HLA-A*11:01-restricted CD8+ T cell immunity against influenza A and influenza B viruses in Indigenous and non-Indigenous people

HLA-A*11:01 is one of the most prevalent human leukocyte antigens (HLAs), especially in East Asian and Oceanian populations. It is also highly expressed in Indigenous people who are at high risk of severe influenza disease. As CD8+ T cells can provide broadly cross-reactive immunity to distinct influenza strains and subtypes, including influenza A, B and C viruses, understanding CD8+ T cell immunity to influenza viruses across prominent HLA types is needed to rationally design a universal influenza vaccine and generate protective immunity especially for high-risk populations. As only a handful of HLA-A*11:01-restricted CD8+ T cell epitopes have been described for influenza A viruses (IAVs) and epitopes for influenza B viruses (IBVs) were still unknown, we embarked on an epitope discovery study to define a CD8+ T cell landscape for HLA-A*11:01-expressing Indigenous and non-Indigenous Australian people. Using mass-spectrometry, we identified IAV- and IBV-derived peptides presented by HLA-A*11:01 during infection. 79 IAV and 57 IBV peptides were subsequently screened for immunogenicity in vitro with peripheral blood mononuclear cells from HLA-A*11:01-expressing Indigenous and non-Indigenous Australian donors. CD8+ T cell immunogenicity screening revealed two immunogenic IAV epitopes (A11/PB2320-331 and A11/PB2323-331) and the first HLA-A*11:01-restricted IBV epitopes (A11/M41-49, A11/NS1186-195 and A11/NP511-520). The immunogenic IAV- and IBV-derived peptides were >90% conserved among their respective influenza viruses. Identification of novel immunogenic HLA-A*11:01-restricted CD8+ T cell epitopes has implications for understanding how CD8+ T cell immunity is generated towards IAVs and IBVs. These findings can inform the development of rationally designed, broadly cross-reactive influenza vaccines to ensure protection from severe influenza disease in HLA-A*11:01-expressing individuals.

Introduction Influenza viruses remain an annual epidemic human pathogen despite progress in vaccine formulation and anti-viral therapies. Three types of influenza viruses infect humans, type A (IAV), B (IBV) and C (ICV). IAVs (H3N2; H1N1) and IBVs co-circulate to cause seasonal epidemics of mild, severe, or fatal respiratory disease, while ICV can cause severe disease in children [1][2][3][4]. A major hindrance to long-term vaccine effectiveness is influenza virus shift and drift, making annual re-formulation of the vaccine a requirement to maintain immunity, although protection is not guaranteed [1].
Much of the focus on influenza virus vaccine formulation has been on inducing humoral immunity, involving influenza-specific antibodies targeting viral surface glycoproteins to neutralize the virus and prevent infection. However, it is well established that cellular immunity, particularly cytotoxic CD8 + T cells, plays a vital role in the clearance of influenza virus infection [5][6][7]. A previous study involving patients infected with highly fatal H7N9 virus revealed that those who recovered from infection had greater numbers of H7N9-specific CD8 + T cells than those who succumbed to infection [8]. Therefore, CD8 + T cell-mediated responses towards influenza virus-infected cells need to be considered for generating protective immunity. CD8 + T cells can confer broad cross-reactivity across all IAVs, IBVs and ICVs [9], having key implications for the design of universal vaccines that do not require annual reformulation. Vaccines eliciting cross-reactive cytotoxic CD8 + T cells could reduce annual rates of IAV and IBV-induced morbidity and mortality as well as protect children from ICV. As current influenza vaccines do not promote cytotoxic T cell memory [10], it is important to understand how to elicit protective CD8 + T cell immunity against seasonal, pandemic and recently emerged IAVs and IBVs.
CD8 + T cells recognize peptides presented on HLA class I (HLA-I) molecules and can form long-lasting memory after being primed by an immunogenic epitope. HLA alleles are highly polymorphic, with some alleles having higher frequencies in certain ethnic groups. Variations in HLA alleles within the population lead to differential immune responses to influenza viruses. The HLA-A � 11:01 allele is a member of the HLA-A � 03 supertype and is prominent in many Asian and Oceanian populations, including Indigenous populations who are a high-risk group for severe disease with influenza virus infection. Morbidity and mortality resulting from the 2009 H1N1 pandemic virus was higher in Indigenous populations globally, including Indigenous Australians [11], Native Americans and Alaskans [12], Pacific Islanders [13] and Māori [14]. To date, only a handful HLA-A � 11:01-restricted CD8 + T cell epitopes have been reported for IAVs [15][16][17][18][19][20][21], and IBV epitopes are yet to be described.

