Immunopeptidome profiling of human coronavirus OC43-infected cells identifies CD4 T-cell epitopes specific to seasonal coronaviruses or cross-reactive with SARS-CoV-2

Seasonal “common-cold” human coronaviruses are widely spread throughout the world and are mainly associated with mild upper respiratory tract infections. The emergence of highly pathogenic coronaviruses MERS-CoV, SARS-CoV, and most recently SARS-CoV-2 has prompted increased attention to coronavirus biology and immunopathology, but the T-cell response to seasonal coronaviruses remains largely uncharacterized. Here we report the repertoire of viral peptides that are naturally processed and presented upon infection of a model cell line with seasonal coronavirus OC43. We identified MHC-bound peptides derived from each of the viral structural proteins (spike, nucleoprotein, hemagglutinin-esterase, membrane, and envelope) as well as non-structural proteins nsp3, nsp5, nsp6, and nsp12. Eighty MHC-II bound peptides corresponding to 14 distinct OC43-derived epitopes were identified, including many at very high abundance within the overall MHC-II peptidome. Fewer and less abundant MHC-I bound OC43-derived peptides were observed, possibly due to MHC-I downregulation induced by OC43 infection. The MHC-II peptides elicited low-abundance recall T-cell responses in most donors tested. In vitro assays confirmed that the peptides were recognized by CD4+ T cells and identified the presenting HLA alleles. T-cell responses cross-reactive between OC43, SARS-CoV-2, and the other seasonal coronaviruses were confirmed in samples of peripheral blood and peptide-expanded T-cell lines. Among the validated epitopes, spike protein S903-917 presented by DPA1*01:03/DPB1*04:01 and S1085-1099 presented by DRB1*15:01 shared substantial homology to other human coronaviruses, including SARS-CoV-2, and were targeted by cross-reactive CD4 T cells. Nucleoprotein N54-68 and hemagglutinin-esterase HE128-142 presented by DRB1*15:01 and HE259-273 presented by DPA1*01:03/DPB1*04:01 are immunodominant epitopes with low coronavirus homology that are not cross-reactive with SARS-CoV-2. Overall, the set of naturally processed and presented OC43 epitopes comprise both OC43-specific and human coronavirus cross-reactive epitopes, which can be used to follow CD4 T-cell cross-reactivity after infection or vaccination, and to guide selection of epitopes for inclusion in pan-coronavirus vaccines.


Introduction
Coronaviruses are single-stranded RNA viruses of the genus Nidovirales, family Coronaviridae that infect vertebrates. Seven species in the Orthocoronavirinae sub-family are known to infect humans, with a wide range of pathogenicity [1]. Human coronavirus (HCoV) 229E and NL63 in the alphacoronavirus genus, and OC43 and HKU1 in the betacoronavirus genus, are associated with mild upper-respiratory-tract infections and common colds. In contrast, SARS-CoV, MERS-CoV, and SARS-CoV-2, all in the betacoronavirus genus, are associated with a severe respiratory syndrome [2]. Common-cold-associated seasonal HCoVs are widespread and infect humans in seasonal waves [3][4][5]. OC43 is closely related to bovine coronavirus (BCoV) and was initially isolated in 1967 from individuals with upper respiratory tract infections [6]. Among the seasonal human coronaviruses, OC43 is believed to have emerged most recently. Molecular clock analysis of the spike gene sequences suggests a relatively recent zoonotic transmission event and dates their most recent common ancestor between 1890 to 1923 [7][8][9]. This led to the proposal that the 1898 pandemic ("Russian Flu"), which caused a worldwide multi-wave outbreak killing preferentially older individuals similar to COVID-19, may have been the result of the emergence of OC43 [10]. The OC43 reference genome (ATCC-VR-759) spans 30,738 kbp, encoding 9 ORFs which are translated into 23 proteins [11,12].
Before the emergence of the pandemic coronavirus SARS-CoV-2, few studies characterized the immune response to the seasonal HCoVs, which account for~10-30% of common colds [13,14]. Studies of T-cell responses to HCoVs and the identification of epitopes driving them are scarce. Before the SARS-CoV-2 pandemic, Nilges et al. reported that T cells to an MHC-I epitope derived from papillomavirus 16 peptide were cross-reactive with a peptide from OC43 NS2 protein [15]. Later, Boucher et al. studied T-cell responses to OC43 and 229E viral antigens and to multiple sclerosis (MS) autoantigens in MS patients. Virus-specific T-cell clones were isolated, including 34 clones responding to OC43, as well as 10 T-cell clones cross-reactive with HCoV and MS autoantigens, but the specific viral epitopes were not identified [16].
More recently, after the rise of SARS-CoV-2, Woldemeskel et al. [17] reported T-cell responses to pools of spike, nucleoprotein, and membrane proteins of the four seasonal coronaviruses. Peptide responses to the spike protein of NL63 were deconvoluted resulting in the identification of 22 target peptides, of which 3 are SARS-CoV-2 cross-reactive and the remaining 19 are HCoV-specific novel epitopes.
Studies of T-cell responses to SARS-CoV and SARS-CoV-2 have reported that responding T-cell populations are present in blood samples collected before the emergence of these viruses [18,19]. This led to the suggestion that pre-existing immunity, potentially elicited by a previous infection(s) with seasonal HCoVs, could be responsible for these responses, and prompted a search for the cross-reactive epitopes responsible. In fact, most OC43 epitopes reported in the Immune Epitope Database [20] were identified in the context of HCoV/ SARS-CoV-2 cross-reactivity studies. Schmidt et al. used a highly conserved peptide derived from the SARS-CoV-2 nucleoprotein to identify cross-reactive MHC-I responses and found that homologous HCoV peptides, including one from OC43, also were recognized [21]. Mateus et al. used overlapping SARS-CoV-2 peptides to screen for cross-reactive responses in unexposed donors and identified six MHC-II epitopes from five source proteins, for which responses to the OC43 homologs could also be observed [22]. Keller et al. . Despite these advances, an unbiased approach to the identification of OC43 T-cell epitopes independent of SARS-CoV-2 reactivity has not been reported.

Characterization of MHC-I immunopeptidome presented in OC43-infected A549 cells
Our experimental approach to the identification and characterization of naturally processed epitopes is diagrammed in Fig 1A. Peptide-MHC complexes carrying naturally processed and presented peptides were isolated by immunoaffinity from OC43-infected cells, and bound peptides were eluted and characterized by mass spectrometry. Next, peptides corresponding to the naturally processed epitopes were synthesized, tested for HLA binding, and used for evaluation of T-cell responses in mononuclear cells from peripheral blood samples.
Initial experiments were performed using the human lung adenocarcinoma line A549, as in an earlier study of the SARS-CoV-2 MHC-I immunopeptidome [39]. Viruses have evolved many mechanisms to evade the immune system, including the down-regulation of MHC proteins [45][46][47]. To assess this for OC43, we infected A549 cells over a range of multiplicity of infection (MOI, S1A Fig) and measured the surface levels of MHC-I using a pan-MHC-I antibody recognizing the three MHC-I proteins (HLA-ABC). The levels of HLA-ABC were only modestly reduced by infection at 0.01 MOI (an average of 11% reduction in median fluorescence intensity (MFI), but increasing the dose of virus reduced the surface expression of HLA-ABC (~22-34% reduction in MFI). So, for immunopeptidome characterization, four independent batches of A549 cells (120 to 160 x10 6 cells) were infected at a MOI of 0.01, in presence of IFN-γ, for 4 days. Levels of infection were variable (<10-55% cells positive for OC43 nucleoprotein (N), S1B Fig). We used a conventional immunoaffinity peptidomics workflow to identify peptides presented by MHC-I molecules in the infected cells. We purified MHC-bound complexes by immunoprecipitation after detergent solubilization of the membrane fraction of OC43-infected A549 cells using the pan-MHC-I antibody W6/32 to collect peptides bound by the six different MHC-I molecules present in this cell line: A*25:01, A*30:01, B*18:01, B*44:03, C*12:03, and C*16:01. The bound peptides were released from the purified complexes by acid treatment, separated from MHC protein subunits, and the resulting peptide mix was analyzed by LC-MS/MS for sequence identification. A database containing human and OC43 protein sequences was used for peptide assignment, with false-discovery rate (FDR) of 4.8%.
The total MHC-I immunopeptidome of infected A549 cells consisted of 1,474 unique peptides, including 9 derived OC43 sequences. The average length of the eluted peptides was 9 residues, as expected for HLA-ABC proteins (S1C Fig and S1A and S1B Table). To help deconvolute the mixture of peptides, we used unsupervised Gibbs clustering [48,49] of the eluted sequences in each sample. This analysis showed the presence of 4 MHC-I motifs (S1D Fig), closely matching the machine-learning predictions for these alleles (S2A Fig), with 40% of the sequences matching to A*25:01, 27% to either B*18:01 or *44:03, 20% to either C*12:03 or *16:01, and 13% to A*30:01. Of the nine eluted OC43 peptides, two derived from unconventional open reading frames within the OC43 genome (S1B Table). Of the conventional epitopes, one peptide each derived from the structural proteins spike, membrane, and nucleoprotein, and four derived from ORF1ab including two from the papain-like proteinase nsp3, one from the transmembrane autophagosome modulator nsp6, and one from the RNAdependent RNA polymerase subunit nsp12 (S1 Fig and Tables 2 and S1B). None of these peptides were observed in uninfected cells.
A549 cells express very low levels of MHC-II as revealed by surface staining with antibody recognizing HLA-DR (S1E Fig), even after treatment with IFN-γ (S1F Fig), which upregulates MHC-II expression in primary airway epithelial cells [50]. Low MHC-II induction has been reported to be due to MAP kinase-dependent antagonism in A549 cells, which harbor a KRAS

