HLA-A∗02:01 restricted T cell receptors against the highly conserved SARS-CoV-2 polymerase cross-react with human coronaviruses

Cross-reactivity and direct killing of target cells remain underexplored for severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2)-specific CD8+ T cells. Isolation of T cell receptors (TCRs) and overexpression in allogeneic cells allows for extensive T cell reactivity profiling. We identify SARS-CoV-2 RNA-dependent RNA polymerase (RdRp/NSP12) as highly conserved, likely due to its critical role in the virus life cycle. We perform single-cell TCRαβ sequencing in human leukocyte antigen (HLA)-A∗02:01-restricted, RdRp-specific T cells from SARS-CoV-2-unexposed individuals. Human T cells expressing these TCRαβ constructs kill target cell lines engineered to express full-length RdRp. Three TCR constructs recognize homologous epitopes from common cold coronaviruses, indicating CD8+ T cells can recognize evolutionarily diverse coronaviruses. Analysis of individual TCR clones may help define vaccine epitopes that can induce long-term immunity against SARS-CoV-2 and other coronaviruses.


Correspondence
In brief Nesterenko et al. identify T cell responses with potential to confer long-term immunity against SARS-CoV-2. The machinery responsible for replicating the viral genome is highly conserved and, as shown by Nesterenko et al., can be effectively targeted by CD8 + T cells. Many individuals are now immune as a result of successful vaccination campaigns and protection afforded by natural infection with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) (Anand et al., 2021;Baden et al., 2021;Lumley et al., 2020;Polack et al., 2020;Sadoff et al., 2021). The virus continues to evolve and may escape immune responses generated against the original sequence (Harvey et al., 2021;Planas et al., 2021). The BNT162b2 mRNA vaccine is 88% effective against the new Delta variant compared with 93.7% for the alpha variant that was circulating previously (Bernal et al., 2021). Increased spread in vaccinated populations necessitates further understanding of the SARS-CoV-2 immune response. This pandemic can only be controlled by herd immunity against contemporary strains of the virus. Vaccination against the wild-type spike protein can prevent COVID-19 (Baden et al., 2021;Polack et al., 2020;Sadoff et al., 2021). SARS-CoV-2 vaccines target the spike protein by generating neutralizing antibodies that prevent host cell infection (Khoury et al., 2021;Lumley et al., 2020). SARS-CoV-2 variants often contain multiple mutations in the spike protein and can resist antibody neutralization, creating the possibility that, upon further diversification, viral variants may escape current vaccine defenses (Hoffmann et al., 2021;Kuzmina et al., 2021;Muik et al., 2021;Planas et al., 2021;Wang et al., 2021). Cytotoxic T cells kill infected cells, thereby directly limiting viral dissemination once the infection occurs (Hall et al., 1986;Harty et al., 2000;Jozwik et al., 2015;McMichael et al., 1983). T cell recognition is not limited to surface proteins like the spike protein; more conserved proteins can be targeted. Internal SARS-CoV-2 proteins are more conserved than the spike and may present a therapeutic opportunity for generating T cell responses that can recognize many coronavirus strains . T cell vaccine strategies, targeting the nucleocapsid protein, are being explored to generate long-term immunity against SARS-CoV-2 (Dutta et al., 2020;Gauttier et al., 2020;Sieling et al., 2021). It remains unknown which epitopes elicit the most effective antiviral responses (Chen and John Wherry, 2020).
