Profound CD8 T cell responses towards the SARS- CoV-2 ORF1ab in COVID-19 patients


 A large global effort is currently ongoing to develop vaccines against SARS-CoV-2, the causative agent of COVID-19. While there is accumulating evidence on the antibody response against SARS-CoV-2, little is known about the SARS-CoV-2 antigens that are targeted by CD8 T cells. To address this issue, we have analyzed samples from 20 COVID-19 patients for T cell recognition of 500 predicted MHC class I epitopes. CD8 T cell reactivity against SARS-CoV- 2 was common. Remarkably, a substantial fraction of the observed CD8 T cell responses were directed towards the ORF1ab polyprotein 1ab, and these CD8 T cell responses were frequently of a very high magnitude. The fact that a major part of the SARS-CoV-2 specific CD8 T cell response is directed against a part of the viral genome that is not included in the majority of vaccine candidates currently in development may potentially influence their clinical activity and toxicity profile.


Introduction
COVID-19, the disease caused by the novel corona virus SARS-CoV-2, is a global emergency. The rst cases of COVID-19 were reported in December 2019, and as of May 22nd 20201 there are more than 5.200.000 con rmed cases and 335.000 deaths. Due to the measures undertaken in attempt to contain the rapid spread of the infection, this global pandemic is having tremendous health and socioeconomic consequences, and there is an urgent need for vaccines. As many as 78 candidate vaccines are currently under development2,3. While there is accumulating evidence on the antibody response against SARS-CoV-24, we are only beginning to acquire knowledge regarding the SARS-CoV-2 speci c T cell response. Therefore, it is an urgent matter to gain a deeper insight into the virus speci c T cell response to both, assist vaccine design and provide tools to evaluate the vaccine-induced T cell responses.
It has been demonstrated in multiple studies that T cells isolated from COVID-19 patients have impaired effector functions and express higher levels of inhibitory receptors in comparison to T cells from healthy individuals, which was found to increase with disease severity5,6. Seminal work from Sette and colleagues has provided rst evidence that T cells speci cally recognize SARS-CoV-2. They demonstrated that both CD4 and CD8 T cells from recovered COVID-19 patients can recognize large pools of SARS-CoV-2 derived peptides7 which is supported by multiple studies currently in preprint archives8. Due to the fact that these rst studies have been focused on large peptide pools, it is currently unknown which exact epitopes are driving the T cell response towards SARS-CoV-2. Identi cation of the exact viral epitopes recognized by T cells, and among the T cell recognized epitopes which are most immunogenic (i.e. the immunodominant epitopes), is of immediate relevance as such knowledge will help guide vaccine design, and allow the development of tools for monitoring SARS-CoV-2 speci c T cell responses. There has been (and still is) a substantial effort ongoing to predict potentially immunogenic SARS-CoV-2 epitopes restricted to both, major histocompatibility complex (MHC) class I and II9-13. However, there is very limited data available con rming which of these epitopes can induce T cell responses. To date, one study tested for CD8 T cell recognition of 30 epitopes derived from the spike protein restricted to HLA-A*02:01. Analyzing material from two COVID-19 patients, a few immunogenic epitopes were identi ed14. However, more systemic efforts are necessary to dissect which of the many potential immunogenic SARS-CoV-2 derived epitopes are in fact recognized by CD8 T cells. In the current study we probed for CD8 T cell recognition towards 500 SARS-CoV-2 derived epitopes restricted to 10 of the most common HLA class I alleles. CD8 T cell reactivity against SARS-CoV-2 was common. Remarkably, a substantial fraction of the observed CD8 T cell responses were directed towards the ORF1ab polyprotein, and ORF1ab speci c CD8 T cell responses were frequently of a very high magnitude.

