Influenza A Virus Infection Induces Viral and Cellular Defective Ribosomal Products Encoded by Alternative Reading Frames

CD8⁺ T cells (T_CD8+) eliminate virus-infected cells by recognizing viral oligopeptides displayed by MHC class I molecules on the infected cell surface. The source of such peptides remains an area of intense interest. Typically, the rapid presentation of viral peptides starkly contrasts with the high stability of their ostensible source proteins (1, 2). This observation led to the defective ribosome hypothesis of Ag presentation, which after a number of updates, posits that errors in transcription, translation, and posttranslational protein maturation result in the synthesis of rapidly degraded polypeptides that are exploited for immunosurveillance (3).

The contribution of translational errors to viral peptide generation via frame-shifting and noncanonical or downstream initiation has been elegantly determined in model systems (4–7), but the relative contribution of mistranslation to induction of antiviral T_CD8+ responses is unclear. Ironically, a systematic survey for downstream misinitiation of influenza A virus (IAV) defective ribosomal products (DRiPs) on Met residues led to the discovery of PB1-F2—at the time, the first IAV gene product discovered in decades (8). A number of immunogenic retroviral peptides of uncertain biological significance have been identified from putative alternative reading frames (ARFs) (9–11), but whether the peptides derive truly from ARFs versus short open reading frames (ORFs) or random frame-shifting is uncertain [the shortest known functional viral proteins is 45 residues (12), whereas 11 residue “proteins” are known to function in Drosophila (13)].

IAV possesses a negative-sense single-stranded genome whose eight segments encode 13 known functional proteins (14, 15) and other in-frame truncated proteins that are likely to be functional (14, 16). The synthesis of IAV mRNA in the nucleus enables segments 7 and 8 to encode M1 and NS1, respectively, and M2 and NS2 are created by the host nuclear RNA splicing machinery. PA-X is generated from the PA mRNA by frame-shifting (15).

The mouse IAV infection model has played a key role in deciphering the rules of T_CD8+ immunodominance (8, 17–22). In studying anti-IAV T_CD8+ responses in BALB/c mice, we observed an extremely robust response to a naturally processed peptide that we were unable to identify. After an arduous and extended hunt, we have finally caught our Moby Dick peptide, which provides the clearest example of a viral peptide generated from a natural ARF.

BALB/c mice were purchased from Walter Eliza Hall Institute of Medical Research (Kew, Melbourne, VIC, Australia). Mice were housed in specific pathogen–free isolators. Experiments were performed with animals aged at 6–12 wk, conducted under the auspices of the Austin Health and La Trobe University Animal Ethics Committee, and conformed to the National Health and Medical Research Council Australian code of practice for the care and use of animals for scientific purposes.

IAV peptides (please see Supplemental Table I) were synthesized by Mimotopes at >80% purity (Clayton, Melbourne, VIC, Australia). PE-labeled anti–IFN-γ, PE-Cy7–labeled anti–TNF-α, FITC-labeled anti-CD107α, and allophycocyanin-labeled anti-CD8α were purchased from eBioscience (San Diego, CA). For flow cytometry, Abs were used at 1/300 dilution in PBS supplemented with 10% FCS. For IAV protein staining, cells were fixed with 1% paraformaldehyde and stained in the presence of 0.4% saponin with either anti-NS1 mAb (1A7), anti-NP mAb (HB65), anti-M1 mAb (M2-1C6), or anti-M2 mAb (O19) at 1/1000 dilution, followed by secondary 1/100 anti-mouse-FITC Ab.

IAV A/Puerto Rico/8/34 (H1N1) (PR8), A/Tasmania/2004/2009 (Pan09, H1N1), and NS1-GFP IAV were grown in 10-d embryonic chicken eggs and used as infectious allantoic fluid. The recombinant NS1-GFP IAV (23) was a kind gift from Dr. Adolfo García-Sastre (Department of Microbiology, Icahn School of Medicine at Mount Sinai, New York, NY). Mice were infected with 100 PFU IAV intranasally (i.n.) or 10⁷ PFU IAV i.p.

Mutant and wild-type (wt) IAVs were generated by plasmid-based reverse genetics, following an adaptation of the method by Hoffmann et al. (24). For the transfection, plasmid clones of the eight genomic segments were transfected into cocultured MDCK (4 × 10⁵ cells per well) and HEK293T cells (8 × 10⁵ cells per well) with Fugene 6 (Promega, Sydney, NSW, Australia), according to the manufacturer’s protocol, and incubated at 37°C, 5% CO₂ in a humidified incubator. Transfection medium was carefully removed 6 h posttransfection and replaced by 1 ml of Easy Flu medium (Opti-MEM I medium, containing 10 U/ml penicillin and 100 μg/ml streptomycin). Following 30 h of incubation, 1 ml of Easy Flu medium containing 1 μg of TPCK Trypsin (Worthington Biochemicals, Lakewood, NJ) was added to the cells (total concentration of 0.5 μg/ml of TPCK Trypsin in the cell supernatant). Cytopathic effects were obvious 48–72 h following addition of TPCK Trypsin. Following cytopathic effect occurrence, supernatant was collected and clarified by centrifugation (5000 rpm for 5 min) and used for infecting 10-d embryonic chicken eggs (100 μl neat/egg) for generating a working viral stock.

