Cryptic MHC-E epitope from influenza elicits a potent cytolytic T cell response

The extent to which unconventional forms of antigen presentation drive T cell immunity is unknown. By convention, CD8 T cells recognize viral peptides, or epitopes, in association with classical major histocompatibility complex (MHC) class I, or MHC-Ia, but immune surveillance can, in some cases, be directed against peptides presented by nonclassical MHC-Ib, in particular the MHC-E proteins (Qa-1 in mice and HLA-E in humans); however, the overall importance of nonclassical responses in antiviral immunity remains unclear. Similarly uncertain is the importance of ‘cryptic’ viral epitopes, defined as those undetectable by conventional mapping techniques. Here we used an immunopeptidomic approach to search for unconventional epitopes that drive T cell responses in mice infected with influenza virus A/Puerto Rico/8/1934. We identified a nine amino acid epitope, termed M-SL9, that drives a co-immunodominant, cytolytic CD8 T cell response that is unconventional in two major ways: first, it is presented by Qa-1, and second, it has a cryptic origin, mapping to an unannotated alternative reading frame product of the influenza matrix gene segment. Presentation and immunogenicity of M-SL9 are dependent on the second AUG codon of the positive sense matrix RNA segment, suggesting translation initiation by leaky ribosomal scanning. During influenza virus A/Puerto Rico/8/1934 infection, M-SL9-specific T cells exhibit a low level of egress from the lungs and strong differentiation into tissue-resident memory cells. Importantly, we show that M-SL9/Qa-1-specific T cells can be strongly induced by messenger RNA vaccination and that they can mediate antigen-specific cytolysis in vivo. Our results demonstrate that noncanonical translation products can account for an important fraction of the T cell repertoire and add to a growing body of evidence that MHC-E-restricted T cells could have substantial therapeutic value. Eisenlohr and colleagues identify a novel influenza A virus peptide that elicits a robust CD8+ T cell response and is restricted by the nonclassical Qa-1 class I molecule.

The extent to which unconventional forms of antigen presentation drive T cell immunity is unknown.By convention, CD8 T cells recognize viral peptides, or epitopes, in association with classical major histocompatibility complex (MHC) class I, or MHC-Ia, but immune surveillance can, in some cases, be directed against peptides presented by nonclassical MHC-Ib, in particular the MHC-E proteins (Qa-1 in mice and HLA-E in humans); however, the overall importance of nonclassical responses in antiviral immunity remains unclear.Similarly uncertain is the importance of 'cryptic' viral epitopes, defined as those undetectable by conventional mapping techniques.Here we used an immunopeptidomic approach to search for unconventional epitopes that drive T cell responses in mice infected with influenza virus A/Puerto Rico/8/1934.We identified a nine amino acid epitope, termed M-SL9, that drives a co-immunodominant, cytolytic CD8 T cell response that is unconventional in two major ways: first, it is presented by Qa-1, and second, it has a cryptic origin, mapping to an unannotated alternative reading frame product of the influenza matrix gene segment.Presentation and immunogenicity of M-SL9 are dependent on the second AUG codon of the positive sense matrix RNA segment, suggesting translation initiation by leaky ribosomal scanning.During influenza virus A/Puerto Rico/8/1934 infection, M-SL9-specific T cells exhibit a low level of egress from the lungs and strong differentiation into tissue-resident memory cells.Importantly, we show that M-SL9/Qa-1-specific T cells can be strongly induced by messenger RNA vaccination and that they can mediate antigen-specific cytolysis in vivo.Our results demonstrate that noncanonical translation products can account for an important fraction of the T cell repertoire and add to a growing body of evidence that MHC-E-restricted T cells could have substantial therapeutic value.
CD8 T cells conventionally recognize pathogen-derived epitopes presented by the hyperpolymorphic major histocompatibility complex (MHC)-Ia molecules, including HLA-A, HLA-B and HLA-C in humans and H2-D, H2-K and H2-L in mice.However, protection from some viral [1][2][3][4][5][6] or bacterial 7 diseases has also been demonstrated by CD8 T cells recognizing epitopes presented by the MHC-Ib proteins, which exhibit limited polymorphism and are typically associated with more specialized functions 8 .An important subgroup of the MHC-Ib family are the MHC-E Article https://doi.org/10.1038/s41590-023-01644-5cells, suggesting minor co-precipitation 36 of peptide-MHC-I complexes and/or free MHC-I ligands along with MHC-II (Extended Data Fig. 1g).
Only one identified PR8 peptide mapped to a potential noncanonical ribosomal product and was named 'M-SL9' based on its origin in the matrix (M) gene segment and its amino acid sequence, SLQGRTLIL.This 9-mer was validated as an epitope by detecting reactive splenocytes from PR8-infected mice by interferon (IFN)-γ enzyme-linked immunosorbent spot (ELISpot) assay (Fig. 1a and Extended Data Fig. 2).M-SL9 is encoded within the +1 RF of the matrix protein 1 (M1)-coding sequence (Fig. 1b and Extended Data Fig. 3) and could not be attributed to any open RF (ORF) known to be translated in wild-type (WT) PR8 infection, and was therefore designated as a cryptic epitope.
Intriguingly, ELISpot analysis revealed M-SL9-specific (IFN-γ-producing) cells within bulk splenocytes but not purified spleen CD4 T cells from C57Bl/6 mice; no M-SL9 response was detectable in BALB/c spleens from a small sample of N = 4 mice (Extended Data Fig. 2).We further characterized the M-SL9-specific response by peptide-stimulating lung lymphocytes from PR8-infected C57Bl/6 mice and intracellular cytokine staining of IFN-γ, tumor necrosis factor (TNF) and IL-2.This clearly showed that M-SL9 peptide activated CD8 T cells but not CD4 T cells (Fig. 1c).In fact, over 10% of stimulated lung CD8 T cells were specific for M-SL9, in comparison to ~20% for the well-described H2-D b -associated PR8 epitopes, nucleoprotein (NP) 366- 374 and polymerase acidic protein (PA) 224-233 (Fig. 1d and Extended Data Fig. 4).In some mice, the M-SL9 response exceeded the NP 366-374 or PA 224-233 responses.Notably, M-SL9-specific CD8 T cells displayed a high level of cytokine polyfunctionality, with three-quarters of cells expressing ≥2 cytokines and one-quarter expressing 3 cytokines.This level of polyfunctionality was roughly double that of the NP 366-374 response and was comparable to that of PA 224-233 .Cytolytic markers were analyzed similarly, and comparable frequencies of cytolytic CD8 T cells (~10%) were identified as specific for M-SL9 or NP 366-374 based on simultaneous expression of perforin, granzyme B and the degranulation marker CD107a (Fig. 1e and Extended Data Fig. 4).The effector expression profile of M-SL9-specific CD8 T cells is thus typical of a type 1-polarized antiviral immune response.

M-SL9 is presented by Qa-1
Given that a CD8 T cell epitope was identified via MHC-II immunopeptidomics, we asked which MHC molecule serves as the restricting element for M-SL9.This investigation was aided by a T cell hybridoma clone, called B6.23, which produces β-galactosidase as a sensitive and specific reporter of M-SL9 presentation (Fig. 2a).This clone was discovered fortuitously by screening PR8-specific T cell hybridoma clones previously generated in house.B6.23 recognizes M-SL9 peptide-pulsed and PR8-infected APCs (Fig. 2b) and is activated by the M-SL9 sequence present in our molecular clone of PR8 (ref.37) as well as the M-SL9-P variant (SLQGRTPIL) (Extended Data Fig. 5) commonly present in other PR8 isolates.To test MHC restriction, we co-cultured B6.23 cells with M-SL9 peptide-pulsed APCs bearing various sets of MHC molecules.B6.23 was activated by the B6-CIITA fibroblast line but not by L929 cells (H2 k haplotype) expressing the C57Bl/6-derived MHC-Ia molecules H2-D b or H2-K b ; these lines were validated by measuring presentation of NP 366-374 /H2-D b and chicken ovalbumin 257-264 (SIINFEKL)/H2-K b to corresponding T hybridomas (Fig. 2c).Lack of MHC-Ia binding by M-SL9 was corroborated by the observation that M-SL9 peptide was not able to stabilize empty H2-K b or H2-D b molecules on the surface of transporter associated with antigen processing (TAP)-deficient RMA-S cells 38 (Extended Data Fig. 6a).
Interestingly, B6.23 was activated by M-SL9 peptide-pulsed bone marrow-derived dendritic cells (BMDCs) from BALB/c mice (Fig. 2c), suggesting usage of an MHC-Ib molecule with low polymorphism.We noted that M-SL9 bears a striking resemblance to Qdm, with matching or favored amino acids at all anchor positions 39,40 (Fig. 2d); therefore, we tested whether Qa-1 was able to present M-SL9 to B6.23 cells.
proteins.These are best known for downregulating the cytolytic activity of natural killer (NK) cells by presenting an MHC-Ia-derived self peptide (called the 'Qa-1 determinant modifier' or Qdm in mice) to inhibitory NKG2/CD94 receptors 9,10 ; thus, NK cells are able to specifically kill cells when the MHC-Ia presentation machinery is antagonized or lost due to mutation.Recently, it has become apparent that MHC-E can also stimulate robust CD8 T cell responses in multiple species against various infections and cancer [11][12][13] .Most notably, the Picker laboratory elicited MHC-E-restricted CD8 T cells in rhesus macaques using a cytomegalovirus (CMV) vector encoding antigens from simian immunodeficiency virus (SIV), and, remarkably, these T cells were uniquely capable of clearing a pathogenic SIV infection 1,2,4 .A protective capacity of MHC-E-restricted CD8 T cells has also been observed in a handful of other infectious disease settings [5][6][7] .Despite its striking therapeutic potential, research into the biology of MHC-E-restricted CD8 T cells has been minimal and is hampered by a lack of tractable small animal models, since, to our knowledge, no virally derived Qa-1-restricted epitope sequence has been published so far.
Another problem in T cell biology is understanding the full diversity of MHC-binding peptides, collectively known as the immunopeptidome.Efforts to map T cell responses to viruses are typically limited to peptides from annotated viral proteins [14][15][16] .However, many reports have shown that T cell epitopes can originate from cryptic sources, including polypeptides created by noncanonical translation [17][18][19][20][21][22][23][24][25][26] and posttranslational peptide splicing [27][28][29][30] .While recent advancements in mass spectrometry have allowed for deeper probing of cryptic epitopes, uncertainties remain regarding the fraction of the immunopeptidome that is derived from noncanonical sources and their clinical relevance 31 .As most, if not all, cryptic epitopes described so far have been MHC-Ia restricted, an additional unknown is whether unconventional epitopes can be presented on MHC-Ib and MHC class II (MHC-II).
In this Article, we set out to investigate the possibility that the T cell response against influenza A virus (IAV) may include cryptic epitopes presented on MHC-II.Serendipitously, we discovered an epitope derived from a cryptic viral translation product that is instead restricted to the nonclassical MHC-Ib molecule, Qa-1.Remarkably, this cryptic Qa-1 epitope drives a major fraction of the CD8 T cell response in IAV-infected C57Bl/6 mice and exhibits potent cytolytic capacity after infection or messenger (m)RNA vaccination.

