key: cord-0989192-kvuxxofb
authors: Yu, Esther Dawen; Wang, Eric; Garrigan, Emily; Goodwin, Benjamin; Sutherland, Aaron; Tarke, Alison; Chang, James; Gálvez, Rosa Isela; Mateus, Jose; Ramirez, Sydney I.; Rawlings, Stephen A.; Smith, Davey M.; Filaci, Gilberto; Frazier, April; Weiskopf, Daniela; Dan, Jennifer M.; Crotty, Shane; Grifoni, Alba; Sette, Alessandro; da Silva Antunes, Ricardo
title: Development of a T cell-based immunodiagnostic system to effectively distinguish SARS-CoV-2 infection and COVID-19 vaccination status.
date: 2022-02-08
journal: Cell Host Microbe
DOI: 10.1016/j.chom.2022.02.003
sha: 8bd4bdbde88984ad1abf0f01289979eaddbcc58b
doc_id: 989192
cord_uid: kvuxxofb

Both SARS-CoV-2 infection and COVID-19 vaccines elicit memory T cell responses. Here, we report the development of two pools of experimentally-defined SARS-CoV-2 T cell epitopes, that in combination with spike, were used to discriminate four groups of subjects with different SARS-CoV-2 infection and COVID-19 vaccine status. The overall T cell-based classification accuracy was 89.2% and 88.5% in the experimental and validation cohorts. This scheme was applicable to different mRNA vaccines, different lengths of time post-infection/post-vaccination, and yielded increased accuracy when compared to serological readouts. T cell responses from breakthrough infections were also studied, and effectively segregated from vaccine responses, with a combined performance of 86.6% across all 239 subjects from the five groups. We anticipate that a T cell-based immunodiagnostic scheme to classify subjects based on their vaccination and natural infection history will be an important tool for longitudinal monitoring of vaccination and establishing SARS-CoV-2 correlates of protection.

Immune memory against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is associated with cellular and humoral adaptive immunity Rydyznski Moderbacher et al., 2020; Sette and Crotty, 2021) . Progress has been made in defining the correlation of protection based on neutralizing antibody titers , but a full mechanistic understanding of protection from symptomatic and severe disease might require comprehensive characterization of both antibody responses, effector and memory B and T cell responses (Feng et al., 2021; Koup et al., 2021; Krammer, 2021a, b) .

Broad measurement of T cell responses is hindered by the lack of immunodiagnostics tools with effective predictive power able to discriminate pre-existing immunity, vaccination, and infection (Ogbe et al., 2021; Peeling and Olliaro, 2021; Sekine et al., 2020; Vandenberg et al., 2021) . While SARS-CoV-2 T cell responses are detected in nearly all individuals who have recovered from symptomatic COVID-19 (Grifoni et al., 2020b; Le Bert et al., 2020; Tarke et al., 2021a) , they can also be found in 20-50% of unexposed individuals (Mateus et al., 2020; Sette and Crotty, 2020; Tarke et al., 2021a) . However, recent evidence suggests that SARS-CoV-2 infection generates a largely novel repertoire of T cells, with over 80% of the epitopes not recognized in unexposed donors (Mateus et al., 2020; Tarke et al., 2021a) . In addition, mRNA or viral vector vaccines boost the spike protein-specific immune responses in both unexposed and convalescent individuals without affecting the responses to non-spike SARS-CoV-2 components (Bertoletti et al., 2021; Lozano-Ojalvo et al., 2021; Mateus et al., 2021) .

Further complexity is associated with evaluating responses in subjects previously infected and subsequently vaccinated, and conversely, previously vaccinated and subsequently infected (breakthrough infection) Lucas et al., 2021; Niessl et al., 2021; Rovida et al., 2021) .

We have shown that SARS-CoV-2 specific T cells can be detected and quantitated using peptide pools in various T cell assays (da Silva Antunes et al., 2021; Dan et al., 2021; Grifoni et al., 2020a; Mateus et al., 2020; Tarke et al., 2021b) which have proven useful to derive information about the kinetics and magnitude of SARS-CoV-2 specific T cell responses in both COVID-19 infection and vaccination Mateus et al., 2021) . Subsequent studies detailed the repertoire of epitope specificities recognized in a cohort of COVID-19 convalescent subjects (Tarke et al., 2021a) . More recently, a meta-analysis of experimental curated data from the Immune Epitope Database (IEDB) revealed a large repertoire of over 1400 epitopes defined in 25 different studies . Here, we used this information to develop SARS-CoV-2-specific peptide pools optimized for broader epitope repertoire and wider HLA coverage for both CD4+ and CD8+ T cell responses. Accordingly, two pools of Experimentally defined epitopes derived from the non-spike Remainder of the SARS-CoV-2 proteome, (CD4RE and CD8RE) were established.

Several platforms and strategies have been developed to assess T cell responses in both vaccinated or infected individuals, using different read-outs and technologies, such as cytokine release assays (ELISPOT, or ELISA) (Krishna, Preprint; Kruse et al., 2021; Martínez-Gallo, 2021; Murugesan, Preprint; Tan et al., 2021; Tormo, Pre-proof) or flow cytometry-based assays (Blast transformation, or intracellular cytokine staining (ICS)) (Lind Enoksson et al., 2021; Zelba et al., 2021) . These assays mainly rely on the characterization of responses to the spike or nucleocapsid antigens, and therefore do not address the entire SARS-CoV-2 proteome and the remarkable breadth of T cell responses generated against this pathogen .

