Identification and relative abundance of naturally presented and cross-reactive influenza A virus MHC class I-restricted T cell epitopes

ABSTRACT

Cytotoxic T lymphocytes are key for controlling viral infection. Unravelling CD8⁺ T cell-mediated immunity to distinct influenza virus strains and subtypes across prominent HLA types is relevant for combating seasonal infections and emerging new variants. Using an immunopeptidomics approach, naturally presented influenza A virus-derived ligands restricted to HLA-A*24:02, HLA-A*68:01, HLA-B*07:02, and HLA-B*51:01 molecules were identified. Functional characterization revealed multifunctional memory CD8⁺ T cell responses for nine out of sixteen peptides. Peptide presentation kinetics was optimal around 12 h post infection and presentation of immunodominant epitopes shortly after infection was not always persistent. Assessment of immunogenic epitopes revealed that they are highly conserved across the major zoonotic reservoirs and may contain a single substitution in the vicinity of the anchor residues. These findings demonstrate how the identified epitopes promote T cell pools, possibly cross-protective in individuals and can be potential targets for vaccination.

KEYWORDS:

• Influenza virus
• CD8 T cells
• HLA class I
• T-cell epitope
• immunopeptidomics

Introduction

Efforts to control influenza virus infection are mainly based on the annual adaptation and reformulation of vaccines according to the predominant circulating strains. Influenza viruses exhibit a relatively high mutational rate since the hetero-trimeric RNA-dependent RNA polymerase (RdRp) complex lacks proof-reading activity [1,2]. Accumulating point mutations, in particular within the major surface protein, hemagglutinin (HA), constantly generate virus mutants and the exchange of vRNA segments upon infection with multiple strains also facilitates emerging variants [1,3]. Influenza vaccines against the H3N2 subtype have been implicated with poor protection, according to subtype measurements [4]. This could be due to the tremendous mismatch between circulating and egg-adapted strains used for vaccine production [5,6]. Broadly-neutralizing antibodies (bnAbs) against the major surface proteins, either the stalk or HA head domain, have great potential to protect against seasonal influenza virus infections and could provide near-universal protection [7,8], however, bnAbs-escape mutant viruses have recently been reported [9,10].

Cytotoxic T lymphocytes (CTLs) are key for controlling infectious diseases and cancer. Human leukocyte antigen (HLA) molecules are a determinant factor of CD8⁺ T cells recognition and can drive long-lasting memory post-priming with specific immunogenic epitopes [11–13]. In 2013, during the H7N9 outbreak in China, it was reported that the rapid recovery of patients coincided with early prominent H7N9-specific CD8⁺ T cell responses, while those patients with late CD8⁺/CD4⁺ T cell recruitment showed prolonged hospitalization [14]. In turn, in the absence of bnAbs, pre-existing immunity mediated by cross-reactive CD8⁺ T cells recognizing conserved epitopes may have the capacity to promote recovery from infection and minimize disease severity upon emergence of novel variants [11,12,15]. Given that current influenza vaccines lack CD8⁺ T and NK cell activation [16], broadly reactive vaccines that can efficiently induce CD8⁺ T cell immunity against different influenza virus strains and subtypes would be desirable particularly for high-risk populations. The optimal design of peptide-based vaccines is mainly attributed to unveiling the factors underlying the immunogenicity of peptides and the magnitude of the immune response. It is seemingly evident that the relative abundance of peptides presented by MHC class I plays a pivotal role, in particular, during the course of natural infection in triggering the immune response hierarchies [17,18].

Herein, we characterize the immunological landscape of CD8⁺ T cells recognizing different influenza A virus (IAV) strains at different time points after infection of Human lung adenocarcinoma cells (Calu-3). Using immunopeptidomics, novel HLA class I-restricted (HLA-I) IAV ligands were identified in the context of HLA-A*24:02, HLA-A*68:01, HLA-B*07:02 and HLA-B*51:01 alleles. The identified HLA-I peptides were further screened for their in vitro immunogenicity using peripheral blood mononuclear cells (PBMCs) obtained from healthy donors expressing the respective HLA alleles. Nine out of sixteen peptides were recognized by T cells from healthy individuals, with multifunctional effector functions (TNF, CD107a, IFN-γ) and five ligands proved immunodominant. A conservation analysis revealed that two epitopes were highly conserved among zoonotic reservoirs (≥95%), while the other epitopes contained minor substitutions, 1–2 amino acids (AA) away from the peptide-anchor residues, which hardly affect binding affinity. The identified epitopes can potentially provide cross-protection across distinct IAV strains and subtypes. Moreover, the incorporation of such highly conserved CD8⁺ T cell epitopes into a broadly reactive T cell-inducing vaccine may achieve long-term heterosubtypic protection and match newly emerging strains.

Materials and methods

Cells and viruses

Calu-3 cells (ATCC® HTB-55) were cultured in Minimum Essential Media (MEM) supplemented with 1% Penicillin/Streptomycin (P/S), 10% fetal bovine serum (FBS), 2 mM L-glutamine, 1 mM sodium pyruvate and 1% non-essential AA. Calu-3 cells express HLA-A*24:02, HLA-A*68:01, HLA-B*07:02, HLA-B*51:01, and HLA-C*15:02 alleles. The HLA typing information of the Calu-3 cell line was retrieved from the TRON Cell Line Portal [19]. Madin-Darby canine kidney cells (MDCK II, ATCC® CRL2936™) were grown in IMDM medium and used for IAV propagation. The following IAV strains were used: A/Puerto Rico/8/34 (H1N1/PR8), A/Victoria/03/75 (H3N2/Victoria), and A/Fukui/20/2004 (H3N2/Fukui). For virus stocks preparation, MDCK II cells were infected with each strain and incubated for 1 h at 37°C. The inocula were removed, and cell monolayers overlaid with infection media (1X DMEM supplemented with 1% P/S, 0.3% bovine serum albumin (BSA), and 1 µg ml⁻¹ TPCK-treated trypsin). After 48–72 h post infection (hpi), cell culture supernatants containing viruses were clarified by centrifugation, aliquoted, and stored at −80°C. Virus titres were determined by standard plaque assay protocol.

Direct infection of cells

Approximately 1 × 10⁸ Calu-3 cells were infected with each of the IAV strains at an MOI of 4 and the infection rate was determined as described below to ensure that a minimum 50% of the cells were infected. After 1 h incubation at 37°C, the inocula were removed, cell monolayers washed with phosphate-buffered saline (PBS), and overlaid with infection media (1X DMEM supplemented with 1% P/S, 0.3% BSA, and 1 µg ml⁻¹ TPCK-treated trypsin). At 3, 6, 9, 12 and 24 hpi, the cells were harvested using enzyme-free cell dissociation buffer (Gibco), washed twice times with cold PBS, and stored at −80°C.