For recombinant vaccinia virus (rVV) infection, cells were resuspended in 0.1% BSA/PBS and infected with rVV, as indicated in the figure legends, at a multiplicity of infection (MOI) of 10. For IAV infection, cells were resuspended in FCS-free, acidified RPMI medium and infected with IAV at 10 MOI. Cells were incubated in a 37°C water bath for 1 h with gentle agitation every 15 min. Infected cells were topped to 2 ml with RPMI-1640 with 10% FCS (RF-10) and incubated for a further 4 h at 37°C. Cells were then washed twice with PBS.

All cells were cultured at 37°C with 5% CO₂ in a humidified incubator. P815, D2SV, D2SV transductants, and HEK293T transfectants were cultured in RPMI 1640 containing 10% FCS, 50 μM 2-ME, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (RF-10). Platinum-E retroviral-producing cell line (Cell Biolabs) was cultured in DMEM with 10% FCS and the above supplements, plus 1 μg/ml puromycin and 10 μg/ml blasticidin. For reverse genetics, MDCK cells were maintained in MEM (Life Technologies, Australia) containing 10% FCS, 10 U/ml penicillin, and 100 μg/ml streptomycin (MEM-10), and HEK293T cells were cultured in Opti-MEM I (Life Technologies, Carlsbad, CA) containing 5% FCS, 10 U/ml penicillin, and 100 μg/ml streptomycin. Cells were passaged, and media were changed when cells became 80% confluent; cells were used during the exponential growth phase.

For generating polyspecific T_CD8+ cultures, P815 were infected with PR8 (PR8-P815) at 10 MOI. The infected cells were then washed and irradiated for 10,000 rads and used as APC to stimulate PR8-primed memory (>30 d) splenic cells. Repeatedly restimulated T_CD8+ cultures were restimulated every 14 d with PR8-P815 cells. For generating peptide-specific T_CD8+ cultures, 1/10 of memory splenic cells were pulsed with 1 nM peptide of interest for 60 min. The cells were then washed with PBS and cocultured with 9/10 of responder memory splenic cells (19).

A 931-bp cDNA containing the wt IAV PR8 (Mount Sinai) NS gene (898 bp) or its cDNA variant with mutated splicing elements (the 5′, 3′ splicing sites and the branch point) with 16-bp upstream and 17-bp downstream nucleotide sequences containing restriction enzyme sites was designed (sequence not shown), synthesized, and cloned into pCU57 vector (GenScript, Nanjing, China). NS-Full and NS-mut fragments were PCR amplified from the pCU57-NS-full and pCU57-NS-mut template vectors, respectively, using the NS-specific primer set (forward primer: 5′-tattAGATCTAGGGAGCAAAAGCAGGGTG-3′ and reverse primer: 5′-acgcCTCGAGTATTAGTAGAAACAAGGG-3′). Both the NS PCR product (Full and mut) and PMIM vector (retroviral vector containing IRES-mCherry) were then subjected to BglII^5′ and XhoI^3′ double digestion to linearize PMIM and to generate matching sticky ends in both the vector and the PCR products for directional cloning the PMIM-NS-Full and PMIM-NS-mut plasmids.

A 1237-bp eGFP splicing reporter fragment was constructed by inserting either the 473-bp wt or splicing mutant NS intronic sequence in between 30 and 31 bp of the 744-bp eGFP fragment flanked by 10-bp sequences at either side containing restriction enzyme sites (sequences not shown). The genes were synthesized (GenScript) and cloned into pCU57 vector. eGFP-NS^intron-WT and ^{PMIM-eGFP-NSintron}-mut fragments were PCR amplified using the eGFP-specific primer set (forward primer: 5′-acgtGAATTCATGGTGAGCAAGG-3′ and reverse primer: 5′-atcgCTCGAGTTATGATCTAGAGTC-3′). Both PCR products (eGFP-NS^intron-WT and PMIM-eGFP-NS^intron-mut) and PMIM vector were then subjected to EcoRI^5′ and XhoI^3′ double digestion to linearize PMIM and generate matching sticky ends in both the vector and the PCR products for directional cloning of the PMIM-eGFP-NS^intron-WT and PMIM-eGFP-NS^intron-mut plasmids.

