Defining Viral Defective Ribosomal Products: Standard and Alternative Translation Initiation Events Generate a Common Peptide from Influenza A Virus M2 and M1 mRNAs

CD8⁺ T cells play a central role in immunosurveillance of transplants, tumors, and intracellular microbes, including viruses. Antiviral CD8⁺ T cells recognize virus-encoded oligopeptides bound to MHC class I molecules and presented on the surface of the infected cell. Peptides are typically generated by proteasomal cleavage of viral proteins in the cytoplasm or nucleus. They are then transported by the TAP peptide transporter into the endoplasmic reticulum, where they are trimmed and loaded onto waiting class I molecules in an exquisitely choreographed molecular ballet (1). After peptide binding, class I molecules are exported from the endoplasmic reticulum through the Golgi complex to the cell surface for immunosurveillance.

The entire process of peptide creation, loading, and MHC complex delivery to the cell surface can occur with surprising speed. Antiviral T cells can recognize viral peptides within 45 min of cellular infection, despite the stability of the corresponding source full-length viral gene product (2). This observation spawned the defective ribosomal product (DRiP) hypothesis of peptide generation, which posits that the class I system exploits a distinct pool of metabolically unstable translation products for immunosurveillance (3).

Considerable experimental evidence from numerous approaches indicates that DRiPs are a major contributor to immunosurveillance (4), including recent mass spectrometry studies that elegantly demonstrate the disparate kinetics between peptide generation and degradation of the full-length source protein (5, 6). Myriad translational mechanisms can generate DRiPs (4, 7–9), including the following: 1) degradation of misfolded or mistargeted full-length proteins; 2) overproduction of a polypeptide relative to the expression of its normal interaction partner(s); 3) truncation due to mistranslation (i.e., frame shifting, alternative initiation on CUG codons or downstream AUG); and 4) noncanonical translation in the nucleus of immature and mature mRNA (10, 11).

Despite the wide variety of studies that support DRiPs as a major source of class I peptide ligands, the mechanisms involved in the generation of DRiPs are poorly defined outside of experimental systems in which DRiPs are artificially created. A problem inherent to characterizing DRiPs is the high ratio of viral proteins synthesized by infected cells (typically in the range of 10⁵ to 10⁷ copies per cell) versus the number of complexes presented on the cell surface (typically in the range of 10¹ to 10³ copies per cell). The roughly 10,000-fold ratio of native protein to class I peptide complex makes biochemical analysis difficult. Because complexes can be generated from viral gene products with an efficiency at least as high as 2.5%, antigenically relevant DRiPs can represent only a small fraction of the synthesis of a given gene product. For example, at an efficiency of 2.5%, 1000 complexes would derive from 40,000 substrates, which represent just 4% of the total pool of 10⁶ native viral proteins synthesized. These circumstances favor experimental strategies based on genetic or chemical manipulation of source Ag expression.

In this work, we study the generation of a specific model peptide (SIINFEKL) from DRiPs encoded by influenza A virus (IAV) mRNA in a context that allows us to dissect the contributions of standard versus alternative reading frames (ARFs) and initiation codons. SIINFEKL forms highly stable complexes with the mouse K^b class I molecule, enabling detection by the 25-D1.16 mAb (12) or by OT-I transgenic T cells (13) via T cell activation assays in vitro and in vivo. Our findings demonstrate the contribution of noncanonical CUG-codon–based translation initiation to antiviral CD8⁺ T cell immunosurveillance.

L929 cells (American Type Culture Collection) and HeLa cells from American Type Culture Collection stably transfected with a cDNA encoding the MHC class I K^b molecule were cultured in DMEM supplemented with 7.5% FBS at 37°C with 9% CO₂. The 293T and MDCK cells from American Type Culture Collection were propagated in DMEM supplemented with 10% FBS at 37°C with 9% CO₂.