Initial evidence for T cell control of respiratory infections was provided by children with genetic defects in T cell development (Hall et al., 1986). Resident memory T cells, which are permanently localized in non-lymphoid tissues, including the lung, are thought to mediate antiviral responses (Jozwik et al., 2015). In a human respiratory syncytial virus (RSV) infection, disease severity was inversely correlated with the preexisting T cells in the lung (Jozwik et al., 2015). Adoptive transfer of highly functional T cell clones can reduce severity of viral diseases as well (Einsele et al., 2002;Feuchtinger et al., 2010a). The mechanism of respiratory viral infection T cell control is thought to happen through Fas and perforin-mediated lysis of infected cells (Topham et al., 1997). The efficiency of lysis correlates with the ability to clear an infection (McMichael et al., 1983).
Both convalescent donors and unexposed individuals have SARS-CoV-2-specific T cell responses (Le Bert et al., 2020;Braun et al., 2020;Grifoni et al., 2020b;Mateus et al., 2020;Peng et al., 2020;Tarke et al., 2021;Weiskopf et al., 2020). CD8 + T cell responses have been identified as correlates of protection in SARS-CoV-2 infection (Chen and John Wherry, 2020;Liao et al., 2020;McMahan et al., 2021). Unexposed individuals may have T cell responses that were generated by common cold human coronaviruses (HCoVs) and may be partially protective against SARS-CoV-2 encounter (Lipsitch et al., 2020;Mallajosyula et al., 2021;Mateus et al., 2020). T cells interact with target antigens through the T cell receptor (TCR), which is a heterodimer of alpha and beta chains. TCRs are inherently cross-reactive to maximize the breadth of ligand recognition; however, a single TCR is not guaranteed to recognize related antigens (Sewell, 2012). Several cross-reactive CD8 + T cell responses are known, but specific TCRab clones that can drive such reactivity are not defined (Lineburg et al., 2021;Lipsitch et al., 2020;Mallajosyula et al., 2021;Mateus et al., 2020). T cell memory is most often defined as ability to recognize synthetic peptide epitopes in functional assays or peptide-major histocompatibility complex (MHC) multimer staining. Recognition of processed epitopes derived from full-length intracellular antigens is underexplored in SARS-CoV-2. Isolation of specific TCR clones permits unambiguous determination of reactivity and detailed characterization of immune responses such as cytotoxic potential and measurement of cross-reactivity against related viruses.
We employ recent technological advances in single-cell sequencing, DNA synthesis, and gene transfer to recover anti-gen-specific TCRab and subsequently characterize them in allogeneic T cells. The viral polymerase (NSP12/RNA-dependent RNA polymerase [RdRp]) was identified as highly conserved within SARS-CoV-2 and other human coronaviruses. RdRp-reactive CD8 + T cells were then selected for TCRab droplet-based sequencing (Drop-seq) based on the intracellular level of tumor necrosis factor alpha (TNFa) and interferon gamma (IFNg) via mRNA sequencing via cross-linker regulated intracellular phenotype (CLInt-seq), which allows for antigen-specific TCR sequencing via commercially available Drop-seq in cells that are stained for intracellular cytokines . TCRs were initially screened for single-epitope recognition in a cell line system via the NFAT-GFP reporter system. Reactive TCRs were overexpressed in human peripheral blood mononuclear cells (PBMCs) and killed antigen-presenting cells that expressed the full-length RdRp. Three TCR constructs were broadly reactive and cross-reacted with epitope homologs from HCoVs.