Epitope selection
To cover as many HLA alleles as possible in a patient-speci c, and in a high throughput manner, we focused our analysis on 10 common HLA alleles of the Italian population including HLA-A*01:01, HLA-  17. In addition, SARS-CoV-2 epitope predictions shared by the science community were considered, and epitopes shared between SARS-CoV and SARS-CoV-2 for which T cell reactivity had previously been reported were included (Table S. 1). So far, most efforts have been focused on identifying epitopes and T cell responses towards in particular the spike protein but also the nucleo-and membrane proteins of SARS-CoV-29-13, however, we decided to broaden our analysis spanning the entire proteome of the virus, which resulted in a high representation of epitopes from the polyprotein encoded by ORF1ab (Fig 1A). Of note, the contribution of epitopes derived from the spike protein and from ORF1ab roughly re ects the difference in size of the proteins.
The SARS-CoV-2 speci c CD8 T cell response To investigate which epitopes from SARS-CoV-2 are recognized by CD8 T cells, PBMC samples collected from 22 patients hospitalized were analyzed (Table 1). All samples were collected during hospitalization and included 10 patients with severe disease that required non-invasive ventilation (NIV, in Infectious Disease Clinics, ID), 9 patients with critical disease (requiring intubation and mechanical ventilation, hospitalized in intensive care unit, ICU), and 3 patients who had recovered from the disease and were off therapies for a week. To probe for T cell recognition of the selected epitopes, we made use of our in-house technology based on multiplexing of peptide HLA (pHLA) multimers conjugated to uorescent dyes. We have previously successfully used this technology to probe for T cell recognition towards shared selfantigens and neoantigens in cancer patients18,19. Utilizing 14 different uorescent dyes to make unique dual uorescent codes made it feasible to probe for T cell recognition of 75 epitopes in parallel. Patients were typed for HLA-A and HLA-B loci, and with the 10 selected HLA alleles we could cover 1 HLA allele for 13 patients, 2 HLA alleles for 5 patients, and 3 alleles for 2 patients (Table 1).
In total we performed 1500 individual antigen-speci c CD8 T cell analyses. We detected a total of 16 SARS-CoV-2 speci c CD8 T cell responses in 9 of 18 patients towards 9 different epitopes (two patients were excluded from the data analysis due to insu cient CD8 T cell count: <1000). Examples of SARS-CoV-2 speci c T cell responses are shown in Fig 1B. Strikingly, a subset of the identi ed CD8 T cell responses were of profound magnitude (above 10% of total CD8+ cells). The average magnitude of the detected responses was 2.4% of total CD8+ cells (range: 0.006 to 18.4% of total CD8+ cells) (Fig 1C).
Eleven of the detected responses were restricted to HLA-A*01:01, 3 responses were restricted to HLA-A*02:01, 1 response was restricted to HLA-A*03:01, and 1 to HLA-A*24:02. Interestingly, among the COVID-19 patients, there was a complete lack of T cell responses restricted to the included HLA-B alleles.
To validate the identi ed CD8 T cell responses, PBMC samples from 2 COVID-19 patients with none of the HLA alleles included in the study were analyzed. All the epitope and HLA combinations for which we had identi ed CD8 T cell responses were included, and no non-speci c signals were observed (data not shown). Furthermore, for 3 patients we had su cient PBMC material to validate the high magnitude responses detected. This validation was carried out using altered dual uorescent codes compared to the initial detection and all responses were con rmed (representative examples are shown Fig S3). Taken together these data strongly indicate that the identi ed responses are indeed true antigen speci c CD8 T cell responses.
It has recently been demonstrated that T cell responses towards SARS-CoV-2 derived epitopes can be detected in PBMCs of unexposed healthy individuals7. To test if there was cross-reactivity towards the SARS-CoV-2 epitopes included in our analysis, PBMC samples from 4 healthy individuals collected prior to October of last year were analyzed. In total, we could cover 7 of the 10 included HLA alleles with these analyses We identi ed a single response in healthy donor 1 (HD1) restricted to HLA-B*15:01 which was of low magnitude (0.008% of total CD8+ cells), and thus observed a considerably lower level of antigen-speci c CD8 T cell responses in healthy donors compared to COVID-19 patients ( Fig 1D). This lack of responses in healthy donors compared to work from Grifoni et al.7 may be due to the low number of epitopes (11 of 500) included with known overlap with coronaviruses causes 'common colds' (Table S1).