Spliceostatin A (SSA) was synthesized as described (14). Cycloheximide (CHX) (Sigma-Aldrich), anti–β-actin Ab (Sigma-Aldrich), and MG132 (Calbiochem) were obtained from vendors. Anti-SIINFEKL C-terminal Ab (1F10.2.2) was obtained from Kenneth Rock (University of Massachusetts Medical School). Anti-IAV mAbs were described previously (15): anti-HA, H36-26; anti-NA, NA2-1C1; anti-M1, M2-1C6; anti-M2, O19 (16). The anti-NA rabbit polyclonal Ab used for Western blotting was described (17). For flow cytometry, FITC anti-mouse H-2K^b (clone AF6-88.5) was from BD Biosciences. MAb 25D1.16 (anti-K^b-SIINFEKL) and M2-1C6 were labeled with Alexa Fluor 647 (Life Technologies). NA2-1C1 was labeled with Pacific Blue (Life Technologies). H36-26 and O19 were labeled with Alexa Fluor 488 (Life Technologies).

PR8-M2-SIIN, PR8-M2 (24)-SIIN, PR8-M2 (45)-SIIN, PR8-M2 (47)-SIIN, PR8-M2-C888T-SIIN, PR8-M2-M72L-SIIN, PR8-M2-G1001A-SIIN, and PR8-M2-C888T-G1001A-SIIN were generated by eight-plasmid transfection into a mixture of 293T and MDCK cells, as described (18). The rescue plasmids pDZ-PR8(M2-SIIN), pDZ-PR8(M2- (24)SIIN), pDZ-PR8(M2- (45)SIIN), pDZ-PR8(M2- (47)SIIN), pDZ-PR8(M2-C888T-SIIN), pDZ-PR8 (M2-M72L-SIIN), pDZ-PR8(M2-G1001A-SIIN), and pDZ-PR8(M2-C888T-G1001A-SIIN) were cloned by PCR mutagenesis of PR8 segment 7. Rescued viruses were propagated in 10-d embryonated chicken eggs and sequenced to confirm their identity. IAV titers were determined by standard 50% tissue culture infective dose assay using MDCK cells. For PR8 infection, cells were infected at a multiplicity of infection of 3–10 for 1 h at 37°C with mixing in DMEM without FBS. Then cells were washed and treated with inhibitors as specified for indicated times in growth medium. For overnight infection, cells were infected at a multiplicity of infection of 0.2–0.5 to minimize cell death.

For measuring cell surface proteins, L-K^b cells were incubated at 4°C for 30 min with fluorescently tagged Abs at various times postinfection (p.i.) After three washes with HBSS/BSA, cells were analyzed using a BD LSR Π flow cytometry (BD Biosciences) and FlowJo software (TreeStar). To measure intracellular M1, cells were fixed in 3% paraformaldehyde with 10 mM HEPES in PBS, washed with PBS, and incubated with Alexa Fluor 647–conjugated M2-1C6 in PBS containing 1% bovine calf serum and 0.5% saponin.

Pathogen-free C57BL/6 mice were acquired from Taconic. OT-I TCR transgenic mice were bred in-house and crossed to mice ubiquitously expressing dsRed (The Jackson Laboratory stock 5441) to create OT-I dsRed mice. Adult 6- to 12-wk-old mice were used in all experiments. All mice were housed under specific pathogen–free conditions (including mouse norovirus, mouse parvovirus, and mouse hepatitis virus) and maintained on standard rodent chow and water supplied ad libitum.

OT-I CD8⁺ T cells were purified using an autoMACS and a negative selection kit (Miltenyi Biotec) from OT-I dsRed mice to ∼95% purity. Approximately 5 × 10⁵ cells were adoptively transferred i.v. into C57BL/6 recipients. Mice were infected i.p. with 1 × 10⁶ PFU of the indicated virus. At 5 d p.i., spleens were removed and restimulated for 4 h with either 100 nM irrelevant or SIINFEKL peptide. Brefeldin A (10 μg/ml; Sigma-Aldrich) was added during restimulation. After incubation at 37°C, cells were stained for CD8, fixed for 20 min with 1% paraformaldehyde, washed, and then stained with anti–IFN-γ Ab (clone XMG1.2; eBioscience) in 0.5% saponin overnight at 4°C.

