key: cord-1053129-978g6hxe
authors: Saini, Sunil Kumar; Hersby, Ditte Stampe; Tamhane, Tripti; Povlsen, Helle Rus; Hernandez, Susana Patricia Amaya; Nielsen, Morten; Gang, Anne Ortved; Hadrup, Sine Reker
title: SARS-CoV-2 genome-wide mapping of CD8 T cell recognition reveals strong immunodominance and substantial CD8 T cell activation in COVID-19 patients
date: 2020-10-19
journal: bioRxiv
DOI: 10.1101/2020.10.19.344911
sha: 86ae3e48d0204dd7c30126c58395898bef8184d9
doc_id: 1053129
cord_uid: 978g6hxe

To understand the CD8+ T cell immunity related to viral protection and disease severity in COVID-19, we evaluated the complete SARS-CoV-2 genome (3141 MHC-I binding peptides) to identify immunogenic T cell epitopes, and determine the level of CD8+ T cell involvement using DNA-barcoded peptide-major histocompatibility complex (pMHC) multimers. COVID-19 patients showed strong T cell responses, with up to 25% of all CD8+ lymphocytes specific to SARS-CoV-2-derived immunodominant epitopes, derived from ORF1 (open reading frame 1), ORF3, and Nucleocapsid (N) protein. A strong signature of T cell activation was observed in COVID-19 patients, while no T cell activation was seen in the ‘non-exposed’ and ‘high exposure risk’ healthy donors. Interestingly, patients with severe disease displayed the largest T cell populations with a strong activation profile. These results will have important implications for understanding the T cell immunity to SARS-CoV-2 infection, and how T cell immunity might influence disease development.

The COVID-19 (Coronavirus disease 2019) pandemic caused by the highly infectious SARS-CoV-2 (severe 2 acute respiratory syndrome coronavirus 2) has challenged public health at an unprecedented scale, 3 causing the death of more than one million people so far (World Health Organization). The T-cell of the 4 immune system is the main cell type responsible for the control and elimination of viral infections; CD8 + 5 T cells are critical for the clearance of viral-infected cells, whereas CD4 + T cells are critical for supporting 6 both the CD8 + T cell response efficacy and the generation of specific antibodies. Characteristics from the 7 ongoing pandemic suggest that T cell recognition will be critical to mediate long-term protection against 8 SARS-CoV-2 (Cañete and Vinuesa, 2020), as the antibody-mediated response seems to decline in 2 To reveal the full spectrum of T cell immunity in COVID-19 infection, we used a complete SARS-CoV-2 3 genome sequence to identify immunogenic minimal epitopes recognized by CD8 + T 4 cells. Using NetMHCpan 4.1 (Reynisson et al., 2020) , we selected 2204 potential HLA binding peptides 5 (9-11 amino acids) for experimental evaluation. These peptides were predicted to bind one or more of Table 3) , and for comparative 20 evaluation, we furthermore included 39 T cell epitopes from common viruses; cytomegalovirus (CMV), 21

Epstein-Barr virus (EBV), and influenza (Flu) virus (CEF, Supplementary Table 4) ( Figure 1B) . 22 We found broad and strong SARS-CoV-2-specific CD8 + T cell responses in COVID-19 patients, 1 contributing up to 27% of the total CD8 + T cells ( Figure 1C) . A few epitopes showed characteristics of 2 immunodominance, and raised T cell responses up to 25% of total CD8 + T cells against two overlapping 3 epitopes; HLA-A01:01 TTDPSFLGRY (11%), and TTDPSFLGRYM (14%), in some patient-specific T cell 4 repertoires ( Figure 1D, Supplementary Figure 2 , and Supplementary Table 5 ). We validated the 5 presence of SARS-CoV-2 CD8 + T cells identified with our pMHC multimer-based T cell detection 6 technology using functional analysis in selected patient samples (Supplementary Figure 3) . 7

In total, we identified T cell responses to 142 pMHC complexes corresponding to 122 unique T cell 8 epitopes across the ten analyzed HLAs ( Figure 1E ). HLA-A01:01, A02:01, and B15:01 dominated in terms 9 of the total number of identified epitopes as well as the immunogenicity score (i.e., the number of T cell 10 responses normalized to the number of probing pMHC multimers and the number of patients analyzed) 11 ( Figure 1F ). HLA-A03:01 and C07:01 specific peptides showed the least T cell reactivity (three epitopes 12 each) despite being analyzed in nine and six patients, respectively ( Figure 1E) . Most of the immunogenic 13 epitopes were mapped to the ORF1 protein, followed by S and ORF3 proteins ( Figure 1G and H, and 14