The analyzed patient cohort included 6 patients hospitalized with severe disease, 9 patients with critical disease (admitted to ICU) and 3 patients who were recovering in the hospital. This allowed us to compare differences in magnitude and number of detected CD8 T cell responses between these groups. We found that the recovering patients had a trend towards higher magnitude T cell responses compared to the two other groups of patients, and that the magnitude of T cell responses decreased with increasing disease stage, although not signi cantly ( Fig 1E). Intriguingly, 13 of the 16 identi ed T cell responses were detected in patients with severe disease or recovering from severe disease while only 3 of 13 responses were detected in patients with critical disease (Fig 1C).
Immunogenicity hierarchy within T cell recognized SARS-CoV-2 epitopes A striking observation was that the vast majority of identi ed responses were restricted to the HLA-A*01:01 allele (11 of 16) ( Fig 1C). Furthermore, a CD8 T cell response towards the ORF1ab epitope TTDPSFLGRY was detected in all 5 HLA-A*01:01 positive patients, and these responses were of signi cantly higher magnitude compared to the magnitude of all other detected CD8 T cell responses ( Fig 1F, p = 0.0002). Furthermore, for 3 out of the 5 HLA-A*01:01+ patients, additional responses were identi ed (likewise restricted to HLA-A*01:01), and the magnitude of these additional responses was at least 20-fold lower compared to the response towards the TTDPSFLGRY epitope within each patient. This observation strongly indicates that this epitope is likely to drive a large part of the CD8 T cell response towards SARS-CoV-2 in HLA-A*01:01 positive patients, and is likely the immunodominant epitope in this subgroup of patients. Of note the bias of detected CD8 T cell responses towards HLA-A*01:01 was not a re ection of differences in the quality of epitope prediction (measured by predicted binding a nity to HLA) ( Fig 1SA).
The origin of the CD8 T cell recognized SARS-CoV-2 epitopes Of the 9 CD8 T cell recognized epitopes identi ed in COVID-19 patients, 4 are unique for SARS-CoV-2 and 5 are shared between SARS-CoV-2 and SARS-CoV (Table S. 1). Of the 5 epitopes shared between SARS-CoV-2 and SARS-CoV, 1 was previously demonstrated to be immunogenic20. Noteworthy, a substantial part of identi ed CD8 T cell recognized epitopes (4 of the 9) were derived from OFR1ab. However, CD8 T cell recognized epitopes derived from the spike protein (S) were also well-represented (3 of 9). In contrast, only 1 epitope was derived from the nucleoprotein (N) and 1 from the membrane protein (M). Importantly, the CD8 T cell responses speci c for the epitopes derived from ORF1ab were of signi cantly higher magnitude compared to the CD8 T cell responses towards S, N and M combined ( Fig 1G, p = 0.0027). This is potentially a very important observation for design of vaccine candidates, as the majority of current vaccine candidates are focused on the spike protein which may result in an induction of a more limited CD8 T cell response compared to the naturally induced CD8 T cell response.
CD8 T cell recognized SARS-CoV-2 epitopes are not located in hotspot regions of the SARS-CoV-2 genome Next, we examined whether the SARS-CoV-2 epitopes recognized by CD8 T cells were spanning the positions of the SARS-CoV-2 genome demonstrated to contain a high level of single nucleotide polymorphisms (SNPs). the so-called 'hotspots'. Such information is of key importance for the utilization of the obtained information for both vaccine development and monitoring efforts. For this purpose, we used the SARS-CoV-2 Alignment Screen tool (version as per 14-05-2020) containing sequencing information of 7667 SARS-CoV-2 isolates21. Importantly, the 9 T cell recognized epitopes identi ed in the COVID-19 patients did not overlap with these 'hotspots' (Fig S4). Furthermore, between 7667 and 6996 of the virus isolates contained information regarding the potential consequence of the SNPs on the amino acid level in the regions encoding for the 9 CD8 T cell recognized epitopes. The median fraction of these isolates encoding the T cell epitopes were 99.98% with a range of 97.13% to 100%. These ndings strongly indicate that the identi ed immunogenic epitopes are highly preserved, making them excellent candidates to be included in the development of tools for monitoring the SARS-CoV-2 speci c T cell responses in COVID-19 patients and evaluation of vaccine-induced immunity.