Cells were lysed in 50 mM Tris pH 7.4, 150 mM NaCl, 1% Nonidet P-40, and 0.1% SDS supplemented with protease inhibitors (Roche). Lysates were separated on 4–12% PAGE gels and transferred to nitrocellulose. Membranes were probed with primary Abs, followed by anti-rabbit or anti-mouse secondary Abs coupled to the infrared dyes 680 or 800CW. Membranes were analyzed using an Odyssey infrared imager (LI-COR).

Statistical analysis was performed using GraphPad Prism software (GraphPad, San Diego, CA). Error bars in graphs show standard deviations.

All animal experiments were conducted in accordance with the U.S. Animal Welfare Act and the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All procedures were approved by the National Institute of Allergy and Infectious Diseases Animal Care and Use Committee.

We generated a recombinant PR8 IAV encoding SIINFEKL appended in-frame to the C terminus of the viral M2 protein (PR8 M2-SIIN) (schematic shown in Fig. 1). Owing to the nuclear localization of IAV mRNA synthesis, the virus can exploit cellular splicing machinery to increase its repertoire of gene products. M2 is a type I integral membrane protein abundantly expressed on infected cell surfaces and is one of two influenza proteins (the other being NEP) generated by splicing (19). The M2 mRNA is generated by splicing of the M1 mRNA to create a protein consisting of the amino-terminal 17 aa of M1 fused to 80 aa coded by the +1 reading frame of M1, extending past the M1 stop codon (Fig. 1A).

PR8 M2-SIIN replicated robustly in eggs or MDCK cells, consistent with a prior study that found that M2 function is not significantly affected by short C-terminal extensions (20). We confirmed M2-SIIN expression using flow cytometry of infected L-K^b cells (L cells permanently transfected to express K^b) after staining with Abs recognizing Kb-SIINFEKL complexes (25-D1.16) or against M2 protein and comparing with cells infected with wild-type IAV PR8 (Supplemental Fig. 1A, 1C). Immunoblotting cell lysates 12 h p.i. with a SIINFEKL-specific mAb also revealed M2 SIINFEKL expression (21) (Supplemental Fig. 1B).

To determine the contribution of DRiPs to K^b-SIINFEKL generation from M2-SIIN, we measured the temporal kinetics of K^b-SIINFEKL expression using the 25-D1.16 mAb (Fig. 1B). This measurement revealed a strong kinetic connection between the cell surface expression of M2 and K^b-SIINFEKL. As expected, presentation was completely inhibited by incubating cells with the protein synthesis inhibitor CHX at the time of infection to prevent viral protein synthesis (Supplemental Fig. 1C). This experiment demonstrated that K^b-SIINFEKL complexes were generated from nascent viral proteins and not the small amounts of M2 present on incoming virions [approximately five copies per virion (22)]. Immunoblotting infected cells treated with CHX showed that the M2-SIIN fusion protein is highly stable (Fig. 1C), similar to M1 (top row) but distinct from the IAV protein PB1-F2 (third row), which possesses a major rapidly degraded cohort (23). If K^b-SIINFEKL complexes were generated from standard turnover of “retirees”—that is, native proteins being degraded after reaching their predetermined life spans—presentation should begin only as degraded M2-SIIN protein accumulates. Instead, maximal K^b-SIINFEKL formation closely tracked with M2 synthesis (Fig. 1B), clearly indicating that K^b-SIINFEKL complexes are predominantly generated from a rapidly degraded cohort of DRiP antigenic precursors.

Because M1 mRNA splicing generates M2 mRNA, SIINFEKL can potentially be produced from either M2 mRNA in the proper reading frame or M1 mRNA in the +1 reading frame downstream of the normal stop codon. To examine the contributions of M1 versus M2 mRNA to K^b-SIINFEKL expression, we treated cells with SSA, a small, cell-permeable natural product that blocks mRNA splicing by binding to splicing factor 3b (14, 24). SSA was highly effective in preventing splicing of M1 mRNA, as indicated by ≥95% inhibition of M2 expression and a concomitant increase in M1 expression (Fig. 1D, 1E).