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 activating mutation, with the defect mitigated by treatment with the MEK inhibitor trametinib [51]. We observed a small increase in MHC-II expression in A549 cells treated with IFN-γ and trametinib, but the expression of MHC-II was not recovered to levels required in our immunoaffinity peptidomics workflow (Fig 1). Immunopeptidome workflow and HLA-ABC, HLA-DR, and HLA-DP immunopeptidomes in OC43-infected HEK293 cells. A. Experimental approach: HEK293 cells transduced with CIITA were infected with OC43. After 3 days, cells were collected and pMHC complexes were purified by immunoaffinity. Peptides were eluted from pMHC and analyzed by LC-MS/MS for identification. Identified peptides were used in biochemical and immunological assays. B. MHC expression on the surface of HEK293 cells. Four panels corresponding to the surface expression of HLA-ABC, HLA-DR, HLA-DQ, and HLA-DP are shown. HLA levels on wild-type cells are shown by grey histograms. HLA levels after transduction with CIITA are shown by colored histograms: HLA-ABC (blue), HLA-DR (purple), HLA-DQ (green), and HLA-DP (yellow). Isotype control staining is shown as an open histogram with dotted lines, following the same color scheme. C. Relative protein quantitation of HLA-DR, HLA-DP, and HLA-DQ proteins in CIITA-transfected HEK293 uninfected cells measured by label-free quantitative proteomics. D. Representative dot plots of intracellular staining for OC43 nucleoprotein in uninfected cells (top) and at 3 days after infection (middle); summary of 6 experiments (bottom). E. Representative histograms showing the comparison of surface levels of HLA-ABC, HLA-DR, and HLA-DP on uninfected (dark-shaded histograms) and infected (light-shaded histograms) cells. Graphs show the MFI in uninfected (non) and infected (oc43) cells from 3-6 independent infections. Statistical analysis in D and E by paired t-test, * p<0.05, ** p<0.01, ns: not significant. F. Length distribution of HLA-ABC, HLA-DR, and HLA-DP eluted immunopeptidomes from uninfected (grey histograms) and OC43-infected cells (color histograms). G. Sequence logos of clusters obtained using the Gibbs clustering analysis of HLA-ABC, HLA-DR, and HLA-DP eluted immunopeptidomes from OC43-infected cells; percentage of peptides in each cluster and probable allele are shown. https://doi.org/10.1371/journal.ppat.1011032.g001

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43

Characterization of MHC-I and MHC-II immunopeptidomes presented in OC43-infected HEK293 cells
Given the low numbers of OC43-derived peptides identified in the MHC-I peptidome of A549 cells, and the low expression of MHC-II, we sought another cell line for additional characterization of OC43-derived naturally-processed T-cell epitopes. HEK293 cells are homozygous in all MHC-I and MHC-II loci, and have been previously used in a study of the SARS-CoV-2 MHC-I immunopeptidome [39]. Moreover, the HLA alleles present in this cell line, A*02:01 (A2), B*07:02 (B7), C*07:02 (C7), DRB1*15:01 (DR2b), DRB5*01:01 (DR2a), DPA1*01:03/ DPB1*04:02 (DP4.2), and DQA1*01:02/DQB1*06:02 (DQ6.2), notably are highly abundant in many populations worldwide [52], as summarized in S3 Fig. We measured the expression of MHC-I and MHC-II on the surface of HEK293 cells using antibodies recognizing the three MHC-I proteins HLA-ABC or the individual MHC-II proteins HLA-DR, HLA-DQ, and HLA-DP. Expression of HLA-ABC was detected, but levels of HLA-DR, HLA-DP, and HLA-DQ were very low or below detection limits (Fig 1B, wild type HEK293). To induce expression of MHC-II, HEK293 cells were transduced with CIITA, as previously used in studies of naturally-processed tumor [53,54], autoantigen [55], and viral [35] peptide antigens. Transduced cells successfully upregulated the expression of HLA-DR and DP (Fig 1B,  HEK293.CIITA), although levels of HLA-DQ remained low. To confirm the low HLA-DQ expression level, the relative amounts of total MHC proteins were measured using a quantitative proteomics analysis (Fig 1C and S2B Table). The levels of HLA-DQ were~20-fold lower than HLA-DR and HLA-DP. Thus, we restricted immunopeptidome analysis to HLA-ABC, HLA-DR, and HLA-DP. HEK293.CIITA cells were infected at 0.1 MOI and harvested on day 3 post-infection. Intracellular staining for OC43 nucleoprotein showed a clear positive population of virus-infected cells at harvest, as compared to uninfected cells ( Fig 1D). In 6 biological replicates, we observed that 11-68% of the HEK293.CIITA cells were positive for OC43 nucleoprotein expression. The levels of HLA-ABC were significantly reduced after infection (an average of 68% reduction in MFI), while the expression of HLA-DR and HLA-DP were mostly not affected (less than 10% reduction in MFI) ( Fig 1E). This suggests that OC43 has a specific effect on the expression of MHC-I, which is more marked than in A549 cells (22% at the same MOI). While no apparent reduction of MHC-II expression in these cells was observed after infection, it is possible that CIITA transfection counteracts any effect of virus infection in our system as reported for SARS-CoV-2 and Ebola viruses [56].
Naturally-processed peptides bound to MHC-I and MHC-II molecules were characterized from the solubilized membrane fraction of infected HEK293.CIITA cells, as described above for A549; in this case, the FDR for peptide identification was 4.2%. We purified MHC-bound complexes of two independent infections (62 and 116 x10 6 cells), using sequential immunoaffinity purification with anti-HLA-DR (LB3.1), anti-HLA-DP (B7/21), and anti-HLA-ABC (W6/ 32) antibodies, collecting three immunoprecipitated samples, one from each antibody, per biological replicate infection. The total immunopeptidome of infected cells consisted of 1,744 unique peptides (613 HLA-ABC, 629 HLA-DR, and 502 HLA-DP, S3 Table). The eluted peptides showed the expected length distribution peaking at 9 aa for HLA-ABC and 15-16 amino acids for HLA-DR and -DP (Fig 1F), although HLA-DP showed a small peak of 8-11 residue peptides that might include non-specifically bound species [57]. The overall OC43-infected HEK293.CIITA immunopeptidome includes 83 virus-derived peptides (~96% host-derived). No viral peptides were identified in the peptidome of uninfected cells (S3E-S3G Table). Peptides were assigned to the MHC molecules present in each eluted pool by Gibbs clustering analysis, with reference to the expected peptide binding motif for each allele [48,49]. This analysis showed the presence of 2 motifs for MHC-I, representing 42 and 41% of the sequences