In this study, we developed an immunodiagnostic T cell assay using a pool of overlapping peptides spanning the entire spike protein in combination with experimentally defined non-spike pools to classify subjects based on their vaccination and infection history.

This tool showed high predictive power to discriminate responses based on distinctive COVID- 

239 participants were enrolled in the study and classified into five groups based on known vaccination and infection history: (50 non-infected, non-vaccinated (I-V-); 50 infected and nonvaccinated (I+V-); 66 infected and then vaccinated (I+V+); 50 non-infected and vaccinated (I-V+); and 23 vaccinated and then infected (V+I+). An overview of the characteristics from all the participants is provided in Table 1 . For the I+V-, I+V+ and V+I+ groups, SARS-CoV-2 infection was determined by PCR-based testing during the acute phase of infection or verified by serological detection of antibodies against the SARS-CoV-2 Spike protein RBD region at the time of blood donation.

The study primarily consisted of subjects recruited in San Diego, California (see material and methods for more details). Among individuals with history of COVID-19 disease, the majority were symptomatic mild disease cases, owing to the nature of the study recruitment design. Specifically, 44 donors (88%) for I+V-, 45 donors (90%) for I+V+, and 23 donors (100%) for V+I+ had mild symptoms, 3 donors (6%) of I+V-and I+V+ groups had moderate symptoms, and 3 (6%) and 2 donors (4%) from the I+V-and I+V+ groups, respectively, had severe symptoms. The median days of blood collection post symptom onset (PSO) were 119 (20-308), 354 (57-508) and 32 (18-93) for I+V-, I+V+ and V+I+ groups respectively. For the I-V+, I+V+ and V+I+ groups, the vaccinated subjects received two doses of mRNA vaccines BNT162b2 (Pfizer/BioNTech) or mRNA-1273 (Moderna), as verified by vaccination records and positive plasma SARS-CoV-2 spike protein RBD IgG titers. Similar distribution of Pfizer or Moderna administered vaccines (45%-55%) were present in vaccinated subjects from either the I-V+ or I+V+ group, while in the V+I+ group, 15 (65%) subjects had received the BNT162b2 vaccine, and 8 (35%) the mRNA-1273 vaccine.

The median days of blood collection post second dose of vaccination (PVD) were 16 , 32 (7-188) and 163 (55-271) for I-V+, I+V+ and V+I+ groups, respectively. All the I-V-subjects were collected before the attributed pandemic period (2013-2019) and confirmed seronegative with undetectable SARS-CoV-2 Spike protein RBD IgG titers. In all cohorts, the median ages were relatively young (25 (17-64), 42 (19-67), 40 (21-74), 38 (21-73), 30 (22-68) for I-V-, I+V-, I-V+, I+V+ and V+I+ groups respectively), with the female gender well represented and different ethnicities represented. In our study, participants were further divided in an exploratory cohort (120 donors, Table S1 ), an independent validation cohort (96 donors, Table S2 ) and a third cohort of breakthrough infections (V+I+; 23 donors, Table 1 ).

To detect SARS-CoV-2 T-cell reactivity, we previously routinely utilized a pool of overlapping peptides spanning the entire spike (S) sequence (253 peptides) and a pool of predicted HLA Class II binders from the Remainder (R) of the genome (CD4R; (221 peptides) (Grifoni et al., 2020b) (Table S3 and S4) . Here to further optimize detection of non-Spike reactivity, we designed epitope pools based on Experimentally (E) defined epitopes, from the non-spike sequences of the SARS-CoV-2 proteome. The CD4RE and CD8RE megapools (MP) consisted of 284 and 621 peptides respectively (Table S3 and S4). A pool of epitopes derived from an unrelated ubiquitous pathogen (EBV) (Carrasco Pro et al., 2015) was used as a specificity control (Table S3) .

T cell reactivity was assessed by the Activation Induced Marker (AIM) assays (da Silva Antunes et al., 2021) and data represented as either absolute magnitude or stimulation index (SI). As shown in Figure 1A SARS-CoV-2-specific CD4+ T cell responses were detected in all convalescent and/or vaccinated individuals and approximately 50% of non-infected, non-J o u r n a l P r e -p r o o f vaccinated individuals. Similar results were observed when responses were plotted as SI ( Figure 1B) . Unexposed subjects were associated with significantly lower reactivity as compared to all the other groups (p-values ranging 1.3e-7 to 1.0e-15) and convalescent and vaccinated (I+V+) subjects exhibited higher responses than convalescent (I+V-) subjects (p=0.02 and p=0.04 for absolute magnitude and SI, respectively) or vaccinated (I-V+) subjects (p=0.01 and p=0.02 for absolute magnitude and SI, respectively) ( Figure 1A,B) . Importantly, CD4RE responses were able to differentiate convalescent subjects (I+V-or I+V+) from unexposed and vaccinated (I-V+) subjects with p-values ranging 5.6e-8 to 5.7e-12 and vaccinated (I-V+) from infected and vaccinated (I+V+) subjects (p=1.4e-11 and p=1.1e-11 for absolute magnitude and SI, respectively) ( Figure 1A ,B). As expected, no statistically significant difference in EBV reactivity was observed when the four groups were compared ( Figure 1A,B) .

SARS-CoV-2 specific CD8+ T cell responses were also broadly detected among all the cohorts studied. CD8+ T cell responses were detected in 90-100% of the convalescent and/or vaccinated individuals and approximately in 1/4 of non-infected, non-vaccinated individuals ( Figure 1C ).