We next examined the effect of SSA treatment on Ag presentation (Fig. 2). As a control for SSA specificity on M2 synthesis inhibition, we infected cells with a recombinant PR8 IAV with SIINFEKL inserted into the stalk region of neuraminidase (NA) (NA-SIIN), an IAV protein that is translated from a nonspliced mRNA (25). Treating NA-SIIN–infected cells with SSA did not inhibit NA expression, as determined by immunoblotting (Fig. 1E), and enhanced cell surface NA expression, as determined by flow cytometry (Fig. 2D).

We next examined the effect of SSA on the generation of K^b-SIINFEKL complexes from M2-SIIN. Despite near-total inhibition of M2 synthesis in SSA-treated L-K^b (Fig. 2A–C) or HeLa-K^b (Fig. 2D, 2E) cells (permanent HeLa cell transfectants expressing K^b), K^b-SIINFEKL was expressed between 25 and 60% of levels in untreated cells. In L-K^b cells, the net effect of SSA was to increase the ratio of K^b-SIINFEKL complexes to M2 ∼10-fold (Fig. 2B, 2C). In contrast, SSA had little effect on K^b-SIINFEKL complex generation in cells infected with the control NA-SIIN, decreasing the ratio of K^b-SIINFEKL complex/NA owing to the enhanced expression of NA (Fig. 2E).

These findings demonstrate that in the presence of SSA, K^b-SIINFEKL generation occurs independently of M2 biosynthesis. On the basis of this finding, we hypothesized that when SSA blocks M2 mRNA biogenesis, SIINFEKL is largely generated from the +1 reading frame of the M1 mRNA via frame shifting or alternative initiation.

Antigenic peptides can be generated from translation within the nucleus (10, 11). Previously, we demonstrated that preventing nuclear export of IAV-encoded NA-SIIN mRNA by treating cells with the RNA polymerase II inhibitor 5,6-dichloro-1-β-d-ribofuranosyl benzimidazole (DRB) inhibited NA biosynthesis but had much less impact on K^b-SIINFEKL complex generation, suggesting that SINFEKL peptides were produced from nuclear translation of NA-SIIN mRNA (10). Could aberrant translation of M1 mRNA in the nucleus account for the generation of K^b-SIINFEKL complexes in SSA-treated cells?

To explore this possibility, we characterized the nuclear versus cytoplasmic localization of M2 and M1 mRNA under various conditions (Fig. 3). Quantitative PCR confirmed that SSA nearly completely prevented M2 mRNA generation, concomitantly increasing both nuclear (detergent insoluble) and cytoplasmic (detergent soluble) M1 mRNA (Fig. 3A), accounting for the effects of SSA on M1 and M2 protein levels.

As reported by Amorim et al. (26), DRB nearly completely inhibited M1 expression, as determined by intracellular flow cytometry (Fig. 3D), which correlated with reduced cytoplasmic M1 mRNA levels (Fig. 3A, left panel). Concomitantly, DRB enhanced M2 mRNA levels in both the nucleus and cytoplasm by ∼4-fold (Fig. 3A, right panel). DRB treatment increased K^b-SIINFEKL expression to a similar extent (Fig. 3B). Curiously, M2 cell surface expression was only slightly increased (Fig. 3C).

DRB did not interfere with the ability of SSA to block M2 mRNA generation and M2 synthesis (Fig. 3A–C). Importantly, K^b-SIINFEKL expression paralleled the decreased amount of M1 protein synthesized, tracking with the diminished cytoplasmic levels of M1 RNA. This finding is consistent with the conclusion that under SSA blockade of M2 mRNA biogenesis, K^b-SIINFEKL complexes are generated from cytoplasmic translation of M1 RNA in the +1 reading frame.

To better understand the generation of K^b-SIINFEKL complexes from M1–M2 mRNAs in SSA-treated cells, we inserted the SIINFEKL coding sequence into different regions of the mRNA in the M2 reading frame. Placing SIINFEKL in-frame after M2 residues 24, 45, or 47 (Fig. 1A) maintained M2 function, allowing robust virus growth with strong M2 surface expression (Fig. 4). In addition, insertion in these positions supported the robust generation of K^b-SIINFEKL complexes, although the ratio of M2 to K^b-SIINFEKL varied, likely owing to the expected influence of flanking sequences on peptide liberation (27, 28). Indeed, locating SIINFEKL after positions 24 and 45 was a more efficient source of K^b-SIINFEKL complexes than C-terminal SIINFEKL (Fig. 4).