Identification of viral peptides presented by HLA-ABC, HLA-DR, and HLA-DP
Among the viral peptides identified in the immunopeptidomes of OC43-infected cells, 12 were eluted from HLA-ABC (9 from A549 and 3 from HEK293.CIITA), 35 from HLA-DR, and 45 from HLA-DP, representing 0.6 (for all HLA-ABCs), 5.6, and 9.4% of the peptides isolated from each type of MHC-II protein. The average length of the viral peptides was consistent with that observed for the total peptides, with a peak at 9 residues for the MHC-I and around 15-16 residues for the MHC-II (Figs 2A and S1C).
The three MHC-I-binding viral peptides derived from HEK293.CIITA cells were identified at relatively low abundances within the overall MHC-I peptidome of infected cells (Fig 2B, HLA-ABC and S3D Table). Peptide P17 (Fig 2C), derived from the spike protein, was assigned to HLA-A2 by motif analysis, with predicted binding in the top 0.5% (S3D Table). Peptides P15 and P16 ( Fig 2C) were derived from the 3C-like proteinase of the ORF 1ab polyprotein and were assigned to HLA-B7 and HLA-A2 respectively, based on predicted binding within the top 0.5% for these alleles, although weak binding of peptide P16 to HLA-C7 was also predicted (1.5%-tile) (S3D Table). Among viral peptides isolated from A549 cells, three were within the top 100 most abundant peptides, (P18, P21, and P20; S1H Fig and S1B Table). Three peptides were derived from the ORF1ab: P18 from nsp3, P19 from nsp6, and P20 from nsp12, and one each from the structural proteins: P21 from spike, P22 from membrane, and P23 from nucleoprotein (S1I Fig). P18, P19, and P22 were assigned to A*25:01, P21 was assigned to B*18:01 / B*44:03, and P23 to C*12:03 with predicted binding in the top 0.5%. The two peptides derived from alternative ORFs were at very low abundance and predicted binding was within 0.5-2% to A*30:01, B18*01, B*44:03, C*12:03 ( Table 2). The low abundance of virus-derived peptides within the overall MHC-I peptidomes might be a result of MHC-I immune-evasion mechanisms, similar to those reported for SARS-CoV-2 [46,63,64].
Eighty MHC-II-binding viral peptides were identified, derived from nucleoprotein, spike, hemagglutinin esterase (HE), and envelope proteins (S3D Table). Some of these were among the most abundant peptides identified in the MHC-II peptidomes: the most abundant peptide for HLA-DR and the third most abundant peptide for HLA-DP were virus-derived peptides To relate the abundance of eluted peptides to the overall abundance of the source proteins, we performed a whole cell proteomics analysis of infected HEK293.CIITA cells. Four viral proteins were detected: nucleoprotein, spike, HE, and the accessory protein N2 (S2A Table). Label-free quantitative analysis showed that the most abundant viral protein was nucleoprotein, followed by spike, and HE ( Fig 2E). Spike, nucleoprotein, and HE proteins were also the major source proteins for the eluted peptides ( Fig 2F).

MHC-II allele restriction of eluted peptides
The nested sets of peptides characteristic of MHC-II peptidomes are comprised of length variants surrounding a 9-residue core epitope that includes the major sites of MHC-peptide interaction. This is believed to result from variable trimming of MHC-bound peptides by endosomal proteases, leaving different numbers of residues flanking the core regions. As expected, for each of the nested sets of peptides, the predicted core epitope (underlined in Fig 2D) was found in the center of the overlapping set. Core epitopes for the eluted peptides For each peptide, the two most abundant species in the nested set are indicated. C. HLA-ABC eluted viral peptides. A schematic representation of each source protein and the location of the eluted sequence is shown (first and last residues indicated). D. HLA-DR and HLA-DP eluted viral peptides. A schematic representation as in C; the predicted core epitope in each sequence is underlined. Nested sets of eluted peptides comprising length variants with the same core epitope are shown by lines below the sequence. The peptide sequences highlighted in red were used for biochemical and immunological assays (see Table 1). In C and D, each eluted sequence or nested set was identified by a "P" followed by a number. E. Label-free quantitation of proteins present in infected cells; proteins were ranked from most to least abundant, with viral proteins highlighted in color. F. Relationship between viral protein abundance and eluted peptides abundance. For each source protein, the sum of intensities of all eluted peptides derived from it was used to calculate the eluted peptides abundance. https://doi.org/10.1371/journal.ppat.1011032.g002

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 were among the top-ranked predicted binders for each protein (S4A-S4C Fig), helping to explain why these particular peptides were selected for presentation. For instance, the topranked predicted peptides for nucleoprotein, spike, and envelope contain the binding core from the HLA-DP-eluted peptides P2, P5, and P14, respectively (S4A Fig). Similarly, the topranked predicted peptides for nucleoprotein, spike, and HE contain the binding core from the HLA-DR-eluted peptides P10, P11, and P8, respectively (S4B and S4C Fig).
For HLA-DR, peptides were tentatively assigned to DR2a or DR2b by motif analysis. In some cases, one allele was clearly preferred, with predicted binding in the top 5 th percentile to DR2b but not DR2a as for P8, P9, P10, and P11 peptides (S3D Table). P12 peptides were predicted to bind in the top 5 th percentile for both DR2b and DR2a, and P13 peptides were not predicted to bind to either DR2b or DR2a. For HLA-DP, predicted binding was in the top 5 th percentile for P2, P3, P5, P6, P7, and P14 peptides, but P1 and P4 were below this threshold. To experimentally assess MHC-II peptide binding for the eluted peptides, we used a fluorescence polarization competition binding assay [65,66] with synthetic peptides and purified recombinant MHC proteins (S4 Table and S4D Fig). For each set of nested peptides, we selected one abundant peptide containing the predicted binding core for the nested set and the allele of interest (S3D Table). These peptides are listed in Table 1. For DR2b, IC 50 values were below 1 μM for all the HLA-DR-eluted peptides except P12, including P13 which was not predicted to bind (S4D Fig and S4A Table). For DR2a, IC 50 values were below 1 μM for P12, as predicted, and also for P9. For DP4.1, only P1 and P5 of eight representative eluted peptides tested showed IC 50 values below 1 μM, although all but P2 and P6 exhibited IC 50 values below a more relaxed 10 μM criterion (S4B Table).

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43

T-cell recognition of eluted HLA-DR and HLA-DP viral peptides
We evaluated whether the naturally processed and presented viral peptides were recognized by circulating CD4 T cells in blood from healthy donors. We selected donors with a partial HLA match to HEK293 cells (donors expressing DRB1*15:01 and DRB5*01:01 for DR peptides, and donors expressing DPA1*01:03/DPB1*04:02 or DPB1*04:01 for DP peptides, S5 Table). We first assessed T-cell responses to the eluted peptides directly ex vivo in PBMC samples using ELISpot assays with the same set of peptides as tested for MHC-II binding. Ex-vivo IFN-γ responses were measured in donors expressing at least one of the alleles of interest, by stimulating PBMCs with a pool of all DP or all DR peptides ( Fig 3A). Positive responses were observed in most donors tested (9/14 for DP and 8/12 for DR). Responding T cells were present at low frequencies, which varied considerably between donors (0.001-0.055% for DP; 0.001-0.012% for DR). Note that in this assay, other HLA alleles are present in the donors besides the HEK293 alleles used for the elution studies, but with very few exceptions, these alleles are the best-predicted binders among the HLA-DR, HLA-DP, and HLA-DQ alleles present in each donor (S6 Table).
To increase the frequency of OC43-responding cells for detailed assessment of the responses to individual peptides, we stimulated PMBC in vitro with selected peptides to expand any responding T-cell populations. Using these expanded T-cell populations, we measured IFN-γ production in response to re-stimulation with the same peptides, individually presented by single-allele antigen presenting cells (DPA1*01:03/DPB1*04:01 for P1-P7 and P14, DRB1*15:01 for P8-P13, and DRB5*01:01 for P9 and P12, Table 1). Eleven peptides (all except P8, P13 and P14) showed individual positive responses by IFN-γ ELISpot in at least one of the donors analyzed ( Fig 3B and 3C, filled symbols), validating the presence of T-cell responses to the peptide. Not every donor responded to every peptide, and different donors showed different patterns of responses. The fraction of donors who are positive for each of the responding peptides ranged from 60-100% (Fig 3B and 3C, pies). In general, responses were more frequently observed (p = 0.006) in DR15 donors (80-100%) than in DP4 donors (60-88%), while responses were slightly stronger for DP peptides (3.7 ± 2.1x10 3 SFU/10 6 cells) than DR peptides (2.2 ± 1.8x10 3 SFU/10 6 cells) when tested at 1 μg/mL peptide concentration, although this difference is not significant ( Fig 3D). There was a weak but significant correlation between the eluted peptide abundance (sum of precursor ion intensities by nested set) and the observed T-cell response (r = 0.64, p = 0.009, Spearman). No correlation was observed between binding (predicted or experimental) and T-cell responses, nor between binding and peptide abundance.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 To explore the overall sensitivity of the different peptide-expanded T cells, dose-response experiments were performed, and the minimal activating peptide concentrations were determined (Fig 3E and 3F). In general, a wide range of minimal concentrations was observed. For instance, for P2 and P3 the minimal concentrations were 10 −6 μg/mL and 10 −7 μg/mL, respectively for expanded cells from donor 61, while for P11 (donor 07) and P9 (donor 40), the minimal concentration was 1 μg/mL. This indicates that T cells responding to P2 and P3 in donor 61 were more sensitive to lower peptide concentrations and may be able to respond more efficiently to infection. Within donors, differences in minimal concentration were observed for