Determination of infection rate and HLA-I surface expression

HLA-I surface expression was evaluated using the QIFIKIT quantitative flow cytometric assay kit (Agilent Dako, Denmark), according to the manufacturer’s instructions. In brief, Calu-3 infected cells were stained with the pan-HLA-I-specific monoclonal antibody (mAb) W6/32 (produced in-house) in triplicates or IgG isotype control (BioLegend, San Diego, CA, USA). Subsequently, secondary staining with FITC-conjugated rabbit-anti-mouse F(ab’)2 fragments (Agilent Dako) was carried out alongside QIFIKIT quantification beads (Agilent Dako). For the infection efficiency rate, infected cells were fixed and permeabilized using Foxp3 Staining Buffer Set (eBioscience™, Invitrogen, CA, USA) followed by intracellular staining of IAV nucleoprotein (NP) step with FITC-conjugated anti-IAV NP monoclonal antibody (clone D67J; Invitrogen). Flow cytometric acquisition was performed on a BD FACSCanto II system (BD Biosciences, Franklin Lakes, NJ, USA), and the acquired data were analysed by FlowJo 10.3 (FlowJo LLC, BD). The complete workflow is illustrated in (Figure 1a).

Figure 1. Infection rates and MHC-I expression in Calu-3 cells. Cells were mock-treated or infected with different IAV strains. (a) Schematic representation of MHC-I quantification and evaluation of IAV infection rate. (b) Cell surface expression of MHC-I molecules was quantified by staining with pan-HLA-I antibody (W6/32) and analysed by flow cytometry at 24 hpi. (c) Alterations of the MHC-I surface expression were evaluated using QIFIKIT quantitative bead-based assay. Statistical differences were assessed using Brown-Forsythe ANOVA with Dunnett’s T3 multiple comparisons correction (d) Post infection, the cells were permeabilized for intracellular staining of IAV nucleoprotein (NP). The results are representative of two independent experiments (n = 2).

Immune precipitation of HLA-I-peptide complexes and peptide isolation

HLA-I molecules from cell pellets were isolated using the standard immunoaffinity purification procedure as previously described [20,21]. Briefly, frozen cells were lysed using 10 mM CHAPS (AppliChem, Darmstadt, Germany) buffer prepared in PBS, containing a complete protease inhibitor (Roche, Mannheim, Germany). The cell lysates were homogenized by applying pulsed sonification. Thereafter, the HLA-I molecules were purified from lysates by immunoaffinity chromatography with the pan-HLA class I specific W6/32 antibody coupled to CNBr-activated sepharose (GE Healthcare, Chicago, IL, USA). HLA-associated peptides were eluted with 0.2% TFA followed by ultrafiltration of the eluate using 3-kDa Amicon filter units (Merck Millipore, Billerica, MA, USA). Desalting and concentration steps were accomplished by ZipTip C18 (Merck Millipore) and 0.1% TFA, and elution was performed with 32% Acetonitrile (AcN)/0.2% TFA. The final volume of the eluate was reduced by vacuum centrifugation.

Characterization of naturally presented HLA-I ligands by LC-MS/MS

Purified HLA ligands were separated by reverse-phase high-performance liquid chromatography (HPLC; UltiMate 3000 RSLCnano system, Dionex, Sunnyvale, CA, USA) using a 75 μm × 2 cm trapping column and a 25 cm length and 50 μm separation column (Thermo Fisher Scientific, Waltham, MA, USA) subsequently, applying a gradient ranging from 2.4–32% AcN over 90 min with a flow rate of 175 nl/min. For MS/MS, a top-five collision-induced dissociation (CID) method generating ion trap MS/MS spectra in the mass range 400–650  m/z was used with an online coupled mass spectrometer (LTQ Orbitrap XL, Thermo Fisher Scientific).

Database search and spectral annotation

Data were processed against concatenated FASTA sequences containing the reviewed UniProt human proteome as well as the IAV proteome of the strains used for this study (viewed January 2020) by applying the Sequest HT algorithm in the Proteome Discoverer 1.4 software (Thermo Fisher Scientific). Precursor mass tolerance was set to 5 ppm, fragment mass tolerance was set to 0.5 Da, and oxidized methionine was allowed as a dynamic modification. The percolator-assisted false discovery rate (FDR) algorithm [22] was implemented at a target value of q ≤ 0.05 (5% FDR), search engine rank  = 1, and peptide lengths were limited to 8–12 AA. All identified peptides were annotated to their respective HLA motifs using both NetMHCpan 4.0 [23] and SYFPEITHI [24] databases. Cutoff values of IC50 ≤ 500 nM, percentile rank ≤ 2% for NetMHCpan 4.0, and a normalized SYFPEITHI score ≥ 50% were applied. Peptides fulfilling the cut-off criteria of either or both prediction tools were designated as predicted and tested in PBMCs from healthy donors with the respective HLA allele.

Label-free quantification of HLA ligands

To determine the abundance of presented IAV-derived ligands over the infection time course, the label-free quantification (LFQ)-method was used as previously described by Nelde et al. [25]. Briefly, prior to the LC-MS/MS analysis, the total number of cells for each infection time point was normalized and LC-MS/MS analysis was performed in five technical replicates for each sample. The relative quantification of HLA ligands was performed by calculating the area under the curve (AUC) of the corresponding precursor-extracted ion chromatograms using Proteome Discoverer 1.4 (Thermo Fisher Scientific). To contend with the common LFQ and data-dependent acquisition (DDA) MS proteomics issues [26–28], the LFQ strategy was implemented by lowering FDR cut-offs and using matching between runs to reduce missing values in quantitation. High-quality peptide spectrum matches were filtered for 5% FDR and subsequently screened for binding affinity to the respective HLA molecules.

Peptide synthesis

Peptides were synthesized and provided by the Department of Immunology, Tübingen, Germany. Peptide synthesis was performed by the standard 9-fluorenylmethyloxycarbonyl/tert-butyl (Fmoc/tBu) method using an automated peptide synthesizer (Liberty Blue, CEM, Matthews, NC, USA). The identity and purity of peptides were evaluated by reversed-phase HPLC (e2695, Waters, Milford MA, USA) coupled online hybrid mass spectrometry (LTQ Orbitrap XL, Thermo Fisher Scientific). Synthesized peptides were used in functional experiments.

