemokines such as CXCL10 and CCL5. Influenza A virus masks aberrant RNA through NS1, giving a minimal transcriptional response in the infected cell. SARS-CoV-2 was known to diminish type I interferon while nonetheless driving high NF-kB activation for reasons the authors state remain unclear, a combination associated with poor local control and high proinflammatory output. At the time of the study, vaccination against spike was established as effective at preventing COVID-19, and virus-specific T and B cells had been detected in human blood for up to eight months after vaccination. What remained unknown was whether that memory blocks transmission or reinfection, including reinfection by variants under positive selection, because human data lacked the controls and designs available in a small animal model. The golden hamster had been shown to largely phenocopy COVID-19 biology, but commercial immunological reagents for the species were scarce.
Central question
Does immune memory established by SARS-CoV-2 infection protect the host against rechallenge, including by a variant of concern with a divergent spike protein, and does that protection prevent onward transmission.
Experimental strategy
The design has two halves, one methodological and one comparative. Because the golden hamster lacks a reagent base, the first task was to assemble a flow cytometry panel from antibodies raised against human and mouse targets that cross-react with hamster orthologs, combined with the few commercial anti-hamster reagents, with titrations chosen to maximize separation between populations. Antigen specificity was then read in two ways, by restimulating lymphocytes with pooled spike, nucleocapsid and matrix peptides and scoring Ki67 and IRF4 co-expression in CD4 T cells, and by labeling B cells with biotinylated spike detected through two differently colored streptavidins, with specificity controlled using splenocytes from influenza-infected animals.
The comparative half uses influenza A/California/04/2009 as a benchmark rather than as an object of study, on the reasoning that a well-characterized respiratory RNA virus with similar peak titer provides the calibration needed to say what is distinctive about the SARS-CoV-2 response in this host. Three perturbations then separate the components of memory. Rechallenge with the homologous strain at more than four months tests durability, with cohousing of naive animals one day after rechallenge added to test transmission, and the one-day delay before moving animals to a new cage is what distinguishes transmission from direct carryover of inoculum. Adoptive transfer of lymphocytes depleted of B cells using anti-MHC class II, anti-IgG and anti-IgM tests whether T cells contribute independently, and it also circumvents the confound of residual antibody in the recovered animals themselves. Rechallenge with the B.1.351 beta variant tests whether memory raised against the founder spike covers a divergent one, in a setting where T cells and other components act together, which a serum neutralization assay in vitro cannot capture.
Key findings
-
Peak lung titers were comparable between viruses, around 10^8 plaque-forming units at 3 days after infection, from inocula of 10^5 plaque-forming units for influenza and 10^3 for SARS-CoV-2. Influenza was undetectable by 7 days, whereas SARS-CoV-2 titers continued with little change until 7 days (Fig. 1A and 1B), and SARS-CoV-2-infected animals showed delayed growth relative to mock and influenza.
-
The innate response to SARS-CoV-2 is delayed. mRNA sequencing of lung at 3 and 5 days showed influenza driving NF-kB-associated and chemokine gene expression at 3 days that returned toward baseline by 5 days, while SARS-CoV-2 required two additional days to reach a comparable response despite similar viral load (Fig. 1C to 1E). Quantitative RT-PCR for Isg15, Irf7, Cxcl10 and Ccl5 gave the same pattern (Fig. 1F). The authors read this as SARS-CoV-2 antagonizing the host response more potently than influenza in this model.
-
The adaptive response to SARS-CoV-2 exceeds the response to influenza. Total lymphoid cell yield from lung rose at 7 days after SARS-CoV-2, as did total CD3 T cell frequency, returning to baseline by 14 days, with the increase attributable to CD8 cells (Fig. 2B to 2D). Both viruses induced CXCR3-positive T helper 1 cells, but SARS-CoV-2 gave significantly lower CXCR3-negative FoxP3-positive regulatory T cells than influenza, a more robust and sustained CXCR5 and Bcl6 double-positive T follicular helper population, and stronger CXCR3-positive CD8 T cells (Fig. 2E to 2G). B cell frequencies did not differ significantly between the two infections.
-
Antigen-specific cells persist as memory. Ki67 and IRF4 double-positive CD4 T cells responding to spike, nucleocapsid and matrix peptides, and spike-specific B cells, were present at 7 and 14 days, with lung frequencies highest early and blood frequencies rising later (Fig. 3B and 3C). Both populations were still present beyond 40 days, with antigen-specific CD4 T cells in lung and spleen but below detection in blood, and spike-specific B cells at high frequency in lung, spleen and blood (Fig. 3D and 3E).
-
Anti-RBD IgG and IgG2 and neutralizing titers peaked within two weeks and then declined to a plateau that remained above mock (Fig. 3F and 3G), with reported 50 percent plaque reduction neutralization values of 30.63 at 5 days, 1550.00 at 7 days, 829.30 at 14 days and 988.30 at 35 days.