A 693-bp cDNA containing the wt PR8 NS1 gene was used as a template for PCR amplification of full-length as well as six 3′-terminal truncated fragments subjected to TOPO cloning into the pcDNA3.1/V5-His TOPO vector (Life Technologies). NS1 truncation variants were PCR amplified using PR8 NS1-specific forward (5′-CACCATGGATCCAAACACTGTGTCAAGCT-3′) primer coupling with various reverse primers: NS1/698 (5′-GGATCCTCAAACTTCTGACCTAATTGTTC-3′) for generation of full-length NS1 cDNA encoding NS1 1–230 aa region, NS1/620 (5′-GGATCCGTAGAGTTTCAGAGACTCGAA-3′) for NS1 1–198 aa, NS1/520 (5′-GGATCCGAAGGCAATGGTGAAATTTCG-3′) for NS1 1–164 aa, NS1/420 (5′-GGATCCCTTTCAGTATGATGTTCTTAT-3′) for NS1 1–131 aa, NS1/320 (5′-GGATCCCATTTCCTCAAGAGTCATGTC-3′) for NS1 1–98 aa, NS1/220 (5′-GGATCCACTATCTGCTTTCCAGCACGT-3′) for NS1 1–64 aa, and NS1/120 (5′-GGATCCTCGGCGAAGCCGATCAAGGAA-3′) for NS1 1–38 aa coding regions. These plasmids were then used to transfect HEK293T-L^d cells.

A 931-bp cDNA of IAV PR8 (Mount Sinai) NS gene (898 bp) consisting of mutated NS-ARF2_1–14 variant (5′-ACGCTCCTTTTCTAGACCGCCTCCGGCGGGACCAAAAGTCTCTGA-3′ with bold mutating nucleotide positions) consisting of 16-bp upstream and downstream nucleotide sequences encompassing BsmbI^5′/3′ restriction enzyme sites was designed (sequence not shown), synthesized, and cloned into pCU57 vector (GenScript). As incorporated mutations are silent in NS1 and NS2 ORFs, their intended translational products remain wt sequences. The NS-ARF2_1–14 knockout (KO) fragments were agarose gel–purified following BsmbI restriction enzyme digestion of pCU57-ARF2_1–14 KO backbone vector and cloned into BsmbI-digested and -purified pHW2000 vector for construction of pHW2000-NS-ARF2_1–14 KO vector, which was then used in reverse genetics IAV production.

One day before transfection, 2 × 10⁶ retrovirus-producing cells were plated onto a 10-cm tissue culture dish in 10 ml of DMEM-10 without antibiotics. When cells reached 70–80% confluence, FuGENE HD (Promega) was used for transfecting Platinum-E cells with the bicistronic mammalian expression retroviral vector (PMIM) containing the gene of interest. Forty-eight hours later, retrovirus-containing supernatant was harvested, filtered (0.45 μm), and added to 3 × 10⁵ D2SV target cells in a 10-cm tissue culture dish. Polybrene was added into the retrovirus-containing supernatant at 8 μg/ml. Twenty-four hours later, the medium was replaced, and target cells were cultured for three more days. Transduced cells were then analyzed for mCherry reporter expression and were sorted using FACSAriaIII flow cytometer (Becton Dickinson).

⁵¹Cr-release assays were performed as described (25). Data are expressed as the percentage of specific lysis = (T_CD8+-induced lysis − spontaneous release)/(release by detergent − spontaneous release) × 100%.

For ex vivo isolates, BAL wash and splenic cells were collected in RF-10. The Ag-specific T_CD8+ were enumerated using intracellular cytokine staining (ICS) for the production of IFN-γ after being stimulated with 1 μM antigenic peptide in the presence of 10 μg/ml brefeldin A (BFA) for 5 h at 37°C, unless otherwise stated (22).

For BFA kinetics assay, P815 cells were infected with 10 MOI of PR8 for 1 h at 37°C, washed, and then added to monospecificity T_CD8+ cultures, with BFA added at various time points. Following 4 h of exposure to BFA, T_CD8+ were transferred onto ice, fixed with 1% paraformaldehyde, and stained in ICS for IFN-γ in the presence of 0.4% saponin (26).

The method was detailed previously (25). Briefly, 5 × 10⁸ cultured cells were infected at 10 MOI with PR8 for 5 h. Cells were washed, resuspended in trifluoroacetic acid (TFA)/H₂O, homogenized, and further sonicated. Peptide-containing supernatant was collected after ultracentrifugation and passed through a 3K cutoff filter. Samples were then dehydrated to <400 μl using a speed-vac, and the extracted peptides were then fractionated on an reverse phase (RP)-HPLC (Agilent 1100). Eluted fractions were collected between 10.0 and 38.8 min at 6-s intervals. Elution times of synthetic peptides were determined under the same running condition after sample collection. To detect naturally presented peptides in the HPLC fractions, P815 cells were low temperature–induced overnight at 26°C, pulsed (10⁵) with either synthetic peptide at log-fold dilutions or fractions for 1 h at 26°C in RPMI 1640. Then, 10⁵ T_CD8+ were added to the pulsed P815 cells in RF-10 containing 10 μg/ml BFA for a further 5 h and Ag-specific T_CD8+ activation was enumerated by ICS.