Discovery of a cryptic CD8 T cell response to influenza
We conducted an immunopeptidomic analysis with the initial aim of identifying cryptic MHC-II epitopes in a mouse model of IAV infection.We hypothesized this was possible given that MHC-II presentation in IAV-infected mice is dominated by endogenous epitopes directly presented by infected antigen-presenting cells (APCs) 32 .As outlined in Fig. 1a, we infected an in-house APC line called B6-CIITA-E d (C57Bl/6 (B6) fibroblasts expressing MHC-II molecules I-A b and I-E d ) 33 with influenza virus A/Puerto Rico/8/1934 (PR8), immunoprecipitated MHC-II, isolated peptide ligands, and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS 2 ).Mass spectra were interrogated against a database containing the PR8 genome fully translated in all six reading frames (RFs), as well as the mouse proteome.We identified 7,864 unique peptide species at an estimated false discovery rate of 3.4%, with 95.7% mapping to mouse proteins.Two-thirds of all peptides were 14-18 amino acids in length, as is typical for MHC-II ligands, and 335 unique peptide species mapped to PR8, showing a similar length distribution (Extended Data Fig. 1a,b).Among both PR8 and non-PR8 peptides, a large portion matched the consensus MHC-II-binding motifs available from NetMHCIIpan 4.0 (Extended Data Fig. 1c-f) 34 .Peptides of nine amino acids formed a small local peak in the overall length distribution (Extended Data Fig. 1a), as seen previously 35 , and the consensus of these peptides resembled that of the MHC-I molecules expressed by these Article https://doi.org/10.1038/s41590-023-01644-5 When HeLa cells were transduced with the H2-T23 gene encoding Qa-1 b (expressed in C57Bl/6), they gained the ability to present synthetic M-SL9 peptide (Fig. 2e and Extended Data Fig. 6b-e).This was not true for HeLa cells expressing H2-T11, a related MHC-Ib protein originating from a duplication of H2-T23.Next, we tested whether Qa-1 b expression was required for C57Bl/6 cells to present M-SL9.We obtained BMDCs from WT C57Bl/6 or various mutant mouse strains, pulsed with peptides, and tested for activation of B6.23 and control hybridomas.M-SL9 presentation was high for WT BMDCs and those deficient for I-A b (MHC-II −/− ) or both H2-K b and H2-D b (K b D b−/− ) but was abrogated for BMDCs that were also deficient for Qa-1 b (K b D b Qa-1 −/− ) or for β 2 microglobulin (β 2 m) (Fig. 2f).These studies revealed that Qa-1 is both necessary and sufficient for M-SL9 presentation and that the β 2 m component of all MHC-I is also required, as expected.This finding was extended in vivo by measuring M-SL9/Qa-1 b tetramer-binding CD8 T cells in the lungs of PR8-infected mice.These cells were robustly detected and accounted for about 5% of all lung CD8 T cells, while NP 366-374 /H2-D b tetramer-specific CD8 T cells accounted for roughly 18% (Fig. 2g).We noted that the exact frequencies of each T cell specificity can differ on the basis of the virus dose and the detection method.
To our knowledge, M-SL9 is the first sequenced viral CD8 T cell epitope shown to be restricted to Qa-1.While we cannot rule out the possibility that M-SL9 has some affinity for MHC-II in addition to Qa-1, this seems unlikely based on weak M-SL9/MHC-II binding prediction (Supplementary Table 2) and the lack of M-SL9-specific CD4 T cells (Fig. 1c and Extended Data Fig. 2b,c).A more likely explanation is that M-SL9 peptide or M-SL9/Qa-1 b complexes co-precipitated during the immunoenrichment of MHC-II, consistent with previous findings that MHC-Ia, MHC-Ib and MHC-II ligands can all immunoprecipitate nonspecifically and can be detectable by MS 36 .This idea is also supported by the fact that the sequence motif of identified 9-mers resembles that of MHC-Ia and, to a lesser extent, Qa-1 ligands.Additionally, we note that our mouse-derived peptide identifications also included the canonical ligand for Qa-1, Qdm (−log 10 P = 31.77).Co-precipitation therefore seems to be a general property of immunopeptidomics, highlighting the need to confirm MHC restriction during such studies.

Phenotype of M-SL9-specific CD8 T cells
We next tested whether the immunophenotype and functional characteristics of M-SL9/Qa-1 b -specific T cells resemble that of classical IAV-specific CD8 T cells.We identified CD8 T cells recognizing M-SL9 or NP 366-374 from PR8-infected mouse lungs by tetramer staining.While some Qa-1-dependent CD8 T cells have been shown to be T cell receptor (TCR) γδ positive 41 , M-SL9-specific CD8 T cells were    uniformly TCRβ + and TCRγδ − , consistent with αβ T cells (Fig. 3a).
Qa-1 has also been associated with regulatory CD8 T cells, which can express Qa-1-binding CD8αα homodimers 42 and even the transcription factor FoxP3 (ref.43), normally a hallmark of regulatory CD4 T cells.Here, we found that M-SL9-specific CD8 T cells only expressed typical CD8αβ heterodimers (Fig. 3b), and all PR8-induced CD8 T cells were negative for FoxP3 (Extended Data Fig. 4e).
Additionally, the M-SL9-specific population was virtually all within the CD44 + CD62L − effector population, as expected of effector CD8 T cells (Fig. 3c).
Given the strong sequence homology between M-SL9 and Qdm (Fig. 2d), we asked whether M-SL9/Qa-1 b might serve as a ligand for NKG2A/CD94 or the related NK receptors NKG2C/CD94 and NKG2E/ CD94, all recognized by the same antibody clone used here.We observed that NKG2A/C/E + cells from naive mouse spleens were efficiently bound by Qdm/Qa-1 b tetramer but not by M-SL9/Qa-1 b or negative control  Frequency (%) Frequency (%) Frequency (%) Gene segment   NP 366-374 /H2-D b tetramers (Extended Data Fig. 7a).No other CD3 − CD19 − spleen cells were stained by M-SL9/Qa-1 b tetramer, indicating that M-SL9/Qa-1 b does not interact with any subset of NK cells.
Next, we tested the related question of whether Qdm/Qa-1 b tetramer would bind to M-SL9-specific CD8 T cells.Consistent with the literature 44 , we found that PR8-induced CD8 T cells stained with Qdm/Qa-1 b tetramer regardless of antigen specificity (Extended Data Fig. 7b).This can be explained by the fact that CD8 T cells normally express NKG2A/CD94 in response to activation 45 , allowing interaction with Qdm/Qa-1 b as part of a negative feedback or 'immune checkpoint' pathway 46,47 .Indeed, we found that Qdm/Qa-1 b tetramer mainly co-stained with anti-NKG2A/C/E.These data suggest that Qdm/Qa-1 b tetramer binds to PR8-induced CD8 T cells of multiple specificities in a TCR-independent manner.Some reports of MHC-Ib-restricted CD8 T cells have shown them to have limited clonality or restricted TCR gene usage 48 .We examined this by sequencing recombined Tcrb genes in M-SL9/Qa-1 b and NP 366-374 /D b tetramer-specific cells sorted from the lungs of PR8-infected mice.The top ten expanded Tcrb clones accounted for ~50% and ~80% of epitope-specific Tcrb sequences for M-SL9-and NP 366-374 -specific CD8 T cells, respectively, with a similar fold difference in the Simpson clonality index, indicating that M-SL9-specific cells were moderately more clonally diverse than NP 366-374 -specific cells (Fig. 3d).Both cell populations showed a variety of Tcrb variable (V) and joining ( J) gene segment usage and recombined V-diversity (D)-J sequences, with a limited number of V genes that were particularly enriched, for example, Tcrb V12-2 in M-SL9-specific cells and V13-1 in NP 366-374 -specific cells (Fig. 3e, Extended Data Fig. 8a,b and Supplementary Table 1).The gene usage was more similar within each T cell population across different mice than within each mouse across different T cell populations (Extended Data Fig. 8c,d).Length distributions of complementary determining region 3 (CDR3) were roughly similar with a mode of 14 amino acids in each case (Fig. 3e).Taken together, we conclude that M-SL9/Qa-1 b -specific CD8 T cells are largely similar to conventionally restricted IAV-specific CD8 T cells in cytokine and cytolytic functionality, major surface markers and clonality.