Similar responses were observed when plotted as SI ( Figure 1D ). As observed for CD4+ T cell responses, CD8+ T cell responses of unexposed subjects (I-V-) were discriminated from all the other groups (p-values ranging 2.6e-5 to 8.8e-13) and I+V+ infected/vaccinated subjects exhibited higher responses than I+V-convalescent (p=0.03 and p=0.16 for absolute magnitude and SI respectively). Identical results were observed parsing spike-specific responses with CD8RE able to differentiate convalescent (I+V-) from unexposed and vaccinated (I-V+) subjects (p-values ranging 0.02 to 5.9e-6) and vaccinated from infected/vaccinated (I+V+) J o u r n a l P r e -p r o o f subjects (p=0.04 and p=0.02 for absolute magnitude and SI, respectively) ( Figure 1C,D) . When the four groups were compared, no statistically significant difference in EBV reactivity was observed ( Figure 1C,D) .

In parallel, an IFN FluoroSpot assay was also employed to evaluate the CD4+ and CD8+ T cell responses using a threshold of 20 IFN spot forming cells (SFC) per million PBMC. Responses were detected in many infected or vaccinated individuals, and similar results were observed for Spike, CD4RE or CD8RE when considering both the absolute magnitude or stimulation index, albeit with predictably lower sensitivity and specificity than AIM ( Figure   S1A ,B).

Results from both AIM and IFN FluoroSpot assay demonstrated that the newly developed CD4RE pool had both improved sensitivity and specificity, compared to the previously used CD4R pool of predicted epitopes ( Figure S1C,D) . In more detail, higher positive CD4+ T cell responses in I+V-(28/30 (93%) vs 26/30 (87%), p = 2.0e-4) and I+V+ (28/30 (93%) vs 23/30 (77%), p = 5.0e-6), and lower non-specific response in I-V-(8/30 (27%) vs 14/30 (47%), p = 0.037) and I-V+ (2/30 (7%) vs 4/30 (13%), p = 0.031) were detected using CD4RE when compared to CD4R in the AIM assay ( Figure S1C) . Similar results were shown by IFN FluoroSpot, assay albeit with lower sensitivity compared to AIM ( Figure S1D ). These results demonstrate that the use of experimentally defined, as opposed to predicted epitopes provides higher signal in SARS-CoV-2 exposed subjects, while lowering responses from non-exposed subjects. The fact that experimentally defined epitopes yield better results is consistent with mass spectrometry studies showing the divergence of predicted from HLA-eluted SARS-CoV-2 immunopeptidome (Knierman et al., 2020; Pan et al., 2021; Weingarten-Gabbay et al., 2021) .

We reasoned that unexposed (I-V-) subjects would be unreactive to experimentally defined SARS-CoV-2 peptide pools, while uninfected vaccinated (I-V+) subjects should react only to the S pool. We further reasoned that infected (I+V-) subjects should recognize both S and CD4RE, but infected and vaccinated (I+V+) subjects would have a higher relative S reactivity than infected only (I+V-), as is often the case with hybrid immunity (Crotty, 2021) , due to exposure to S twice, once during infection and the other during vaccination. Figure 2A , spike-and CD4RE-specific CD4+ T cell responses derived from the AIM assay were arranged in a two-dimensional plot. Each dot represents a single subject from a total of 120 donors (30 for each of the 4 groups, Table S1 ). Optimal cutoffs were established to discriminate the four groups and the positive predictive value (PPV), negative predictive value (NPV), sensitivity and specificity were calculated for each individual group.

Subjects with spike responses lower than 0.025% were classified predictively as unexposed (I-V-) (Figure 2A) . 29 out of 29 subjects with responses matching this criterion were correctly classified (100% of PPV), while nearly all the actual I-V-subjects (29 out of 30) were found to be associated with responses below the threshold, corresponding to a sensitivity of 96.7 % (Figure 2A , grey circles). Subjects with spike responses greater than 0.025% and CD4RE responses lower than 0.015% were classified predictively as I-V+. Twenty-eight out of 30 subjects with responses matching this threshold were correctly classified (93.3% of PPV), and 28 out of the 30 I-V+ subjects detected within this threshold (93.3% of sensitivity) ( Figure   2A , blue circles).

Lastly, subjects with spike and CD4RE responses above 0.025% and 0.015% respectively, and above or below a diagonal line (log(y)=0.454log(x)-0.18) were classified as I+V+ or I+V-respectively. 24 out of 27 subjects with responses matching the lower J o u r n a l P r e -p r o o f compartment (I+V-) were correctly classified (88.9% of PPV) while 24 out of the 30 I+Vsubjects were found to be associated with this threshold (80 % of sensitivity) (Figure 2A , red circles). Conversely, the majority of subjects (26 out of 34) with responses matching the upper compartment (I+V+) were correctly classified (76.5% of PPV), while 26 out of the 30 I+V+ subjects studied were found to be associated with this threshold, corresponding to a sensitivity of 86.7 % (Figure 2A , yellow circles). Further statistical examinations to assess the robustness of the classification scheme as a potential diagnostic test were performed, specifically assessments of specificity and negative predictive value (PPV). High specificity and NPV were observed for each individual group with a range of 91.1-100% and 93.5-98.9% respectively ( Figure 2A ). In summary, good PPV, NPV, sensitivity and specificity values were observed across all the groups with an overall classification accuracy of 89.2%.