Cells were homogenized using the QIAshredder (Qiagen). Total RNA was extracted from 3 × 10⁶ cells per sample using the Qiagen RNeasy Mini Kit, according to manufacturer instructions, with on-column DNase I digestion incorporated. Following extraction, RNA integrity number was quality controlled using the Agilent 2100 bioanalyzer, and samples with a RIN score above 7 were used for sequencing. Following mRNA purification with Poly-T oligo-attached magnetic beads, strand-specific cDNA libraries were prepared using the NEBNext Ultra RNA Library Preparation Kit, according to manufacture instructions, Illumina sequencing was conducted on the Hiseq4000 by Novogene Bioinformatics Technology Co. (Beijing, China), and 150-bp paired-end reads were generated.

Quality control was first performed on raw data (FastQ files) using the FastQC software, and any reads containing adapter sequences were trimmed using TrimGalore. Reads were then aligned with HISAT2 (27) to a modified reference genome comprising the GRCm38.p5 genome and 12 wt PR8 influenza virus gene segments (PB1 NC_002021.1, PB2 NC_002023.1, PB1-F2 NC_002021.1, PA NC_002022.1, PAx NC_002022.1, HA NC_002017.1, NP NC_002019.1, NA NC_002018.1, M1 NC_002016.1, M2 NC_002016.1, NS1 NC_002020.1:27-719, and NEP NC_002020.1:27-864). Gene counts were quantified with featureCount, and strand-specific antisense strands were included in the analysis. For bioinformatic analysis, genes with valid expression were defined as having log count per million >0.3 in at least two different replicates. After multidimensional scaling analysis, sample KO infected 3 was removed as an outlier. Differentially expressed genes were calculated using DESEQ2 (28) and by setting log fold change ≥1 or ≤−1, p value ≤0.01, and false discovery rate (FDR) ≤0.01.

P815 cells were infected with 10 MOI wtPR8 or ARF2_1–8 KO PR8 IAV for 5 h; washed, stained, and cell sorted for IAV HA expression; and 3 × 10⁶ sorted cells were dried by vacuum centrifugation. Cell pellets were resuspended in digestion buffer (8 M urea, 50 mM ammonium bicarbonate, 10 mM DTT) before incubation for 5 h at 25°C. Fifty-five millimolar iodoacetamide was then added to alkylate thiol groups at 20°C for 35 min in the dark. The alkylated preparation was diluted to 1 M urea with 25 mM ammonium bicarbonate (pH 8.5) before sequencing-grade trypsin (Promega) was added to 5 μM final concentration. Digests were performed overnight at 37°C. The digests were acidified with 1% (v/v) TFA and the peptides desalted on SDB-XC (Empore) StageTips. Peptides were modified by stable isotope dimethyl labeling for quantitative proteomics. After mixing the peptides 1:1, the samples were fractionated offline by high-pH reversed-phase fractionation on 24 fractions. Peptides from each fraction were reconstituted in 0.1% TFA and 2% acetonitrile (ACN) and loaded onto C₁₈ PepMap 100-μm ID × 2-cm trapping column (Thermo Fisher Scientific) at 5 μl/min for 6 min and washed for 6 min before switching the precolumn in line with the analytical column (Vydac MS C₁₈, 3 μm, 300 Å, and 75-μm ID × 25 cm; Grace Pty.). The separation of peptides was performed at 300 nl/min using a nonlinear ACN gradient of buffer A (0.1% formic acid, 2% ACN) and buffer B (0.1% formic acid, 80% ACN), starting at 5% buffer B to 55% over 60 min. Data were collected on an Orbitrap Elite (Thermo Fisher Scientific) in data-dependent acquisition mode using mass/ions 300–1500 as mass spectrometry (MS) scan range; collision-induced dissociation tandem MS (MS/MS) spectra were collected for the 20 most intense ions per MS scan. Dynamic exclusion parameters were set as follows: repeat count 1, duration 90 s, and the exclusion list size was set at 500 with early expiration disabled. Other instrument parameters for the Orbitrap were as follows: MS scan at 120,000 resolution, maximum injection time of 50 ms, automatic gain control target of 1 × 10⁶, and collision-induced dissociation at 35% energy for a maximum injection time of 150 ms with an automatic gain control target of 5000. The Orbitrap Elite was operated in dual analyzer mode with the Orbitrap analyzer being used for MS and the linear trap being used for MS/MS.

Identification and isotopic quantification of proteins were performed on raw output files from liquid chromatography electrospray ionization MS/MS using MaxQuant (Version 1.5.1.6; Cox and Mann, 2008) together with its built-in search engine Andromeda. Uniprot using the Mouse Uniprot FASTA database (September 2015) together with common contaminants was used for this analysis. Carbamidomethylation of cysteines was set as a fixed modification, whereas acetylation of protein N termini and methionine oxidation were included as variable modifications, and dimethLys0, dimethLys4, dimethNterm0, and dimethNterm4 were included as labels. Parent mass tolerance was set to 5 ppm (after refinement by MaxQuant), and fragment mass tolerance was set to 0.5 Da. Trypsin was set as the digestion enzyme with up to two missed cleavages allowed. Prior to searching, MS/MS spectra were filtered by retaining only the top eight peaks per 100 Da. The match between runs feature of MaxQuant was used to transfer peptide identifications from one run to another based on retention time and mass to charge ratio. Both peptide and protein identifications were reported at an FDR of 1%.