RESULTS
RdRp is highly conserved among human coronaviruses and within SARS-CoV-2 Antigens derived from highly conserved SARS-CoV-2 proteins should generate immune responses effective against multiple variants. Human coronaviruses are separated by hundreds of years of evolution and serve as a model of the evolutionary constraints that may restrict variant emergence in SARS- CoV-2 (Forni et al., 2017;Killerby et al., 2018). A group of coronavirus proteins involved in RNA synthesis, immune modulation, and structural machinery was selected for further analysis. Sequence identity was compared across all known human coronaviruses. All proteins showed conservation within sub-classes: alpha coronaviruses (229E and NL63), beta coronaviruses (HKU1 and OC43), and both SARS viruses ( Figure 1A). The RdRp was most conserved across all coronaviruses ( Figure 1A). Across 893,589 SARS-CoV-2 samples sequenced, RdRp was well conserved and had few mutations compared with the spike protein ( Figure 1B; Table S4).

TCRs from unexposed individuals recognize RdRp epitopes
To generate TCR clones, we screened pooled peptide epitopes predicted to bind human leukocyte antigen (HLA)-A*02:01 against HLA-matched PBMCs collected prior to December of 2019 ( Figure 2A). We refer to these samples as unexposed to SARS-CoV-2. CD8 + T cells that responded by production of TNFa and IFNg were sorted from four different PBMC donors via fluorescence-activated cell sorting (FACS) (Figure 2A). Responses were low in all donors, around the level of background set based on DMSO control stimulation, as would be expected for donors who were not exposed to a specific pathogen.
Reactive cells were sorted for single-cell TCRab sequencing via a highly sensitive technique called CLInt-seq. Clonally expanded TCR clones were synthesized and tested in an allogeneic cell-based system for evaluation of immune receptor activation ( Figure 2B). A high-throughput system for TCR reactivity profiling was established ( Figure S1A). We utilized a Jurkat cell line that expressed the NFAT-zsGreen T cell activation reporter construct and the CD8 molecule to stabilize MHC class I interactions. This cell-based reporter system was then optimized with a well-characterized TCR, clone 1G4, which is specific for the cancer antigen NY-ESO-1(D'Angelo et al., 2018) ( Figure S1B). Comparison of TCR delivery by electroporation or viral integration resulted in similar extent of T cell activation ( Figure S1C). SARS-CoV-2-specific TCRs were then electroporated or transduced into the Jurkat cell line and activation was measured by FACS measurement of zsGreen. SARS-CoV-2reactive epitopes were identified via epitope deconvolution using an array of sub-pools ( Figures 2C and S2). Of 44 TCR constructs tested in this system, 10 recognized the cognate peptide pool ( Figure 2D). TCR clones that did not score as reactive in this assay either did not reach the threshold of the reporter system or were originally expressed in T cells that did not recognize the queried peptide pool. Because the responses sorted were around the level of background, nonreactive TCRs likely represent the background signal. Nine TCRs clearly recognized four unique epitopes of the RdRp (Figure S2; Table S1).
MERS SARS S1-NTD S1-CTD  Isolated TCRs recognize and kill RdRp-expressing cells Processed antigen recognition is critical for vaccine-induced priming of naive T cell responses as well as for lysis of infected cells. To establish potential antiviral efficacy, seven RdRp-specific TCRs were overexpressed in HLA-A*02:01-positive human PBMCs via retroviral delivery ( Figure 3A; Table S2). Engineered PBMCs were co-cultured with a target cell line engineered to overexpress the full-length SARS-CoV-2 RdRp protein and HLA-A*02:01 ( Figure 3B). The engineered T cells were able to produce TNFa and IFNg in response to recognition of processed antigens ( Figures 3C and 3D). Full-length antigen recognition was significantly lower than peptide pulsing, as measured by T cell cytokine production, most likely due to the concentration  of peptide during pulsing assays being supraphysiological ( Figure 3D). CD4 + T cells that overexpressed the coronavirus specific TCR (CoVTCRs) also responded to peptide pulsing but did not recognize processed antigen. Production of TNFa and IFNg in CD4 + T cells ranged from 0.058% to 10.3%, depending on the TCR. T cells control viral spread by killing virus-infected cells. Cytotoxicity assays showed five out of seven TCRs can direct T cells to kill target cell lines ( Figure 3E). Recognition of processed epitopes was confirmed by supernatant IFNg ELISA assay ( Figure 3F). At 48 h, processed antigen recognition was equivalent to or better than the peptide pulsing control ( Figure 3F).
We sought to determine how common RdRp-specific T cells are. Recently, a set of more than 160,000 TCRb genes specific for SARS-CoV-2 was made publicly available (Nolan et al., 2020). This data were generated by peptide pool stimulation of PBMCs from 118 donors and subsequent TCRb gene sequencing in reactive T cells (Klinger et al., 2015). Unique epitopes were ranked by the count of cognate TCRb sequences ( Figure S3A). Three of the four epitopes we identified were frequently targeted by the SARS-CoV-2-specific TCRb (Figure S3A). GLIPH2, an algorithm for grouping TCRs that recognize the same antigen (Huang et al., 2020), showed three TCRab constructs we defined grouped with other TCRs against ORF1ab, which contains the RdRp ( Figure S3B; Table S3). The epitope FV9 was frequently targeted, but its cognate TCR, CoVTCR 18, did not share sequence similarity with any ORF1ab-specific TCRb (Figures S3A and S3B). This TCR also lacked killing ability in the prior assay ( Figure 3E). Peptide titration showed that this  TCR only recognized antigen at high concentration of 10 mg/mL, confirming that this TCR has low affinity for this specific target ( Figure S4). Two of the four TCRs against the RV9 epitope grouped with ORF1ab-specific TCRb ( Figure S3B). CoVTCR 34, specific against RV9, was strongly cytotoxic but did not group with any TCRb by GLIPH2 analysis.