Prevalence of HLA-A � 11:01 in Asian and Oceanian populations
HLA-A � 11:01 is one of the most prevalent HLAs, especially in East Asia and Oceania. Compared to the 7.3% global distribution of HLA-A � 11:01, the detected frequencies of HLA-A � 11:01 were the highest in South-East Asia (24.5%), Oceania (21.9%), South Asia (13.9%) and Australia (11.8%) ( Fig 1A). Previous work by our group using the Looking into InFluenza T cell immunity (LIFT) cohort determined HLA-A � 11:01 as one of the most prominent HLA-I alleles present in Indigenous Australians [22]. The Allele Frequency Net Database (allelefrequencies.net) also revealed high frequency of HLA-A � 11:01 in other Indigenous and Asian populations. High HLA-A � 11:01 prevalence was found in Papua New Guinea Madang people (63.6%), China Yunnan Hani (61.3%), Taiwan Hakka (40.0%), Pakistan Brahui (25.2%), Vietnam Hanoi Kinh (22.9%), Cape York Peninsula Aboriginal people (18%) and New Zealand Māori (16.7%). In contrast, the lowest frequency of HLA-A � 11:01 was detected in Caucasian people from the USA and Australia (7% and 6.7%, respectively) ( Fig 1A). HLA-A � 11:01 prevalence in our LIFT cohort of Indigenous Australians was at 16.1% (Fig 1A). Despite such prominence of HLA-A � 11:01 in East Asia and Indigenous people in Oceania, only a handful of CD8 + T cell epitopes have been previously identified for IAV and there are currently no epitopes identified for IBV.
A total of 159 peptides (non-redundant by sequence) were assigned to A/X31 at a 1% FDR. Although peptides assigned with scores below the 1% FDR threshold should be treated with increased caution without further validation, especially those with poor/lower scoring spectra, we also considered an additional 37 peptides identified at scores below this threshold and assessed their predicted binding to HLA-A � 11:01 (S1 Dataset). Of the 196 A/X31 peptides, 58 and 34 were predicted to be strong (SB) and weak binders (WB) of HLA-A � 11:01 respectively (using NetMHC4.0), while a further 38 were considered potential binders (PB) based on evidence of pull down and overlap with predicted binders. Similarly, a total of 109 peptides (nonredundant by sequence) were assigned to B/Malaysia/2506/04 at 1% FDR, with an additional 46 peptides identified having scores below this threshold. Of these, 27, 21 and 19 were predicted to be SB, WB or PB of HLA-A � 11:01 respectively (S1 Dataset). These peptides were predominantly 9-11 amino acids in length, as for peptides assigned to the human proteome ( Fig  3A, 3C and 3D). Most prominent proteins represented for A/X31 were PB2>PB1>NP, M1, while for B/Malaysia/2506/04 the hierarchy was HA, NP>M1, PB2 (Fig 3E, F). This is similar to previous analyses where A/X31 PB2 and PB1 peptides were prominent in the immunopeptidome of HLA-A � 24:02, and B/Malaysia/2506/04 HA and NP peptides were prominent in the immunopeptidomes of HLA-A � 02:01 and -A � 24:02 [9,23]. Due to the use of the pan-class I antibody W6/32 for HLA I isolation, influenza peptides with the capacity to bind HLA-B � 35:03 and HLA-C � 04:01 were also observed, as well as peptides with predicted capacity to bind more than one of the expressed HLA (S1 Dataset).
Of note, numerous potential alternative reading frame peptides mapping to the translation of the A/X31 genome alone were also assigned ( Fig 3E and S1 Dataset). For simplicity, these peptides were named according to the frame starting at the start of the viral UTR, with positional numbering considering stop codons as equal to an amino acid (S1 Fig). The majority of these peptides were assigned with scores above the 1% FDR threshold (Fig 3E and S1 Dataset). Interestingly, two peptides mapping to the M2 intronic region of segment 7 (M(+3) [39][40][41][42][43][44][45][46][47][48] and M(+3) 94-103 ) are immediately preceded by potential alternative start codons within the RNA sequence (S1 Fig). Indeed, M(+3) 39