The normalized protein group ratios values obtained from MaxQuant were then used to perform differential expression analysis with the limma package in R. A moderated t statistic was calculated for each protein group to test for differences in expression between samples. Resulting p values were corrected for multiple testing and converted to an FDR. Missing values were excluded from the analysis, with df adjusted accordingly.

We previously reported NP_147–155 as the most immunodominant of five immunodominant IAV peptides recognized by T_CD8+ from IAV-infected BALB/c mice (22). Interestingly, when we used IAV-infected P815 cells (PR8-P815) as APC to maintain a polyspecific T_CD8+ line, the line gradually lost specificity for each of the five peptides yet remained highly specific for IAV (Fig. 1A). What did these T cells recognize?

To identify this potentially novel IAV epitope, we again raised a polyspecific T_CD8+ line in the same manner. After the first round of stimulation, we determined recognition of 10 potential IAV gene products by infecting L929 cells (H-2^k) expressing individual H-2^d allomorphs with rVVs expressing individual IAV mRNA (as a strictly cytoplasmic virus, rVV-encoded mRNAs are not spliced). HA-, NP-, and PA-specific responses were stimulated by L929-K^d cells, most likely representing the previously identified HA_518–526-, NP_147–155-, and PA-specific T cells, respectively (18). A modest response to rVV-PB2 infected L929-D^d cells was observed (Fig. 1B), as originally reported (29, 30), and is likely PB2_289–297-specific (21).

Critically, polyspecific T_CD8+ also recognized L929-L^d cells infected with rVV-NS1, indicating the presence of novel NS1 epitope(s) (Fig. 1B). The L^d-restricted recognition of rVV-encoded NS1 was confirmed by a repeatedly PR8-P815 restimulated T_CD8+ line using H-2L^d expressing HEK293T cells infected with PR8 or rVV-NS1 (Fig. 1C).

To identify the novel NS1 T cell epitope(s), we initially cloned six truncated NS1 gene fragments into the pcDNA3.1/V5-His TOPO vector to express NS1/1-38, NS1/1-64, NS1/1-98, NS1/1-131, NS1/1-164, NS1/1-198, and full-length NS1/1-230. Following transfection of HEK293T-L^d cells (31), none of the fragments displayed antigenicity (Supplemental Fig. 1A).

As an alternative, we turned to H-2L^d peptide binding motif prediction of potentially antigenic peptides. Accordingly, we selected peptides with a position 2 serine or proline and position 8, 9, 10, and 11 isoleucine, phenylalanine, valine, leucine, arginine, and aspartic acid. Of the 17 peptides tested, none demonstrated antigenicity using a repeatedly PR8-P815–restimulated T_CD8+ line (Supplemental Fig. 1B).

Many well-characterized antigenic peptides do not match predictive algorithms (32). Taking a broader approach, we synthesized a set of 13mer peptides with 10 overlapping amino acids to create a peptide panel spanning all possible 11mers and shorter linear epitopes and likely all potential 12mer epitopes, as most T_CD8+ recognize cognate peptides missing a single residue at either end when the peptide is provided at high concentration (micromolar level). Using a restimulated polyspecific T_CD8+ line, none of the 76 13mer peptides were antigenic at even micromolar concentrations (Supplemental Fig. 1C).

Although MHC class I ligands are typically ∼9–10 aa and rarely longer than 12 aa, exceptions are known [e.g., 13mers from EBV BZLF1 Ag (33) or NY-ESO-1 (34) and a 14mer from M-CSF (35)]. To exclude the possibility of an unusually long epitope being recognized by our T_CD8+ line, we screened a set of 18mer peptides with 12 aa overlapping. Once again, failure ensued (Supplemental Fig. 1D).

These findings were consistent with the idea that the L^d-restricted T_CD8+ recognize an unusual IAV peptide.

Failing to find the epitope in the standard NS1 reading frame, we searched for predicted L^d-binding peptides in the +1 or +2 reading frames downstream of potential Met initiating residues. This revealed two peptides that we designate as NS1-ARF1 (MSANELQTK) and NS1-ARF2 (MPHSLIGFAEI). Testing synthetic peptides corresponding to the sequences, we found that synthetic NS1ARF2 activated a major population of the PR8-P815–stimulated polyclonal T_CD8+ (Fig. 1D).

Within ARF2_1–11, coincidentally, is an 8-mer peptide that matches the H-2L^d binding motif xPxxxxxF. We compared the abilities of synthetic ARF2_1–8 versus ARF2_1–11 to generate T_CD8+ lines from splenocytes derived from PR8-primed BALB/c mice (Fig. 1E, 1F). Interestingly, each peptide could elicit T_CD8+ that greatly preferred their activating peptide compared with truncated versions, suggesting that each is naturally immunogenic in the context of PR8 infection in vivo.