Human biomaterials and ethics statement

Samples from healthy blood donors with matching HLA alleles were kindly provided by the Institute of Clinical and Experimental Transfusion Medicine (University of Tübingen) after obtaining informed consent documented in writing, according to the principles of the Declaration of Helsinki and conforming with applicable laws and regulations. This project has been reviewed and approved by the Ethics Committee at the Medical Faculty of the Eberhard Karls University Tübingen and the University Hospital Tübingen (Project no. 887/2020BO2 and 200/2021BO2).

T cell culture

Peripheral Blood Mononuclear Cells (PBMCs) were isolated by SepMate™ tubes for layering Lymphoprep™ density gradient media (STEMCELL Technologies, Vancouver, Canada). Obtained cells were cryopreserved and stored in the vapour phase of liquid nitrogen. Vials of frozen PBMCs were thawed at 37°C and cells rested overnight before stimulation. Cell culture was performed with IMDM media supplemented with 5% heat-inactivated human serum (Sigma-Aldrich, Steinheim, Germany), 1% P/S, 25 µg ml⁻¹ gentamicin, and 50 µM β-mercaptoethanol with 5% CO₂ at 37°C in a humidified incubator.

IFN-γ ELISpot assay

IFN-γ ELISpot assays were performed after 12-day peptide stimulation (as described below) along with suitable controls. In contrast, ex vivo ELISpot was carried out 24 h after thawing the cells. Briefly, for IFN-γ ELISpot PBMCs were stimulated for 24 h after thawing with 1 µg ml⁻¹ IAV peptides of interest and control peptides. IL-2 (R&D Systems, Minneapolis, MN, USA) was added at a final concentration of 20 IU ml⁻¹ on days 2, 5, and 7. On day 12, cells were harvested and IFN-γ ELISpot was performed by seeding PBMCs at a final count of 10⁵ cells/well onto MultiScreen_HTS-HA nitrocellulose plates (Millipore). Subsequently, cells were incubated for 24 h following stimulation with candidate peptides and controls. Phytohemagglutinin (PHA) (Sigma-Aldrich), at a final concentration of 15 µg ml⁻¹ was used as a positive control. The following peptides were used as negative controls: DPYKATSAV human MUC16_6326–6334 (B*51); TPGPGVRYPL HIV Nef_128–137 (B*07), GSEELRSLY HIV POL_71–79 (A*01), RLRPGGKKK HIV GAG_20–28 (A*03). DMSO was used as a negative control for HLA-A*68:01 and HLA-A*24:02. The spot counts corresponding to IFN-γ producing cells were determined using ImmunoSpot® S5 analyzer (CTL, Cleveland, OH, USA) and ImmunoSpot software® ver. 5 (CTL). T cell responses were considered positive when results fulfilled the following criteria: each test well contains at least 10 spots, and the mean number of replicates is threefold higher than the mean spot count of the negative control. According to the recognition frequency among tested donors, the tested peptides were grouped into 3 different categories: negative group (0% response rate in any donor), subdominant group (<50% recognition frequency), and dominant group (≥50% recognition frequency).

Intracellular cytokine staining

The functionality of peptide-specific T cells was analysed by ICS. PBMCs were stimulated with individual peptides at a final concentration of 10 µg ml⁻¹ and incubated for 12 h in the presence of GolgiStop (BD Biosciences), Brefeldin A (Sigma-Aldrich), and FITC mouse anti-human CD107a mAb (1:100, clone H4A3, BD Bioscience). Ionomycin and PMA (Sigma-Aldrich) served as a positive control. Thereafter, cells were stained for surface markers with PerCP anti-human CD8a (1:100 dilutions, clone RPA-T8, BioLegend) and APC-Cy™7 mouse anti-human CD4 (1:100, clone RPA-T4, BD Bioscience) mAb in addition to using LIVE/DEAD™ Fixable Aqua Stain to discriminate viable cells (1:400 dilutions, Invitrogen, Waltham, MA, USA). After incubation, a fixation/permeabilization step with Cytofix/Cytoperm solution (BD Biosciences) was added and the cells were further stained with PE anti-human IFN-γ (1:200 dilutions, clone B27, BioLegend) and Pacific Blue™ anti-human TNF-α (1:120, clone MAB11, BioLegend) mAb. Data were acquired by BD FACSCanto II flow cytometry (BD Biosciences) using FACS DIVA software (BD Biosciences). The gating strategy applied for the evaluation of flow cytometry-acquired data is provided in Supplementary materials (suppl. Figure 1).

HLA-Tetramer staining

The frequency of peptide-specific CD8⁺ T cells was evaluated by tetramer staining after incubation of PBMCs with respective PE-labeled HLA-peptide tetramer complexes. Briefly, PBMCs were cultured, as described above, and the HLA-multimers (ImmunAware, Hørsholm, Denmark) were added at a final concentration of 30 nM, according to the manufacturer’s instruction and incubated for 20 min at room temperature. Cells were further washed twice with PBS and stained with LIVE/DEAD™ Fixable Aqua Stain (1:400 dilutions, Invitrogen) for 20 min at 4°C. The antibody cocktail of cell surface markers PerCP anti-human CD8a (1:100 dilutions, clone RPA-T4, BioLegend) and APC-Cy™7 mouse anti-Human CD4 (1:100, clone RPA-T4, BD Bioscience) mAb was added for 20 min at 4°C. Following mAb incubation, cells were analysed by flow cytometry (FACSCanto II, BD Biosciences). The gating strategy applied for the evaluation of flow cytometry-acquired data is provided in Supplementary materials (suppl. Figure 2). Due to sample constraints, staining was carried out only for peptides that elicited a high recognition frequency in the ELISpot assay.

Figure 2. Identification and presentation kinetics of IAV-derived ligands. (a) Schematic representation of epitope mapping workflow. Calu-3 cells were infected with different IAV strains representing both H1N1 and H3N2 subtypes. HLA-peptide complexes were isolated from the cell lysates of infected cells at 3, 6, 9, 12, and 24 hpi by immunoaffinity chromatography with the pan-HLA class I-specific W6/32 antibody coupled to CNBr activated sepharose. HLA-peptide complexes were eluted by acidic elution followed by desalting and concentration steps. The HLA ligands were further analysed by HPLC tandem mass spectrometry (MS/MS). (b-d) The presentation kinetics of HLA-I-bound peptides isolated from infected Calu-3 cells during the infection time course are shown in the upper panel and the area under the curve (AUC) is presented as a percentage of maximum levels detected over the time course (lower panel). Dashed lines indicate the immunodominant epitopes. The MS/MS data were processed using proteome discoverer software V. 1.4, analysed, and visualized using GraphPad prism software ver. 9.3.