Identification of novel IAV-specific HLA-A � 11:01-restricted CD8 + T cell epitopes
To determine CD8 + T cell reactivity to the selected 79 IAV-derived peptides identified by LC-MS/MS, peptides were divided into 8 pools of increasing predicted binding affinity determined by the NetMHCpan 4.0 algorithm (S3 Table). CD8 + T cell lines were generated by stimulation with one of the 8 peptide pools for 10 days followed by re-stimulation with the respective pool to measure IFN-γ production. Frequencies of IFN-γ + CD8 + T cells were significantly higher than the respective DMSO negative control for pools A-C ( Fig 4A). Dissection of individual peptides in pools A and B revealed that PB2 320-331 and PB2 323-331 were the most immunogenic, while none of the peptides in pool B resulted in significantly higher frequencies of IFN-γ + CD8 + T cells than background (Fig 4B and 4C). While stimulation with PB1 659-669 did not significantly increase CD8 + T cell activation, two donors had responses that were above the basal IFNγ + CD8 + T cell frequencies observed in the negative control, suggesting this peptide may be a subdominant epitope within HLA-A � 11:01 ( Fig 4C). Stimulation with PB2 323-331 , PB2 320-331 , or PB1 659-669 did not induce strong polyfunctionality in responsive CD8 + T cells, with the majority of cells expressing one or two activation markers, namely MIP1-β and/or CD107a ( Fig 4D).
Both PB2 epitopes were predicted strong binders of HLA-A � 11:01 and the fragmentation spectrum of the synthetic peptides were well matched with the discovery spectrum of the eluted peptide (S1 Dataset). PB2 320-331 and PB2 323-331 are variants of a previously published HLA-A � 11:01-restricted IAV epitope, PB2 322-331 [15,17] which did not elicit IFN-γ responses when tested using our CD8 + T cell probing method (Fig 2A). Although none of the alternative reading frame peptides tested were identified as major epitopes, fragmentation spectra of the synthetic peptides were consistent with the discovery spectra of the eluted peptides (S1 Dataset). Overall, our IAV CD8 + T cell epitope identification studies found alternative variants of a previously reported epitope A11/PB2 320-331 and A11/PB2 323-331 , and confirmed their presentation on HLA-A � 11:01 during IAV infection using mass spectrometry.

PLOS PATHOGENS
HLA-A*11:01-restricted CD8 + T cells in IAV and IBV can cause severe influenza disease, especially in young children [1,3]. To identify novel IBVderived CD8 + T cell epitopes restricted HLA-A � 11:01, we determined the immunogenicity of 57 LC-MS/MS-identified IBV peptides by probing memory CD8 + T cells in PBMCs of

PLOS PATHOGENS
HLA-A*11:01-restricted CD8 + T cells in IAV and IBV HLA-A � 11:01-expressing individuals. The 57 peptides were divided into three pools containing 19 peptides, each based on predicted binding affinity according to the NetMHCpan 4.0 algorithm (S4 Table). CD8 + T cell lines were generated by stimulating PBMCs with IBVinfected C1R-A � 11:01 cells, which were expanded in vitro for 12 days. On day 13, PBMCs were re-stimulated with IBV peptide-pools to assess intracellular IFN-γ and TNF production.
Analysis of IFN-γ production by CD8 + T cells demonstrated that pools 1 and 2 were the main source of CD8 + T cell activation, as the frequencies of IFN-γ + CD8 + T cell were significantly higher than the DMSO negative control (Fig 5A). Individual peptides from pools 1 and 2 were then dissected to identify novel immunogenic IBV CD8 + T cell epitopes. Re-stimulation of PBMCs with individual IBV-derived peptides demonstrated that peptides in pool 1 did not produce robust CD8 + T cell responses, while M1 41-49 from pool 2 was the only peptide to elicit significantly higher IFN-γ + CD8 + T cell frequencies than the DMSO negative control (Fig 5B  and 5C). While stimulation with NP 511-520 and NS1 186-195 did not result in significantly increased CD8 + T cell activation, some donors had lower magnitude responses that were still above the basal IFN-γ + CD8 + T cell frequencies observed in the negative control (n = 5 and n = 4, respectively), which suggests that these peptides form subdominant epitopes with HLA-A � 11:01 ( Fig 5B). CD8 + T cells stimulated with M1 41-49 and NS1 186-195 were found to have significantly higher proportions of polyfunctionality when compared to the negative control ( Fig 5D). All three peptides were predicted to bind HLA-A � 11:01 with an affinity stronger than 200nM (NetMHC4.0 and NetMHCpan 4.0, S4 Table and S1 Dataset) and the fragmentation spectrum of the synthetic peptides were consistent with the discovery spectrum of the eluted peptides (S1 Dataset).