Supplementary Table 5) . Given the size difference of the viral proteins, the 'immunogenicity score' was 15 used to evaluate their relative contribution to T cell recognition. Based on such evaluation, we observe 16 that peptides derived from ORF3 displayed the highest relative immunogenicity (in terms of T cell 17 recognition), followed by ORF1 protein (Figure 1H ). Of the 122 epitopes recognized by T cells in the 18 patient cohort, 13 were determined as 'immunodominant' based on the presence of T cell recognition in 19 two or more patients, and prevalence of >25% in the tested samples with the given HLA molecule 20 ( Figure 1I) . Among these, a very robust HLA associated immunodominance was observed for two of the 21 epitopes: HLA-A01:01-TTDPSFLGRY-specific (and its variant peptides TTDPSFLGRYM and HTTDPSFLGRY), 22

with specific T cells detected in all five analyzed patients (estimated frequency reaching up to 25% of 23 total CD8 + T cells); and HLA-B07:02-SPRWYFYYL, with specific T cells observed in four of the five patients 24

1 patients 2 Since our study design integrated T cell phenotype characterization in combination with the SARS-CoV-3 2-specific T cell identification, we evaluated and compared phenotypic characteristics of the SARS-CoV-2 4 reactive T cell populations in COVID-19-infected patients and healthy donors. This also allowed us to 5 evaluate whether the multimer-specific T cell profile of the high-risk COVID-19 healthy cohort (HD-2) has 6 any distinct features compared to the unexposed cohort (HD-1). Dimensionality reduction visualization 7 of SARS-CoV-2 multimer positive T cells using UMAP (Uniform Manifold Approximation and Projection) 8 showed distinct clustering for the patient cohort compared to the two healthy donor cohorts ( Figure  9 3A). Patient-derived SARS-CoV-2 multimer-positive T cells were distinguished with higher expression of 10 activation markers CD38, CD69, CD39, HLA-DR, CD57, and reduced expression of CD8 and CD27 (Figure  11 3B). These features were found unique to SARS-CoV-2-specific T cells, as no difference was observed 12 between the three cohorts in similar analysis for CEF-specific multimer positive T cells (Supplementary 13 Figures 6A and B) . Additionally, SARS-CoV-2 reactive T cells in patients and healthy donor cohorts 14 showed a similar distribution of memory subsets, effector memory (E M ) CCR7 -CD45RA -; central memory 15 (C M ), CCR7 + CD45RA -; and terminally differentiated effector memory (T EMRA ), CCR7 -CD45RA + ; and naïve 16 (CCR7 + , CD45RA + ) phenotype (Supplementary Figures 7A and B) . However, SARS-CoV-2-specific T cells of 17 the patient cohort showed unique clustering (UMAP) of EM and TEMRA and these two subsets 18 particularly showed strong expression of the T cell activation markers (Supplementary Figures 7C and D Phenotype analysis demonstrated a highly activated state of SARS-CoV-2-specific T cells, with a 21 significantly higher fraction of such T cells expressing the inflammation marker (CD38) and early-stage 22 activation markers (CD39, CD69, and HLA-DR), and showed a late-differentiated effector memory profile 23 (reduced CD27) together with increased CD57 expression (not significant) compared to the two healthy 1 donor cohorts ( Figure 3C) . We did not observe activation of SARS-CoV-2 specific multimer-positive T 2 cells in the high-risk COVID-19 healthy cohort, except for non-significant trends for reduced CD27 and 3 increased CD57 expression ( Figure 3C) . Furthermore, the highly activated and differentiated T cell 4 phenotype in the COVID-19 patients was associated with the SARS-CoV-2-specific T cells and not to the 5 CEF-specific T cells detected in this cohort ( Figure 3D ). To further characterize the SARS-CoV-2-specific T 6 cell phenotype, we compared the expression of the T cell activation markers in combination with 7 inflammatory response marker CD38 on multimer positive CD8 + T cells across the three cohorts, which 8 showed significantly enhanced expression of activation molecules (CD39, CD69, and HLA-DR) and PD-1 9

inhibitory receptor on CD38 + T cells only in the patient cohort ( Figure 3E and F). 10