Conservation analysis

To assess the conservation of the immunogenic peptides among human, swine, and avian influenza viruses, the exact sequence match of each peptide was evaluated in unique protein sequences of HA, MP, NP, PB1, and PB2 of all IAV subtypes retrieved from influenza research database (IRD) and National Center for Biotechnology Information (NCBI). Sequence homology and AA substitution rates were estimated using the IRD [29]. To visualize the AA substitutions, full-length sequences of MP, PB2, PB1, NP, and HA genes were aligned using MAFFT ver. 7 [30,31]. The phylogenetic relationship was evaluated with the Neighbor-Joining method. The evolutionary distances were computed using the Jukes-Cantor model, and the variation rate among sites was modelled with a gamma distribution. Bootstrap values were calculated out of 1000 replicates. Trees were visualized and annotated using Evolview ver. 3 [32]. Sequence logos were generated using Seq2Logo-2.1 [33] by applying the default setting; logo type Kullback–Leibler with colouring scheme by physicochemical properties and Hobohm1 clustering method.

Software and statistical analysis

Proteome Discoverer 1.4 (Thermo Fisher Scientific) software was used to analyse the MS/MS data. Mapping of the immunogenic T cell epitopes onto IAV proteins was performed using UCSF ChimeraX software [34]. ELISpot data were analysed with ImmunoSpot software® ver. 5 (CTL). Flow cytometric data were analysed using FlowJo version 10 (FlowJo LLC, BD). Workflow illustrations were created with BioRender.com Alterations of the HLA-I surface expression were evaluated using Brown-Forsythe ANOVA with Dunnett’s T3 multiple comparisons correction. Statistical analysis was conducted using GraphPad Prism ver. 9.3 and p values <0.05 were considered statistically significant. Where applicable, statistical details for each experiment are described in the corresponding figure legends.

Results

Identification of IAV-derived HLA-I ligands

Although many IAV-derived CD8⁺ T cell epitopes have been previously identified, their identification was primarily achieved by epitope prediction and screening of T cell responses. Here, a mass spectrometry-based (LC-MS/MS) immunopeptidomics approach was applied to increase the spectrum of MHC-I-presented IAV-derived peptides after infection of a human lung adenocarcinoma cell line (Calu-3).

To assess the features and kinetics of the naturally presented HLA-I peptides, the immunopeptidomes of Calu-3 cells were examined at different time points post infection with different IAV strains. Different viruses downregulate MHC-I surface expression [36–39]. Thus, as a first step, the HLA-I surface expression profile and the infection rates were determined (Figure 1a–d). The expression of HLA-I molecules was similar after infection with H1N1/PR8 and H3N2/Fukui strains with mean expression levels of 246,847 (PR8) and 242,753 (Fukui) molecules/cell compared to the untreated Calu-3 cells (263,502 molecules/cell), despite the variance in infection rates. Whereas a significant (p = 0.0046) downregulation (221,341 molecules/cell) was observed after infection with H3N2/Victoria strain. We assume that the viral interference with the HLA-I expression may be virus/subtype-specific.

Calu-3 cells were infected with either H1N1/PR8, H3N2/Victoria and H3N2/Fukui strains followed by isolation of naturally presented HLA-I peptides and LC-MS/MS assessment at time points 3, 6, 9, 12, 24 hpi (Figure 2a). A total of 16 IAV-derived MHC-I peptides were identified as predicted binders for HLA-A*24:02, HLA-A*68:01, HLA-B*07:02, and HLA-B*51:01 alleles (suppl. Table 1). Among these newly characterized peptides, four were derived from the HA protein, one each from M1, NS1, NS2, and PB1-F2 proteins, two from NP protein, and three each from PB1 and PB2 (suppl. Table 1). The M1_EAM peptide was identified both in H1N1/PR8 and H3N2/Victoria infected cells. HA_VYR was determined in H3N2/Victoria and H3N2/Fukui infected cells, whilst the rest of the HLA-I peptides were solely identified in either of the IAV subtype-infected cells.

Kinetics of IAV peptide presentation following infection of Calu-3 cells

Next, we evaluated the kinetics of HLA presentation and relative abundance of the sixteen IAV-derived peptides during infection (Figure 2a). After H1N1/PR8 infection, eight peptides were determined. Three peptides were identified at 6 hpi (HA_LPY, NS1_TMA, and PB2_GTA), suggesting that neither structural nor non-structural IAV protein-derived peptides have presentation preferences at early time points, whilst most of the remaining peptides could only be determined at 9 hpi, e.g. the NP_NLN epitope was first identified at 12 hpi (Figure 2b).

In Calu-3 cells, infected with the H3N2/Victoria strain, eight peptides were identified. Already at 3 hpi, PB1-F2_STE, PB2_STS, and PB2_TTV peptides were presented, while most of the remaining peptides were first identified at 9 hpi. All peptides showed peak presentation at 12 hpi, except for the PB2_TTV peptide, which peaked at 9 hpi. Interestingly, the M1_EAM peptide was presented by the H1N1/PR8 and H3N2/Victoria infected cells but showed different presentation profiles in both earlier time points and altered peak expression (Figure 2b, c).

Only two peptides were identified after infection with strain H3N2/Fukui. The NP_DAT peptide was first observed at 6 hpi and reached its peak level of presentation at 12 hpi. Notably, the HA_VYR epitope that was also presented by Calu-3 cells infected with H3N2/Victoria was observed only at 24 hpi (Figure 2d).

Overall, most peptides were first identified between 6–9 hpi and peaked at 12 hpi. Not all viral protein segments were represented among the identified peptides. Taken together, the frequency and presentation kinetics showed a disparity between different peptides.

IFN-γ ELISpot assay establishes IAV-derived HLA-I ligands as T cell epitopes

All IAV peptides were analysed with either SYFPEITHI or NetMHCpan 4.0 to assign their putative HLA-I restriction (suppl. Table 1). Thereafter, the peptides were tested for immunogenicity in at least eight different HLA-matched healthy blood donors using a 12-day IFN-γ ELISpot assay.