M-SL9-specific T cells home to the lung during PR8 infection
Next, we compared the frequency of CD8 T cells specific for M-SL9/ Qa-1 versus the classical epitopes NP 366-374 /D b and PA 224-233 /D b within various tissues from 6 to 56 days after respiratory PR8 infection using tetramer staining, and we observed striking differences in both anatomical distribution and kinetics.At 6 days, M-SL9 was the dominant specificity in lungs, representing a median of 2.5% of lung CD8 T cells, while the responses to PA 224-233 and NP 366-374 were significantly lower, at 0.9% and an undetectable level, respectively (Fig. 4a and Extended Data Fig. 9a).By day 9, the NP 366-374 and PA 224-233 responses had surged to arrive at or near their peak levels of 15% and 10%, respectively, while the M-SL9 response increased more slowly and did not peak until day 14, with a median of 10%.Between days 14 and 56, all IAV-specific CD8 T cells waned in the lung until reaching a similar level of roughly 5% of CD8 T cells for each specificity.
The CD8 T cell kinetics followed a similar pattern in bronchoalveolar lavage fluid (BALF) (Fig. 4b) and the lung-draining mediastinal lymph node (MLN) (Fig. 4c), although the responses in MLN were about one order of magnitude lower than in lungs or BALF.At day 6, M-SL9 was either the dominant response (BALF) or co-dominant with PA 224-233 (MLN), while the NP 366-374 responses were negligible.At both sites, classical CD8 T cell responses reached their peak at approximately day 9, while the M-SL9 response peaked at day 14.Thereafter, the frequency of each CD8 T cell specificity remained stable or waned slightly over time, and the infiltration of all T cells into BALF dropped significantly by day 31 (Extended Data Fig. 9b).
In blood (Fig. 4d) and spleen (Fig. 4e), classical CD8 T cells peaked at day 9, when the NP 366-374 response accounted for 15% (blood) and 6% (spleen) of CD8 T cells, and the PA 224-233 response was at about half these levels.These frequencies diminished rapidly by day 14, in comparison to the slow waning of responses in the respiratory tract and draining lymph node.In marked contrast, M-SL9-specific T cells were largely absent from the blood or spleen.The M-SL9 response on days 6-9 was barely above background at either site and peaked at low frequencies of 1.5% (blood) and 0.7% (spleen) on day 14 before returning to background levels by day 31.Of note, we originally detected M-SL9-specific T cells in spleen (Fig. 1a) by IFN-γ ELISpot, a much more sensitive method of T cell detection than tetramer staining.

Induction of M-SL9-specific tissue-resident memory T cells
Given the remarkable degree of skewing of M-SL9-specific CD8 T cells toward the lungs and BALF (~10% of CD8 T cells at peak) and nearly complete absence from the blood and spleen after PR8 infection, we analyzed the expression of trafficking molecules by different CD8 T cell specificities.We measured CD103, an integrin associated with homing to the respiratory epithelium and possibly the lamina propria of mucosal sites 49,50 , and CD69, an inhibitor of lymphocyte egress from the lung and other organs 51,52 .When present on lung T cells ≥30-40 days after IAV infection, CD103 and CD69 serve as markers for noncirculating tissue-resident memory T cells (T RM ), which are associated with protection from heterosubtypic IAV challenge 53,54 .In the lung and BALF of PR8-infected mice, CD103 and CD69 expression increased over time on all T cell specificities, but to the greatest extent on M-SL9-specific cells (Fig. 5a,b and Extended Data Fig. 9c).At day 56, T cells specific for M-SL9 showed a significantly greater level of differentiation (67%) into T RM compared with PA 224-233 (54%) and NP 366-374 (24%) (Fig. 5c).This supports a model in which M-SL9-specific CD8 T cells skew preferentially to the lung due to increased expression of trafficking molecules that prevent egress from the PR8-infected lung and, eventually, enhanced differentiation into T RM cells.

Recall of the M-SL9 response by heterosubtypic infection
Prior studies have detailed the disparate ways in which NP 366-374 -specific and PA 224-233 -specific CD8 T cells are recalled upon secondary antigen exposure in C57Bl/6 mice.Specifically, 'PR8→X31' mice that are infected with PR8, rested for 1 month, and then infected with X31 (a reassortant virus encoding H3N2 surface proteins and PR8 internal proteins) experience a profound boost in their NP 366-374 response but little to no increase in the PA 224-233 response 55 .Here, we reproduced these findings and tested whether the M-SL9 response can be efficiently recalled in the same scenario (Fig. 6a).In all examined anatomical sites, M-SL9-specific CD8 T cells were elevated in PR8→X31 mice in comparison to mice that received PR8 infection alone (Fig. 6b-f).However, PR8→X31 M-SL9 responses in the lung, BALF and MLN were not elevated in comparison to the peak response after X31 infection alone, suggesting that the additional expansion or 'boost' from the secondary infection was minimal in these sites.A trend toward a larger boost was observed in blood and spleen, where the median frequencies of M-SL9-specific CD8 T cells in PR8→X31 mice were increased at least threefold relative to either virus alone, although P values (up to 0.15) did not reach statistical significance (Fig. 6e,f).This is in contrast to the NP 366-374 response, which was significantly increased by two-to fivefold in PR8→X31 mice compared with mice infected with either virus alone.This was also distinct from the PA 224-233 response, which fell to a significantly lower percentage after secondary infection compared with primary infection with either virus.Interestingly, we noted that X31 infection alone was able to consistently elicit M-SL9-specific T cells in circulation; these responses were low compared with NP 366-374 and PA 224-233 but substantially greater than after PR8 alone.This suggests that properties of the antigen exposure can affect the tissue homing of nonclassical CD8 T cells.Accordingly, an alternative explanation for the slight boost in M-SL9-specific T cells in the blood and spleen of PR8→X31 mice is that secondary infection with X31 induced greater circulation of M-SL9-specific T cells rather than a true expansion of memory T cells.analyses were performed within 1-2 days of the reported times.Line graphs display the median ± interquartile range, and dot plots show the underlying data for individual mice and median.Data are combined from two independent experiments, with a mix of male and female mice.BALF from naive mice did not contain sufficient cell numbers for quantitation.P values reflect two-way ANOVA.

Article
https://doi.org/10.1038/s41590-023-01644-5 Taken together, these results demonstrate that the M-SL9-specific CD8 T cell response is modestly enhanced by heterosubtypic reinfection, reaching a similar or slightly higher level compared with primary infection, depending on the tissue.Compared with the classical CD8 T cell specificities, the extent of M-SL9 recall during secondary infection was intermediate between that of NP 366-374 and PA 224-233 .