-
Homologous rechallenge at more than 120 days gave near-complete protection, with no infectious virus recoverable from nasal wash at 2 and 5 days or from lung at 5 days (Fig. 4B). Despite this, all eight cohoused naive animals became infected, with robust titers in both nasal wash and lung (Fig. 4C). The authors interpret the discrepancy as transmission of virus present below the limit of detection, citing prior work that very low amounts suffice for transmission in this model. This is interpretation, since no infectious particle was detected in the transmitting animals.
-
A caution the authors raise about their own rechallenge result. Antibody amounts in rechallenged animals were similar to those at 72 days, and spike-specific B cell frequency did not correlate with anti-RBD titer after homologous reinfection, which they read as indicating that the antibody and B cell measurements do not reflect a memory recall and that protection in this experiment may instead have come from residual antibody from the primary infection.
-
Memory T cells contribute independently. Adoptive transfer was first validated by CellTrace Violet labeling, showing donor lymphoid cells persisting in recipient organs for up to 4 days. Transfer of unfractionated cells from recovered animals produced a significant boost in anti-RBD antibody at 3 days in recipients, indicating functional transferred B cells, with the effect lost by 8 days as the primary response took over. Transfer of B cell-depleted lymphocytes from lung, blood and spleen of recovered animals into naive recipients, one day before challenge, significantly lowered nasal wash titers at 1, 3 and 5 days relative to matched control transfers (Fig. 5B), with lung titers trending downward (Fig. 5C), and spike-specific B cells were readily detectable in all recipients of T cells from recovered donors (Fig. 5D). The authors note baseline differences in viral load between cohorts receiving different cell lineages and state that comparisons were therefore made only within treatment cohorts.
-
Memory raised against the founder strain clears a beta variant challenge. Serum from animals infected with the founder strain neutralized the beta variant far less well in vitro, with 50 percent plaque reduction values of 1030.00 and 896.20 against the founder strain at 7 and 14 days against 12.01 and 5.53 against the variant (Fig. 6A). In vivo, previously exposed and naive animals rechallenged with the beta variant at 120 days showed comparable nasal wash titers at 1 and 2 days, but previously exposed animals had cleared infectious virus by 4 days while the control group remained positive to 5 days, a pattern mirrored in lung (Fig. 6C). Spike-specific B cells and anti-RBD antibodies were robustly induced, their frequencies and titers correlated, and neutralizing titers rose against both the founder strain and the variant (Fig. 6D to 6G).
Mechanistic model
The study establishes several causal relationships and stops short of a full mechanism, and the paper is explicit about which is which.
What the data support directly. Memory T cells are sufficient, in the near absence of B cells, to lower SARS-CoV-2 titers in the upper respiratory tract of a naive recipient, which the adoptive transfer of B cell-depleted lymphocytes shows. Their presence is followed by rapid appearance of spike-specific B cells and virus-specific antibody in recipients that received no memory B cells, which the authors read as memory T cells driving early induction of virus-specific antibody. Memory established against the founder spike is sufficient to clear a variant whose spike escapes neutralization by the same animals' serum in vitro, and the difference between the in vitro and in vivo outcomes is the basis for the paper's argument that serum neutralization assays understate protection because they omit T cells and the rest of the response.
What is not established. The study does not identify which effector functions of the transferred T cells reduce viral load, does not distinguish CD4 from CD8 contributions, and by the authors' own account of their reagent limitations could not resolve antigen-specific CD8 T cells at all. The mechanism by which SARS-CoV-2 delays the innate response in this host is not addressed here, only observed. The explanation for transmission from animals in which no infectious particle was detected, that the transmitted dose lay below the assay's limit of detection, is a proposal consistent with prior observations in this model rather than a measured quantity. The relationship between the homologous rechallenge protection and memory is left deliberately unresolved by the authors, who raise residual antibody as an alternative account for that particular experiment. The clearance of the beta variant is therefore the stronger of the two rechallenge results, since it occurred in animals whose serum neutralized the challenge virus poorly.
Conceptual or technical advance
Two things became possible because of this work. The first is technical. A flow cytometry approach for the golden hamster was assembled from cross-reactive antibodies, with the paper reporting CXCR3, CXCR5 and Bcl6 as newly usable in this species, together with peptide restimulation and biotinylated antigen probes for detecting antigen-specific T and B cells and a validated adoptive transfer procedure that exploits the inbred status of the animals. This turns a model previously valued for reproducing disease into one in which the cellular components of the immune response can be dissected.
The second is conceptual. By running rechallenge and transmission in the same animals, the study separates protection of the host from interruption of spread, and finds them dissociated. By running a variant rechallenge in vivo alongside serum neutralization in vitro on the same animals, it shows the two measures giving different answers, which bears directly on how reduced neutralization titers against variants should be interpreted. The authors extend both findings to the human situation in their discussion, noting that recovered or vaccinated individuals may transmit without knowing it and that immune memory nonetheless appears sufficient against the variant tested. Those extensions are the authors' reading of a hamster experiment and are not demonstrated in humans here.