To confirm the accuracy of this classification scheme, we assessed CD4+ T cell responses in an independent validation cohort of 96 donors (20 for I-V-, I+V-, I+V+, and 36 for I-V+; Table   S2 ). As shown in Figure 2B , using the same cutoffs as described above for spike and CD4RE responses, similar PPV, NPV, sensitivity and specificity to the experimental cohort was observed across all the groups in the validation cohort with an overall classification accuracy of 88.5%. To further validate the robustness of this classification scheme, the same data ( Figure   2 ) was plotted as a function of the stimulation index ( Figure S2A,B) . Strikingly, these results paralleled the observations using the absolute magnitude, with a similar overall classification accuracy (86.7% and 85.4% for the exploratory and validation cohorts, respectively).

Applying the same classification scheme using either absolute magnitude or stimulation index for IFNresponses yielded an overall classification accuracy of 72.5% and 60.0% respectively ( Figure S2C,D) . A lower accuracy was observed when CD8+ T cell responses J o u r n a l P r e -p r o o f from AIM assay were analyzed, as compared to CD4+ T cell responses (data not shown).

Overall, these results demonstrate the feasibility of an integrated classification scheme in assessing CD4+ T cell responses as a clinical immunodiagnostic tool. Importantly, it also displays the potential to discriminate previously undetected infection, including in vaccinated individuals.

To gain further insights into the applicability of the classification scheme, we sought to further test and validate this tool across vaccine platforms, and longer timepoints post-symptom onset (PSO) or post-vaccination. First, we looked at the response classification as a function of whether vaccinated subjects received BNT162b2 or mRNA-1273 vaccines. As shown in Figure   3A the overall classification accuracy when using the different mRNA vaccines was of 89.7%. Specifically, both vaccines showed similar magnitude for both total CD4+ and CD8+ T cell responses in the I-V+ or I+V+ groups ( Figure S3A and B). The accuracy of the classification scheme for the different types of vaccines in the combined I-V+ or I+V+ groups was almost identical (88.5% and 90.9% for the mRNA-1273 and BNT162b2 vaccines, respectively) ( Table   2 ).

Next, we looked at the response classification as a function of the length of time PSO.

The overall classification accuracy was of 84.0% ( Figure 3B ). No differences were observed in the magnitude of both total CD4+ and CD8+ T cell responses between early (180 days) and late (>180 days) timepoints from PSO in either the I+V-or the I+V+ groups (Figure S3C and D). CD4+ T cell reactivity associated with different time from PSO was also plotted as a continuous variable ( Figure S4A ). The accuracy of the classification scheme when considering J o u r n a l P r e -p r o o f the different PSO timepoints was 82.0% and 81.8% in the I+V-group and 90.0% and 85.0% in the I+V+ group for the early and late timepoints, respectively ( Figure 3B ).

We also looked at the responses as a function of the length of time from the 2 nd dose of vaccination. The overall classification accuracy was of 89.7% ( Figure 3C ). No differences were observed in the magnitude of both total CD4+ or CD8+ T cell responses between early (30 days) or late (>30 days) timepoints from the last dose of vaccination in either the I-V+ or the I+V+ groups ( Figure S3E and F). CD4+ T cell reactivity associated with different post vaccination dates was also plotted as a continuous variable ( Figure S4B ). The accuracy of the classification scheme when considering the different vaccine timepoints was 93.5% and 90.0%

in the I-V+ group and 86.4% and 85.7% in the I+V+ group for the early and late timepoints respectively ( Figure 3C) .

Lastly, as an alternative to the T cell classification scheme, we classified subjects based on spike RBD and nucleocapsid (N) antibody responses. An overall classification accuracy of 69% ( Figure S5A ) was observed when previously described standard clinical cutoffs were employed Grifoni et al., 2020b; Tarke et al., 2021a) . The attempt to classify infected individuals at late PSO timepoints resulted in even lower accuracies (Figure S5B) , consistent with reports that N positivity is relatively short lived Ibarrondo et al., 2020; Ortega et al., 2021) . We next examined the possibility that this low classification accuracy might be reflective of suboptimal thresholds. By setting more stringent cutoffs based on the optimal classification of the exploratory cohort, we achieved an overall classification accuracy of 84.2% ( Figure S5C) . However, when the same classification scheme was applied to the validation cohort, the overall accuracy decreased to 52.1% (Figure S5D) , indicating that the previous value was likely a result of data overfitting. Overall, the use of antibody responses failed to yield a useful classification scheme, unlike the classification scheme using CD4+ T J o u r n a l P r e -p r o o f cell responses, which proved to be a robust tool that can accurately classify subjects regardless of the days post-infection/post-vaccination or vaccine administered ( Table 2 ).

Breakthrough infections are defined as cases of previously COVID-19 vaccinated individuals associated with positive SARS-CoV-2 PCR tests (Bergwerk et al., 2021; Kustin et al., 2021; Mizrahi et al., 2021) . Studies of antibody or T cell responses associated with breakthrough infection are scarce (Collier et al., 2021; Rovida et al., 2021) . Breakthrough infection might be associated with increased immune responses as a result of the re-exposure (hybrid immunity) (Collier et al., 2021) . In other cases, subjects experiencing breakthrough infections might be associated with general weaker immune responsiveness or decrease of vaccine effectiveness (Klompas, 2021; Mizrahi et al., 2021) .