M-SL9 presentation depends on AUG2
Given the cryptic origin of M-SL9, the mode in which this epitope is translated is not obvious.The in-frame codon immediately preceding M-SL9 is the second AUG (AUG2) in the matrix gene segment, located 85 nucleotides after the primary start codon (AUG1) shared by M1 and the matrix protein 2 (M2) ion channel (Extended Data Fig. 3).Therefore, we hypothesized that translation of M-SL9 is initiated at AUG2.Support for this theory was published by Machkovech et al., who performed ribosomal profiling of human cells infected with an IAV strain closely related to PR8 and found evidence that the ribosome can initiate on AUG2 (ref.56).If translation starts at AUG2 and proceeds to the nearest in-frame stop codon, the resulting 16-mer peptide would be a novel gene product that we term M-MG16 (Fig. 7a).A less likely possibility is raised by a reported RNA splicing event that produces an alternative ion channel, called M42, which contains M-SL9 in its N-terminus joined to the second exon of M2; however, the M42 transcript was previously shown to not be expressed by WT PR8 (ref.57).
To investigate these possibilities, we encoded M-MG16 and M42 using mRNA vectors 58 and tested their potential to present M-SL9 when introduced into B6-CIITA fibroblasts.Interestingly, a similar efficiency of M-SL9 presentation per mole of input mRNA was observed for M-MG16 and M42, and these were both slightly more efficient than a positive control mRNA encoding the full matrix gene segment (Fig. 7b), suggesting both constructs would be processed into M-SL9 if they were transcribed and translated during infection.To examine M-SL9 presentation during viral infection, we designed a panel of infectious PR8 variants with mutations in the matrix gene segment that are synonymous in the M1 ORF (Fig. 7c).The 'M42-up' virus was mutated to make M42 splicing more efficient, as confirmed by primer extension assay here (Fig. 7d) and in a prior study 57 .We found that an increase in M42 transcript from undetectable to abundant levels had no effect on the efficiency of M-SL9 presentation, suggesting that M42 is not an important source of M-SL9 presentation in PR8 infection (Fig. 7e).In fact, the sole RNA species encoding M-SL9 that our assay detected was the unspliced M1 transcript.PR8 was also mutated to ablate AUG2 ('ΔAUG2') (ref.57) to test its role in M-SL9 presentation, and this largely eliminated activation of B6.23 cells but not of a control hybridoma specific for another epitope from PR8 (Fig. 7e).A stop codon was then introduced three codons upstream of M-SL9 ('preSTOP') to test whether M-SL9 is translated via ribosomal frameshifting 20 from the M1 ORF into the +1 RF between AUG1 and the mutated codon.Frameshifting in this region was ruled out, as M-SL9 presentation was not abrogated by this mutation (Fig. 7e); if anything, presentation was increased globally, possibly due to slightly increased M2 transcript expression 59 by this mutant (Fig. 7d).
To test the requirement for AUG2 in vivo, C57Bl/6 mice were infected with PR8 WT or ΔAUG2 viruses and the CD8 T cell response was evaluated.By tetramer staining and intracellular cytokine staining, M-SL9-specific responses were robust in WT-infected mice and undetectable in ΔAUG2-infected mice, indicating that AUG2 is strictly required for M-SL9 immunogenicity (Fig. 7f).As a control, NP 366-374 responses were comparably high in both cases.Collectively, these findings support a model in which M-SL9 presentation principally derives from 'leaky ribosomal scanning' 17,60 where the ribosome binds to the unspliced M1 transcript but scans past AUG1 and initiates at AUG2, resulting in M-MG16 peptide as the likely precursor to M-SL9.
To understand what other IAV strains might express and present M-SL9, we examined the conservation of this epitope and the preceding AUG2 codon across publicly available human and avian IAV sequences 5.24%  from 1980 until the present.The AUG2 codon was intact in ~100% of human H3N2 and pre-2009 H1N1 sequences but was replaced with valine in all post-2009 (pandemic) H1N1 strains (Extended Data Fig. 10a,b), and all lineages encoded a sequence with homology to M-SL9.Representative sequences since 1999 showed slow amino acid evolution but are all predicted to bind Qa-1 and HLA-E with moderate or strong confidence (Supplementary Table 2).Strikingly, avian H5N1 sequences from the current outbreak uniformly encode an intact AUG2 codon followed by a peptide with 100% identity to the M-SL9-P variant (Extended Data Fig. 10c) present in some isolates of PR8.These findings raise the possibility that M-SL9-like epitopes could be immunogenic in H3N2 and H5N1 strains of IAV currently circulating in humans and birds.

Cytolytic M-SL9 response is induced by mRNA vaccination
The mechanisms governing the induction of nonclassical CD8 T cell responses by infection and vaccination are poorly understood.In recent years, mRNA vaccines have proven to be a versatile and potent vaccine platform with outstanding clinical efficacy against coronavirus disease 2019 (COVID-19) 61 .However, the ability of this vaccine class to elicit nonclassical CD8 T cells has not yet been explored.Here, we asked whether an mRNA vaccine would be able to elicit an M-SL9/Qa-1 b -specific CD8 T cell response and whether this would exhibit physiologically relevant effector functions in vivo.M-SL9 was genetically fused to the C-terminus of a highly immunogenic 62 antigen, the surface glycoprotein (GP) of lymphocytic choriomeningitis virus (LCMV), and nucleoside-modified mRNA encoding this fusion protein ('GP:M-SL9') was encapsulated in lipid nanoparticles (LNP) following the design used by the Moderna COVID-19 vaccine 61 (Fig. 8a).C57Bl/6 mice were vaccinated with GP:M-SL9 mRNA-LNP or severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike mRNA-LNP as a negative control, and a robust circulating M-SL9/ Qa-1 b -specific CD8 T cell response was detected by tetramer staining (Fig. 8b).Interestingly, M-SL9-specific T cells were abundant in the blood following intraperitoneal mRNA vaccination, in contrast to intranasal PR8 infection.Ten days after vaccination, in vivo cytolysis activity was tested by intravenously infusing vaccinated mice with naive donor spleen cells pulsed with M-SL9 or irrelevant peptide, which were labeled with high or low concentrations of carboxyfluorescein succinimidyl ester (CFSE) dye (Fig. 8c).After 18 h, spleens were recovered from recipient mice and analyzed by flow cytometry.M-SL9-pulsed cells were cleared in an epitope-and vaccine-specific manner at a mean rate of 46 ± 5.1% (s.e.m.) (Fig. 8d), and clearance of M-SL9-pulsed cells was significantly correlated with the magnitude of the M-SL9/Qa-1 b tetramer + response (Fig. 8e).To our knowledge, this is the first demonstration of a nonclassical cytotoxic T cell response elicited by an mRNA vaccine.

Discussion
In this study, we employed an immunopeptidomic approach and identified a dominant CD8 T cell response against an IAV PR8 epitope, M-SL9, with cryptic genomic origin and unexpected Qa-1 restriction in C57Bl/6 mice.The discovery of M-SL9 in some ways resembles that of another PR8 epitope, NS1-ARF2 1-8 , which was found to elicit a CD8 T cell response in BALB/c mice and was mapped to a noncanonically translated alternative RF of the NS1 protein-coding sequence, believed to be translated into a nonfunctional 14-mer peptide 26 .In both cases, the cryptic CD8 T cells accounted for a major fraction of the T cell response against IAV and demonstrated a cytolytic phenotype.Interestingly, NS1-ARF2 1-8 was also discovered serendipitously, as it was the dominant specificity of CD8 T cells from PR8-infected mice after repeated in vitro stimulation and expansion by PR8-infected APCs.This work and ours raise the intriguing possibility that noncanonically translated epitopes, defective ribosomal products 63 and other kinds of cryptic epitopes 64 may drive a larger fraction of the T cell response than previously thought, arguing for the use of comprehensive methods to explore the immunopeptidome rather than relying on traditional approaches such as screening with overlapping peptide libraries.This may be especially important in autoimmune disease and cancer, where T cell-based therapies are particularly valuable but where the epitopes Schematic of possible M-SL9 precursors: M-SL9 presentation from possible precursors are often unknown and may be cryptic in origin [65][66][67] .It is even possible that cryptic epitopes are overrepresented in the immunopeptidome 68 , since nonfunctional translation products are expected to skew toward short, unstructured polypeptides that are rapidly degraded, and MHC-I presentation appears to be dominated by rapidly degraded polypeptides [69][70][71] .Several aspects of the M-SL9-specific T cell response appear to be exceptional among IAV-specific CD8 T cell responses.In particular, M-SL9 was by far the dominant CD8 T cell specificity in the lung at 6 days postinfection-a critical time point associated with falling viral loads and rising disease symptoms in mice 72,73 .This finding prompts the hypothesis that M-SL9-specific and possibly other Qa-1-restricted CD8 T cells may play a role as 'early responders' in the lung during IAV infection, giving the classical CD8 T cell specificities such as NP 366-374 and PA 224-233 sufficient time to reach protective levels.There is precedent for just such a niche being filled by other nonclassical CD8 T cells restricted to the MHC-Ib molecule H2-M3 responding to Listeria monocytogenes 74 .Another notable property of M-SL9-specific CD8 T cells is their retention in the lung and efficient differentiation into lung T RM cells.As noted previously, T RM cells are considered an important component of cross-strain immunity to IAV 53,54 , and therefore it will be valuable to ascertain whether Qa-1-restricted CD8 T cells contribute to IAV protection both at early time points (day 6) and at memory time points.Interestingly, M-SL9-specific cells also appeared to be amenable to 'flexible programming' in terms of their trafficking properties, since infection with X31 and an mRNA vaccine encoding M-SL9 were both able to induce M-SL9/Qa-1-specific cells to enter the circulation.This suggests the possibility that a vaccine or immune therapy harnessing MHC-E-restricted T cells might be able to direct tissue trafficking with some degree of specificity.For all the unique properties discussed above, it remains to be determined whether these are generalizable properties of Qa-1-restricted CD8 T cells or unique to M-SL9.It is also unclear whether Qa-1 restriction is a common feature of cryptic T cell responses or whether it was just a coincidence that M-SL9 is both cryptic and Qa-1-restricted.Future immunopeptidomic studies that analyze the full breadth of Qa-1 ligands presented on virus-infected cells will be critical to shed light on these questions, offering the possibility to analyze multiple Qa-1 epitope specificities and thus discern more general properties of MHC-E-restricted CD8 T cells.
The importance of Qa-1 (MHC-E) restriction by M-SL9 is augmented by the fact that MHC-E-restricted CD8 T cells are the only type of immune response so far that has been shown to protect against a pathogenic primate lentivirus infection.The Picker laboratory found that MHC-E-restricted CD8 T cells elicited by a CMV vector were able to control and eventually clear a severely pathogenic SIVmac239 infection in 50% of rhesus macaques, while classical CD8 T cells were totally unprotective 3 .The regulation of classical versus nonclassical CD8 T cell responses is largely undefined, but recent data implicate myeloid cell tropism of the vector as a contributor to the nonclassical response 4 .Since M-SL9 represents the first published viral epitope restricted to Qa-1 in mice, this provides a much-needed mouse model in which to investigate underlying mechanisms.We note that CD8 T cells are necessary but not sufficient for protection from IAV challenge in mice 75,76 , and therefore it was outside the scope of this study to establish an experimental system in which to test for protection by M-SL9-specific T cells, but this will be an important area of follow-up investigation.
T cell-based vaccines and immune therapies that harness MHC-E-restricted CD8 T cells present at least two major advantages over classically restricted CD8 T cells.First, HLA-E is ubiquitously expressed and minimally polymorphic across the human population, having only two major allomorphs (HLA-E*01:01 and HLA-E*01:03) with identical peptide-binding grooves 77 .Therefore, vaccines and therapies could be designed for the general population as opposed to the personalized approaches that are often necessary when raising MHC-Ia-restricted T cells 78,79 .Second, the evidence from Picker and colleagues suggests that MHC-E-restricted T cells have unique protective properties compared with classically restricted T cells, leading to interest in targeting MHC-E epitopes in vaccine strategies against infectious pathogens and cancer 80 .In the present study, we demonstrated that a strong, circulating cytotoxic T cell response could be raised against M-SL9/Qa-1 b by an mRNA vaccine using the same approach as the highly successful COVID-19 vaccine from Moderna.This provides a proof-of-concept that mRNA vaccines can be used to explore the regulation and protective capabilities of MHC-E-restricted CD8 T cells, and it provides optimism that this safe and effective platform could be used to elicit these therapeutically promising T cell responses in humans.