NP_DAT and HA_LPY were found to be the most immunogenic ligands for HLA-B*51 with 100% and 85.7% recognition rates, respectively (Figure 3a-b; suppl. Table 1). Moreover, HA_VYR, the only identified HLA-A*24-restricted ligand showed a 76.9% recognition rate (Figure 3c, d; suppl. Table 1), while the M1_EAM and PB2_TTV were proven immunodominant epitopes for HLA-A*68:01 and revealed a similar recognition frequency 66.7% (Figure 3e-g; suppl. Table 1). PB2_GTA and PB2_STS, spanning the same AA positions with two AA substitutions, were found to be subdominant epitopes for HLA-A*68:01 and presented quite similar recognition rates accounting for 41.7% and 33.3% respectively, as well as the PB1_ETM exhibiting a recognition rate at 33.3%. It was observed that the NP_NLN epitope was least recognized at a rate of 16.7% among the other HLA-A*68:01 epitopes, whilst the remainder of the identified peptides did not show immunogenic properties.

Figure 3. Immunological characterization of naturally presented IAV-derived T cell epitopes by IFN-γ ELISpot assay. Representative ELISpot assays after 12-day in vitro stimulation of PBMCs isolated from HLA-B*51 (a-b), HLA-A*24 (c-d), and HLA-A*68 (e-g) -matched donors. Bars indicate mean spot counts of technical duplicates. The cumulative analysis of screening results was summarized in scatter plots (b), (d), and (g). Each data point represents one single donor tested within one single experiment. Shown are the mean IFN-γ spot forming cells (SFCs) from two technical replicates of each tested donor normalized to the negative control. Irrelevant HLA-matched peptides, either HIV-derived peptides, human self-peptides, or DMSO served as a negative control. Horizontal lines represent the mean values of all tested donors. Positively tested donors are depicted by red circular shapes and negatively tested donors are shown by grey circular shapes. (h) Comparison between IFN-γ SFCs from ELISpot assays in an ex vivo setting and after 12-day stimulation. Four to five donors that showed a high T cell response in the 12-day ELISpot assay for the dominant epitopes NP_DAT, HA_LPY (B*51), and HA_VYR (A*24) were retested in the ex vivo system. Paired data points represent one single donor tested within one single experiment. HLA-A*68 restricted dominant epitopes were not included due to sample limitations. HA_LPY, as HLA-B*07 binder, was also included in the analysis. Shown are the mean IFN-γ SFCs from two technical replicates of each tested donor normalized to the respective negative control. Data points in blue colour represent donors that showed T-cell responses in both experimental settings. Data were analysed and visualized using GraphPad Prism software ver. 9.3.

To preclude potential competing effects during the 12-day stimulation assay among the tested peptides, ex vivo IFN-γ ELISpot assay was performed (Figure 3h). Four to five donors that showed a strong T cell response in the 12-day ELISpot assay for the immunodominant epitopes NP_DAT, HA_LPY (B*51) and HA_VYR (A*24) were reassessed with the ex vivo system. However, we were unable to test the HLA-A*68:01 restricted immunodominant epitopes due to limited availability of sample materials. The results showed that the memory T cell response was either lacking or marginal, when compared with the 12-day setting upon pre-stimulation (Figure 3h). Moreover, the NP_DAT epitope only elicited low but detectable responses in one HLA-B*51:01 positive donor when compared to the respective results after 12-day stimulation. Altogether, in contrast to the minimal IFN-γ production towards the immunogenic peptides in the ex vivo system, the 12-day culture intensified the CD8⁺ T cell immune response substantially, insofar as the epitope-specific memory CD8⁺ T cells were shown as expandable.

Functional assessment of IAV-specific memory T cells

Next, HLA tetramer staining alongside intra-cellular cytokine staining (ICS) was carried out to assess the multi-functionality and peptide-specificity of T cells previously investigated by ELISpot assays.

Using ICS, TNF, IFN-γ, and CD107a were determined to assess the multifunctionality of memory T cells, following stimulation with IAV-peptides (Figure 4a-c, e-h; Suppl. Fig. 3 a). IFN-γ-specific T cell populations responding to the HLA-A*68-restricted peptides, were 0.1−2.1%, (Figure 4e). Likewise, the HLA-A*24-restricted peptides elicited T cells in a similar frequency (Figure 4f). Moreover, the HLA-B*51:01 immunodominant epitopes revealed specific T cell populations ranging from 0.1–3.6% for the HA_LPY peptide and reaching 17.2% for the NP_DAT peptide (Figure 4g). Notably, the NP_DAT epitope elicited the strongest CD8⁺ T cell responses, corroborating respective ELISpot results.

Figure 4. IAV-derived T cell epitopes are multifunctional. (a-d) Representative intracellular IFN-γ, TNF-α, and CD107a staining of PBMCs isolated from healthy HLA-matched donors exhibiting HLA-B*51 (a, b), HLA-A*24 (c), after 12-day stimulation with IAV-derived HLA class I peptides evaluated by flow cytometry. (d) representative tetramer staining after 12-day amplification of CD8⁺ T cells, derived from HLA-matched donors; HLA-B*51 and HLA-A*24. Irrelevant tetramers were used as a negative control. (e-g) Box and whisker plots represent a cumulative percentage of CD8⁺ IFN-γ⁺, CD8⁺ IFN-γ⁺ TNF-α⁺, or CD8⁺ CD107a⁺ (h) and the frequency of epitope-specific CD8⁺ T cells evaluated by tetramer staining. Boxes extend from the 25th to 75th percentiles, whiskers represent minimum to maximum, and the horizontal plotted lines are the median values. The indicated percentages represent the frequency of T-cell responses post stimulation with the test peptide minus the negative control of the respective donor. Each data point represents one single donor tested within one single experiment. Data were analysed and visualized using GraphPad Prism software ver. 9.3 and FlowJo software ver. 10.3. The gating strategies applied for the flow cytometry-based analysis presented in this figure are depicted in the supplements (suppl. Figures 1 and 2).

HLA multimer staining was carried out for all but one epitope, NP_NLN, due to sample constraints (Figure 4d, h; Suppl. Fig. 3 b). Robust epitope-specific recognition was demonstrated for all tested tetramers except for PB2_GTA and PB2_STS. The strongest response was consistently observed for NP_DAT, reaching up to 19.3% tetramer-positive CD8+ T cells, followed by M1_EAM, HA_VYR and HA_LPY (eliciting 3.4−4.4%).

Generally, the predicted HLA restriction was confirmed by tetramer staining for all immunogenic peptides. The multifunctional T cell recognition frequencies for IAV-derived epitopes varied profoundly between donors as well as peptides. Moreover, it was shown that the T cell immune response evoked by IAV-specific epitopes was mainly driven by CD8⁺ T cells in all tested donors.