Here, we assessed spike and CD4RE T cell responses in a group (n=23) of breakthrough infected individuals (V+I+). Responses were compared to the vaccinated (I-V+), infected (I+V-) or infected and then vaccinated (I+V+) groups matching the V+I+ intervals of vaccination and infection (55-271 and 18-93 days, respectively). As shown in Figure 4A , CD4+ T cell responses from V+I+ subjects were associated with significant higher levels compared to I+V-(p=0.04) and I-V+ (p=2.3e-3) subjects and similar magnitude as the I+V+ subjects. CD8+ T cell responses had comparable levels across all the groups ( Figure 4B ). Similar to CD4+ T cell responses, spike RBD IgG titers from V+I+ subjects were equivalent to I+V+ subjects and significantly higher than I+V-(p=4.2e-7) and I-V+ (p=4.0e-15) subjects ( Figure 4C) . Thus, at the population level breakthrough infections are associated with CD4+ T cell and spike IgG responses that resemble hybrid immunity.

At the level of the T cell response classification scheme, individuals who had COVID-19 were effectively segregated from non-infected groups (unexposed and vaccinated). (Figure 4D) . We further expected that the V+I+ breakthrough infections would be classified in the same manner of I+V+ hybrid immunity samples. Approximately two thirds (15/23 subjects) were identified by the same thresholds associated with responses from the I+V+ group ("High responders"), while the remaining third were classified similarly to I+V-subjects ("Low responders"). No obvious difference in terms of age, PSO, PVD, disease severity or length of infection from vaccination was detected between these donors and the high responders sub-group of 15 donors ( Figure S6 and Table 1 ).

In summary, while T cell responses following breakthrough infections (V+I+) are effectively segregated from the responses of uninfected donors (vaccinated or not) and follow the same pattern of responses of individuals vaccinated following natural infection (I+V+) in the majority of the cases, the classification scheme revealed heterogeneity in the CD4+ T cell responses of breakthrough donors.

Finally, we summarized the overall accuracy of the classification scheme across the five cohorts used in this study including breakthrough infections. For this purpose, we clustered individuals that had been infected and vaccinated, irrespectively of the event that occurred first, into a single group, i.e. I+V/V+I+ (Figure 5) . When the 239 subjects with distinct COVID-19 status of infection and/or vaccination were combined, the classification scheme achieved a high overall accuracy, either as function of absolute magnitude (86.6%) or SI (82.4%). Also, high specificity and NPV were retained for each individual group with a range of 92. 2-98.4% and 88.6-98.4% respectively. These results further illustrate the highly predictive power of this classification scheme and its broad clinical applicability.

There is a need to understand the roles of SARS-CoV-2 T cell responses as potential correlates of disease outcome, and/or correlates of vaccine protection from infection or severe disease.

Herein, we present the results of T cell quantitation based on the determination of relative activity directed against spike and the rest of the genome, by the use of optimized pools of experimentally defined epitopes (CD4RE and CD8RE). We report successful classification of subjects with different COVID-19 vaccination or natural infection history in the 85-90% range of accuracy. We further show that the strategy is applicable to characterizing immune responses in a group of infected vaccinees (i.e. breakthrough infections).

Although previous reports studied responses to SARS-CoV-2 in either unexposed, COVID-19 infected or vaccinated individuals (da Silva Antunes et al., 2021; Dan et al., 2021; Goel et al., 2021; Grifoni et al., 2020b; Le Bert et al., 2020; Mateus et al., 2021) , this is the first demonstration, to the best of our knowledge, that a simple assay strategy can classify T-cell responses measured simultaneously in five different groups of known COVID-19 status of infection, and/or vaccination. The improved sensitivity and specificity resulted from the concept of considering the relative magnitude of responses against the spike and "rest of the genome" components, which overcomes issues related to the fact that magnitude of responses may wane over time, and also by the inclusion of experimentally defined epitopes, which we show are associated with improved signal and selectivity as compared to previously utilized predicted epitopes.

We suggest that the combined use of overlapping spike and CD4RE pools can be used to detect differential and relative reactivity to different SARS-CoV-2 antigens and therefore classify individuals based on SARS-CoV-2 infection and COVID-19 vaccine status. More importantly, this approach allows to identify bone fide exposition to SARS-CoV-2 even in individuals that have been vaccinated and thus effectively distinguishing COVID-19 vaccine J o u r n a l P r e -p r o o f and infection history. This is of importance, as current COVID-19 diagnostic practices rely heavily on subjectively reported history, clinical records and lab modalities with imperfect performance, leading to limited reliability. For example, in longitudinal vaccination studies it will be important to monitor whether subjects enrolled in the studies might have been associated with asymptomatic infection (Kustin et al., 2021; Mizrahi et al., 2021; Pouwels et al., 2021) , or even associated with abortive seronegative infections (Swadling et al., 2021) . Also, diagnosis certifications (e.g., "health passes") and to further characterize individuals that might have been exposed but have not tested positive or had false negative results for COVID-19 using a molecular or antigen diagnostic test.

Our study builds on the well-known fact that infected individuals mount a T cell response against multiple SARS-CoV-2 antigens and that individuals vaccinated with mRNA vaccines are mounting only a T cell response to Spike. A detailed classification of T cells response in different categories of vaccinated/ infected individuals have not been described and compared as in the current study. Indeed, the use of our developed pools, spanning all the antigens from SARS-CoV-2, allowed for detection of SARS-CoV-2 responses with increased J o u r n a l P r e -p r o o f sensitivity and specificity compared to other studies performing T cell assays using only spike or other SARS-CoV-2 antigens (Krishna, Preprint; Kruse et al., 2021; Martínez-Gallo, 2021; Murugesan et al., 2020; Tan et al., 2021; Zelba et al., 2021) .