To determine the extent to which cells naturally present ARF2_1–8 versus ARF2_1–11, we HPLC fractionated acid-soluble peptides from PR8-infected cells and assayed fractions for antigenicity against T_CD8+ lines (25). Using an IAV polyspecific T_CD8+ line, fractions eluting at 18.9, 21.2, 23.9, 24.3, and 25.0 min exhibited antigenic activity (Fig. 2Ai). Using monospecific T_CD8+ restimulated with synthetic peptides, it was clear that the two major peaks at 18.9 and 24.9 min are recognized by NP_147–155 and ARF2_1–8, respectively (Fig. 2Aii, iii). Corroborating this finding, this correlates perfectly with the elution of the cognate synthetic peptides. We fail to obtain evidence for natural presentation for ARF2_1–11 (Fig. 2Aiv), which is expected to elute at 26.6 min based on the behavior of the synthetic version. This is mostly likely due to inefficient ARF2_1–11 epitope presentation by PR8-P815 cells.

Titration of peak antigenic fractions (Fig. 2B) revealed that ARF2_1–8 is generated at least 3-fold more abundantly than NP_147–155, using the synthetic peptides as standards. We corroborated this finding with peptide generation kinetics, using BFA to block further transport of class I peptide complexes to the surface of PR8-infected P815 cells in conjunction with T_CD8+ lines raised to individual peptides (26). As the two T_CD8+ lines clearly had similar avidity shown by their cognate peptide titration (Fig. 2C), ARF2_1–8 is presented ∼2- to 3-fold faster than NP_147–155 (Fig. 2D), which is consistent with the results of peptide elution and fraction titration results (Fig. 2B) and likely explains why multiple in vitro restimulated T_CD8+ lines become ARF2_1–8 specific.

Having defined NS1-ARF2_1–8 as the major peptide recognized by L^d-restricted IAV-specific T_CD8+, we next used the synthetic peptide to measure the magnitude of the in vivo response 7 d post-i.p. infection. In mouse groups infected with IAV doses lower than 5 × 10⁷ PFU, NS1-ARF2_1–8 T_CD8+ responses were slightly smaller than that of T_CD8+ specific to the immunodominant NP_147–155. Remarkably, NS1-ARF2_1–8 response ascended to the α-position in the immunodominance hierarchy at higher virus doses (Fig. 3A). In mice 10 d post-i.n. infection with 100 PFU of IAV, in both spleen and lungs, NS1-ARF2_1–8–specific cells rank second in the immunodominance hierarchy to NP_147–155 (Fig. 3B, 3C).

To be certain that recognition of NS1-ARF2_1–8 is a bona fide measure of its in vivo immunogenicity, we identified a 2009 pandemic H1N1 virus, A/Tasmania/09 (Pan09) (H1N1), possessing a phenylalanine → serine substitution in the C-terminal anchor residue of NS1-ARF2_1–8 that obviates high-affinity binding to L^d. Pan09 induced a vigorous response to all of the major epitopes except NS1-ARF2_1–8, confirming its status as a naturally processed epitope in vivo (Fig. 3B, 3C).

We next compared the effector function of T_CD8+ specific for NS1-ARF2_1–8 versus NP_147–155 after their ex vivo activation with cognate peptide. This revealed that T_CD8+ populations expressed similar amounts of effector cytokines IFN-γ and TNF-α and were equally cytolytic, as judged by expression of the degranulation marker, CD107a (Fig. 3D).

IAV generates NS2 via splicing of the NS mRNA. As the NS1-ARF2_1–8 epitope is encoded within the region that is normally spliced out to generate NS2-mRNA, we wanted to know whether the epitope is translated from NS1-mRNA as an ARF peptide or translated from the spliced-off intronic mRNA. Because vaccinia virus mRNAs are synthesized in the cytoplasm and do not traffic to the nucleus, rVV-expressed NS1 is not spliced. Consequently, its recognition by NS1-ARF2_1–8–specific-T_CD8+ (Fig. 1D) demonstrates that the peptide can be made from the NS1 mRNA. This does not, however, exclude a role for splicing in generating the peptide in IAV-infected cells, particularly given the evidence for nuclear translation as a source of peptides from IAV (36) and from intronic regions of plasmid-encoded genes (37).

To examine the role of mRNA splicing in NS1-ARF2_1–8 generation, we cloned the PR8 NS gene into the retroviral vector PMIM as wt sequence (NS-full) or modified to abrogate splicing (NS-mut) (Fig. 4Aii). D2SV cells transduced with either NS-full or NS-mut equally activated an ARF2_1–8–specific T_CD8+ line, demonstrating that mRNA splicing does not enhance ARF epitope generation (Fig. 4B). We noted that in our NS1 constructs, ARF2_1–8 epitope presentation is clearly weaker compared with IAV-infected cells. The likely explanation is that NS1 expression levels are lower from our constructs compared with that from IAV infection (Supplemental Fig. 2).