Remarkably, after treating infected cells with SSA, few to no cell surface K^b-SIINFEKL complexes were generated from M2 with SIINFEKL at any of the upstream locations, despite enhanced generation in the absence of SSA. This finding implies that in SSA-treated cells, translation initiation in the +1 reading frame of the M1 mRNA downstream of M2 residue 47 can generate SIINFEKL when the peptide is encoded at the M2 C terminus.

Viral peptides can be produced from noncanonical downstream AUG initiation sites in both standard and ARFs (29–32). Furthermore, the elegant studies of Shastri and colleagues (33–35) demonstrate the contribution of class I peptide–ligand generation from CUG-codon initiation. This work presaged recent ribosome profiling papers that clearly show the general importance of CUG-mediated initiation to mammalian cell translation initiation (36, 37).

Because of the importance of nonstandard translation in peptide generation, we next investigated the role of noncanonical +1 initiation from M1 mRNA in generating K^b-SIINFEKL in SSA-treated cells. Examination of the +1 reading frame of the downstream M1 RNA sequence revealed one AUG (encoding Met₇₂) and two CUG in-frame codons (encoding Leu₅₉ and Leu₉₆) (see Fig. 1A). We therefore generated IAV viruses with synonymously altered CUG codons (C₈₈₈T or G₁₀₀₁A). For Met₇₂, we altered the codon to encode Leu (UUG, M72L).

Infection with the mutated viruses revealed that alteration of the AUG or individual CUG codons did not inhibit SSA-resistant K^b-SIINFEKL complex generation relative to presentation in the absence of the drug (Fig. 5A). Altering both CUG residues, however, nearly abrogated K^b-SIINFEKL complex generation in SSA-treated cells (Fig. 5A, lower right panel), despite maintaining the expected increase in M1 expression levels (Fig. 5B, Supplemental Fig. 2).

In the absence of SSA, each of the CUG mutations reduced the number of K^b-SIINFEKL complexes relative to M2-SIIN cell surface expression, with the greatest reduction occurring with the double CUG mutant (Fig. 5C). These findings suggest that even when M2 is generated by an intact splicing mechanism, CUG initiation provides a significant fraction of the K^b-SIINFEKL complexes generated.

To corroborate the participation of CUG-based initiation in peptide generation, we infected cells with M2-M72L-SIIN to maximize the relative contribution of CUG codons to +1 initiation, preventing peptide generation from M2 mRNA by treating cells with SSA at the time of infection. At 3 h p.i., we added NSC119893, a specific inhibitor of Met-based translation initiation (33) (Fig. 6). Although NSC119893 reduced M1 synthesis by ∼50% in SSA-treated cells (Fig. 6C), consistent with partial inhibition of canonical Met initiation, it had no effect on K^b-SIINFEKL generation in SSA-treated cells (Fig. 6A). By contrast, NSC119893 + SSA inhibited K^b-SIINFEKL generation from NA-SIIN (Fig. 6B) by 1.5- to 3-fold, while inhibiting NA cell surface expression by 10–30%.

Taken together, these findings support the conclusion that a subset of K^b-SIINFEKL complexes is generated via CUG initiation in the +1 frame of the IAV M1 mRNA.

To more directly demonstrate that SIINFEKL can be generated by cytoplasmic translation from the M1 mRNA in the +1 reading frame, we inserted a cDNA encoding the IAV segment 7 from PR8 M2-SIIN (encoding the unspliced M1/M2 mRNA) into vaccinia virus (VV) to create VV-M1-SIIN_ARF (Supplemental Fig. 3). As a strictly cytoplasmic virus, VV-encoded mRNAs are not spliced and are translated exclusively in the cytoplasm in viral factories (38, 39). As a positive control for VV-mediated K^b-SIINFEKL expression, we generated a recombinant VV (rVV) expressing the spliced M2-SIIN mRNA (SIIN located at the M2 COOH terminus).