Altogether, these results demonstrate highly active and proliferative SARS-CoV-2-specific T cell 11 responses in COVID-19-infected patients, and distinguishes them from the potential cross-reactive T cell 12 repertoire detected in healthy donors. 13 15 To dissect the association of T cells with COVID-19 disease severity, we next evaluated the phenotype 16 characteristics of SARS-CoV-2-specific CD8 + T cells in the patient cohort related to their requirement for 17 hospital care. Our patient cohort consists of severely diseased patients requiring hospitalization 18 (hospitalized; n = 11) and patients with mild symptoms not requiring hospital care (outpatient; n = 7). 19

Substantially higher frequency of SARS-CoV-2-specific CD8 + T cells and increased total number of 20 epitopes were detected in samples from hospitalized patients compared to outpatient samples ( Figure  21 4A). Comparing the SARS-CoV-2 specific T cell population (multimer + ) between hospitalized and 22 outpatients for phenotype markers, we observed a clear trend in increased expression of CD38, CD39, 23 CD69, HLA-DR (non-significant), and PD-1 (significant) ( Figure 4B) . Furthermore, measuring co-1 expression of immune activation marker CD38 together with CD39, CD69, PD-1, and HLA-DR showed a 2 strong elevation of these combinations of activation markers in hospitalized patients ( Figure 4C and D To further elucidate the potential origin of such a cross-reactive T cell population in the healthy donors 16 cohort, we next evaluated the sequence homology of SARS-CoV-2 MHC-I binding peptides with the four 17 common cold coronaviruses; HCoV-HKU1, HCoV-NL63, and HCoV-229E. With a variation limit of up to 18 two amino acids in each peptide sequence, 15% of the total predicted peptides showed sequence 19 similarity with one or more HCoV peptide sequence ( Figure 5C, grey pie) . Among the T cell recognized 20 peptides, in both the patient and healthy donor cohorts, respectively, this fraction was comparable with 21 19% and 16% of T cell recognized peptides sharing sequence homology with one or more HCoV ( Figure  22 5C). As an alternative approach the similarities were calculated by kernel method for amino acid 23 sequences using BLOSUM62, indicating comparable sequence similarity of peptides recognized by T cells 24 and those not recognized in reference to HCoV. However, interestingly, peptides of lowest similarity 1 were not recognized by T cells in the patient cohort (Supplementary Figure 8) . the sequence similarity to the core of the peptide that would most likely interact with the TCR (Bentzen 5 et al., 2018) . Based on the protein-core only, up to 74% of all the identified epitopes showed sequence 6 homology to HCoV (one or more) (Figure 5C ), suggesting these common cold viruses as a potential 7 source of the observed low-avidity interactions in healthy donors. Further, when evaluating peptides 8 frequently recognized by T cells in both the patients and the healthy cohort, we find evidence of 9 substantial homology, as exemplified with the peptide sequences listed in Figure 5D . However, similar 10 sequence homology is observed for the peptide sequences that are recognized only in the patient 11 cohort ( Figure 5D) . Thus, at present, our data points to substantial T cell cross recognition being 12 involved in shaping the T cell response to SARS-CoV-2 in COVID-19 patients, however, we find no specific 13 enrichment of T cell recognition to peptide sequences with large sequence homology compared to the 14 total peptide library being evaluated, and the responses that are identified in the patient samples only 15

does not hold a more SARS-CoV-2 unique signature than those recognized in both cohorts. Interestingly, 16 however, ORF1 being one of the elements displaying the highest T cell recognition immunogenicity, also 17 display the highest sequence identity to HCoV (40%, as oppose to 22-34% for all other SARS-CoV- 2 18 proteins, calculated using direct sequence alignment). Seeking to fully understand the role and origin of 19 the underlying T cell cross-recognition will likely require an in-depth evaluation of pre-and post-20 infection samples. 1 We identified CD8 + T cell responses to 122 epitopes in 18 COVID-19 patients after screening for T cell 2 recognition based on 3141 peptides derived from the full SARS-CoV-2 genome, and selected based on 3 their predicted HLA-binding capacity. Of these, a few dominant T cell epitopes were recognized in the 4 majority of the patients. Strikingly, both dominant and subdominant T cell epitopes were cross-5 recognized by low-level existing T cell populations in SARS-CoV-2 unexposed healthy individuals. We 6 have observed that the SARS-CoV-2 dominant epitopes mount very strong T cell responses, with up to 7 25% of all CD8 + lymphocytes specific to a single epitope (two overlapping peptides with same peptide 8 core). 9