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 different peptides, suggesting a heterogeneous population that responds to different antigens with different efficiencies. In some cases, different donors showed similar sensitivity to a particular peptide, as is the case of P10 in donors 18, 22, and 40, which all responded at 10 −5 μg/ mL. However, in other cases, there was heterogeneity in the responses to a given peptide. For instance, for P4 the minimal concentration varied between 0.1 and 10 −5 μg/mL in 4 donors. All these results may reflect the different history of exposure to OC43 and other coronaviruses and the evolution of the responding T-cell repertoire in each individual, which translates to a lack of a clear hierarchy of functional avidity and immunodominance for most of the eluted peptides.
To characterize the T cells producing these responses, we performed intracellular cytokine staining (ICS) assays using the single-peptide-expanded T-cell lines. As in the ELISpot assays, peptides P8, P13, and P14 did not produce a response. For the remaining 11 peptides, IFN-γ responses were observed exclusively in CD4+ T-cell populations. Results from one representative cell line per peptide are shown in Fig 3G, with a summary of all results in Fig 3H and 3I. For 9 of these peptides, we were able to measure CD107a mobilization along with IFN-γ production (S5C Fig). For one peptide (P3), production of low levels of TNF-α by a small responding population was observed (S5D Fig). We did not observe IL-2 or IL1-production for any of the eluted peptides. Analysis of Boolean gates for IFN-γ, CD107a, and TNF-α showed polyfunctional responses for some donor/peptide combinations (S6 Fig, pies and arcs). Trifunctional cells were observed for P3 and P4. Bifunctional IFN-γ/CD107a positive cells were more common, and analysis in 4 donors showed significant enrichment of this population in P2, P6, and P10 (S6 Fig, bar graphs). This suggests that the CD4 T cells responding to the eluted OC43 peptides can be polyfunctional and may have cytotoxic potential.
Altogether, these results present clear evidence of CD4+ T cells that recognize and respond to OC43-derived, DR2b, DR2a, and DP4.1/4.2-presented peptides, confirming the immunogenicity of these peptides in natural settings, showing that some of these peptides may be recognized by T cells at very low antigen concentrations in some donors, and highlighting the complexity of these responses.

T-cell cross-reactivity between OC43 and other human coronaviruses
The substantial sequence homology between OC43 and the other HCoVs (S7A Fig) raises the question of whether responding T cells could cross-react between the different orthologs. Sequence alignments of the naturally processed OC43 peptides with homologous sequences from other HCoVs are shown in S7B Fig, and a heatmap of conservation indices is shown in S7C Fig. Overall, the highest conservation is between OC43-and HKU1-derived peptides, with less for the other betacoronaviruses MERS-CoV, SARS-CoV, and SARS-CoV-2, and even less for the alphacoronaviruses 229E and NL63. Among the eluted peptides, P4, P6, and P11 are the most conserved across the 7 viruses and would be expected to have a high potential for cross-reactivity. The remaining peptides (P1, P2, P3, P5, P7, P8, P9, P10, P12, P13, and P14) were less conserved. Note that the HE protein, the source of the P1, P8, and P9 epitopes, is expressed by OC43 and HKU1 but does not have a homolog in any other HCoVs [7].
To evaluate experimentally the potential for cross-reactivity we initially focused on OC43 and SARS-CoV-2. We measured responses to the eluted OC43 peptides and their SARS-CoV-2 homologs, using T-cell populations expanded with individual OC43 peptides from PBMC samples banked pre-pandemic before the outbreak of SARS-CoV-2 into the human population. Peptides with no homolog in SARS-CoV-2 (P1, P8, P9), or with no response in our donor pool (P8, P13, P14) were excluded. We measured T-cell responses in single-peptideexpanded T-cell lines using IFN-γ ELISpot assays, using partial-HLA-matched donors as

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 before. Only the P4 and P11 SARS-CoV-2 homologs induced cross-reactive T-cell responses in the single-peptide expanded lines (Fig 4A). Across a larger set of donors, similar cross-reactive responses were observed, with somewhat lower responses to the heterologous SARS-CoV-2 homologs than the OC43 peptides used for expansion (average 2-fold, p = 0.044 for P4 and average 3.5-fold, p = 0.011 for P11; paired t-test; Fig 4B). Responses to SARS-CoV-2 homologs in T-cell lines from pre-pandemic donors expanded by OC43 P4 and P11 strongly suggest that individual T cells are able to respond to both OC43 and SARS-CoV-2 peptides. To rule out the possibility that the observed SARS-CoV-2 responses were unrelated to expansion with the OC43 peptides, we assessed responses to P4 and P11 peptides and their SARS-CoV-2 homologs using samples from the same donors before and after in vitro expansion. All but one line expanded with SARS-CoV-2 peptides exhibited positive response to the heterologous stimulation with OC43 peptide (Fig 4C). The numbers of responding cells were similar to that observed in the same line for cognate stimulation with SARS-CoV-2 peptide, and indeed for a corresponding line raised from the same donor using in vitro expansion with OC43 peptide. These results indicate that a substantial proportion of T cells responding to the OC43-P4 and OC43-P11 peptides can cross-react with their SARS-CoV-2 homologs. To evaluate the sensitivity of these T-cell lines to cross-reactive stimulation, we measured the dose-response to cognate and heterologous peptides. Robust cross-reactivity to heterologous stimulation was observed across the dose-response range for both P4 and P11 homologs in all donors tested, including pre-pandemic ( Fig 4D) and those with recent COVID-19 infection (Fig 4E), with minimal stimulatory peptide concentrations in a wide range but similar for OC43 and SARS-CoV-2 homologs (Fig 4F).
To explore factors that could have resulted in the observed pattern of OC43 and SARS-CoV-2 cross-reactive responses, we measured MHC binding of the SARS-CoV-2 homologs and compared them to the OC43 peptides (Fig 4G). We found weaker binding for most of the SARS-CoV-2 homologs, except for P4, for which DP4.1 binding was 10-fold greater for the SARS-CoV-2 homolog. In addition to altering MHC binding affinity, amino acid substitutions can cause shifting of the preferred binding register, which would interfere with the T-cell recognition of homologous peptides. Of the nine peptides tested, only P3, P4, and P11 retain the predicted binding register in the SARS-CoV-2 homologs (Fig 4H), and only for P4 and P11 are the predicted T-cell contacts completely or mostly conserved (shaded in Fig 4H).
We extended this analysis to the other seasonal human coronaviruses, using the T-cell lines expanded in vitro with P4 and P11 peptides from pre-pandemic and COVID-19 donors. The P4 and P11 homologs from the seasonal coronaviruses mostly retained binding to DP4.1 (for P4) and DR2a/DR2b (for P11) (Fig 4I), and we measured the cross-reactive T response to these peptides. In general, all the P4-and P11-expanded T-cell lines recognized each of the homologs, except for P11 from 229E, which was recognized poorly by T-cell lines expanded with SARS-CoV-2 or OC43 homologs (Fig 4J).