PLOS PATHOGENS
HLA-A*11:01-restricted CD8 + T cells in IAV and IBV homology for their main anchor residues (Fig 6E). Despite these striking differences, the antigen binding cleft of HLA-A � 11:01 for both complexes showed little change with an average r. m.s.d. of 0.44 Å.
Overall, all three structures of HLA-A � 11:01 presenting peptides derived from NP, M1 and NS1 IBV proteins provide an understanding into peptide repertoire specificity in the P2 and P9 pockets. They also show distinct peptide presentations (Fig 6F), which would enable the activation of CD8 + T cells with distinct TCR repertoires, providing broad CD8 + T cell coverage.
While the PBMCs were initially stimulated with virus-infected autologous PBMCs, they were restimulated by peptide-pulsed C1R cells transduced with either HLA-A � 11:01 or -A � 02:01 to isolate the presenting HLA allotype, confirming the observed CD8 + T cell responses observed to the seven immunogenic HLA-A � 11:01-restricted epitopes are indeed restricted to HLA-A � 11:01 in our assay.
HLA-A � 11:01 restriction across all prior assays is supported by the highly dissimilar peptide binding preferences of HLA-A � 02:01 and HLA-A � 11:01. Whilst HLA-A � 11:01 favours peptides possessing Val, Ser and Thr at P2 and basic (Arg and Lys) residues at the C-terminus ( Fig  3B), HLA-A � 02:01 favours those possessing Leu and Met at P2 and hydrophobic (Val and Leu) residues at the C-terminus [9]. As such the peptides anticipated to bind these two allotypes are highly distinct, with the majority of the peptides tested here (including the 7 immunogenic peptides) predicted to bind HLA-A � 11:01 at lower nM concentrations than HLA-A � 02:01, as calculated by NetMHC4.0 [29,30] (S5 Fig). Moreover, while it is possible that enzymatic processing of these peptides might generate improved binders for HLA-A � 02:01 (as compared to the full-length sequences), assessment with NetMHC4.0 did not identify any shorter sequences within the immunogenic peptides that were predicted to bind HLA-A � 02:01. We therefore have no evidence that the 7 HLA-A � 11:01-restricted immunogenic peptides are presented by HLA-A � 02:01 during infection, suggesting it is unlikely that the responses observed against these peptides are restricted to HLA-A � 02:01.

Immunogenic peptide conservation and predictability
To determine the sequence conservation of the immunogenic peptides identified, conservation analysis was performed. IAV and IBV protein sequences were sourced from the NCBI

PLOS PATHOGENS
HLA-A*11:01-restricted CD8 + T cells in IAV and IBV  (Table 1). The lower conservation observed in NP 511-520 is predominantly due to variations in position 3 residue 513N, which was more commonly observed in years prior to 2008, and from 2008 onwards, 513S became the dominant strain.
Peptide prediction algorithms have been used previously to identify immunogenic epitopes. Prediction algorithms factor in HLA-specific binding motifs and the predicted binding affinity of a peptide to the HLA of interest. Using NetMHCpan 4.0, we found that three of the five immunogenic peptides described in this study, or a variant thereof, were in the top five predicted epitopes from their respective influenza virus protein (Table 2). Interestingly, M1 125-134 represented a predicted peptide which ranked 3 rd in the prediction of HLA-A � 11:01-restricted IAV M1 protein epitopes, however our experimental data suggest that the minimal peptide presented by HLA-A � 11:01 was M1 126-134 .
Overall, high conservation of IAV-and IBV-derived CD8 + T cell epitopes identified in our study suggest that these viral peptides could potentially be targeted in a broadly cross-reactive T cell vaccine against IAV and IBV strains to provide coverage for both Indigenous and non-Indigenous people expressing HLA-A � 11:01.