HLA cross-reactivity and mismatch analysis

The NetMHCpan 4.0 prediction results revealed that some peptides were assigned to different HLA-I allotypes (suppl. Table 2). To avoid potential false positives, the following restrictions were applied for donor selection: HLA-B*07⁺/B*51⁻, for peptide HA_LPY, and HLA-A*03⁺/A*68:01^- for peptides HA_NSN and PP2_TTV, whilst peptide PB1-F2_STE was tested in two different settings HLA- A*01⁺/A*03^-/A*68:01^- and HLA-A*01^-/A*03⁺/A*68:01^-. The 12-day ELISpot and ICS results revealed that all peptides with putative HLA-A*68 restriction proved unable to bind to more than one HLA allotype (Figure 5a,b). Interestingly, HLA-B*07:02⁺ donors showed strong IFN-γ immune response in ELISpot with 45% T cell recognition frequency against HA_LPY peptide (Figure 5c). Moreover, the ICS data corroborated the 12-day ELISpot assay results (Figure 5d,e) and demonstrated that the immune response was mediated by CD8⁺ T cells. The HA_LPY tetramer staining results were positive in eight of the tested donors (Figure 5f,g). HA_LPY epitope was further tested by ex vivo ELISpot assay in 5 donors, who showed a strong T cell response after 12-day stimulation. The result revealed that none of the donors elicited immune response (Figure 3h). Ultimately, the ELISpot as well as the ICS results demonstrated that peptides that induced T cell responses are restricted to one distinct HLA-I allotype with the exception of HA_LPY peptide that is restricted to two prominent HLA-I allelic products (B*51 and B*07), providing broader coverage in population harbouring at least one of these alleles.

Figure 5. Immunological characterization of naturally presented IAV-derived T cell epitopes by IFN-γ ELISpot assay. Representative ELISpot assay after 12-day in vitro stimulation of PBMCs isolated from HLA-A*01 (a), HLA-A*03 (b), and HLA-B*07 (c) -matched donors. Bars indicate mean spot counts of technical duplicates. The cumulative analysis of screening results was summarized in scatter plots (a-c). Each data point represents one single donor tested within one single experiment. Shown are the mean IFN-γ spot forming cells (SFCs) from two technical replicates of each tested donor normalized to the respective negative control. Horizontal lines represent the mean values of all tested donors. Positively evaluated donors are depicted by red circular shapes and negatively tested donors are shown by grey circular shapes. (d-e) Representative intracellular IFN-γ, TNF-α, and CD107a staining of PBMCs obtained from healthy HLA-matched donors (HLA-B*07), and tetramer staining (f-g) after 12-day amplification of CD8⁺ T cells derived from HLA-matched donors. Box and whisker plots represent a cumulative percentage of CD8⁺ IFN-γ⁺, CD8⁺ IFN-γ⁺ TNF-α⁺ or CD8⁺ CD107a⁺ (e) and the frequency of epitope-specific CD8⁺ T cells evaluated by tetramer staining (g). Boxes extend from the 25th to 75th percentiles, whiskers represent minimum to maximum, and the horizontal plotted lines are the median values. The indicated percentages represent the frequency of T-cell responses post stimulation with the test peptide minus the negative control of the respective donor. Each data point represents one single donor tested within one single experiment. The gating strategies applied for the flow cytometry-based analysis presented in this figure are depicted in the supplements (suppl. Figures 1 and 2).

Epitope conservation analysis across the zoonotic reservoir

Sequence variation of IAV internal proteins exists among the zoonotic reservoirs, even though they exhibit a higher degree of conservation compared to the surface HA and NA proteins.

Therefore, a point mutation analysis was conducted for each epitope in more than 30,000 human IAV sequences, 6000 swine IAV sequences, and 10,000 avian IAV sequences. It was observed that the most highly conserved epitope was PP2_TTV; accounting for 100%, 96.8%, and 99% similarity in human, swine, and avian IAV sequences respectively, with total sequences conservation in 62,503 of 62,938 (99.3%) (Figure 6a–f).

Figure 6. HLA-I epitope conservation analysis in human, swine, and avian zoonotic reservoirs. Bar graphs represent the number of identical sequences (blue), in all available IAV strains sequences (black) in avian (a), swine (b), and human (c). The conservation score (%) is depicted as superimposed red triangles. The distribution of conserved sequences across IAV subtypes is shown in heatmaps for avian (d), swine (e), and human sequences (f). The colour scale represents the conservation index (%), which is indicated as 0% (white colour) to 100% (Dark slate grey colour); the grey colour indicates no sequences were available. Sequences with unknown hemagglutinin (H) or neuraminidase (N) were not included in the heatmaps. Red boxes indicate avian strains associated earlier with potential pandemics. Data were visualized using GraphPad Prism software ver. 9.3.

Peptide NP_NLN showed the second-leading conservation rate; at 92.4% in human, and 98% in both swine and avian IAV strains (Figure 6a–f). In terms of immunogenicity, it was determined to be the least immunogenic peptide by ELISpot assay (Figure 3g, suppl. Table 1). The most immunogenic peptides, NP_DAT and HA_VYR, (suppl. Table 1, Figure 3d–g) were poorly conserved, especially concerning avian IAV sequences (1.8% and 2.4%, respectively), but they had dissimilar conservation scores accounting for 42.5% and 98.6% in human IAV strains; 82% and 42.3% in swine IAV strains, respectively (Figure 6a–f). Conversely, M1_EAM had nominal sequence identity in human and swine IAV strains, while it was highly conserved in avian IAV strains with a 79.4% homology score (Figure 6a–f). Moreover, HA_LPY comprised at least one amino acid substitution leading to a minimal conservation rate of < 0.1% in each reservoir. Similarly, PB1_ETM was not conserved at all among swine IAV strains, while it showed conservation scores of 70% and 46.9% in both avian and human IAV strains, respectively.

Although PB2_GTA and PB2_STS showed comparable response rates by ELISpot assay to HLA-A*68:01, both peptides had identical sequences with two interchangeable AAs (suppl. Table 1). Moreover, PB2_GTA proved to be highly conserved in avian and swine IAV strains but had a lower conservation score in human strains. Inversely, PB2_STS had a minimal number of identical sequences in avian and swine IAV strains but exhibited an increased conservation rate in human sequences (Figure 6a–f).