We first tested the ability of rVVs to generate K^b-SIINFEKL complexes in L-K^b cells. VV-M2-SIIN generated a robust signal detectable by 25-D1.16 staining within 3 h p.i. (Fig. 7A, 7C). However, VV-M1-SIIN_ARF failed to generate a significant signal, even by 5 h p.i. We were able, however, to detect a specific K^b-SIINFEKL signal (relative to cells infected with a rVV expressing M1 without SIINFEKL) if we pretreated cells with IFN-γ to increase expression of the class I processing machinery and thereby enhance Ag presentation (Fig. 7B, 7D).

Although the number of K^b-SIINFEKL complexes generated from M1-SIIN_ARF may seem small, 25D-1.16 staining is much less sensitive (threshold of hundreds of complexes) than T cell recognition (threshold of tens of complexes or less). To confirm SIINFEKL generation from M1-SIIN_ARF and demonstrate its in vivo relevance, we examined the capacity of rVVs to activate CFSE-labeled OT-I TCR transgenic CD8⁺ T cells (specific for K^b-SIINFEKL) adoptively transferred into B6 mice. At 2.5 d after i.p. infection with VV-M2-SIIN (positive control), splenic OT-I CD8⁺ T cell numbers were significantly increased over cell numbers in mice infected with a control rVV lacking SIINFEKL (3.2 × 10⁶ versus 7.8 × 10⁴), indicating Ag-induced T cell division (Fig. 7E). Infection with VV-M1-SIIN_ARF likewise produced increased numbers of OT-I cells (3.7 × 10⁵). Furthermore, the fraction of undivided OT-I cells (Fig. 7F) was clearly lower in mice infected with VV-M1-SIIN_ARF than in control rVV, although to a lesser extent than VV-M2-SIIN–infected mice (83.4% undivided in control VV-infected mice; 23.4% for VV-M1-SIIN_ARF; 1.0% for VV-M2-SIIN).

To confirm OT-I activation by VV-M1-SIIN_ARF, we examined T cell effector function 5 d p.i. (Fig. 7G, 7H). As expected from the 2.5-d time point, we recovered far greater numbers of splenic OT-I cells from mice infected with either VV-M2-SIIN (1.1 × 10⁷) or VV-M1-SIIN_ARF (5.9 × 10⁵) relative to the control rVV (2.1 × 10⁴). In addition, infection with either VV-M1-SIIN_ARF or VV-M2-SIIN greatly increased the number OT-I cells synthesizing IFN-γ, an indicator of fully activated effectors (control = 1.7 × 10⁴; M2-SIIN = 1.6 × 10⁷; M1-SIIN_ARF = 5.2 × 10⁵).

Together these data demonstrate that SIINFEKL is generated from cytoplasmic mRNA in vitro and in vivo despite its location in a +1 reading frame after the stop codon of the standard M1 open reading frame (Supplemental Fig. 3). The study provides in vivo relevance for translation initiation on CUG for the purpose of viral immunosurveillance.

We have investigated the sources of viral-encoded class I peptide ligands by inserting the model peptide SIINFEKL into the IAV segment 7 gene at the C terminus of M2. Flow cytometry of infected cells revealed that K^b-SIINFEKL complexes were generated in a close kinetic relationship with M2-SIIN synthesis, despite the high metabolic stability of M2-SIIN. Such tight linkage between viral protein synthesis and peptide generation is the rule and not the exception, as shown by a mounting number of studies (reviewed in Ref. 4). This observation is particularly well shown by applying mass spectrometry to kinetic measurements of viral protein versus peptide generation, which enables the simultaneous characterization of dozens of gene products (5, 40), revealing the dominant contribution of DRiPs to peptide generation from viral gene products.

We show that SSA, which targets cellular splicing machinery, blocked M2 synthesis. Remarkably, despite SSA-mediated inhibition of M2 synthesis, K^b-SIINFEKL complexes continued to be generated at a robust rate. Cotreating cells with DRB and SSA revealed that K^b-SIINFEKL generation paralleled cytoplasmic levels of M1 mRNA, suggesting that K^b-SIINFEKL was generated from unspliced M1 mRNA in the +1 reading frame. Using VV-M1-SIIN_ARF, we conclusively show that SIINFEKL was synthesized from the M1 mRNA by cytoplasmic translation, as VV mRNAs are not spliced and translation of VV-infected cells is limited to cytoplasmic viral factories (39).