Viruses and infections
The PR8 virus was an infectious molecular clone derived from a set of eight ambisense plasmids in the pDZ vector 37 , which encode each of the PR8 gene segments and are transcribed into both negative sense genomic RNA and positive sense mRNA when transfected into cells.WT PR8 virus stock and pDZ plasmids were a kind gift from Scott E. Hensley (University of Pennsylvania).Mutant PR8 viruses were created by site-directed mutagenesis of the pDZ M gene segment, making the following mutations based on the numbering used in Extended Data Fig. 3: U148A for M42-up, G145A for M42-down, U115C for ΔAUG2 (ref.57) and U109A for preSTOP.These viruses and WT PR8 were launched in a co-culture of highly transfectable 293T cells and infection-susceptible MDCK cells and expanded in MDCK cells with OptiMEM medium containing 0.3% bovine serum albumin (BSA), 0.01% FBS, 0.1 mg ml −1 CaCl 2 , penicillin-streptomycin and 3 μg ml −1 of trypsin treated with l-(tosylamido-2-phenyl) ethyl chloromethyl ketone (Worthington LS003750).PR8 plasmids and matrix RNA were sequence verified, and WT PR8 was identical to GenBank accession AF389121.Influenza virus B/Lee/1940 was propagated in the allantoic fluid of day 10 embryonated chicken eggs.
Mice were infected with 25-200 FFU of WT or mutant PR8 by instilling 20-40 μl of PBS-diluted virus dropwise in the right nostril under isoflurane anesthesia.Early experiments used lower virus doses (25-40 FFU), while later experiments used higher doses (160-200 FFU) that yielded more consistently strong M-SL9-specific CD8 T cell responses, particularly when comparing male and female mice.Mice were monitored between days 5 and 10 after infection and euthanized if their clinical condition became severe (humane endpoint per the IACUC protocol).

Immunoprecipitation of peptide/MHC complexes
B6-CIITA-E d cells (200 × 10 6 ) were either mock infected or PR8 infected in suspension in 0.1% BSA/PBS for 1 h and incubated at 37 °C, 6% CO 2 for ~18 h.Cells were collected with PBS containing 2 mM ethylenediaminetetraacetic acid (EDTA).Cell lysis and immunoprecipitation of peptide/MHC-II complexes were carried out largely as previously described 84 , and all chemicals used after cell collection were analytical grade or equivalent when available.Briefly, anti-MHC-II antibody (clone M5/114.5,BioXCell) was cross-linked to Protein G Sepharose beads (GE 17061801).Lysates and beads were co-incubated and rotated overnight at 4 °C and then transferred to cleaned glass columns (Bio-Rad 7371012).After extensive washing, peptide/MHC-II complexes were eluted in 10% acetic acid, and the collected material was lyophilized.

LC-MS 2
Peptides were analyzed by LC-MS 2 using an UltiMate 3000 RSLCnano System supplemented with a PepMap C18 column (2 μm particle size, 75 μm × 50 cm, Thermo Fisher) directly interphased with an Orbitrap Fusion Lumos Tribrid™ Mass Spectrometer (Thermo Fisher).A 60 min linear gradient from 3% to 25% ACN in 5% dimethyl sulfoxide (DMSO)/0.1% formic acid at a flow rate of 250 nl min −1 was applied for peptide elution.Peptide ions were introduced to the mass spectrometer using an Easy-Spray Source at 2,000 V. Detection was performed with a resolution of 120,000 for full MS (300-1,500 m/z scan range), and precursors were selected using TopSpeed ion selection within a 2 s cycle time, and a quadrupole isolation width of 1.2 a.m.u. for fragmentation.For MS 2 acquisition, the resolution was set at 30,000 and high-energy collisional dissociation energy was set at 28 for peptides with two to four charges and 32 for peptides that were singly charged.

LC-MS 2 data analysis
LC-MS 2 datasets were analyzed using PEAKS v8.5 with the following parameters: precursor mass tolerance: 5 ppm; fragment mass tolerance: 0.03 Da; digestion: none; fixed modifications: none; variable modifications: 313 (PEAKS posttranslational modification search).The search database included the full reference sequences of the pDZ PR8 gene segments translated in all six RFs (positive and negative sense), irrespective of start and stop codons, concatenated with the reviewed mouse entries in Uniprot (Swissprot version 17 July 2019).For immunopeptidomics data, a threshold of −10logP =15 was applied across all datasets, and the false discovery rate for peptide/sequence spectrum matches was reported as 3.4% for PR8-infected samples and 2.8% for uninfected control samples.Subsets of peptide identifications were analyzed by sequence logo diagrams (Extended Data Fig. 1).

Preparation of mouse tissues
Mouse bone marrow was collected by flushing it from the femur and tibia with R10 medium.BMDCs were prepared by plating 2 million cells in 10 ml R10 medium in 10 cm tissue culture plates with 50 ng ml −1 granulocyte macrophage colony-stimulating factor ('GM-CSF', Shenandoah), with complete replacement of medium on day 3 and partial medium replacement as needed between day 6 and collection on day 9 or 10.
Mouse T cell responses were measured using cell suspensions prepared from various tissues as follows.BALF was collected immediately after killing by carbon dioxide inhalation (before lung collection) by exposing the trachea, inserting a catheter and securing with a suture tie, and using a syringe to gently lavage the lower airway with 1 ml of PBS/0.1 mM EDTA three times.BALF was collected into PBS 1% FBS, and ammonium-chloride-potassium (ACK) lysis was performed when red blood cells were visually detectable.
Lung samples were prepared by first injecting the right cardiac ventricle of mice with 3 ml of PBS 1% FBS to perfuse the lungs; then lungs were collected in gentleMACS C tubes containing 1% FBS in PBS on ice.Digest media was added to achieve a final concentration of 2.25 mg ml −1 sterile-filtered Collagenase D (Sigma 11088866001) and 0.15 mg ml −1 DNase I in 4 ml of 1% FBS in PBS.Lungs were disrupted using gentleMACS Dissociator program m_spleen_01.01 and incubated for 45 min at 37 °C with shaking.Then 10 ml of R10 medium was added to each tube, followed by further homogenization using gentleMACS Dissociator program m_lung_02.01.Digested lungs were passed through a 70 μm strainer, incubated in ACK lysis buffer to remove red blood cells, resuspended in R10 medium and then passed through a 40 μm strainer to obtain a single-cell suspension.
The MLN was collected after perfusing/expanding the lung with PBS 1% FBS (see above) but before removal of the lung.This was confirmed as the lung-draining lymph node based on marked expansion upon IAV infection.The MLN was collected with forceps into PBS 1% FBS and homogenized similarly to spleens (below), and ACK lysis was performed if red blood cells were visually detectable.
Spleens were collected in PBS and manually homogenized through a 70 μm cell strainer using the hard end of a syringe plunger, followed by ACK lysis and passage through a 40 μm strainer to prepare a single-cell suspension.
Blood was collected from the submandibular vein of live mice into a microcentrifuge tube containing PBS with 25 mM EDTA.Red blood cells were lysed using two incubations in ACK lysis buffer and the remaining cells were resuspended in R10 medium.
All primary cell samples were maintained at 0-4 °C during preparation (except as otherwise specified) and until analysis.