Pre-existing immunity based on cross-reactive T cells can influence how our immune system reacts upon cell cross-recognition is very difficult to predict, without knowing the precise interaction required for the 15 given TCR, as even T cell epitopes with as low as 40% sequence homology can be recognized by a given 16 TCR (Bentzen et al., 2018) . Therefore, the underlining source of T cell cross-reactivity might arise from a 17 larger set of epitopes within the HCoV viruses, including sequences with larger variation than those 18 evaluated here (i.e., max. two amino acid variants per peptide sequence/peptide core). 19

While the T cell recognition itself was largely overlapping in identity between patients and healthy 20 donors, the magnitude for the T cell responses and, in particular, the T cell phenotype of SARS-CoV-2-21 specific T cells was substantially different. A unique phenotype characterization demonstrated a strong 22 activation profile of SARS-CoV-2-specific T cells only in patients. This strong 'activation signature' (high 23 expression of CD38, CD39, CD69, PD1) was further enhanced in patients requiring hospitalization. Such 24 strong and highly activated T cell responses should be able to clear the virus, and hence our data further 1 support the notion that some severely affected patients might suffer from hyperactivation of their T cell 2 compartment as a consequence of their primary viral infection, which may even be cleared. 3

Taken together, the data presented here demonstrate a substantial role for T cell recognition in 4 COVID-19, and in-depth evaluations in larger cohorts over time will provide essential insight to the role 5 of such T cells in disease severity and how pre-existing T cell immunity can be leveraged to fight novel 6 infections. The patient cohort included samples from individuals with severe symptoms who required hospital care 7 (hospitalized; n = 11) and patients with mild symptoms not requiring hospital care (outpatient; n = 7). 8

For hospitalized patients, we collected full data from the medical record regarding disease course, age, 9

gender, travel history, performance status, symptoms, comorbidity, medications, laboratory findings, 10 diagnostic imaging, treatment, need of oxygen, need for intensive care, and an overall estimate of 11 disease severity (Supplementary Table 2 ). For outpatients, we used a questionnaire to collect data on 12 comorbidity, travel history, medications, and performance status. Additionally, we included 20 health care employees from Herlev Hospital during the COVID- 19 3 pandemic, who were at high risk of COVID-19 infection but not positive, as a cohort to follow immune 4 responses in a potentially exposed population. 5

PBMCs from all three cohorts were isolated immediately after sampling using Ficoll-Paque PLUS (GE 6

Healthcare) density gradient centrifugation and were cryopreserved thereafter at a density of 2-20 × 7 10 6 cells/mL. 8 9 Potential HLA class I binding peptides were predicted from the complete set of 8-11mer peptides from ORF1 protein, a percentile rank binding threshold of 0.5% was used, and for peptides derived from 16 all other proteins, a threshold of 1% was used. Altogether, 2203 peptides were selected, binding to one 17 or more HLA molecules, generating 3141 peptide-HLA pairs for experimental evaluation (Supplementary 18 Table 1 ). All peptides were custom synthesized by Pepscan Presto BV, Lelystad, The Netherlands. 19

Peptide synthesis was done at a 2 µmol scale with UV and mass spec quality control analysis for 5% 20 random peptides by the supplier. 21 22 All ten MHC-I monomer types were produced using methods previously described (Saini et al., 2013) . 1

Briefly, MHC-I heavy chain and human ß2-microglobulin (hß2m) were expressed in Escherichia coli using 2 pET series expression plasmids. Soluble denatured proteins of the heavy chain and hß2m were 3 harvested using inclusion body preparation. The folding of these molecules was initiated in the presence 4 of UV labile HLA specific peptide ligands (Hadrup et al., 2009a) . HLA-A02:01 and A24:02 molecules were 5 folded and purified empty, as described previously (Saini et al., 2019) . Folded MHC-I molecules were 6 biotinylated using the BirA biotin-protein ligase standard reaction kit (Avidity, LLC-Aurora, Colorado), 7

and MHC-I monomers were purified using size exclusion chromatography (HPLC, Waters Corporation, 8 USA). All MHC-I folded monomers were quality controlled for their concentration, UV degradation, and 9 biotinylation efficiency, and stored at -80°C until further use. 10

The DNA-barcoded multimer library was prepared using the method developed by Bentzen et al. 1 To process the sequencing data and automatically identify the barcode sequences, we designed a 2 specific software package, 'Barracoda' (https://services.healthtech.dtu.dk/service.php?Barracoda-1.8). 3