Discussion
The importance of CD8 + T cells in the clearance of influenza virus infections is well documented in mice [31][32][33] and humans [34][35][36]. HLA-I molecules on antigen presenting cells prime epitope-specific CD8 + T cells to mount an adaptive immune response which aids in viral clearance and forms long-lived immunological memory. Given that CD8 + T cell recognition is determined by a spectrum of HLAs expressed in an individual, and that HLAs differ between ethnic groups, broadly protective cross-reactive CD8 + T cell immunity needs to be investigated specifically for HLAs that are dominant across different ethnicities. To date, CD8 + T cell epitopes for both IAVs and IBVs have been identified only for two class I HLAs, HLA-A � 02:01 [9] and HLA-A � 24:02 [23]. As shown in our study, HLA-A � 11:01 is one of the most prevalent HLAs, especially in East Asia and Oceanian populations, with high enrichment  [11][12][13][14], especially when a new influenza virus emerges, it is of critical importance to identify CD8 + T cell epitopes for prominent HLAs in Indigenous people to understand how to effectively protect these populations.
Here, HLA-A � 11:01-specific CD8 + T cell responses were dissected to further understand immunity to IAV and IBV infections. Previous reports described a handful of HLA-A � 11:01-restricted IAV epitopes [15][16][17][18][19][20][21], although most epitope-specific CD8 + T cell responses were detected in low frequencies or used a low number of donors. When these reported epitopes were tested using our CD8 + T cell expansion and ICS methods, the majority of published HLA-A � 11:01-restricted peptides did not induce detectable IFN-γ by CD8 + T cells. This suggests that the HLA-A � 11:01-restricted epitopes reported previously were either at low frequencies in our donors, or CD8 + T cells elicited towards those epitopes are found only in a select number of individuals, perhaps dependent on their TCR repertoires. The precursor frequencies of epitope-specific CD8 + T cells determines the magnitude of the response towards each of the peptides screened. The method used to identify the HLA-A � 11:01-restricted epitopes could affect the interpretation of CD8 + T cell responses. The lack of IFN-γ response towards published IAV epitopes was unexpected and emphasized the need to identify HLA-A � 11:01-restricted influenza virus epitopes that are presented by HLA-A � 11:01 during infection and activate an influenza virus-specific CD8 + T cell response.
The crystal structures generated provide a basis for the preference of anchor residues that sit within the B and F pockets of HLA-A � 11:01 [37] observed by mass spectrometry. The B pocket of HLA-A � 11:01 contains Tyr7, Tyr9, Tyr159 and Met45, that have large side chains making the B pocket relatively shallow optimal to bind small peptide residues such as Thr, Val, and Ser as shown by the mass spectrometry data. Meanwhile, the F pocket of HLA-A � 11:01 contains small buried hydrophobic residues (Ile95, Ile97 and Leu81) as well as negatively charged Asp77 and Asp116 that form a salt bridge with the preferred Lysine at the C-terminal part of the peptide as shown by mass spectrometry. Therefore, the crystal structures are characteristic of the HLA-A � 11:01 B and F pockets and give a rational for the preferred residues observed for 9-, 10-and 11-mer peptides eluted from HLA-A � 11:01 molecule. These structures represent the first insight into IBV epitope presentation by HLA-A � 11:01 molecule, and their diverse conformations suggest that a diverse T cell repertoire might be able to recognise them.
Identification of novel T cell epitopes for IBV is of particular importance as immunity to IBVs is greatly understudied. As there are no established animal reservoirs for IBVs, there is no potential for zoonotic transmission to humans, hence research has largely focused on IAV, leaving IBV understudied [3,38]. However, IBVs still have substantial clinical importance, accounting for~25% of influenza cases annually [39,40], causing mild to severe disease especially in 5-17 year-old children. IBVs can also lead to severe neurological, cardiovascular and muscular complications [41] and secondary bacterial pneumonia [41,42]. Because the effectiveness of the current vaccines against IBV is modest at~50% [43], a universal T cell-based