SARS-CoV-2 RdRp-targeted TCRs broadly recognize human coronaviruses
We then queried RdRp TCR cross-reactivity against the HCoV epitopes. Epitope homologs were identified by alignment of RdRp sequences from all human coronaviruses. Each of the homologous epitopes was synthesized and TCR reactivity against each of the epitopes was profiled via peptide titration assay. RdRp-specific TCRab constructs exhibited a diverse pattern of coronavirus reactivity (Figures 4A-4F). Three TCRs were highly specific for SARS-CoV-2 ( Figures 4A, 4B, and 4E). This TCR reactivity may represent the naive T cell repertoire or could be an immune response to unknown antigen. Only two of the four RV9-reactive TCRs recognized one HCoV ( Figures 4D and 4F). The TL9-reactive TCR had strong cross-reactivity with SARS, MERS, 229E, and NL63 coronaviruses ( Figure 4C).

DISCUSSION
This study provides a strong basis for considering the development of vaccines against either specific epitopes or the fulllength RdRp. Current vaccines provide strong protection against COVID-19 caused by circulating variants of SARS-CoV-2. Continuous evolution of SARS-CoV-2 may necessitate updates to the vaccine's spike sequence, selection of a more conserved antigen, or a combination of both. One of the challenges of developing booster shots is the need to predict which variant will be the most common when the vaccine is administered. Failure to predict this accurately may decrease the efficacy of the booster. SARS-CoV-2 infection can be recognized by RdRp-specific T cells as indicated by strong RdRp CD8 + T cell responses in convalescent donors (Tarke et al., 2021). The RdRp sequence is particularly well conserved within SARS-CoV-2 and among other human coronaviruses. Sequence conservation suggests that the critical functional role of this protein places restriction on its capacity to evolve. We show that RdRp-specific T cells are cytotoxic against cells that express full-length antigen, which suggests T cell responses against RdRp should help control SARS-CoV-2 infection and prevent COVID-19 disease.
Inducing broadly reactive T cell responses may be particularly important for generating lifelong immunity against SARS-CoV-2. T cells can recognize the target antigen even after accumulation of point mutations (Sewell, 2012). While we identified RdRp as the most conserved protein, it too is likely to change, as evident from the accumulation of point mutations. Here, we defined two RdRp epitopes that can elicit broadly coronavirus-reactive T cell responses. T cells that recognize different human coronaviruses are likely to recognize novel mutation variants as they emerge, due to strong affinity for the antigen. Epitope TL9-reactive T cells have been previously identified as cross-reactive and associated with reduced disease severity; however, a specific TCR clone driving this response was not identified (Mallajosyula et al., 2021). For the RV9 epitope, some TCRs were cross-reactive but others only recognized SARS-CoV-2, showing that TCRs against the same antigen can have distinct reactivity. The crossreactive TCRs against RV9 used the same V alpha chain TRAV38-2DV8, implicating that the usage of a common alpha chain may allow cross-reactivity. Specific TCR sequences that are known to allow for broad reactivity can be used as benchmarks for induction of such immunity. TCR-based disease severity correlation will require more TCR characterization to expand the scope of TCRs and HLA allele restrictions. Induction of broadly reactive T cell responses that are not affected by point mutations in the epitope sequence, as well as benchmarks for measurement of such responses, can help guide development of T cell vaccines.
Several reports proposed the use of adoptive transfer of antigen-specific T cells from convalescent donors to treat severe COVID-19 disease (Basar et al., 2021;Ferreras et al., 2020;Keller et al., 2020). Viral infections such as cytomegalovirus (CMV) and Epstein-Barr virus (EBV) have previously been treated by transfer of highly functional cytotoxic T cells (Einsele et al., 2002;Feuchtinger et al., 2010b;Papadopoulou et al., 2014). It remains to be shown whether adoptive transfer of T cells can control SARS-CoV-2 infection in pre-clinical models, which are complicated by the requirement to be done in the Biological Safety Level (BSL) 3 setting. Therapeutic T cell engineering is now routinely done for cancer treatment both in the context of clinical trials as well as US Food and Drug Administration (FDA)-approved therapeutics (D'Angelo et al., 2018;Depil et al., 2020;Johnson et al., 2006). There are several advantages to adoptive cell therapy with engineered T cells: (1) a large number of antigen-specific T cells can be readily produced; (2) well-validated TCR specificity; (3) T cells have a younger phenotype. TCR-engineered T cell also enlist additional CD4 + T cells, which are critical for establishing long-term CD8 + T cell memory and antibody production (Sant and McMichael, 2012;Sun and Bevan, 2003). Current approaches for adoptive T cell therapy are expensive and cumbersome (Depil et al., 2020). Technological advances in gene delivery may make T cell engineering a practical approach for viral disease treatment in specific groups of patients (Frank et al., 2020).
A note added in proof: replication complex reactive T cells were recently identified to be protective against seroconversion in health care workers who were likely exposed to SARS-CoV-2 (C) FACS analysis of intracellular staining for TNFa and IFNg in T cells. Cells are gated on light scatter, CD3 + , CD8 + . (D) FACS results from (C) Data from PBMC samples expressing the different TCR constructs are pooled into groups based on the activation stimulus and differences are compared across groups. Student's t test is used to compare the groups: **p > 0.01, ***p > 0.001, ****p > 0.0001. (E) PBMC cytotoxicity assay against RdRp-overexpressing target cell line. (F). IFNg production after 48 h of co-culture of PBMCs and target cell lines expressing RdRp. HLA-A*02:01-only parental cell line is used as negative control. Error bars represent standard deviation.
6 Cell Reports 37, 110167, December 28, 2021 Report ll OPEN ACCESS (Swadling et al., 2021). The authors suggest that parts of the replication complex, which contains the polymerase, can be used as a vaccine to induce pan-coronavirus cellular immunity. This work is complementary to our findings here.