Discussion
The immune response to seasonal human coronaviruses is largely understudied and few T-cell epitopes have been identified, although interest in this area has increased with the COVID-19 pandemic. To help fill this gap we identified naturally processed and presented viral epitopes expressed in OC43-infected cells using immunoaffinity purification of MHC-peptide complexes followed by mass spectrometry of eluted peptides. Nine viral peptides presented by MHC-I molecules were identified within the overall immunopeptidome of A549 cells infected with OC43, and only 3 viral peptides presented by MHC-I molecules were identified within the overall immunopeptidome of CIITA-transfected OC43-infected HEK293 cells, possibly

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 due to virus-induced down-regulation of MHC-I expression. A total of 80 viral peptides presented by MHC-II molecules were identified in the overall immunopeptidome of CIITAtransfected OC43-infected HEK293 cells, representing 14 distinct core epitopes present in nested sets characteristic of MHC-II processing. Eleven of these OC43-derived epitopes were recognized by recall responses in partially-HLA-matched donors. Almost all of the OC43-derived MHC epitopes identified in this work are reported here for the first time, although T responses to the two highly-cross-reactive epitopes P4 and P11 have been reported previously in studies characterizing seasonal coronavirus cross-reactivity to identified SARS-CoV-2 epitopes [22,26,28,29,67].
We identified only twelve OC43-derived peptides presented by MHC-I molecules, and these were present at very low abundance within the overall MHC-I peptidome. In infected HEK293 cells we observed one peptide from the spike protein and one from the 3C-like proteinase encoded by the ORF1ab polyprotein, both likely presented by HLA-A2, and a second 3C-like proteinase peptide likely presented by HLA-B7. In infected A549 cells we identified one peptide each from the spike, membrane and nucleoprotein structural proteins and four peptides from non-structural proteins encoded by the ORF1ab, tentatively assigned to one or two of the MHC-I alleles present in A549. These epitopes have not been previously reported, although a different OC43 spike epitope presented by HLA-24 [68,69] and two OC43-derived epitopes from other ORF1ab-derived proteins, both presented by HLA-A2 [70], have been described in studies of SARS-CoV-2 cross-reactive CD8 T-cell responses. We observed potent MHC-I down-regulation after OC43 infection in HEK293.CIITA cells, which may have limited presentation of viral epitopes on MHC-I molecules. A less dramatic effect was observed in A549 cells. MHC-I down-regulation has not been previously reported for OC43, but is a common feature of many viruses [45][46][47], including SARS-CoV-2 [46,47,64,71]. Current understanding of SARS-CoV-2-induced MHC-I down-regulation points to a complex mechanism, with the involvement of several gene products: ORF3a reduces global trafficking of proteins including MHC-I [46], ORF6 inhibits induction of MHC-I by targeting the STA-T1-IRF1-NLRC5 axis [71], ORF7a reduces cell-surface expression of MHC-I [46,47] by acting as β2-microglobulin mimic to interact with MHC-I heavy chain and slow its egress through the endoplasmic reticulum [46], and ORF8 also has been reported to down-regulate surface MHC-I through a direct interaction, although the specific mechanism is unclear [64]. However, none of these SARS-COV-2 gene products have significant homology with OC43, and elucidating the mechanism by which OC43 down-regulates MHC-I expression will require further investigation.
By contrast, eighty OC43-derived peptides presented by MHC-II molecules were found at high abundance within the overall MHC-II peptidome. Indeed, three of the top four most intense ions in the HLA-DR peptidome mass spectrum, and the third and fourth most intense ions in the HLA-DP peptidome mass spectrum, correspond to OC43-derived peptides. Most of the OC43-derived MHC-II-bound peptides were from spike and nucleoprotein, the major coronavirus structural proteins, consistent with the over-representation of these proteins we observed in the whole-cell proteome of infected cells. Several peptides derived from the hemagglutinin-esterase protein, which is believed to be required for cleavage of sialic acid residues to promote the release of progeny virus from infected cells, similarly to hemagglutinin-esterase proteins from influenza C and certain toroviruses and orthomyxoviruses [72]. Finally, one set of low-abundance peptides is derived from the small envelope protein. All the OC43-derived relevant single allele APC. In A-E and J, ELISpot statistical analysis by DFR method [98]; positive responses shown as filled symbols and negative responses as empty symbols. In B and E, statistical analysis was done by unpaired t-test. * p<0.05). https://doi.org/10.1371/journal.ppat.1011032.g004

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 MHC-II-bound peptides were found as nested sets, except for three very low abundance peptides found as singletons. In each case, the nested sets surrounded the predicted nine-residue core epitope, with 1-9 residue extensions, consistent with endosomal protease trimming of MHC-bound peptides as expected in the MHC-II antigen-presentation pathway. We selected one representative peptide from each nested set to confirm binding to MHC-II, and to assign presenting MHC allotypes to the HLA-DR peptides, which could derive from either DR2a (DRB5*01:01) or DR2b (DRB1*05:01), both of which are expressed by HEK293 cells and copurified with the LB3.1 antibody that we used for immunoaffinity. Each of the eight representative HLA-DP eluted peptides bound to DP4.1, although with varying affinity not entirely predicted by NetMHCIIpan4.1. Of the six representative HLA-DR peptides, one (P12) bound exclusively to DR2a, four exclusively to (P8, P10, P11, and P13) DR2b, and one to both allotypes (P9). As previously observed in another study of naturally processed MHC-II peptides in virus-infected cells [35], the eluted peptides generally were among the top predicted binders for each viral protein, one exception being P1 from the hemagglutinin-esterase protein.
We tested representative eluted peptides for recognition by T cells from HLA-matched donors. Of fourteen peptides tested, we observed robust T-cell responses to eleven. In other systems, characterization of naturally-processed, MHC-bound peptides by mass spectrometry of infected cells has proven to be an efficient route for T-cell epitope discovery [32-40,73,74]. We observed a correlation between the observed T-cell response and epitope abundance in the overall immunopeptidome, whereas a significant correlation was not observed for the predicted or even observed peptide binding affinity. Thus, characterization of naturally-processed peptides from virus-infected cells can be a highly efficient epitope discovery approach, particularly compared to screening comprehensive overlapping peptides libraries or large sets of predicted MHC binders, where typically T-cell responses are observed to only a small fraction of the candidate epitopes. A similar trend relating T-cell response to epitope abundance has been observed in some [40] but not all [33, 73,74] previous studies, although it should be noted that all of these previous studies involved CD8 T-cell responses. Three eluted peptides (P8, P14, and P13) were not recognized by T cells from HLA-matched donors. These peptides were present at relatively low abundance in the peptidomes, although in some cases (P6, P7, P9) peptides with even lower abundance were recognized. We examined whether these peptides might not be immunogenic because of homology to self-peptides [75]. The peptides that were not recognized had similar homology scores to the closest matching self-peptides as did peptides that were recognized, although the number of exact matches in the core epitope region was somewhat larger for peptides that were not recognized (mean 6.3 vs 4.6, p = 0.016).
Among human and animal coronaviruses, the approach of characterizing naturally-processed peptides presented by MHC proteins in infected cells to date has only been applied to SARS-CoV-2 [39,76]. Weingarten-Gabbay et al. [39] eluted MHC-I bound peptides from SARS-CoV-2-infected A549 and HEK293 cell lines, and identified 28 canonical epitopes from spike, nucleoprotein, membrane, ORF7a, and several ORF1ab-derived nonstructural proteins, together with 9 unconventional epitopes derived from out-of-frame transcripts in spike and nucleoprotein. Nagler et al. [76] similarly identified two MHC-I epitopes derived from out-offrame viral transcripts together with 11 conventional epitopes from spike, nucleoprotein, nsp1, and nsp3. We searched for such out-of-frame peptides in the OC43-derived immunopeptidome but did not find convincing evidence (see methods). As an alternative to infection, Pan et al. [77] transfected cell lines with membrane or nsp13 genes and identified five MHC-I epitopes. In addition to the infection studies mentioned above, Nagler et al. [76] also characterized MHC-bound peptides derived from cell lines transfected with individual nucleoprotein, envelope, membrane, and nsp6 genes, and identified additional MHC-I and also HLA-DR epitopes. Using a somewhat different experimental approach, Knierman et al. [78] and Parker