PLOS PATHOGENS
HLA-A*11:01-restricted CD8 + T cells in IAV and IBV IBV vaccine that provides broadly cross-reactive and long-lasting protection is of a great interest. In our study, we identified three prominent IBV CD8 + T cell epitopes restricted by the prevalent HLA-A � 11:01, thus extending the epitope knowledge beyond already reported immunodominant IBV epitopes for HLA-A � 02:01 [9] and HLA-A � 24:02 [23].
Importantly, all the IAV-and IBV-derived peptides constituting IAV and IBV epitopes were >90% conserved among respective influenza viruses, indicating their suitability as vaccine targets to provide broad cross-reactive immunity across IAVs and IBVs.
The lack of variation in these peptide regions indicate their functional importance and that mutation might be detrimental to viral fitness. With conservation at >91% for each IAV and IBV peptide over the course of the past 100 years, the epitopes described here make ideal candidates for a CD8 + T cell peptide-based influenza virus vaccine. High conservation of the peptide sequence will allow for CD8 + T cells to provide long-lasting immunity. Once an epitopespecific CD8 + T cell memory pool is established, subsequent infections will be less severe and resolve faster than without pre-existing CD8 + T cell immunity. Overall, our findings provide insight for the development of rationally designed, broadly cross-reactive, influenza vaccines to protect HLA-A � 11:01-expressing individuals from severe influenza disease, and have the potential to contribute to a universal influenza vaccine providing global coverage for prominent HLA types across different ethnicities.

Ethics statement
Experimental work involving the use of human blood was conducted in line with Declaration of Helsinki Principles and according to the Australian National Health and Medical Research Council Code of Practice. Human blood samples were collected after obtaining signed informed consent from all participants. PBMCs were obtained from buffy packs (Australian Red Cross Lifeblood, West Melbourne, Australia) or whole blood from consenting donors. LIFT cohort participant PBMCs were obtained in collaboration with the Menzies School of Health Research (Charles Darwin University, NT, Australia), as previously described [22]. Human experimental work was approved by the University of Melbourne Human Ethics Committee (ID 1955465, 1443389), the Australian Red Cross Lifeblood Ethics Committee (ID 2015#8) and HREC of Northern Territory Department of Health and Menzies School of Health Research (ID 2012(ID -1928. Human PBMCs were isolated and cryopreserved in liquid nitrogen until later use.