Relocating SIINFEKL to various positions within the M2 coding sequence revealed the plasticity of M2 to accept this peptide at the amino (position 24) or carboxy (position 45 or 47) terminus of its transmembrane helix [which approximately encompasses residues 22–43 (41)]. This plasticity is consistent with a previous study in which IAV genes were screened for their ability to function after random transposon insertional mutagenesis (42). The mutational flexibility of M2 allowed us to determine that although each of these locations supported robust SIINFEKL presentation in untreated cells, SSA reduced presentation to near-background levels, implying that peptide location is critical in generating antigenic peptides from the M1 mRNA.

An explanation for this finding is the absence of potential alternative AUG or CUG start codons in the +1 (M2) reading frame upstream of position 47. M2 encodes 2 Met residues, at positions 1 and 72, and two CUG codons, located downstream of position 47 at residues 59 and 96. CUG was described as an initiation codon for mammalian (43) and viral (44) proteins more than 25 y ago. A substantial body of work from Shastri and colleagues (8) established that CUG initiation can generate small open reading frames in transgenes for immunosurveillance.

Simultaneously modifying CUG codons encoding Leu₅₉ and Leu₉₆ greatly reduced peptide generation in SSA-treated M2-SIIN–infected cells, strongly indicating that CUG-based initiation is important for SIINFEKL synthesis. In support of this conclusion, NSC119893, a specific inhibitor of standard AUG Met-initiated translation, had little effect on K^b-SIINFEKL generation from M1 mRNA in SSA-treated cells, while inhibiting M1 synthesis ∼50%. Although we cannot eliminate a contribution of frame shifting to SIINFEKL generation from M1 mRNA, these findings suggest that peptides are largely derived from CUG-based initiation.

Together, our findings support alternative initiation as a source of IAV DRiPs. Ironically, our initial search for IAV ARF DRiPs (45) led to the discovery of PB1-F2, generated by AUG initiation in the +1 reading frame of the PB1 gene, downstream of the standard PB1-initiating AUG (23, 45). Since then, additional bona fide IAV gene products have been shown to arise from downstream Met initiation in the standard reading frame [PB1-N40 (46)] and by frame shifting [PAX (47)]. Although not impossible, it is unlikely that the translation we detect at the 3′ end of the M1 mRNA produces a functional gene product or otherwise functions to increase viral fitness in some manner. Translation initiation at Leu⁵⁹ would create a polypeptide of 40 aa, which would make it the shortest described IAV polypeptide, but still possibly functional. M42, a splice variant of M2, is 54 residues in length (48), and a functional 45-residue protein has been defined in Luteoviridae (49). Our findings imply that translation can also be initiated at Leu⁹⁶_, which would create a 2 aa gene product before the stop codon at position 98—surely too short to be functional. Philosophically, it is basically impossible to disprove that a given genetic element or gene product is functional in some evolutionary context. In terms of practicality, having an open mind about potential function is a good idea. “Junk” peptides (like junk DNA) can only increase in importance as knowledge about a system grows.

In studying the generation of SIINFEKL, we artificially altered the IAV genome, which could potentially influence the occurrence of alternative translation initiation. This seems unlikely to have greatly influenced our results, which strongly implicate initiation on CUG encoding Leu⁵⁹, nearly 100 nucleotides distant from the SIINFEKL coding sequence. How widespread is alternative nonfunctional initiation in creating viral DRiPs? We previously reported that 14% of newly synthesized NP is truncated from the N terminus in IAV-infected cells, possibly owing to downstream initiation (30). It seems likely that many more viral peptides are generated by nontraditional translation of gene products that are inconsequential to viral transmission (the driving force in viral evolution), except as serving as targets for host immunity.

We thank Tong-Ming Fu, Merck Research Laboratories, and Kenneth Rock (University of Massachusetts Medical School) for the generous gifts of Abs as well as Tom Kristie (National Institute of Allergy and Infectious Diseases) for critical comments on the manuscript.

The authors have no financial conflicts of interest.