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 et al. [79] added purified recombinant SARS-CoV-2 spike protein to monocyte-derived dendritic cells, which might simulate physiological antigen uptake by professional antigen-presenting cells at sites of infection. Peptides containing SARS-CoV-2 homologs of the OC43 P4 and P11 epitopes that we characterized here were among the many MHC-II-bound peptides that derived from the added recombinant proteins [78,79].
Several previous studies of the T-cell response to SARS-CoV-2 in pre-pandemic donors have identified T-cell responses that are cross-reactive with homologous epitopes from seasonal coronaviruses including OC43 [17,22,27,28,67,[80][81][82][83]. However, there is still not a consensus on the involvement of the cross-reactive response in the clinical outcome, although recent studies have pointed to a role for cross-reactive CD8 T-cell responses in protection from SARS-CoV-2 infection [70] and severe COVID-19 [84]. To identify additional crossreactive epitopes, we tested the reactivity of T-cell lines expanded with the eluted OC43 peptides for cross-reactivity with SARS-CoV2 homologs. Among the nine naturally processed CD4 T-cell epitopes that were robustly recognized by donors in our cohorts, only two (P4 S 903-917 and P11 S 1085-1099 ) were targeted by T cells cross-reactive with SARS-CoV-2. Doseresponse curves were similar for both SARS-CoV-2 and OC43 versions of the cross-reactive P4 and P11 epitopes, in both pre-pandemic and COVID-19 donors. This suggests that T cells might respond similarly during infections with either virus. Notably, these same epitopes were observed previously in an unbiased screen of SARS-CoV-2-derived peptides targeted by HCoV cross-reactive T cells [26], as well as in other studies of T-cell responses cross-reactivity between SARS-CoV-2 and HCoVs [22,27-29, [85][86][87][88]. For both the P4 and P11 epitopes, the OC43 and SARS-CoV-2 homologs are predicted to bind to the respective MHC-II proteins using the same binding frame, and peptide residues at the predicted T-cell contact positions are identical or conserved. For the seven OC43-derived naturally processed T-cell epitopes with SARS-CoV-2 homologs that were not targeted by cross-reactive responses, six had predicted shifts of the MHC-II binding frame caused by peptide substitutions at MHC-II contact positions. The one epitope for which the predicted MHC-II binding frame was preserved (P3 S [97][98][99][100][101][102][103][104][105][106][107][108][109][110][111] ) has substitutions at each of the TCR contact positions, which would be expected to abrogate cross-reactive T-cell binding. Thus, the pattern of observed CD4 T-cell cross-reactivity can be explained by a simple model in which the key parameters are the preservation of the MHC-II binding frame and conservation of T-cell receptor contact residues as we have previously reported [26].
For studies of the differential response to SARS-CoV-2 and seasonal coronaviruses, epitopes specific to the seasonal coronaviruses are required. Among the OC43-eluted peptides for which cross-reactive T-cell responses to SARS-CoV-2 homologs were not observed, P10 N 54-68 elicited recall responses in all donors tested. Responding CD4 T cells showed a high sensitivity, with minimal peptide concentrations of about 10 pg/mL. This epitope is not strongly conserved among the HCoVs (S7C Fig) and may be a good candidate to study and follow OC43-specific responses. In addition, epitopes P1 HE 259-273 , and P9 HE 128-142 both are recognized by strong responses in a large majority of donors tested. Of the human coronaviruses, only OC43 and HKU1 express HE proteins, consistent with their use of 9-O-acetylated sialic acids as an entry receptor. Neither SARS-CoV-2 nor MERS-CoV, SARS-CoV, 229E, or NL63 express HE homologs. No HE-derived T-cell epitopes have been reported from any other organism (although neutralizing antibodies to influenza C HE have been reported [89,90]. Thus, T-cell responses to P1 HE 259-273 and P9 HE 128-142 would be expected to mark specific exposure to HCoVs (OC43 and/or HKU1) and might be useful in evaluating the contribution of HCoV exposure in SARS-CoV-2 incidence or pathogenesis.
There are some limitations to this study. The HEK293 cells used for immunopeptidome characterization were manipulated to ensure stable expression of MHC-II proteins by

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 introducing the CIITA gene, which may favor the processing and presentation in the MHC-II compartment. In addition, these cells may not be representative of the natural targets of infection in the respiratory tract. Pre-pandemic donors testing positive for OC43-specific T-cell or antibody responses could have been exposed to a different virus or other immune stimulus cross-reactive with OC43. It is possible that some T-cell responses to synthetic peptides might consist, in part, of responses to non-native modifications introduced by the synthesis chemistry. We did not consider T-cell responses restricted by the mismatched MHC molecules. Finally, T-cell responses not associated with IFN-γ, not able to expand with peptide stimulation in vitro, or below our detection level would have been missed by our approach.
In summary, we characterized the spectrum of naturally-processed viral peptides presented by MHC molecules in HEK293.CIITA cells infected with the human seasonal coronavirus OC43. MHC-II presented peptides dominated the OC43-derived viral immunopeptidome, possibly due to the potent down-regulation of MHC-I molecules in infected cells. The spike protein is the major source of OC43-derived epitopes, with contributions from nucleoprotein and hemagglutinin-esterase. Most of the naturally-processed peptides are recognized by T cells from HLA-matched donors. Three seasonal-coronavirus-specific CD4 T-cell epitopes and two SARS-CoV-2-cross-reactive CD4 epitopes were identified. These epitopes provide a basis for studies of the cellular immune response to OC43, and for evaluating the role of pre-existing seasonal coronavirus immunity in SARS-CoV-2 infection and vaccination.

Cell lines
A549 cells were kindly provided by Dr. Rene Maehr (UMass Chan Medical School) and maintained in DMEM medium supplemented with L-glutamine (2 mM), sodium pyruvate (1 mM), non-essential amino acids (1 mM), and 10%FBS at 37˚C/5% CO 2 . HEK293 cells were kindly provided by Dr. Kenneth Rock (UMass Chan Medical School). Cells were maintained in DMEM medium supplemented with L-glutamine (2 mM), sodium pyruvate (1 mM), nonessential amino acids (1 mM), and 10%FBS at 37˚C/5% CO 2 . HEK293 cells were transduced using the LentiORF clone of CIITA (OriGene RC222253L3). The cells were selected using puromycin selection marker for 2 passages over the period of 7 days. The cells were further transduced using human ace2 containing lentiviral particles, a kind gift from Dr. Rene Maehr (UMass Chan Medical School), to facilitate future work with other coronaviruses. The cells were stained for anti-HLA-DR, HLA-DP and HLA-DQ to confirm the MHC-II expression. These cells were further enriched by flow-based sorting for ACE2 expression and HLA-DR expression.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43

Virus production and cell infection
Human coronavirus OC43 strain VR-759 was obtained from ATCC (betacoronavirus-1, #VR-1558). The virus was propagated in the lung fibroblast cell line MRC-5 (ATCC# CCL-171) at a multiplicity of infection (MOI) of 0.01 and the virus was collected after 5 days. Virus stocks were titrated using a standard TCID 50 assay. A549 cells were infected at 0.01-1 MOI for 3-4 days, in presence of IFN-γ (100 ng/mL); for some experiments trametinib (50 ng/mL) was added. HEK293.CIITA cells were infected at a MOI of 0.1 for 3 days. Cells were collected, washed with PBS, and the cell pellets were frozen at -80˚C until use. Percentage of infected cells were measured by intracellular staining for the nucleoprotein (mouse anti-coronavirus OC43 nucleoprotein clone 542-70, EMD Millipore, Burlington, MA).