Liquid Chromatography-tandem mass spectrometry (LC-MS/MS) analysis of HLA-bound peptides
C1R-A � 11:01 cells were lysed by cryomilling and detergent-based lysis in buffer consisting of 0.5% IGEPAL CA-630, 50 mM Tris-HCl pH8.0, 150 mM NaCl and protease inhibitors (cOmplete Protease Inhibitor Cocktail Tablet; Roche Molecular Biochemicals), and the HLA class I immunoaffinity purified using the pan class I antibody W6/32 as described previously [24]. Peptide/MHC complexes were dissociated (10% acetic acid) and fractionated by reversed phase high performance liquid chromatography (RP-HPLC), collecting 500μL fractions throughout the gradient as described [44]. 9 pools of peptide containing fractions were generated, vacuum-concentrated, and reconstituted in 15μL 0.1% formic acid (Honeywell) in Optima LC-MS water, containing 0.25pmol iRT internal standard peptides [45]. Fraction pools were analyzed by LC-MS/MS on a Q-Exactive Plus Hybrid Quadrupole Orbitrap (Thermo Fisher Scientific) coupled to a Dionex UltiMate 3000 RSLCnano system (Thermo Fisher Scientific). 6 μl injections were loaded onto an Acclaim PepMap 100 Trap column (100 μm x 2 cm, nanoViper, C18, 5 μm, 100Å; Thermo Scientific) in 2% acetonitrile, 0.1% formic acid at 15 μl/min. Peptides were eluted over an Acclaim PepMap RSLC Analytical column (75 μm x 50 cm, nanoViper, C18, 2 μm, 100Å; Thermo Scientific) with an increasing gradient of buffer B (80% acetonitrile, 0.1% formic acid) of 2.5-7.5% over 1 min, 7.5-32.5% over 55 min, 32.5-40% over 5 min, 40-99% over 5 min, 99% over 6 min and returning to 2.5% buffer B over 1 min, before re-equilibration at 2.5% for 20 min at a flow rate of 250 nL/min. Data were collected in positive mode: MS1 resolution, 70,000; scan range, 375-1,600 m/z; MS2 resolution, 35,000; dynamic exclusion, 15 s. The top 12 ions of +2 to +6 charge per cycle were subject to MS/MS. For motif analysis, a peptide spectrum match (PSM) FDR threshold of 1% was set. The HLA-A � 11:01 length distribution and binding motif was generated based on assignments to the human proteome in the human + B/Malaysia/2506/04 database search, filtered of peptides identified in endogenous HLA data sets (both class I and II) or within the contaminant database alone. Redundancy based on ambiguity of Ile/Leu assignment was minimal (<1%) and both sequence assignments were maintained within the analysis. Sequence Logos were generated with Seq2Logo 2.0 using default settings [46], graphs were generated using GraphPad Prism 8.0.2 for Windows (GraphPad Software, San Diego, California USA, www.graphpad. com). To identify potential influenza-derived HLA ligands, assignments to the influenza proteome/translation were considered for PEAKS peptide scores (-10lgp) both above and below the 1% FDR threshold. Peptide assignments (length <30 amino acid) mapping to the influenza proteome or 6 frame genome translation were considered valid if detected in the infected samples but not the uninfected samples. If a peptide assignment mapping to influenza was found in both infected and uninfected samples, spectra and retention times were compared. If spectra/retention times were distinct, and the higher scoring assignment was within the infected samples, identifications were maintained within the analysis. Assignments included within the analysis are contained in S1 Dataset.

Bioinformatic analysis of mass spectrometry data
Predicted binding affinities of stripped sequences of 8-14-mers to HLA-A � 11:01, -B � 35:03 and -C � 04:01 were calculated using NetMHC4.0 [29,30], and Strong binders (SB) and Weak Binders (WB) assigned based on the default cut-offs of % Rank (0.5 and 2, respectively). Potential binders (PB) were assigned based on evidence of pull down, overlap with predicted binders, and observation in endogenous HLA data sets, including class II peptides which can contaminate isolations.
The mass spectrometry HLA-A � 11:01 immunopeptidome data sets have been deposited to the ProteomeXchange Consortium via the PRIDE [47] partner repository with the dataset identifier PXD028985 and 10.6019/PXD028985.

Infection of C1R cells with influenza viruses
C1R-A � 11:01 or C1R-A � 02:01 cells were infected with IAV (A/X31) or IBV (B/Malaysia/2506/ 2004) for use in CD8 + T cell stimulations. C1Rs were infected using a MOI of 5 in RPMI medium for 1 h at 37˚C/5% CO 2 before adding complete RPMI and incubating a further 11 h. Infected C1Rs were washed twice before use for CD8 + T cell stimulation.

Expansion of virus-or antigen-specific CD8 + T cells
PBMCs were removed from storage in liquid nitrogen, thawed and washed twice in RPMI. Virus-specific expansions involved mixing PBMCs with IAV-or IBV-infected C1R cells at a 10:1 ratio. Antigen-specific expansions were performed as previously described [23] by peptide-pulsing one-third of PBMCs with pooled or individual influenza virus peptides for 1 h before washing and mixing with non-stimulated PBMCs. CD8 + T cell cultures were maintained in complete RPMI and incubated at 37˚C/5% CO 2 for 4 days before adding and maintaining a concentration of 20 U/mL of recombinant human IL-2. Virus-specific expansions were re-stimulated with influenza virus-infected C1Rs at a 1:10 stimulator-to-responder ratio on day 8 and maintained as described above.

Peptide conservation analysis
Protein sequences for IAVs or IBVs were sourced from the NCBI influenza virus resource (https://www.ncbi.nlm.nih.gov/genomes/FLU/Database/nph-select.cgi?go=database). IAV sequences from the dominant circulating strain for the relevant years were used to assess conservation, while all IBV sequences from 1936-2019 were analyzed together. BioEdit (version 7.2.5) [48] was used to remove incomplete sequences from the dataset. Conservation of each amino acid residue of the relevant peptide sequences were determined using Unipro UGENE software [49]. The conservation frequencies of each residue were averaged to determine the conservation of the entire peptide sequence.