Considering the potential for a pandemic threat of the H5N1, H9N2, and H7N9 subtypes, regions on heatmaps were marked by red boxes indicating the conservation of the tested epitopes in these highly relevant strains (Figure 6d–f).

Amino acid substitutions close to anchor residues do not affect the binding of the IAV peptides under consideration

Next, we extended the conservation analysis to assess AA substitutions aiming to reveal the mutational dissimilarities and to estimate the influence of respective substitutions on peptide binding affinity. The positions of the immunogenic peptides were also mapped onto the crystal structure of IAV proteins (Figure 7a–f).

Figure 7. Mapping of the immunogenic T cell epitopes onto IAV proteins. The location of T cell epitopes is mapped on the crystal structure of IAV proteins obtained from the Protein Data Bank (PDB). (a) HA_LPY (1RU7), (b) HA_VYR (6CEX), (c) NP_DAT and NP_NLN (2IQH), (d) M1_EAM (7JM3), PB1_ETM (6RR7), and (f) PB2_GTA/STS and PB2_TTV. Molecular graphics and analyses were performed with UCSF ChimeraX software.

Notably, peptides NP_NLN and PB2_TTV carried one single mutation in a nominal number of human H2N2 and swine IAV sequences respectively, which resulted in different peptide sequence variants, although those peptides showed the highest degree of homology (Figure 8a,b). In both peptides, most substitutions were I30 V or K33R for peptide PB2_TTV and A146 T for peptide NP_NLN (Suppl. Fig. 4 a-b). These substitutions were in the vicinity of the N- and C-terminal anchor residues and did not affect the peptide binding affinity to respective HLA molecules as predicted by the NetMHCpan algorithm.

Figure 8. Amino acid substitutions do not affect the peptide anchor residues. Details of the substitutional divergence of all identified epitopes are depicted in the phylogenetic trees (a-i). Sequence homology and amino acid substitution rates were estimated using the influenza research database by applying the point mutation analysis algorithm. The phylogenetic relationship was evaluated with the Neighbor-Joining method. The evolutionary distances were computed using the Jukes-Cantor model, and the variation rate among sites was modelled with a gamma distribution. Epitopes with identical sequences are designated in green; the number of amino acid substitutions is colour-coded, and red colour represents sequences that harbour more than 3 substitutions. Major IAV lineages and strains associated earlier with potential pandemics are indicated.

For the highly immunogenic NP_DAT peptide, sequence variations are mainly restricted to one AA substitution in human subtypes, (H3N2 and H2N2) and most avian IAV strains, while two AA substitutions were also observed, to a lesser extent, in the avian subtypes (Figure 8c). The predominant substitutions were A131S as well as the I136L/M at the C-terminal anchor residue, but these amino acids are tolerated at the P9 position of the peptide binding groove and the ligand sequence variants harbouring these substitutions exhibit strong binding affinity. A limited number of deleterious AA changes A128 T and I136R were observed at both anchor residues of the NP_DAT peptide (Suppl. Fig. 4 c). Likewise, a comparable substitution pattern was observed for PB1_ETM attributed to single AA changes in human (H3N2, H2N2), avian (H5N1, H7N9, H9N2), and most swine strains (Figure 8d). The prevailing substitutions were M111I and V113 K/A without influence on binding affinity (Suppl. Fig. 4 d).

Peptide M1_EAM showed one to two AA changes in human, avian and swine IAV strains (Figure 8e). These substitutions were mainly V205I, S207N, A209 T or Q208 K and did not affect the anchor residues. Additionally, deleterious residues A202M and R210M/S/T were notable at P2 and P9 anchor positions in a scant number of peptide binding motifs (Suppl. Fig. 4 e).

This was also particularly striking for the HA_VYR peptide in which the majority of avian, swine IAV sequences, as well as human H1N1 and H2N2, encompassed more than three mutations (namely non-identical sequences) (Figure 8f). The predominant substitutions were not at the anchor residues, except, F517I/L/M and I517 V/Y/S mutations, which were found at the P9 C-terminus anchor residue, while the latter was deleterious for binding (Suppl. Fig. 4 f). The peptide HA_LPY sequence showed diverse mutations in swine, human, and avian strains (Figure 8g). To the most extent, substitutions were rather distant from the anchor residues, nevertheless, preferred substitutions L306M/K at the P1 position and I316 V/M at the P9 C-terminal anchor residue were also observed (Suppl. Fig. 4g). These substitutions within HA_VYR and HA_LPY peptides marginally affected their binding affinity.

PB2_GTA and PB2_STS ligands showed one AA substitution in H1N1, H2N2, and H3N2 human subtypes (Figure 8h–i). The PB2_STS sequence was attributed to two mutations in avian and swine sequences, while PB2_GTA showed a high conservation level in both reservoirs except for a limited number of avian H7N9 sequences harbouring single mutations. These substitutions, mainly S682G and A684S, were particularly found in IAV human subtypes (Suppl. Fig. 4h).

Together, these analyses revealed that most epitopes are highly conserved in the major zoonotic reservoirs and can be attributed to a single mutation except for HA_VYR and HA_LPY epitopes. The results demonstrated that, in terms of conservation, epitope-specific CD8⁺ T cells can confer broad recognition and thus potential reactivity against different IAV subtypes across the zoonotic reservoirs.

Discussion

The relevance of Influenza virus clearance driven by CD8⁺ T cells following infection is underscored and has been well-investigated in mouse models and humans [11,12,40]. In humans, antigen recognition mediated by CD8⁺ T cells is dependent on the expression of HLA alleles, which are widely divergent between individuals. So far, different IAV-restricted CD8⁺ T cell epitopes have been described, for the most prevalent HLA alleles [11,12,41]. Nonetheless, further studies are imperative to fully explore the spectrum of CD8⁺ T cell recognition towards influenza viruses for possible inclusion in T cell-based vaccine approaches.

This immunopeptidomic analysis aimed to identify IAV-derived peptides restricted to different HLA allotypes, HLA-A*24:02, HLA-A*68:01, HLA-B*07:02, and HLA-B*51:01, presented on the cell surface of human lung adenocarcinoma cells (Calu-3) following infection with H1N1/PR8, H3N2/Victoria and H3N2/Fukui strains at different infection time points. While a number of viruses, e.g. HIV and Herpesviruses are implicated to modulate MHC-I expression and interfere with MHC-I ligand presentation [37,39,42,43], herein, we investigated the modulation associated with IAV infection. The data revealed that altering the MHC-I surface expression post IAV infection likely depends on the virus/subtype and/or the cell line used. This inference is substantiated by the fact that two IAV strains, H1N1/PR8, and H3N2/Fukui, do not significantly change the MHC-I surface expression, whilst infection with H3N2/Victoria alters the expression profile. An earlier report demonstrated that IAV and influenza B viruses (IBV) alter the expression profile at a late stage of infection [44].