Isolation of MHC Class I and Class II bound peptides
Detergent-solubilized fractions isolated from uninfected A549 or HEK293.CIITA, and OC43 infected HEK293.CIITA (n = 2) or A549 (n = 4) cells were used for elution experiments. Cells were suspended in ice-cold hypotonic buffer (10 mm Tris-HCl, pH 8.0, containing protease inhibitors) and lysed using bath sonicator (Misonix S-4000 Ultrasonic Liquid Processor) maintained at 4˚C with the amplitude of 70. The sonication was done for 3 mins with a cycle of pulse for 20 secs followed by resting cells on ice for 10 secs. Unlysed cells, nuclei, cytoskeleton, and cell debris were removed by centrifuging the lysate at 2000 ×g for 5 min at 4˚C. The supernatant was collected and further centrifuged at 100,000 ×g for 1 h at 4˚C to pellet the membrane/microsome fraction. This fraction was solubilized in ice-cold 50 mM Tris-HCl, 150 mM NaCl, pH 8.0 and 5% β-octylglucoside in a dounce homogenizer and incubated on ice for 1 hour. Benzonase (50 U/mL), 2 mM MgCl 2 , and protease inhibitor cocktail, were added to inactivate virus, and the mixture was rotated slowly overnight at 4˚C. Solubilized membranes were centrifuged at 100,000 ×g for 1 hour at 4˚C and the supernatant used for MHC-peptide isolation and immunopeptidome characterization. The supernatant was equilibrated with protein A agarose beads and isotype antibody conjugated beads sequentially for 1 hour each at 4˚C and allowed to mix slowly to remove nonspecific binding proteins. The precleared membrane fraction was then incubated sequentially with immunoaffinity beads of protein A agarose-LB3.1 antibody (HLA-DR), protein A agarose-B7/21 antibody (HLA-DP), and protein A agarose-W6/32 (HLA-ABC) antibody sequentially for 2 hours each at 4˚C and allowed to mix slowly. The beads were washed with several buffers in succession as follows: (1) 50 mM Tris-HCl, 150 mM NaCl, pH 8.0, containing protease inhibitors and 5% β-octylglucoside (5 times the bead volume); (2) 50 mM Tris-HCl, 150 mM NaCl, pH 8.0, containing protease inhibitors and 1% β-octylglucoside (10 times the bead volume); (3) 50 mM Tris-HCl, 150 mM NaCl, pH 8.0, containing protease inhibitors (30 times the bead volume); (4) 50 mM Tris-HCl, 300 mM NaCl, pH 8.0, containing protease inhibitors (10 times the bead volume); (5) PBS (30 times the bead volume); and (6) HPLC water (100 times the bead volume). Bound complexes were acideluted using 2% TFA. Detergent, buffer components, and MHC proteins were removed using a Vydac C18 microspin column (The Nest Group, Ipswich, MA). The mixture of MHC and peptides were bound to the column, and after washes with 0.1% TFA, the peptides were eluted using 30% acetonitrile in 0.1% TFA. Eluted peptides were lyophilized using a SpeedVac and were resuspended in 25 μL of 5% acetonitrile and 0.1% TFA.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 4.0 μL/min for 4.0 min onto a 100 μm I.D. fused-silica precolumn packed with 2 cm of 5 μm (200 Å) Magic C18AQ (Bruker-Michrom, Auburn, CA) and eluted using a gradient at 300 nL/ min onto a 75 μm I.D. analytical column packed with 25 cm of 3 μm (100 Å) Magic C18AQ particles to a gravity-pulled tip. The solvents were A) water (0.1% formic acid); and B) acetonitrile (0.1% formic acid). A linear gradient was developed from 5% solvent A to 35% solvent B in 60 min. Ions were introduced by positive electrospray ionization via liquid junction into a Orbitrap Fusion Lumos Tribrid Mass Spectrometer. Mass spectra were acquired over m/z 300-1,750 at 70,000 resolution (m/z-200), and data-dependent acquisition selected the top 10 most abundant precursor ions in each scan for tandem mass spectrometry by HCD fragmentation using an isolation width of 1.6 Da, collision energy of 27, and a resolution of 17,500.

Peptide Identification
Raw data files were peak processed with Proteome Discoverer (version 2.1, Thermo-Fisher Scientific) prior to database searching with Mascot Server (version 2.5, Matrix Science, Boston, MA) against a combined database of UniProt_Human, UniProt_hCoV-OC43 and an out-offrame OC43 unconventional ORF database constructed according to Stern-Ginossar et al. [93]. Search parameters included no-enzyme specificity to detect peptides generated by cleavage after any residue. The variable modifications of oxidized methionine and pyroglutamic acid for N-terminal glutamine were considered. The mass tolerances were 10 ppm for the precursor and 0.05 Da for the fragments. Search results were then loaded into the Scaffold Viewer (Proteome Software, Inc., Portland, OR) for peptide/protein validation and label-free quantitation. Scaffold assigns probabilities using PeptideProphet or the LDFR algorithm for peptide identification and the ProteinProphet algorithm for protein identification, allowing the peptide and protein identification to be scored on the level of probability. An estimated FDR of 5% was achieved by adjusting peptide identification probability. Peptides identified in a blank run were excluded from the peptidomes. Peptides with Mascot Ion score below 15 were also excluded. Only one match to the OC43 unconventional ORF database was identified for an HLA-DP-bound peptide. This sequence (LTILYLWVGIILSVIVL), derived from an out-offrame ORF in the membrane gene, did not match the HLA-DP binding motif and the singleion spectrum was poor, so this sequence was not considered further. Samples from infected and uninfected cells were processed using the same parameters, and OC43 peptides were detected only in infected cells.

Label-free proteomic analysis
Relative protein quantitation in infected and uninfected cells, including MHC and viral proteins, was performed using whole-cell lysates and a label-free proteomics analysis. Samples containing 1 μg of total protein were trypsin digested using S-Trap™ Mini Spin Column (PRO-TIFI). An injection of~200 ng was loaded by a Waters nanoACQUITY UPLC in 5% acetonitrile (0.1% formic acid) at 4.0 μL/min for 4.0 min onto a 100 μm I.D. fused-silica precolumn packed with 2 cm of 5 μm (200 Å) Magic C18AQ (Bruker-Michrom). Peptides were eluted at 300 nL/min from a 75 μm I.D. gravity-pulled analytical column packed with 25 cm of 3 μm (100 Å) Magic C18AQ particles using a linear gradient from 5-35% of mobile phase B (acetonitrile + 0.1% formic acid) in mobile phase A (water + 0.1% formic acid) for 120 min. Ions were introduced by positive electrospray ionization via liquid junction at 1.5kV into a Orbitrap Fusion Lumos Tribrid Mass Spectrometer. Mass spectra were acquired over m/z 300-1,750 at 70,000 resolution (m/z 200) with an AGC target of 1e6, and data-dependent acquisition selected the top 10 most abundant precursor ions for tandem mass spectrometry by HCD fragmentation using an isolation width of 1.6 Da, max fill time of 110 ms, and AGC target of 1e5.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 Peptides were fragmented by a normalized collisional energy of 27, and fragment spectra acquired at a resolution of 17,500 (m/z 200). Raw data files were peak-processed with Proteome Discoverer (version 1.4, Thermo Scientific) followed by identification using Mascot Server (version 2.5, Matrix Science) against an UniProt_Human, UniProt_hCoV-OC43 and out-of-frame hCoV-OC43 databases. Search parameters included Trypsin/P specificity, up to 2 missed cleavages, a minimum of two peptides, a fixed modification of carbamidomethyl cysteine, and variable modifications of oxidized methionine, pyroglutamic acid for glutamine, and N-terminal acetylation. Assignments were made using a 10-ppm mass tolerance for the precursor and 0.05 Da mass tolerance for the fragments. All nonfiltered search results were processed by Scaffold (version 4.4.4, Proteome Software, Inc.) utilizing the Trans-Proteomic Pipeline (Institute for Systems Biology) with a 1% false-discovery rate. The data was processed using MaxQuant as well which uses Andromeda search engine and search parameters were kept the same as Mascot Server. The search was performed against a concatenated targetdecoy database with modified reversing of protein sequences. For MHC protein quantitation, HLA-ABC heavy (alpha) chains and HLA-DR, HLA-DQ, HLA-DP beta chains were considered. Intensities of HLA-DRB1*15:01 and HLA-DRB1*01:01 were summed to provide an HLA-DR value. Peptides unique to HLA-C or HLA-E, a non-classical class I MHC bound by W6/32 along with HLA-ABC [94], were not detected, although two peptides identical in HLA-C and HLA-A were detected and assigned to HLA-A, and two peptides identical in HLA-E and HLA-B were detected and assigned to HLA-B.

Gibbs clustering
GibbsCluster-2.0 [49] within DTU Health Tech server, was used to align the eluted peptide sequences and analyze the motifs, which were displayed with Seq2Logo 2.0 [95]. We allowed the software to include cluster sizes of 1-5 with a motif length of 9 amino acids and clustering sequence weighting. Default values were used for other parameters: number of seeds = 1, penalty factor for inter-cluster similarity = 0.8, small cluster weight = 5, no outlier removal, iterations per temperature step = 10, Monte Carlo temperature = 1.5, intervals for indel, single peptide and phase-shift moves = 10, 20, and 100, respectively, and Uniprot amino acid frequencies were used. For each sample, we selected the cluster that included the largest number of peptides analyzed. For HLA-DR and HLA-DP peptides, a preference for hydrophobic residue at P1 was used to align the motifs at the P1 position. For HLA-ABC peptides, MHC-I ligands of length 8-13 residues parameters were loaded. The fraction of sequences that contributed to each cluster is shown in the figures.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 free HLA-DR1, but with 1 U/μg thrombin (DP4.1) or 3C protease (DR2b and DR2a) added to cleave the CLIP linker and HLA-DM included to initiate peptide exchange. Thrombin or 3C protease enzymes was inactivated after 3 hours of reaction using protease cocktail inhibitor, and the reaction was continued for 24 hours at 37˚C before FP measurement using a Victor X5 Multilabel plate reader (PerkinElmer, Shelton, CT). DP4.1-Clip (250 nM), DR2b-CLIP (500 nM) and DR2a-CLIP (250 nM) concentrations were selected to provide 50% maximum binding of 25 nM probe peptide in the presence of 500 nM soluble HLA-DM. Binding reactions also contained serial dilutions of test peptides with 5-fold dilutions. The capacity of each test peptide to compete for binding of probe peptide was measured by the fluorescence polarization (FP) after 24 hours at 37˚C. FP values were converted to fraction bound by calculating [(FP_sample-FP_free)/(FP_no_comp-FP_free)], where FP_sample represents the FP value in the presence of test peptide; FP_free represents the value for free Alexa488-conjugated respective peptide; and FP_no_comp represents values in the absence of competitor peptide. We plotted fraction bound versus concentration of test peptide and fit the curve to the equation y = 1/(1 + [pep]/IC 50 ), where [pep] is the concentration of test peptide, y is the fraction of probe peptide bound at that concentration of test peptide, and IC 50 is the 50% inhibitory concentration of the test peptide.