We further examined the contribution of NS1 splicing to peptide generation using NS1-GFP IAV (23), a recombinant IAV whose NS segment is modified to express NS1-GFP, followed by an FMDV 2A ribosome stop-and-go site (38) that releases NS2 (also called NEP) as it is synthesized. Two synonymous mutations are also present to prevent NS mRNA splicing. P815 cells infected with NS1-GFP IAV efficiently presented ARF2_1–8 epitope (Fig. 4B).

Together, these results demonstrate that the ARF2_1–8 epitope is generated through the translation of the ARF2 of the NS1 mRNA and that mRNA splicing plays, at most, a minor role in generating this peptide.

Given the location of ARF2_1–8 epitope in the NS gene intron, we investigated whether its synthesis is influenced by regulatory sequences outside the intronic sequence by transplanting the NS-intronic sequence into the eGFP gene (Fig. 4Aiii). The construct was then used to transduce D2SV cells, and transduced cells were enriched based on mCherry expression. Interestingly, the ARF2_1–8 epitope was generated more efficiently by eGFP-NS^intron–transduced cells than those transduced with NS-full (Fig. 4B), showing that ARF2_1–8 is generated in its local translational context. Again, when the splicing sites were eliminated in the eGFP-NS^intron, the generation of ARF2_1–8 epitope was not affected.

Although the ARF2_1–8 epitope was generated from D2SV cells transfected with full-length NS gene or transduced with eGFP-NS^intron, in which either the NS or NS-intronic sequence is essentially converted to a self-gene by chromosomal integration, its presentation was clearly less efficient than in the context of IAV infection. We therefore questioned whether such ARF DRiP generation is enhanced with IAV infection.

To explore this, we generated the above-mentioned transfectant and transductant in the P815 cell line expressing either full-length NS or eGFP-NS^intron and then superinfected these cells with Pan09 IAV. As this IAV strain does not generate ARF2_1–8 from its own genome, all detected ARF2_1–8 is derived from the autoantigens expressed in these cell lines. As shown in Fig. 4C, Pan09 IAV infection clearly enhanced ARF2_1–8 presentation, especially from cells that expressed eGFP-NS^intron. Again, such enhanced presentation of endogenous DRiPs did not require mRNA splicing as the epitope presentation was equally efficient in the splicing mutants. The infection by Pan09 IAV was as efficient as those by PR8, as shown by equal NP_147–155 epitope presentation (Fig. 4D).

As very short peptides can have significant biological functions (39), to formally exclude the possibility that ARF2_1–14 functions in IAV infection, we generated ARF2_1–14 KO IAV viruses using the reverse genetics approach (24). We then compared the mutant IAV to a control wt recombinant generated in parallel for in vitro infectivity and ability to trigger immunopathology and cellular immunity in two mouse strains. First, we infected the mouse lung epithelial adenoma cell line LA-4 (40) and measured the production of NP, M1, M2, and NS1 via flow cytometry with the appropriated mAbs. As shown in Fig. 5A (i–iv), LA-4 cells infected with wt or the ARF2 mutant virus generated equal amounts of each of the proteins measured, indicating that the small ARF is not required for viral mRNA synthesis or translation. Further, when we infected P815 cells with wt versus mutant viruses and characterized the transcriptome and proteome, we detected only very minor differences. Only 12 out of 13,167 genes and two out of 4419 proteins were differentially expressed. Moreover, the differentially detected genes and proteins did not overlap between replications, and no biological signaling pathways were overrepresented (Supplemental Fig. 3A, 3B).

Second, when we infected C57BL/6 and BALB/c mice i.n. with 100 PFU with wt or mutant viruses, mice showed identical weight loss profiles, indicating that short ARF is not required for viral pathogenesis.

Finally, we assessed the cellular immune responses following i.n. infection of mice. In C57BL/6 mice, we found a nearly identical T_CD8+ response to seven different peptides in spleen, with relatively minor variation in the BAL response to some peptides (Fig. 5B, 5D, 5F). Similarly, in BALB/c mice, with the obvious exception of the ARF2_1–8 response, there were only minor differences in splenic or BAL responses to six other peptides (Fig. 5C, 5E, 5G). Higher infecting doses of 500 PFU of ARF2_1–14 KO virus did not demonstrate any alteration in lethality or the resulting immune response (data not shown).

Taken together, our findings do not support an important biological role for ARF2_1–14 in the IAV infection of cultured cells or infected mice.

In this study, we show that BALB/c mice mount a robust T_CD8+ response to NS1-ARF2_1–8, which under some conditions, ascends to the top of the immunodominance hierarchy. We previously systematically searched for potential ARF epitopes presented by H-2K^b, D^b, or K^d. Although 9 of the 35 synthetic peptides screened efficiently stimulated T_CD8+ elicited with the autologous peptide, only a single peptide could stimulate or induce IAV-specific T_CD8+ (41). Ironically, the corresponding 87 residue ARF is an ORF encoding PB1-F2 (8), which has proven to exert myriad functions contributing to viral fitness (42, 43). This highlights a critical issue in identifying ARF-encoded peptides: how can we be certain that the gene product is not a short functional ORF? Indeed, it is not certain that the ARF peptides identified to date for HIV, hepatitis B virus, and human CMV are truly nonfunctional (9, 44–47). In each of these published examples, the ARF/ORF is >30 residues, which is of sufficient length to form a membrane spanning oligomer, or at the very least, associate with a viral or cellular protein to modulate its function.