IFN-γ ELISpot for T cell activation
Spleens were collected from C57Bl/6 and BALB/c mice on day 10 postinfection with PR8 (25 FFU).CD4 T cells were purified from the splenocytes by negative isolation using Dynabeads Untouched Kit for mouse CD4 T cells, following the manufacturer's protocol (Thermo Fisher).Mouse IFN-γ ELISpot assays were carried out following the manufacturer's protocol (BD Biosciences).DC2.4 (C57Bl/6) or A20 (BALB/c) cells were used as APCs and cultured in R10 supplemented with 2 units ml −1 murine IFN-γ (BEI NR-3081) for 2 days before the assay.5 × 10 4 APCs were co-cultured with 1 × 10 5 purified CD4 T cells, or 3 × 10 5 bulk splenocytes were cultured alone, with 10 μg ml −1 of peptide or DMSO vehicle, overnight at 37 °C before development of the ELISpot plate.IFN-γ spots were imaged and counted on a CTL ImmunoSpot S6 Universal Analyzer.

Intracellular cytokine and cytolytic marker staining for T cell activation
To perform intracellular cytokine staining, 1 million cells per well were stimulated in U-bottom 96-well tissue culture plates in R10 medium in the presence of synthetic peptides (10 μg ml −1 final concentration) and anti-CD28 co-stimulation (2 μg ml −1 ; Tonbo 40-0281).Secretion was inhibited with 5 μg ml −1 brefeldin A (Biolegend 420601), and cells were incubated for 14 h at 37 °C, 6% CO 2 .A viability stain was performed, followed by Fc receptor blockade, surface stain for CD8α, fixation and permeabilization and intracellular stains for CD3ε, CD4, IFN-γ, TNF and IL-2.Cytolytic markers were analyzed similarly except secretion inhibition was delayed by 1 h after antigen stimulation, 2 μM monensin and 2.5 μg ml −1 anti-CD107a were included along with brefeldin A in the 14 h incubation, and intracellular stains included granzyme B and perforin.Mice that did not show ≥2% of granzyme B + CD8 T cells in unstimulated (DMSO vehicle) lung samples were excluded from antigen-specific lung T -cell analyses as it was not clear that virus infected the lower respiratory tract.All reported cytokine frequencies are background subtracted, with the DMSO condition serving as background for peptide-stimulated conditions.

Hybridoma activation assay
Hybridoma recognition of cognate peptide/MHC complexes results in β-galactosidase production, detected using a fluorometric substrate, 4-methyl-umbelliferyl-β-D-galactopyranoside (Sigma M1633).Briefly, APCs were treated with antigen in various ways and then co-cultured overnight with hybridoma cells at a 1:2 ratio.Antigen stimulation with peptide was performed by pulsing APCs with mixtures of synthetic peptides (37.5-50 μg ml −1 per peptide) for 2 h in PBS and washing three times.In other cases, APCs were infected with influenza virus for 45 min in serum-free DMEM.In other cases, 50,000 APCs per well were transfected with mRNA using TransIT-mRNA (Mirus, MIR 2225), with 0.34 μl of TransIT-mRNA reagent and 0.22 μl of mRNA Boost reagent in a final volume of 200 μl per well.After ~18 h of co-culture, cells were treated with one-fifth volume of lysis/substrate buffer containing PBS with 1.25% Triton X-100, 33 μg ml −1 4-methyl-umbelliferyl-β-D-galactopyranoside, 38.5 μM 2-mercaptoethanol and 9 mM MgCl 2 , incubated for 3 h at 37 °C, 6% CO 2 and then ~24 h at 4 °C.Fluorescence (excitation 365 nm and emission 445 nm) was detected using an Infinite M200 Pro Plate Reader.APCs were validated for the expected expression of MHC molecules by flow cytometry.All antigen presentation assays contained three technical replicates per experiment, averaged together, and at least three independent experiments.

MHC stabilization assay with RMA-S cells
Peptide binding to MHC-Ia was tested using a method based on RMA-S cells deficient in TAP 38 .Briefly, RMA-S cells were incubated for 18 h at 26 °C to accumulate unstable 'empty' surface MHC-Ia molecules, lacking bound epitopes.The cells were then incubated at 37 °C for 1 h to allow internalization of unstable MHC and then stained for surface H2-D b and H2-K b and analyzed by flow cytometry.

TCR analysis of sorted cells
C57Bl/6 mice (n = 3 males and 3 females) were infected with 40 FFU (females) or 200 FFU (males) and confirmed to lose weight as a marker of infection.Lungs were collected and three populations of CD8 T cells were isolated from each lung by fluorescence-activated cell sorting (FACS): naive (CD44 − CD62L + ), M-SL9-specific (CD44 + M-SL9/Qa-1 b -APC tetramer + ) and NP 366-374 /H2-D b -PE tetramer + ).Roughly 89,000-390,000 cells were collected per specificity from each lung.Genomic DNA was isolated by the QIAamp DNA Micro Kit (Qiagen 56304), and the hypervariable region of the recombined Tcrb genes was sequenced, identified and analyzed for abundance by immunoSEQ (Adaptive Biotechnologies).Simpson's clonality was calculated by immunoSEQ as 1−Simpson diversity, a measure of alpha diversity.Further analyses of Tcrb V and J gene usage and CDR3 lengths were performed using the Immunarch software package in R, including spectratype graphs, a principal component plot and gene usage correlation plot.

mRNA production
Preclinical-grade mRNA was produced using previously described methods [85][86][87] .The constructs of interest (full-length PR8 matrix gene segment, M-MG16, M42, enhanced green fluorescent protein, LCMV GP:M-SL9 fusion protein or SARS-CoV-2 spike) were encoded by an mRNA production plasmid used previously 85 .The SARS-CoV-2 spike sequence contained a mutation ablating the furin-cleavage site and was published previously 86 .The GP:M-SL9 sequence was prepared using the LCMV GP protein sequence from GenBank accession NP_694851.1,immediately followed by the M-SL9-coding sequence and then a stop codon.To maximize translation of mRNA vectors, GP:M-SL9 and spike coding sequences were codon optimized.All mRNAs were produced by in vitro transcription (MEGAscript T7 Transcription Kit, Thermo AMB13345) using N1-methylpseudouridine (TriLink N-1081) in place of uridine, co-transcriptionally capped with the cap1 structure via CleanCap (TriLink N-7413) and purified of double-stranded RNA contaminants by adsorption to cellulose 87 (Sigma 11363).The mRNA length and integrity were confirmed by agarose gel electrophoresis.

In vivo cytolysis assay
C57Bl/6 mice were immunized with a dose of spike or GP:M-SL9 mRNA-LNP vaccine containing 18 μg of mRNA by the intraperitoneal route.Ten days later, blood was collected from the submandibular vein and the T cell response was evaluated by M-SL9/Qa-1 b tetramer stain.Also on day 10, a single-cell suspension was made from the spleens of naive donor mice (sex-matched C57Bl/6).Cells were divided and (1) labeled with a low concentration of 0.6 μM CFSE (Thermo Fisher C34554) and pulsed with SIINFEKL peptide, or (2) labeled with a high concentration of 5 μM CFSE and pulsed with M-SL9 peptide.CFSE labeling was performed in PBS for 15 min at 37 °C, followed by washing.Peptide pulsing was performed with 20 μg ml −1 peptide in R10 medium for 40 min at 37 °C, followed by washing.These two cell populations were mixed at a 1:1 ratio, suspended in PBS and injected into the vaccinated mice via the retro-orbital sinus.After 18 h, spleens were recovered from these mice and analyzed by flow cytometry.In vivo cytolysis was read out as the percent specific clearance of M-SL9-pulsed (CFSE high ) cells relative to SIINFEKL-pulsed (CFSE low ) cells, normalized to the control spike-vaccinated mice, according to the formula: %Specific https://doi.org/10.1038/s41590-023-01644-5killing = 100 × (1 − (%CFSE high Vaccine /%CFSE low Vaccine )/(%CFSE high ControlAvg /%C FSE low ControlAvg )), as previously described 88 , where %CFSE low and %CFSE high represent percents of all singlet lymphocytes.