This software tool identifies the barcodes used in a given experiment, assigns sample ID and pMHC 4 specificity to each barcode, and calculates the total number of reads and clonally reduced reads for each 5 pMHC-associated DNA barcode. Furthermore, it includes statistical processing of the data. False-discovery rates (FDRs) were estimated using the Benjamini-Hochberg method. Specific barcodes 13 with an FDR < 0.1% were defined as significant, determining T cell recognition in the given sample. At 14 least 1/1000 reads associated with a given DNA barcode relative to the total number of DNA barcode 15 reads in that given sample was set as the threshold to avoid false-positive detection of T cell populations 16 due to the low number of reads in the baseline samples. T cell frequency associated with each 17 significantly enriched barcode was measured based on the % read count of the associated barcode out 18 of the total % multimer-positive CD8 + T cells population. In order to exclude potential pMHC elements 19 binding to T cells in a non-specific fashion, non-HLA-matching healthy donor material was included as a 20 negative control. Any T cell recognition determined in this samples was subtracted from the full data 21 set. 22 23 PBMCs from healthy donors were expanded in-vitro using pMHC-dextran complexes conjugated with 1 SARS-CoV-2-derived peptides and cytokines (IL-2 and IL-21) for 2 weeks either with single pMHC 2 specificity or with a pool of up to ten pMHC specificities. PBMCs were expanded for 2 weeks in X-vivo 3 media (Lonza, BE02-060Q) supplemented with 5% human serum (Gibco, 1027-106). Expanded cells were 4 used to measure peptide-specific T cell activation or stained using pMHC tetramers to detect T cells 5 recognizing SARS-CoV-2 epitopes. 6

In-vitro expanded healthy donor PBMCs were examined for SARS-CoV-2 reactive T cells using 16 For phenotype analysis, all samples were analyzed using FlowJo data analysis software (FlowJO LLC). 17 -MFI of multimercells)/2 × standard deviation (SD) of multimercells)). MFI of multimer + and multimer -10 CD8 + T cells and the SD of the multimer -CD8 + T cells from FlowJo analysis for patient and healthy donor 11 samples. 12 13 Data and code availability 14 The data that support the finding of this study, in addition to the supplementary supporting data, and 15 the code used to generate the plots and analyses can be accessed from the corresponding author upon 16 reasonable request. 17 18 

Structural interplay between germline interactions and 3 adaptive recognition determines the bandwidth of TCR-peptide-MHC cross-reactivity

T-cell-receptor cross-recognition and strategies to select safe T-6 cell receptors for clinical translation

Large-scale detection of antigen-specific T cells using 9 peptide-MHC-I multimers labeled with DNA barcodes

T cell receptor fingerprinting enables in-depth characterization of 12 the interactions governing recognition of peptide-MHC complexes

SARS-CoV-2-specific T cell immunity in cases of COVID-19 and SARS, and 15 uninfected controls

Deconstructing the peptide-MHC specificity of t cell 18 recognition

SARS-CoV-2-reactive T cells in healthy donors and patients with COVID-19

SARS-CoV-2-derived peptides define heterologous and COVID-19-induced T 1 cell recognition

Detection of SARS-CoV-2-Specific Humoral and Cellular Immunity in COVID-19 Convalescent Individuals

Prevalence of Asymptomatic SARS-CoV-2 Infection

Long-lived memory T lymphocyte responses against SARS coronavirus nucleocapsid protein in 8 SARS-recovered patients

Broad and strong memory CD4+ and CD8+ T cells induced by SARS-CoV-2 in UK 11 convalescent individuals following COVID-19

NetMHCpan-4.1 and NetMHCIIpan-13 4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and 14 integration of MS MHC eluted ligand data

Not all empty 16 MHC class I molecules are molten globules: Tryptophan fluorescence reveals a two-step mechanism of 17 thermal denaturation

Empty peptide-receptive MHC class I molecules for efficient 20 detection of antigen-specific T cells

Respiratory Syndrome: A Six-Year Follow-Up Study

Antibody responses to SARS-CoV-2 short-lived

The 3 laboratory tests and host immunity of COVID-19 patients with different severity of illness

Phenotype and kinetics of SARS-CoV-6 2-specific T cells in COVID-19 patients with acute respiratory distress syndrome

Preexisting influenza-specific CD4 + T cells correlate with 9 disease protection against influenza challenge in humans

A new coronavirus associated with human respiratory disease in China

Cytokine release syndrome in severe COVID-13 19: interleukin-6 receptor antagonist tocilizumab may be the key to reduce mortality

We thank all patients and healthy donors for participating and contributing the 1