We also show that similar results were observed when relative versus absolute determinations were employed to measure T cell responses (i.e. using stimulation index or absolute magnitude), which allows for a more generalized use of the classification tool in different flow-cytometer platforms. The robustness of the T cell-based classification scheme was validated in an independent cohort exhibiting identical performances and was applicable to different types of mRNA vaccines, even when considering extended periods of time elapsed from infection and/or vaccination. T cell responses might differ according to the vaccine platform. Also, despite the wide range of time interval following 2nd vaccine dose between groups, and even when considering extended periods of time elapsed from infection and/or vaccination, the classification scheme performance remained unchanged.

The strength of the approach is further demonstrated by the fact that T cell responses act as a better classifier than antibody responses, consistent with the notion that antibody responses to N protein are short lived Ibarrondo et al., 2020; Ortega et al., 2021) . Also, while applicable to data generated by FluoroSpot cytokine assays, despite the lower intrinsic sensitivity of this assay, we anticipate that this assay strategy will be broadly applicable to other readouts, such as ICS Mateus et al., 2021) , and whole blood in an interferon-gamma release assay (IGRA) (Murugesan et al., 2020; Petrone et al., 2021; Tan et al., 2021) .

T cell responses from breakthrough infections were also evaluated, and high levels of CD4+ and CD8+ T cell reactivity was observed. Elevated T cell responsiveness was paralleled by high levels of spike RBD IgG. Interestingly, these responses were of similar magnitude as responses from a group of individuals infected and then vaccinated (I+V+ in our study), whose J o u r n a l P r e -p r o o f features are commonly associated with hybrid immunity (Crotty, 2021) . Notably, breakthrough infections were also associated with higher CD4+ T cell and spike RBD IgG responses compared to infected only or vaccinated only subjects. These results suggest that T and B cell reactivity associated with breakthrough infections is increased as a result of re-exposure.

However, the classification tool system, also revealed significant heterogeneity in responses in some subjects, possibly linking some breakthrough infections to lower adaptive responses. A more detailed analysis at T cell epitope level could better define whether differences in T cell responses occur in this category of "low responders" compared to vaccinated and infected only individuals.

Although our findings were validated in several different cohorts, further validation in larger and more ethnically diverse populations, and with different HLA backgrounds is warranted. Further studies using cohorts of asymptomatic subjects associated with PCR positive test and lack of clinical symptoms are required to address the performance of a T cell-based immunodiagnostic scheme in the identification of asymptomatic infections. Further additional studies will also have to address the performance of the classification scheme in assessing responses associated with VOCs such as Delta and Omicron, and responses observed after 3 vaccine administrations. The study of additional cohorts representative of different vaccine platforms (e.g., Ad26.COV2.S, ChAdOx1 nCoV-19, CoronaVac) will be important to generalize these findings. Hispanic/Latino (n (%)) 8 (16%) 7 (14%) 11 (17%) 6 (12%) 5 (22%) Asian (n (%)) 8 (16%) 4 (8%) 16 (24%) 3 (6%) 3 (13%)

Black (n (%)) 2 (4%) 2 (4%) 2 (3%) 2 (4%) 0 (0%) Range -20-308 -57-508 18-93 

Asymptomatic - 0 (0%) - 0 (0%) 0 (0%) Mild - 44 (88%) - 45 (90%) 23 (100%) Moderate - 3 (6%) - 3 (6%) 0 (0%) Severe - 3 (6%) - 2 (4%) 0(

Epitope pools used in this study will be made available to the scientific community upon request, and following execution of a material transfer agreement (MTA), by contacting A.S.

(alex@lji.org) and R.d.S.A (rantunes@lji.org). Likewise, biomaterials archived from this study may be shared for further research with MTA.

 The datasets generated and analyzed in this study will be shared by the lead contact  Any additional information required to reanalyze the data reported in this work paper is available from the Lead Contact upon request

The Adults of all races, ethnicities, ages, and genders were eligible to participate, but the association of gender on the results of the study was not explicitly measured. Study exclusion criteria included lack of willingness to participate, lack of ability to provide informed consent, or a medical contraindication to blood donation (e.g., severe anemia). In all cases, PBMCs were isolated from whole blood by density gradient centrifugation according to manufacturer instructions (Ficoll-Hypaque, Amersham Biosciences, Uppsala, Sweden). Cells were cryopreserved in liquid nitrogen suspended in FBS containing 10% (vol/vol) DMSO (Sigma-Aldrich). Plasma was obtained by centrifugation (400g for 15 minutes at 4°C) of whole blood and collection of the upper layer, prior to PBMC isolation and cryopreserved at -80°C.

To study T cell responses against SARS-CoV-2, we used a megapool (MP) of 15-mer peptides overlapping by 10 spanning the entire spike protein sequence (253 peptides; Table S3 ) as previously described (Grifoni et al., 2020b) . For the rest of the SARS-CoV-2 proteome, and in order to design epitope pools with increased HLA coverage and broadly recognized by demographically and geographically diverse populations, experimental defined epitopes from non-spike (R) region of SARS-CoV-2 were selected based on our recent meta-analysis . Briefly, peptides were synthetized and pooled to include both dominant (recognized in 3 or more donors/studies) and subdominant epitopes. To improve specificity, overly short or long ligands which could cause "false positive" signals (Paul et al., 2018) , were J o u r n a l P r e -p r o o f excluded and only peptides of sizes ranging 15-20 and 9-10 amino acids, respectively in CD4RE and CD8RE pools were included, resulting in the generation of CD4RE and CD8RE

MPs with 284 and 621 peptides, respectively (Table S3) . Epitopes were further classified in dominant and subdominant based on the frequency of individual responses as previously described . In addition, detailed information of the MPs composition with peptide sequences, length, ORFs of origin, and HLA coverages are specified in Table S4 .