SARS-CoV-2 speci c T cells display high expression levels of inhibitory receptors
Finally, we evaluated the expression level of 4 inhibitory receptors associated with T cell activation and exhaustion (NKG2A, PD-1, TIM-3, and 2B4). The fraction of CD8 T cells expressing NKG2A has been shown to be increased on bulk CD8 T cells in COVID-19 patients compared to healthy donors6. We found no signi cant differences in the frequency or expression levels of these inhibitory receptors in bulk CD8 T cells of COVID-19 patients in comparison to healthy donors which may re ect low sample numbers (Figure S. 5). However, when focusing the analysis on the SARS-CoV-2 speci c CD8 T cells in comparison with bulk CD8 T cells, we found that the fraction of NKG2A positive cells was signi cantly higher among SARS-CoV-2-speci c CD8 T cells (Fig 2A and B). patients.

Discussion
In this study, we report on a systemic effort to identify CD8 T cell recognized epitopes from SARS-CoV-2.
This is to our knowledge a rst effort dissecting which of the many potential HLA class I restricted SARS-CoV-2 epitopes are recognized by CD8 T cells.
A rst observation based on our data is that CD8 T cell reactivity towards SARS-CoV-2 was common in our patient cohort. These ndings are in line with the data from Sette and colleagues demonstrating SARS-CoV-2 reactive CD8 T cell responses in approximately 70% of recovered COVID-19 patients26. Strikingly, 3 of the 9 identi ed T cell recognized epitopes in our analysis yielded responses in up to 5 patients.
Speci cally, we have identi ed 2 epitopes restricted to HLA-A*01:01, TTDPSFLGRY and PTDNYITTY, which induced detectable T cell responses in 5 and 3 patients, respectively. In addition, we have also identi ed an where we only detected 3 of 16 responses. Together with the observation that the patients in the rst group also had CD8 T cell responses of higher magnitude compared to the latter group, it is tempting to speculate that CD8 T cell responses have a protective role. However, our data set is limited in size (and only one timepoint was available per patient), and it is not possible to differentiate between cause and consequence.
Based on our data it seems that certain HLA alleles can induce strong CD8 T cell responses. The magnitude of 6 of 11 CD8 T cell responses restricted to HLA-A*01:01 was above 0.5% of total CD8+ cells, whereas the magnitude of T cell responses restricted towards all other HLA alleles (5 responses in total) was between 0.006 and 0.085% of total CD8 T cells. This observation suggests that immunodominant epitopes may not exist for all HLA alleles (or were simply not included in our epitope selection), and that patients with a speci c (set of) HLA alleles may have a strong CD8 T cell response towards SARS-CoV-2.
Such a scenario is well described for other viral infections including HIV where patients positive for HLA-B*57:01 are more likely to be long term nonprogressors29. If this is also the case for COVID-19 patients, one would expect that certain subgroups of patients with speci c HLA-alleles are likely to experience milder disease courses (given that a strong CD8 T cell response is indeed protective), and that HLA-A*01:01 is one such allele. However, the current data is limited in size and therefore merely hypothesisgenerating, and this needs to be addressed in larger patient cohorts in the future.
Remarkably, a substantial fraction of the identi ed CD8 T cell responses were directed towards ORF1ab, and these responses were of profound magnitude, and signi cantly higher compared to the CD8 T cell responses directed towards spike, nucleo-and membrane proteins combined. These ndings suggest that the ORF1ab epitopes can be highly immunogenic. In particular, the TTDPSFLGRY epitope from the ORF1ab, which is unique to SARS-CoV-2, induced CD8 T cell responses of profound magnitude. These observations are in line with data from Grifoni et al. demonstrating that CD8 T cells from patients who recovered from COVID-19 respond to peptide pools spanning the ORFs 3a, 8 and part of 1ab, and that T cell responses towards ORF3a accounted for 7% of the total SARS-CoV-2 speci c CD8 T cell response26. Together, these data clearly demonstrate the potent immunogenicity of SARS-CoV-2 epitopes derived from viral proteins other than the spike protein, which is currently the focus for vaccine development2,3.
In conclusion, we have provided compelling evidence supporting that SARS-CoV-2 speci c CD8 T cell responses can be of profound magnitude, in particular in patients experiencing milder disease course. The fact that a major part of the SARS-CoV-2 speci c CD8 T cell response is directed against a part of the viral genome not included in the majority of the currently developed vaccine candidates may potentially in uence their clinical activity and toxicity pro le.

Materials
Blood collection and PBMC isolation

SARS-CoV-2 peptide selection and synthesis
Fifty SARS-CoV-2 peptides were selected for each of the top ten most prevalent HLA alleles in Italy. The selection was primarily based on SARS-CoV-2 epitopes that had the highest predicted binding a nity to the MHC according to NetMHCpan-4.016, as well as receiving a prediction score higher than 0.5 using NetChop-3.117. The SARS-CoV-2 proteome was obtained from UniProt (Proteome ID: UP000464024).
Thirteen proteins (pp1ab (ORF1ab), protein 3a, non-structural protein 6, protein 7a, non-structural protein 7b, non-structural protein 8, protein 9b, ORF10 protein, uncharacterized protein 14, envelope small membrane protein (E), membrane protein (M), nucleoprotein (N) and spike glycoprotein (S)) were considered as sources of potential epitopes. All possible 9-11mer peptide sequences were derived from the thirteen SARS-CoV-2 proteins, and epitopes were ranked and selected as described above. In addition, SARS-CoV-2 epitopes that were predicted to be most immunogenic by the science community9-11,37 were included for analysis (Table S.

Statistical analysis
Differences between two or multiple patient groups were assessed using the non-parametric Mann-Whitney U-test or the ordinary one-way ANOVA test, respectively. The data cut-off for all analyses was 21 May 2020. Statistical analysis was performed using Excel 16

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.