Relationship to the broader research program
This is a tenOever laboratory study with the corresponding author and the senior position held by tenOever, and it extends that laboratory's established use of the golden hamster for SARS-CoV-2, cited here as prior work. It also continues a recurring pattern in the corpus, the use of influenza A virus as a calibrated comparator against which the behavior of another virus is measured, and the treatment of the interferon and NF-kB transcriptional response as the primary readout of what a virus does to its host, with Isg15, Irf7, Cxcl10 and Ccl5 used as the index genes.
Category 3 synthesis, visible only across papers. The framing that morbidity follows from a virus antagonizing host defense, stated in the introduction here through the contrast between NS1-mediated masking by influenza and the delayed response to SARS-CoV-2, is the same view of host and virus that organizes the 2013 Nature Reviews Microbiology article, where the chordate protein-based defense is presented as the system viruses must evade. The interferon-stimulated gene readouts used here sit downstream of the machinery dissected in the 2007 Science report on IKKε. Neither connection is claimed by the paper and both require the other papers to see.
Related publications
- Blanco-Melo et al. 2020 and the earlier hamster work cited in this paper from the same laboratory, predecessor, establishing the transcriptional response to SARS-CoV-2 and the golden hamster model that this study builds on.
- Bouhaddou et al. 2020, Cell, shares the pathogen and includes tenOever among its authors, but no data, method or claim connects it to this study, so no relationship is asserted beyond a shared subject.
- tenOever 2019, Cold Spring Harbor Perspectives in Medicine, and tenOever 2013, Nature Reviews Microbiology, from the same author. Conceptually adjacent through the shared framing of virus and host antagonism and the use of influenza A virus as a reference system, but neither is cited here and neither supplies method or data.
Limitations and boundaries
The authors set out their own reagent-driven limits. The absence of commercial hamster antibodies meant that B cell-specific and CD8 T cell-specific populations could not be identified directly, that CD3 staining required fixation and so live cells of that population could not be sorted, that chemokine markers were not detectable after stimulation and culture, and that Ki67 and IRF4 background was too high in the absence of CD8 staining to resolve antigen-specific CD8 T cells or CD4 subsets. The T cell adoptive transfers therefore test a B cell-depleted lymphocyte population rather than a defined T cell subset, and the paper notes baseline viral load differences between cohorts receiving different cell lineages, restricting comparisons to within-cohort. Beyond these, the work is confined to the golden hamster, a species chosen for phenocopying COVID-19 but not equivalent to humans in immune repertoire or reagent-verified cell definitions. One founder strain and one variant of concern were tested, so the conclusion about variant-independent protection is bounded by B.1.351 and does not extend to variants that arose later. Challenge was by a single intranasal dose of 10^3 plaque-forming units under anesthesia, which is not equivalent to natural exposure. Rechallenge intervals were four months, so durability beyond that is untested. The transmission result rests on cohousing at a one to one ratio in eight pairs with no infectious virus detected in the donors, so the transmitted dose was never measured. The homologous rechallenge protection cannot be attributed to memory rather than residual antibody, as the authors themselves state. Immunity was generated by infection rather than by vaccination, so conclusions do not transfer directly to vaccinated hosts, a point the discussion addresses by analogy rather than by experiment.
Audience summaries
25 words
Hamsters recovered from SARS-CoV-2 cleared a beta variant rechallenge and transferred memory T cells alone lowered viral load, yet protected animals still infected cohoused naive partners.
75 words
Reagents for immune profiling in golden hamsters were assembled from cross-reactive antibodies, then used to follow SARS-CoV-2 against influenza as a benchmark. The innate response to SARS-CoV-2 was delayed but the adaptive response was stronger, and antigen-specific T and B cells persisted past 40 days. Recovered animals rechallenged four months later showed no detectable virus yet still infected cohoused naive animals. Memory raised against the founder strain cleared a beta variant their serum neutralized poorly in vitro.
150 words
Human COVID-19 immunology lacked controlled rechallenge and transmission experiments, and the golden hamster lacked reagents for immune phenotyping. Cross-reactive antibodies, peptide restimulation and biotinylated spike probes resolved antigen-specific T and B cells in this species. Compared with influenza at matched peak lung titer, SARS-CoV-2 delayed NF-kB and chemokine induction by about two days and cleared more slowly, but elicited more CD8 and T follicular helper cells and fewer regulatory T cells, with antigen-specific populations persisting beyond 40 days. Adoptive transfer of B cell-depleted lymphocytes from recovered animals lowered nasal titers in naive recipients and was followed by rapid spike-specific B cell appearance. Animals rechallenged with the homologous strain at four months yielded no detectable virus yet infected all eight cohoused naive partners. Rechallenge with the B.1.351 beta variant gave equal early titers but complete clearance by day four, despite poor in vitro neutralization of that variant.
Documented publication relationships
- The Global Phosphorylation Landscape of SARS-CoV-2 Infection — related.
- Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19 — predecessor.
- The Global Phosphorylation Landscape of SARS-CoV-2 Infection — related.
- A diminished immune response underlies age-related SARS-CoV-2 pathologies — methodological foundation.