Limitations of the study
We show that RdRp-specific TCRs recognize processed epitopes in a reconstructed system of viral infection; however, we do not show direct control of live SARS-CoV-2. Such an experiment is complicated by the requirement to be done in BSL3 setting. In addition, viruses developed complicated mechanisms to escape T cell effector function, which can make it difficult to detect activity and recognition. T cell function in SARS-CoV-2 is still being investigated, and direct T cell suppression of viral replication has yet to be established. We did not show correlation between T cell responses we identified and disease severity. Such analysis requires a large patient cohort and could indicate the importance of specific T cell responses.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:  Peptide dilution assay to measure reactivity against peptide homologs from HCoVs. PBMCs were engineered to overexpress the TCRab constructs, expanded, and then pulsed with the indicated concentrations of each of the indicated peptide epitopes in the presence of the K562-A*02:01 to support antigen presentation. Intracellular staining for IFNg is used to measure T cell activation.

DECLARATION OF INTERESTS
O.N.W., J.M., and P.A.N. are inventors of a patent application in progress that will be filed prior to manuscript publication. J.R.H. is a board member of PACT Pharma and Isoplexis. O.N.W. currently has consulting, equity, and/or board relationships with Trethera Corporation, Kronos Biosciences, Sofie Biosciences, Breakthrough Properties, Vida Ventures, Nammi Therapeutics, Two River, Iconovir, Appia BioSciences, Neogene Therapeutics, and Allogene Therapeutics. None of these companies contributed to or directed any of the research reported in this article.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Owen N. Witte (owenwitte@mednet.ucla.edu).

Materials availability
We generated unique TCR sequences. The full-length TCR clone sequences are provided in this paper. Any cell line that we created and used is available to other investigators.

Data and code availability
Reactive TCR alpha/beta nucleotide sequences are provided in this paper. We have not created any original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Cell culture Cryo preserved peripheral blood mononuclear cells (PBMCs) were commercially purchased (Allcells and Hemacare) or obtained from the CFAR Virology Core Laboratory at the UCLA AIDS Institute. PBMCs were thawed in a water bath set to 37C, transferred to 50 mL