Primer extension analysis of viral RNA
Total cellular RNA was isolated from virus-infected (multiplicity of infection (MOI) of 20) or mock-infected B6-CIITA fibroblasts at 9 h postinfection.TRIzol and chloroform were added and RNA was extracted and precipitated from the aqueous layer.Then, 10 μg of each RNA sample was subjected to primer extension assay with 32 P-labeled primers targeting the IAV matrix segment (5′-GTATTTAAAGCGACGATAAATGCATTTG-3′) and mouse U1 snRNA (5′-TCAGCACATCCGGAGTGCAATGG-3′).Primer extension and sample denaturing were performed as previously described 89 , and samples were resolved on a denaturing 5% polyacrylamide gel (5% polyacrylamide, 1 M urea, 0.5× Tris/borate/EDTA buffer, 0.2% ammonium persulfate and 0.1% N,N,N′,N′-tetramethylethylenediamine).Gels were dried on a gel dryer (Bio-Rad, model 573), exposed on a phosphor screen (GE Healthcare), and imaged on an Amersham Typhoon (GE Healthcare).Relative abundances were quantified using ImageJ software.The primer extension gel was performed twice.

IAV sequence analysis
Variation of the M-SL9 sequence was investigated by accessing the Global Initiative on Sharing All Influenza Data (GISAID) database 90 to download matrix gene segment nucleotide sequences isolated from human or avian species during various intervals since 1980, and preparing sequence logo diagrams using the ggseqlogo 91 software package in R (Extended Data Fig. 10).Sequences from IAV strains recommended by the World Health Organization for seasonal influenza vaccines were downloaded from GISAID, Bacterial and Viral Bioinformatics Resource Center (BV-BRC) or the Influenza Research Database and analyzed as representative examples.Binding predictions for Qa-1 b and HLA-E allomorphs are provided as percentiles in Supplementary Table 2 and were obtained with the Immune Epitope Database (IEDB) MHC-I Binding Prediction tool using the NetMHCpan EL 4.1 algorithm.

Statistics
Intracellular cytokine, cytolytic marker, tetramer frequencies and fluorescence units in antigen presentation assays were normally distributed and were analyzed by one-way analysis of variance (ANOVA) (Brown-Forsythe and Welch test with Dunnett T3 method to correct for multiple comparisons) or two-way ANOVA for time course experiments (with Tukey test to correct for multiple comparisons).In vivo cytolysis results were analyzed by a two-tailed Student's t-test as there were only two groups.Microsoft Excel was used to manage data.GraphPad Prism v.9.5.1 was used to compute ANOVA and t-test results and to plot the four-parameter logistic sigmoidal curve in Extended Data Fig. 5. Immunarch 0.9.0 was used in R 4.3.0 to perform the principal component and Pearson correlation analyses of TCR gene usage, and ggseqlogo version 0.1 was used to create sequence logo diagrams in R.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.Extended Data Fig. 7 | Qdm/Qa-1 b tetramer co-stains with NKG2A/C/E.a, Lung lymphocytes from naïve C57Bl/6 mice were stained with an anti-NKG2A/C/E mAb and Qdm/Qa-1 b , M-SL9/Qa-1 b , and control NP 366-374 /D b tetramers at 37 °C and gated on CD3 − CD19 − cells to interrogate natural killer (NK) cells.NK cells expressing NKG2A/C/E (the natural receptors for Qdm/Qa-1 b ) were the only population that stained with Qdm/Qa-1 b tetramer, but neither these nor other NK cells stained with M-SL9/Qa-1 b tetramer.b, C57Bl/6 mice were intranasally infected with 160 FFU of PR8 and 9 days later lung lymphocytes were stained with anti-NKG2A/C/E and the indicated tetramers at 37 °C.Qdm/Qa-1 b tetramer generally stained PR8-induced CD8 T cells in a manner that was dependent on NKG2A/C/E but independent of TCR specificity.isolates from all avian species, downloaded between April and June 2023 for the indicated sample collection time periods.The BV-BRC database was used for sequences from 1980-1999, while the GISAID database was used for all others.Diagrams were created using the ggseqlogo package in R, and the y-axis units are the probability of each amino acid from 0 to 1. H1N1 sequences after 2009 correspond to the swine-origin pandemic H1N1 lineage, while H1N1 sequences prior to 2009 are from the earlier seasonal H1N1 lineage; sequences from 2009 were omitted to avoid ambiguity.Amino acids are numbered so that position 1 corresponds to the initial serine residue of the M-SL9 epitope, and the preceding residue was designated as position −1 and shown to assess the presence of an initiation codon.The two forms of M-SL9 encoded by PR8 isolates are shown at bottom; note that the avian H5N1 consensus sequence exactly matches the M-SL9-P amino acid sequence.

Fig. 1 |Fig. 2 |
Fig. 1 | Discovery of co-immunodominant CTL response directed against M-SL9, an epitope encoded by a noncanonical IAV matrix ORF.a, Outline of immunopeptidomics and first identification of M-SL9-reactive spleen cells by ELISpot.b, Schematic of M-SL9 position within the PR8 flu matrix gene segment.c-e, C57Bl/6 mice were intranasally infected with 40 FFU (c and d) or 160 FFU PR8 (e); lung lymphocytes were stimulated with indicated peptides and intracellular cytokines (c and d) or cytolytic markers were detected within the CD8 + CD4 − compartment (e).Pie charts represent mean proportions in infected mice, and flow plots are representative.In c and d, N = 34 infected mice and N = 21 naive mice across six independent experiments.In e, N = 11 infected mice and N = 4 naive mice across two independent experiments.In e, percentages on flow plots describe the frequency of CD107a + (blue) cells in each quadrant out of the whole CD8 + T cell population.Error bars are s.e.m.A one-way ANOVA was performed on all cytokine + (d) or triple-marker + cells (e).IP: immunoprecipitate, UTR: untranslated region, GzmB: granzyme B.

Fig. 3 |
Fig. 3 | Immunophenotype and clonal diversity of M-SL9/Qa-1-specific CD8 T cells.a-c, Lung lymphocytes from mice 9 days after intranasal infection with 160 FFU of IAV PR8 or from naive controls were evaluated for expression of TCRβ versus TCRγδ (a), CD8α versus CD8β (b) and CD44 (marker of antigen experience) versus CD62L (marker of lymphoid homing) (c).Antigen specificities were revealed by tetramer staining.d,e, Tcrb genes from sorted naive or tetramerspecific lung CD8 T cells were sequenced and ranked by frequency within each of N = 6 mice (d).The mean Simpson clonality index (C) is shown above each group.'Spectratype' histograms (e) show the average frequency of Tcrb V gene usage and CDR3 lengths for each T cell specificity.In a, N = 10 and in b and c, N = 6 across ≥2 independent experiments each.All studies used female and male mice.In a-c, the dots show individual mice and bars show the mean.P values are shown from one-way ANOVAs.tet: tetramer, EFF: effector, CM: central memory.

Fig. 5 |
Fig. 5 | Efficient lung-resident memory cell differentiation of M-SL9/Qa-1-specific CD8 T cells.C57Bl/6 mice were intranasally infected with 160 FFU of IAV PR8 (the same mice examined in Fig. 4) and lungs were analyzed for expression of the trafficking molecules and lung T RM cell markers, CD103 and CD69.a, Representative gating strategy showing tetramer + antigen-specific

Fig. 7 |
Fig.7| M-SL9 presentation depends on the second AUG codon of the matrix gene segment.a, Schematic of possible M-SL9 precursors.b, B6-CIITA fibroblasts were transfected with mRNA encoding the constructs in a and co-cultured with B6.23 hybridoma to read out M-SL9 presentation.c-e, PR8 was mutated as shown (c), then used to infect B6-CIITA cells to visualize matrix RNA species (d) and tested for M-SL9 and control epitope presentation to hybridomas (e).f, C57Bl/6 mice were infected with 160 FFU of the same viruses and day 9 lung CD8 T cell responses were quantified by tetramer and intracellular cytokine staining.N = 10 or 12 infected mice per group across two independent experiments, with female and male mice shown as circles and diamonds, respectively.Horizontal bars are mean ± s.d.In a and c, the M-SL9 epitope is highlighted in yellow.In b and e, data points represent the means of three independent experiments normalized to background ± s.e.m.In b, e and f, a one-way ANOVA was performed for doses of mRNA (b) or virus (e) yielding the greatest range of values, excluding negative controls, and P values are shown for comparisons of interest.