Immunodominance/PBMC infection
PBMCs were removed from storage in liquid nitrogen, thawed and washed twice in RPMI. One-tenth of PBMCs were infected with either IAV or IBV, while the remaining cells were added to the tissue culture plate and incubated at 37˚C/5% CO 2 . PBMCs for infection were mixed with IAV (A/X31) or IBV (B/Malaysia/2506/04) at a MOI of 5 in serum-free RPMI and incubated at 37˚C/5% CO 2 for 1 h before fetal-calf serum was added and incubated for a further 3 h. Infected PBMCs were washed with RPMI to remove any remaining virus and mixed with the uninfected PBMCs in the tissue-culture plate. PBMCs were maintained in complete RPMI medium and incubated at 37˚C/5% CO 2 for 4 d before adding and maintaining a concentration of 20 U/mL of recombinant human IL-2 (Roche). On day 10 PBMCs were re-stimulated with IAV-or IBV-infected C1R-A � 02:01 or -A � 11:01 cells (as described above), or HLA-A � 02:01-or HLA-A � 11:01-restricted peptides derived from either virus and ICS was performed.

Crystallisation, data collection and structure determination
Crystals of the pHLA-A � 11:01 complexes were grown by the hanging-drop, vapour-diffusion method at 20˚C with a protein/reservoir drop ratio of 1:1 with seeding at a concentration of 6 mg/mL in the following conditions: M1 41-49 : 2M Ammonium Sulfate, 0.1M Tris-HCl pH 8, 0.2M Lithium Sulfate, 2% PEG400, NP 511-520 : 20% PEG 3350, 0.1M BisTris pH 6.5 and NS1 [186][187][188][189][190][191][192][193][194][195] : 26% 3350, 0.1M BisTris pH 6.3, 0.2M Lithium Sulfate. The crystals were soaked in a cryoprotectant solution containing mother liquor solution with the PEG concentration increased to 30% (w/v) and then flash frozen in liquid nitrogen. The data were collected on the MX1 and MX2 beamlines [51]. The data was processed using XDS [52] and the structures were determined by molecular replacement using the Phaser program [53] from the CCP4 suite (1994) with a model of HLA-A � 11:01 without the peptide (derived from PDB ID: 4MJ5 [21]. We observed for HLA-A � 11:01-M1 [41][42][43][44][45][46][47][48][49] complex that the diffraction was anisotropic, as a result the density map quality was poor. To improve the quality of the initial map, we used the Diffraction Anisotropy Server from UCLA [54] that apply ellipsoidal truncation and anisotropic scaling to the diffraction data. As per our observation, the analysis showed that the data has strong anisotropy with a score of 30.1 out of 100. Manual model building was conducted using the Coot software [55] followed by maximum-likelihood refinement with the Buster program [56]. The final model has been validated using the Protein Data Base validation web site and the final refinement statistics are summarized in S6 Table. All molecular graphics representations were created using PyMol [57]. The final crystal structure models for the HLA-A � 11:01 complexes have been deposited to the Protein DataBank (PDB) under the following accession codes: HLA-A � 11:01-NP 511-520 : 7S8Q, HLA-A � 11:01-M1 41-49 : 7S8R and HLA-A � 11:01-NS1 186-195 : 7S8S.

Thermal stability assay
Thermal shift assays were performed to determine the stability of each pHLA-A � 11:01 complex using fluorescent dye, SYPRO Orange, to monitor protein unfolding. The thermal stability assay was performed in the Real Time Detection system (Corbett RotorGene 3000), originally designed for PCR. Each pHLA complex was measured in 10 mM Tris-HCl pH8, 150 mM NaCl, at two concentrations (5 and 10 μM) in duplicate (n = 1), and was heated from 25 to 95˚C with a heating rate of 1˚C/min. The fluorescence intensity was measured with excitation at 530 nm and Emission at 555 nm. The Tm, or thermal midpoint, represents the temperature for which 50% of the protein is unfolded. The results are reported in the S5 Table. Supporting information