The Immunopeptidomic analyses revealed 16 naturally presented peptides exhibiting a high predicted binding affinity for HLA-A*68:01 (13 peptides), HLA-A*24:02 (1 peptide), and HLA-B*51:01 (2 peptides). The presentation kinetics demonstrated also that most peptides were detectable between 6 and 9 hpi and reached their peak at 12 hpi and that a substantial presentation early post infection for immunodominant epitopes is not always persistent. Our findings are in line with an earlier report indicating that the presentation kinetics of different ligands peaked between 6.5 - 9.5 hpi and were independent of antigen expression kinetics [45]. IAV manifest temporal regulation of gene expression at the translational level and this is mirrored by the role of these proteins in the viral replication cycle [46,47]. Although immunopeptidomics was performed at different time points following infection, still several viral proteins were lacking as source proteins among the identified HLA-I ligands. These observations corroborate immunopeptidomics studies that identified different HLA-A*24–restricted peptides mostly annotated to PB2 and PB1 viral proteins without any presentation of either M1 or NA protein [48]. In contrast, Wu and colleagues identified naturally presented HLA-I-restricted IAV ligands comprising all IAV protein segments [45]. Collectively, the significance of viral ligand presentation following direct infection remains controversial. Apparently, the antigen presentation is correlated to the degradation of viral proteins rather than protein expression.

Nine out of sixteen putative epitopes were functionally assessed and proved to be immunogenic, among which 5 epitopes were determined as immunodominant. The sporadic or weak T cell responses to certain peptides could be correlated to either the low avidity TCRs or the TCR-pMHC-I being in reversed docking mode, limiting the ability of naïve T cells to drive strong immune responses [49,50].

So far, in a preceding immunopeptidomics study, Habel and colleagues described two ligands (PB2_TTV and PB2_GTA) as weak binders to HLA-A*11 and nonimmunogenic [51]. However, in our study both peptides as well as PB2_STS, the PB2_GTA-variant, were identified as strong binders to HLA-A*68, based on the in silico prediction algorithms and elicited CD8⁺ T cell responses as confirmed by functional validation experiments. Of note, there are discrepancies between the accuracy of prediction algorithms and can result in ambiguity in the predicted MHC restriction. Also, in silico prediction tools may disregard several peptides that were experimentally validated on T cell level. So, it should be noted that ligands binding prediction does not necessarily correlate with their immunogenicity and cannot be used as a sole parameter. Similarly, Hensen and colleagues identified the HLA-A*24-restricted epitope, HA_VYR, as a weak binder of HLA-A*24:02 and did not reveal immunogenic characteristics during validation whereas we defined this epitope as a strong binder and an immunodominant epitope with polyfunctional characteristics [48].

The presentation of CTL epitopes can be affected by AA substitutions within the HLA ligand sequence. Mutations in CTL epitopes, in particular at an anchor residue, can result in loss of CD8⁺ T cell recognition since the epitope no longer match the specific TCR [52–55]. In this context, our data showed that most epitopes are highly conserved in the major zoonotic reservoirs and alterations can be attributed to single AA substitutions, in particular for the NP_DAT epitope, in the vicinity of the anchor residues without affecting the binding affinity. In addition, HA-derived epitopes showed the lowest conservation score with more than 3 AA changes and predominant substitutions located distant from the anchor residues. Nevertheless, other mutations were found at either P2 or the PΩ terminus, but these substitutions are preferable at this position. Earlier reports examined the variation of AA sequences within the relatively conserved NP protein and showed that substitutions within NP-derived CD8⁺ T cells epitopes at the anchor position P2 with deleterious AA resulted in the loss of epitope recognition and affected the virus-specific CTL response [54,56]. Further reports explored additional AA variations within other NP-derived epitopes that show signs of antigenic drift and revealed that the CTL clones elicit cross-reactive immune responses against heterosubtypic variants of the epitope [57]. Taken together, our data indicate that, in terms of conservation, epitope-specific CD8⁺ memory T cells can confer cross-reactivity against multiple IAV subtypes across zoonotic reservoirs.

In summary, several IAV-derived epitopes were identified, thereby broadening the knowledge beyond the few previously characterized IAV-derived epitopes. These epitopes represent promising targets for T cell-based vaccination alongside other epitopes that could confer heterosubtypic protection in individuals, particularly against newly emerging variants and in the absence of any preexisting neutralizing antibody response.

Limitations of the study

The results of this study should be interpreted in the context of their technical limitations. The immunopeptidome profiling has been carried out in vitro, in an infected tumour cell line. Nevertheless, testing the immunogenicity of the identified peptides in HLA-matched donors in an ex-vivo setting supports their in vivo relevance and natural HLA peptides presentation. Second, the LC-MS/MS-based analysis has technical limitations and is prone to intrinsic biases. Hence, the lack of detection of other possibly immunogenic peptides is conceivable. We assume that performing the analysis with different IAV subtypes and at multiple infection time points may facilitate the identification of additional IAV epitopes. However, further studies with different IAV subtypes as well as cell lines expressing different HLA alleles are a prerequisite for the identification of additional epitopes.

Author contributions

Conceptualization: H.H. and O.P.; methodology: H.H. and M.G.; formal analysis: H.H.; investigation: H.H. and O.P.; resources: O.P.; H.-G.R.; writing – original draft: H.H. and O.P.; writing – review & editing: H.H., M.G., M.W.L., H.-G.R. and O.P; visualization: H.H; supervision: H.H. and O.P.; project administration: H.H and O.P, and M.W.L.; funding acquisition: O.P.

Acknowledgments

We would like to thank Prof. Dr. Cécile Gouttefangeas, Department of Immunology, University of Tübingen, for helpful discussions, and Martin Laure, Claudia Falkenburger, Ulrich Wulle and Beate Pömmerl for technical support. H. Hamza received a Ph.D. fellowship from the German Academic Exchange Service (DAAD) and the Egyptian Ministry of Higher Education and Scientific Research.

Data availability

The MS/MS data have been deposited to the ProteomeXchange consortium repository (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository [35] with the dataset identifier PXD035241.

Disclosure statement

No potential conflict of interest was reported by the author(s).