Alternatively, a MP for the remainder genome consisting of dominant HLA class II predicted CD4+ T-cell epitopes (221 peptides), as previously described (Grifoni et al., 2020b) was also used as control (Table S3 ). In addition, an EBV pool of previously reported experimental class I and class II epitopes (Carrasco Pro et al., 2015) with 301 peptides was used as positive control.

All peptides were synthesized by TC peptide lab (San Diego, CA), pooled and resuspended at a final concentration of 1 mg/mL in DMSO.

The SARS-CoV-2 ELISAs have been described in detail previously . Briefly, 96-well half-area plates (ThermoFisher 3690) were coated with 1 ug/mL of antigen and incubated at 4 o C overnight. Antigens included recombinant SARS-CoV-2 RBD protein obtained from the Saphire laboratory at LJI or recombinant nucleocapsid protein (GenScript Z03488). The next day, plates were blocked with 3% milk in phosphate-buffered saline (PBS) containing 0.05% Tween-20 for 1.5 hours at room temperature. Plasma was heat inactivated at 56°C for 30 to 60 min. Plasma was diluted in 1% milk containing 0.05% Tween-20 in PBS starting at a 1:3 dilution followed by serial dilutions by three and incubated for 1.5 hours at room temperature. Plates were washed five times with 0.05% PBS-Tween-20. Secondary antibodies were diluted in 1% milk containing 0.05% Tween-20 in PBS. Anti-human IgG peroxidase antibody produced in goat (Sigma A6029) was used at a 1:5,000 dilution.

Subsequently, plates were read on Spectramax Plate Reader at 450 nm, and data analysis was performed using SoftMax Pro. End-point titers were plotted for each sample, using backgroundsubtracted data. Negative and positive controls were used to standardize each assay and normalize across experiments. Limit of detection (LOD) was defined as 1:3 of IgG. Spike RBD IgG or nucleocapsid IgG thresholds of positivity (TP) for SARS-CoV-2 infected or COVID-19 vaccinated individuals were established based on uninfected and unvaccinated subjects (I-V-).

The AIM assay was performed as previously described (Mateus et al., 2020 Mateus et al., 2020) . The gating strategy utilized is shown in Figure S7 , as well as reactive CD4+ and CD8+ T cell responses to SARS-CoV-2, EBV and PHA positive control from a representative donor.

The FluoroSpot assay was performed as previously described (Tarke et al., 2021a) . PBMCs derived from 80 subjects from 4 clinical cohorts (20 each for I-V-, I+V-, I-V+, and I+V+ cohorts) were stimulated in triplicate at a single density of 2x10 5 cells/well. The cells were stimulated with the different MPs analyzed (1ug/mL), PHA (10mg/mL), and DMSO (0.1%) in 96-well plates previously coated with anti-cytokine antibodies for IFN, (mAbs 1-D1K;

Mabtech, Stockholm, Sweden) at a concentration of 10ug/mL. After 20-24 hours of incubation at 37 o C, 5% CO2, cells were discarded and FluoroSpot plates were washed and further incubated for 2 hours with cytokine antibodies (mAbs 7-B6-1-BAM; Mabtech, Stockholm, Sweden). Subsequently, plates were washed again with PBS/0.05% Tween20 and incubated for 1 hour with fluorophore-conjugated antibodies (Anti-BAM-490). Computer-assisted image analysis was performed by counting fluorescent spots using an AID iSPOT FluoroSpot reader (AIS-diagnostika, Germany). Each megapool was considered positive compared to the background based on the following three criteria: 20 or more IFN spot forming cells (SFC) per 10 6 PBMC after background subtraction (Threshold defined as 2 times standard deviation of background signals), a stimulation index (SI) greater than 2, and statistically different from the background (p < 0.05) in either a Poisson or t test as previously described (Oseroff et al., 2005) . 

Health under IRB approved protocols (UCSD; 200236X), or under IRB approval (LJI; VD-214) at the La Jolla Institute for Immunology. All donors were able to provide informed consent, or had a legal guardian or representative able to do so. Each participant provided informed consent and was assigned a study identification number with clinical information recorded.

J o u r n a l P r e -p r o o f

Supplemental Information can be found in attached files: Supplemental figures and tables.pdf and Table S4_Spike_CD4R (E) and CD8R(E) megapools sequences.xlsx. Figure S1 . SARS-CoV-2-specific CD4+ and CD8+ T cell IFNresponses and comparison between CD4RE and CD4R MP performances, Related to Figure 1 . 