Fig. 8 |
Fig. 8 | M-SL9/Qa-1-specific CD8 T cells are raised by mRNA vaccination and mediate cytolysis in vivo.a, C57Bl/6 were immunized intraperitoneally on day 0 with mRNA-LNP vaccines encoding M-SL9 fused to the C-terminus of the LCMV surface GP (GP:M-SL9; N = 8 mice) or SARS-CoV-2 spike protein (N = 7 mice), as a negative control, across two independent experiments.b, On day 10, M-SL9/ Qa-1-specific CD8 T cells in blood were quantified by tetramer staining, with a representative plot shown.c,d, In vivo cytolysis assay: later on day 10, vaccinated mice were intravenously injected with ~5 × 10 6 spleen cells from naive donor mice in a 1:1 ratio of M-SL9 peptide-pulsed and SIINFEKL peptide-pulsed cells that were

Fig. 1 |
Peptide analysis by mass spectrometry (MS).a,b, Histograms of peptide lengths of unique peptide species identified from (a) any origin and (b) IAV origin.c-f, NetMHCIIpan 4.0 was used to predict peptide:MHC-II affinity (K D ) values for core epitope sequences of 9 amino acids from MS-identified peptides of all lengths.Sequence logo diagrams were prepared using unique core epitopes predicted to bind to (c,d) I-A b or (e,f) I-E d with a K D < 2,000 nM for peptides of (c,e) any origin or K D < 10,000 nM for peptides of (d,f) IAV origin.These sequence logos clearly exhibit the expected MHC-II ligand sequence motifs (based on ref.34).g, The sequence logo diagram for all unique 9-mer peptide identifications (appearing as a small local peak in panel 'a') shows strong similarity to the H2-K b and H2-D b sequence motifs, but not to the Qa-1 sequence motif (motifs available from NetMHCpan 4.0 at ref.82).In sequence logo diagrams, 'bit' is a unit of relative amino acid frequency that is inversely related to the Shannon entropy of each position.Extended DataFig. 2 | Identification of M-SL9 by mass spectrometry and validation of immunogenicity in C57Bl/6 mice.a, MS 2 spectrum resulting in M-SL9 identification, with b and y ion fragments indicated, along with mass/ charge (m/z), retention time (RT), and P-value.b, c, Spleens were recovered from C57Bl/6 (N = 3) or BALB/c (N = 4) mice 9 days after infection with IAV PR8, and either bulk spleen cells or isolated spleen CD4 T cells were stimulated overnight with synthetic M-SL9 peptide, positive control peptides, or DMSO vehicle, and secreted IFN-γ was detected by ELISpot.b, For C57Bl/6 mice, the MHC-Ia control peptide was NS2 109-121 ; the MHC-II control peptides were NP 264-280 and NP 306-322 ; and DC2.4 cells were used as the APC to stimulate CD4 T cells.c, For BALB/c mice, the MHC-II control peptides were NP 55-71 and HA 121-137 , and A20 cells were used as the APC to stimulate CD4 T cells.Data points represent individual mice in one independent experiment; bars are the mean +/− s.e.m.Extended Data Fig. 3 | Annotation of M-SL9 coding sequence within the full IAV PR8 matrix gene segment.The positive-sense RNA sequence is shown on top.Primary and secondary AUG codons are underlined and labeled, M-SL9 amino acids are highlighted in yellow, the M1 protein sequence is in light blue, M-MG16 nucleotides including stop codon are colored red, and relevant splice sites are labeled.The first nucleotide of each codon is aligned with the singleletter amino acid code and the first digit of the nucleotide number.This sequence was used to generate PR8 for this study and matches the sequence in GenBank accession AF389121.Extended Data Fig. 4 | Intracellular cytokines, cytolytic markers, and FoxP3 expression in individual mice.a-d, Representative gating strategy for intracellular cytokine and cytolytic marker staining (lower boxes from two different experiments/stains). b-e, Lymphocytes were isolated from naïve (N = 21) or PR8 flu-infected (N = 34) C57Bl/6 mouse lungs, stimulated with indicated peptides, and stained for the indicated markers.b-d, Data for individual mice are shown in the same order for each epitope.c, Comparison of intracellular cytokine responses following infection with 40 FFU PR8 (N = 7), 160 FFU PR8 (N = 7), or no virus (naïve; N = 5), showing more consistent M-SL9 responses to 160 FFU.Female and male mice are indicated by purple and orange bars underneath the graphs.e, CD8 T cells (both total unstimulated as well as peptide-restimulated IFN-γ + cells) from PR8-infected mice (N = 11) stain do not upregulate FoxP3 expression relative to the naïve (N = 4) mouse baseline.Gray events are all CD3 + cells; blue events and blue percentages represent CD3 + CD8 + cells.Bars show the mean +/-s.d. and P-values of interest are shown from a two-way ANOVA with Sidak's multiple comparisons test comparing naïve and PR8-infected conditions.GzmB: granzyme B. Extended Data Fig. 5 | Hybridoma clone B6.23 recognizes two forms of M-SL9 present in isolates of PR8.Amino acid sequences are shown for the originally identified M-SL9, present in pDZ PR8, and M-SL9-P, present in other PR8 isolates (for example GenBank V01099).B6-CIITA fibroblasts served as APCs and were co-cultured overnight with B6.23 cells in the presence of the indicated peptide concentrations.A sigmoidal curve was fit to the data points above (mean +/− s.d.), representative of three independent experiments.The geometric mean half-maximal effective concentration (EC 50 ) values across all three experiments were computed as 940 ng/ml for M-SL9 and 51 ng/ml for M-SL9-P.Extended Data Fig. 6 | Evidence supporting Qa-1 restriction of M-SL9.a, The MHC-Ia molecules H2-D b and H2-K b are not stabilized on RMA-S cells by M-SL9 peptide.RMA-S cells bearing unstable empty MHC-I molecules (due to TAP deficiency) were incubated in the presence of the indicated synthetic peptides, and surface expression of H2-D b and H2-K b was measured by flow cytometry.Mean fluorescence intensities of each stain were normalized to the negative control condition using HA 91-107 , an I-A b -binding epitope with no known binding to H2-D b or H2-K b , and shown as averages +/-s.dfrom 3 independent experiments.H2-D b -binding NP 366-374 and H2-K b -binding SIINFEKL were used as positive controls.b-e, Validation of HeLa cell lines and BMDCs showing Qa-1 restriction of M-SL9.b, The sufficiency of Qa-1 b expression for M-SL9 presentation to its cognate T hybridoma was confirmed using a HeLa cell line transduced with fulllength, wild-type Qa-1 b and an untransduced parental HeLa cell line as a control.Bars are mean +/-s.e.m. from triplicate technical replicates, representative of 3 independent experiments, and P-values were calculated by Welch's t-test (twotailed).c, Qa-1 expression on cell lines used in b was validated by flow cytometry.d, The expected staining pattern was confirmed for HeLa cell lines used in Fig. 2; these lines were transduced with retroviruses encoding chimeric MHC-Ib molecules containing the α3 domain (D3) from H2-D b to allow efficient staining with the H2-D b D3-specific mAb 28-14-8.e, The expected staining pattern was also confirmed for BMDCs used in Fig. 2.
Flow plots are representative and show the gating strategy used, and bars show mean +/-s.e.m. for (a) N = 4 mice and (b) N = 5 to 6 mice per group across 2 independent experiments each.P-values are shown from two-way ANOVA with Dunnett's multiple comparisons test.Extended Data Fig. 8 | Analysis of TCRβ V and J gene usage in sorted CD8 T cell populations.a-d, CD8 T cell populations were sorted by FACS into three populations: naïve (CD44 − CD62L + ), M-SL9-specific (CD44 + M-SL9/Qa-1 b tetramer + ), and NP 366-374 -specific (CD44 + NP 366-374 /H2-D b tetramer + ).Genomic DNA was isolated, the VDJ region of recombined TCRβ-coding genes was sequenced, and gene usage was analyzed by (a,b) the immunoSEQ Analyzer and (c,d) Immunarch.a-b, The frequencies of the top 10 most-used (a) V genes and (b) J genes, on average across all mice, are shown as stacked bar graphs, where each bar represents one mouse.c, Principal component analysis (PCA) of the Tcrb V and J gene usage showing clustering by T cell population.d, Pearson correlation analysis of M-SL9-and NP 366-374 -specific T cells showing greater correlation between mice within each T cell specificity rather than between specificities within each mouse.N = 6 mice, half males and half females.Extended Data Fig. 9 | Tracking CD69 and CD103 expression in PR8-infected mice.a-d, C57Bl/6 mice were intranasally infected with 160 FFU of PR8 and were euthanized at day 6 (N = 7), 9 (N = 5-6), 14, 31 (N = 7), or 56 (N = 9) to collect the indicated tissues/fluids.Uninfected mice were used as day 0 controls (N = 7-11).a, Gating strategy.b, Frequency of all CD3 + T cells in lung and BALF over time, showing the lack of T cell infiltration in uninfected mice (plotted as day 0).Data points were omitted when there were <20 live singlet CD3 + CD8 + T cells collected in total.P-values are calculated from Brown-Forsythe and Welch one-way ANOVA with Dunnett T3 multiple comparisons test comparing each condition to day 0 controls.c, Frequency of CD103 + and CD69 + CD8 T cells (analyzed separately) in lung and BALF starting from the approximate peak of the T cell response on day 9. d, Frequency of CD103 + CD69 + double positive T RM cells in lung at 9 days after PR8 only, X31 only, or PR8 prime and X31 boost.c, d, P-values were calculated by two-way ANOVA with Tukey's multiple comparisons test.Extended Data Fig. 10 | Sequence evolution and variation of the M-SL9 epitope and open reading frame in IAV strains over time since 1980.a-c, Sequence logo diagrams were produced from M-SL9-homologous sequences from (a) human H1N1 isolates, (b) human H3N2 isolates, and (c) H5N1