NS1-ARF2 is but 14 residues. Although not impossible, it is unlikely that such a short peptide could be functional. A Drosophila 11-mer seems to have a function (13), and plant 18- and 20-mer microproteins were recently described to possess biological activity (48, 49). The shortest viral gene products with known biological activity are >40 residues. Out of >5000 NS1 sequences available online, only one (A/chicken/Jiangsu/JT34/2011[H9N2]) lacks the start codon of the ARF2_1–8 epitope. All others have sequences highly similar (or identical) to PR8 or Pan09. Interpreting this sequence conservation is muddied by the high conservation of the corresponding overlapping NS1 sequence, which encodes residues ∼30–43 of the RNA-binding domain critical for sequestering IAV dsRNA during its replication (50, 51). To formally exclude the possibility that NS1-ARF2_1–14 might be an intended viral product with biological function, we generated NS1-ARF2_1–14 deletion mutant and its control wt IAV virus by reverse genetics. In our parallel in vitro, in vivo, and systems biology experiments, we were not able to detect any function or influence of NS1-ARF2_1–14 during IAV infection. We therefore conclude that it is likely that NS1-ARF2_1–14 is a bona fide viral DRiP.

Although 14 residues may be too short to generate a functional gene product, we have not completely ruled out that NS1-ARF2 is synthesized as something more than a 14-mer. Our evidence conclusively demonstrates NS1-ARF2_1–8 being generated from NS1 mRNA and not a splice variant. This follows from NS1-ARF2_1–8–specific T_CD8+ recognition of rVV-infected cells expressing NS1 mRNA, cells expressing unspliceable forms of NS1 mRNA expressed by IAV, or from a transgene. IAV generate at least one functional frame-shifted protein (PA-X) (15), but there is very little room upstream of NS1-ARF2 for frame-shifting, with only a Val residue before encountering a stop codon and no known frame shift abetting sequence. Although it seems unlikely, it is possible that the NS1-ARF2 is extended by readthrough of the stop codon or by shifting back into the NS1 frame between residue 8 and the stop codon after residue 14.

Our findings clearly demonstrate the potential importance of unusual translation events in generating biologically important viral peptides, extending the recent exciting evidence for the potential involvement of nuclear translation in generating viral (36, 52) and cell peptides, in the latter case from intronic regions (37). More importantly, we further demonstrated that IAV infection enhances cellular ARF DRiP generation, a finding very similar to that recently reported by Prasad et al. (53), in which the generation of a model T cell epitope translated via alternative start codon was enhanced by viral infection. These findings may help us to further understand the relationship between virus infection and autoimmunity. Perhaps virus infection could trigger autoimmunity by simply increasing DRiP Ag presentation?

Pragmatically, it is important to reconsider strategies for identifying immunogenic and antigenic peptides. In our own specific circumstances, we would have found NS1-ARF2_1–8 had we synthesized all potential ARF peptides at their predicted full length, using serum proteases to liberate antigenically active fragments. Surely, many potentially biologically relevant peptides lurk in unexpected locations in the genomes of medically important viruses. We note that NS1-ARF2_1–8 is predicted to bind HLA-B*5101, HLA-B*3505, and HLA-B*5502 with sufficient affinity to be immunogenic, an obvious area of future research.

ARF2_1–8 elicits a robust response in vivo. Given predicted low levels of synthesis of the ARF combined with predicted rapid degradation, it is extremely unlikely that there is a sufficient amount of protein for cross-priming (54). This provides a clear example of the importance of direct priming in eliciting T_CD8+ to a natural viral gene product. The position of NS1-ARF2_1–8 in the immunodominance hierarchy varies depending on the route and dose of infection. A similar phenomenon occurs with the T_CD8+ response to vaccinia virus in B6 mice (55). Multiple factors contribute to immunodominance (18, 22, 56), whose complexities are the natural product of the highly complex interaction between cells, viruses, and all other constituent parts (19, 57). We note that NS1-ARF2_1–8 is a reasonably abundant peptide (copy number of ∼2500 complexes per cell), ∼3-fold more abundant than NP_147–155, which is present at 800 copies per cell (22). Yet there is no clear relationship between peptide abundance and immunodominance, with very abundant peptides sometime occupying low rungs on the immunodominance ladder (22).

In summary, we show that what is very likely to be mistranslation of a nonfunctional ARF encoded by IAV leads to the rapid presentation of an epitope at or near the top of the antiviral T_CD8+ immunodominance hierarchy. Along with the findings of Yang et al. (58) demonstrating CUG initiation of a +1 RF model peptide after the M1 mRNA stop codon, this provides clear biological evidence for the importance of DRiPs in peptide generation, extending the voluminous evidence from studies of peptide generation in cultured cells.

The authors have no financial conflicts of interest.