Blood, PBMCs and HLA typing
Leukopaks (enriched apheresis product) were obtained from healthy donors by New York Biologics, Inc. (Southampton, NY) before the outbreak of SARS-CoV-2 (2014-2019). Whole blood from COVID-19 convalescent donors was collected under a protocol approved by the Medical School Institutional Review Board of the University of Massachusetts. Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque (Cytiva, Marlborough, MA) by density gradient centrifugation and frozen until use. The HLA type of leukopak (prepandemic) donors was determined in-house using the Protrans HLA typing kits (Protrans Medizinische Diagnostische Produkte GmbH, Hockenheim, Germany) or by The Sequencing Center (Fort Collins, CO). HLA type of COVID-19 donors was determined using a Nanopore protocol [97] or by the Histocompatibility Laboratory at UMass Memorial Medical Center (Worcester, MA). To help evaluate the likely exposure of the pre-pandemic donors to OC43, we evaluated T-cell and antibody responses to OC43 spike-derived antigens. Antibody responses were assessed in Ficoll-Paque upper layer by ELISA, using immobilized recombinant OC43 spike protein (SinoBiologicals, 40607-V08B) and goat anti-human IgG Fc-HRP (Bethyl, A8-104P). Titers for adult pre-pandemic donors are reported relative to a normal serum total IgG concentration of 4.7 mg/mL, along with values for a reference population of healthy children of 8-9 months of age. T-cell responses to overlapping peptide pools covering the spike proteins of OC43, 229E, NL63, HKU1, and SARS-CoV-2 (21 st Century Biochemicals, Marlborough, MA) were assessed by IFN-γ ELISpot assay as described below. Of 23 donors tested, 22 were positive for either OC43-specific IgG antibody or T-cell response to OC43 (S8 Table and

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 self-peptides (Self-1 [35]), and PHA-M (Gibco, Grand Island, NY) was used as a positive control. For ex vivo assays, PBMC were incubated with peptides or controls for~48 hours. We used 4x10 5 cells per well. For assays with cells expanded in vitro,~5x10 4 cells per well were incubated with an equal number of irradiated single allele APCs in the presence of peptides or controls for~18 hours. Two to four wells of each peptide, pool of peptides, or PHA-M, and at least 6 wells for DMSO were usually tested. Secreted IFN-γ was detected following the manufacturer's protocol. Plates were analyzed using the CTL ImmunoSpot Image Analyzer (Immu-noSpot, Cleveland, OH) and ImmunoSpot 7 software. Statistical analysis to determine positive responses was performed using the distribution-free resampling (DFR) method described by Moodie et al. [98].

Intracellular cytokine secretion assay (ICS)
ICS was performed using in vitro expanded T cells as previously described [35] with minor modifications. Briefly, single allele APCs were resuspended in CRPMI (w/o phenol red) +10% fetal bovine serum (FBS, R&D Systems) containing 1 μg/mL of each peptide and incubated overnight. On the day of the assay, T-cell lines were collected, washed, and resuspended in the same medium and added to the pulsed APCs (1:1 ratio); at this time, anti-CD107a-CF594 was added, followed by the addition of brefeldin A and monesin at the suggested concentrations (Golgi plug / Golgi stop, BD Biosciences, San Jose, CA). After 6 hours of incubation, cells were collected, washed, and stained using a standard protocol, which included: staining for dead cells with Live/Dead Fixable Aqua Dead Cell Stain Kit (Life Technologies, Thermo-Fisher Scientific, Waltham, MA); blocking of Fc receptors with human Ig (Sigma-Aldrich, St. Louis, MO); surface staining with mouse anti-human CD3-APC-H7, CD4-PerCPCy5.5, CD8-APC-R700, CD14-BV510, CD19-BV510, CD56-BV510; fixation and permeabilization using BD Cytofix/Cytoperm; and intracellular staining with mouse anti-human IFN-γ-V450, TNF-α-PE-Cy7, IL-2-BV650, (all from BD Biosciences, San Jose, CA). Data were acquired using a BD LRSII flow cytometer equipped with BD FACSDiva software (BD Biosciences, San Jose, CA) and analyzed using FlowJo v.10.7 (FlowJo, LLC, Ashland, OR). The gating strategy consisted in selecting lymphocytes and single cells, followed by discarding cells in the dump channel (dead, CD14+, CD19+, and CD56+ cells), and selecting CD3+ cells in the resulting population. Polyfunctional analysis was performed in FlowJo, defining Boolean combinatorial gates for all the markers in the CD3+/CD4+/CD8-population. These results were visualized in SPICE software v6.0 [99].

Peptides and HLA binding predictions
Peptides for these studies were obtained from 21 st Century Biochemicals (Marlborough, MA) and BEI Resources (Manassas, VA). Peptide sequences using in the assays are shown in Tables  1 and 2. HLA-peptide binding prediction was performed with NetMHCpan4.1 or NetMHCII-pan4.0 [58] for peptides eluted from MHC-I and MHC-II proteins, respectively. Sequence logo of predicted motifs obtained using Motif Viewer in NetMHCpan or NetMHCIIpan. The Immune Epitope Database IEDB [20] was used to search for T-cell responses to seasonal and pandemic coronavirus epitopes.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 Amsterdam I (NC_005831).Sequence alignment of spike, nucleoprotein, hemagglutinin esterase, and envelope proteins were generated using Clustal Omega v1.2.4 [100]. Conservation indices for each position of the alignment were calculated using the AL2CO algorithm [101] using the alignment previously generated and the default settings. Human Peptides sequences Eluted peptides sequences were searched against the whole human proteome to find potential human homologs. Whole OC43 sequence (with core epitope underlined), and differences in the other sequences are shown. For each alignment, the conservation score at each position was obtained using AL2CO algorithm [101] and presented as a bar graph, with the core epitope positions in black. C. Summary of conservation scores for each eluted peptide to their corresponding homolog peptides in other human coronaviruses. Scores normalized to 100% identity to OC43 peptide as 1, and no conservation as 0. An average per peptide is shown at the bottom of the heatmap. NA indicates no homolog protein between OC43 and the corresponding virus. (TIF)

S8 Fig. Assessment of T-cell reactivity to seasonal human coronaviruses in pre-pandemic donors.
Ex vivo T-cell responses to pools of S protein from OC43, HKU1, 229E, NL63, and SARS-CoV-2 in 21 pre-pandemic donors were measured using IFN-γ ELISpot. DMSO and Self-1 [35] were used as negative controls. Statistical analysis by DFR [98]; positive responses are shown by plus signs (red = DFR2x, blue = DFR1x). (TIF) S1 Table. Immunopeptidome of OC43-infected A549 cells. S1A. HLA-ABC immunopeptidome; S1B. OC43 immunopeptidome. For each peptide, mass spectrometry identification parameters are shown (eluted sequence, length, source protein, intensity, Scaffold identification probability, and Mascot Ion and Identity scores). In addition, NetMHCpan 4.1 or NetMHCIIpan 4.0 predictions were performed and predicted core for each relevant allele, score, and rank are shown for each peptide.

PLOS PATHOGENS
Naturally-processed T-cell epitopes of human coronavirus OC43 (S3F), and HLA-DP, (S3G) immunopeptidomes of uninfected cells. For each peptide, mass spectrometry identification parameters are shown (eluted sequence, length, source protein, intensity, Scaffold identification probability, and Mascot Ion and Identity scores). In addition, NetMHCpan 4.1 or NetMHCIIpan 4.0 predictions were performed and predicted core for each relevant allele, score, and rank are shown for each peptide.