Covid-19 Breakthrough Infections in Vaccinated Health Care Workers

SARS-CoV-2-specific T cells in infection and vaccination

Automatic Generation of Validated Specific Epitope Sets

Longitudinal analysis shows durable and broad immune memory after SARS-CoV-2 infection with persisting antibody responses and memory B and T cells

Immune Responses in Fully Vaccinated Individuals Following Breakthrough Infection with the SARS-CoV-2 Delta Variant in

Hybrid immunity

Differential T-Cell Reactivity to Endemic Coronaviruses and SARS-CoV-2 in Community and Health Care Workers

Immunological memory to SARS-CoV-2 assessed for up to 8 months after infection

Correlates of protection against symptomatic and asymptomatic SARS-CoV-2 infection

mRNA vaccines induce durable immune memory to SARS-CoV-2 and variants of concern

-2 human T cell epitopes: Adaptive immune response against COVID-19

A Sequence Homology and Bioinformatic Approach Can Predict Candidate Targets for Immune Responses to SARS-CoV-2

Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with COVID-19 Disease and Unexposed Individuals

Rapid Decay of Anti-SARS-CoV-2 Antibodies in Persons with Mild Covid-19

Neutralizing antibody levels are highly predictive of immune protection from symptomatic SARS-CoV-2 infection

Understanding Breakthrough Infections Following mRNA SARS-CoV-2 Vaccination

The Human Leukocyte Antigen Class II Immunopeptidome of the SARS-CoV

A government-led effort to identify correlates of protection for COVID-19 vaccines

A correlate of protection for SARS-CoV-2 vaccines is urgently needed

Correlates of protection from SARS-CoV-2 infection

Retrospective diagnosis of SARS-CoV-2 infection in patients with Long COVID by measuring specific T cell mediated IL-2 release

Performance of the T-SPOT(R).COVID test for detecting SARS-CoV-2-responsive T cells

Evidence for increased breakthrough rates of SARS-CoV-2 variants of concern in BNT162b2-mRNA-vaccinated individuals

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

A flow cytometry-based proliferation assay for clinical evaluation of T-cell memory against SARS-CoV-2

Impact of circulating SARS-CoV-2 variants on mRNA vaccine-induced immunity

Commercialized kits to assess T-cell responses against SARS-CoV-2 S peptides. A pilot study in health care workers

Change in Symptoms and Immune Response in People with Post-Acute Sequelae of SARS-Cov-2 Infection (PASC) After SARS-Cov-2 Vaccination

Preprint-b). Breakthrough Symptomatic COVID-19 Infections Leading to Long Covid: Report from Long Covid Facebook Group Poll

Low-dose mRNA-1273 COVID-19 vaccine generates durable memory enhanced by cross-reactive T cells

Correlation of SARS-CoV-2-breakthrough infections to time-fromvaccine

Long term accuracy of SARS-CoV-2 Interferon-gamma release assays and its apllication in household investigation

Interferon-gamma release assay for accurate detection of SARS-CoV-2 T cell response

T cell immunity to SARS-CoV-2

T cell assays differentiate clinical and subclinical SARS-CoV-2 infections from cross-reactive antiviral responses

Seven-month kinetics of SARS-CoV-2 antibodies and role of pre-existing antibodies to human coronaviruses

HLA class I-restricted responses to vaccinia recognize a broad array of proteins mainly involved in virulence and viral gene regulation

Rapid induction of antigenspecific CD4(+) T cells is associated with coordinated humoral and cellular immunity to SARS-CoV-2 mRNA vaccination

Mass spectrometric identification of immunogenic SARS-CoV-2 epitopes and cognate TCRs

Determination of a Predictive Cleavage Motif for Eluted Major Histocompatibility Complex Class II Ligands

A whole blood test to measure SARS-CoV-2-specific response in COVID-19 patients

Effect of Delta variant on viral burden and vaccine effectiveness against new SARS-CoV-2 infections in the UK

SARS-CoV-2 vaccine breakthrough infections with the alpha variant are asymptomatic or mildly symptomatic among health care workers

Antigen-Specific Adaptive Immunity to SARS-CoV-2 in Acute COVID-19 and Associations with Age and Disease Severity

Robust T Cell Immunity in Convalescent Individuals with Asymptomatic or Mild COVID-19

Pre-existing immunity to SARS-CoV-2: the knowns and unknowns

Adaptive immunity to SARS-CoV-2 and COVID-19

Pre-existing polymerase-specific T cells expand in abortive seronegative SARS-CoV-2

Rapid measurement of SARS-CoV-2 spike T cells in whole blood from vaccinated and naturally infected individuals

Six-month sequelae of post-vaccination SARS-CoV-2 infection: a retrospective cohort study of 10

Comprehensive analysis of T cell immunodominance and immunoprevalence of SARS-CoV-2 epitopes in COVID-19 cases

Pre-proof). Commercial interferpon-gamma release assay to assesss the immune response to first and second doses of mRNA vaccine

Considerations for diagnostic COVID-19 tests

Profiling SARS-CoV-2 HLA-I peptidome reveals T cell epitopes from out-of-frame ORFs

A Highly Specific Assay for the Detection of SARS-CoV-2-Reactive CD4(+) and CD8(+) T Cells in COVID-19 Patients

Highlights are 3-4 bullet points of no more than 85 characters in length, including spaces, and they summarize the core results of the paper in order to allow readers to quickly gain an understanding of the main take-home messages. A T cell-based assay allows discrimination of SARS-CoV-2 infection and vaccination  The classification scheme has high sensitivity, specificity and broad applicability  The use of SARS-CoV-2 epitope pools yield higher accuracy than serological readouts  Breakthrough infections can be effectively segregated from vaccine responses

No longer than 50 words describing the context and significance of the findings for the broader journal readership.The blurb must be written in the third person and refer to "First Author et al."Yu et al. developed an assay using epitope pools to effectively discriminate T cell responses of subjects based on their SARS-CoV-2 infection and COVID-19 vaccination history. This T cell-based classification scheme could potentially be used as an immunodiagnostic tool for longitudinal monitoring of vaccination responses and establishing correlates of protection.