SRS, JS, SM, and NS conceived and designed the experiments. SRS, JS, TH, and SM performed the experiments. SRS, JS, TH, SM, and NS analyzed the data. SRS contributed reagents/materials/analysis tools. SRS and NS wrote the paper.
The authors have declared that no conflicts of interest exist.
Major histocompatibility class I molecules display tens of thousands of peptides on the cell surface for immune surveillance by T cells. The peptide repertoire represents virtually all cellular translation products, and can thus reveal a foreign presence inside the cell. These peptides are derived from not only conventional but also cryptic translational reading frames, including some without conventional AUG codons. To define the mechanism that generates these cryptic peptides, we used T cells as probes to analyze the peptides generated in transfected cells. We found that when CUG acts as an alternate initiation codon, it can be decoded as leucine rather than the expected methionine residue. The leucine start does not depend on an internal ribosome entry site–like mRNA structure, and its efficiency is enhanced by the Kozak nucleotide context. Furthermore, ribosomes scan 5′ to 3′ specifically for the CUG initiation codon in a eukaryotic translation initiation factor 2–independent manner. Because eukaryotic translation initiation factor 2 is frequently targeted to inhibit protein synthesis, this novel translation mechanism allows stressed cells to display antigenic peptides. This initiation mechanism could also be used at non-AUG initiation codons often found in viral transcripts as well as in a growing list of cellular genes.
Proteins have been identified for which a unique translational machinery makes use of unconventional start codons.
Immune surveillance by cytotoxic T cells (CTLs) is a key mechanism for detecting and eliminating abnormal cells. These include cells infected with viruses or bacteria, and those that have suffered tumorigenic transformations (
The antigen-presenting cells (APCs), which include almost all nucleated cells, are also very efficient in generating peptides for display by MHC class I (
How APCs generate peptides using non-AUG codons remains obscure. It is believed that cells express only one class of initiator tRNA, RNAiMet, which is specific for AUG and is always charged with the methionine residue (
In this study, we used T cells as probes to analyze the translation mechanism that allows the generation of CUG-initiated antigenic peptides and decodes CUG as the leucine rather than the methionine residue. We found some similarities, but also key differences, between the translation mechanisms mediating initiation at conventional AUG versus CUG codons.
We had previously shown both in transfected cell lines and in a transgenic mouse model that CUG can be decoded as leucine when it serves as the initiation codon for the (CTG)-TFNYRNL peptide (the initiation codon is in parentheses and the remaining amino acids are in single-letter code) (
To address this question, we identified the first amino acid of CUG-initiated peptides that were presented by MHC class I molecules and thus detectable by appropriate T cells. In this assay, we cotransfected cells with cDNA constructs encoding the peptide as well as the MHC molecule that binds and presents the peptide on the cell surface. Binding to the MHC protects the initiation residue from being removed by cellular proteases, which can trim antigenic peptides in the cytoplasm or the endoplasmic reticulum (
To determine if decoding of the CUG initiation codon as leucine was influenced by 5′ or 3′ untranslated regions (UTRs) of the mRNA, we transfected COS-7 cells with cDNA encoding the Kb MHC molecule and the *(CTG)-TFNYRNL* peptide ([CTG]YL8). We refer to this as the xYL8 model. We placed the peptide in three different contexts: first, in the pcDNA1 vector; second, upstream of the IRES in the pIRES2-eGFP vector; and, finally, in the 5′ UTR of green fluorescent protein (GFP) in the pcDNA1 vector. The peptides translated in the transfected cells were extracted and fractionated by HPLC. Each fraction was tested for the presence of the leucine-initiated LTFNYRNL (LYL8) peptide and the methionine-initiated MTFNYRNL (MYL8) peptide using BCZ103 T cells and Kb-expressing L cells (a fibroblast cell line) as APCs. With each construct we found a single peak of antigenic activity that eluted in the same fraction as the synthetic LYL8 peptide (
(A) The indicated synthetic peptides mixed with extracts from untransfected COS-7 cells were separated by RP-HPLC. Each fraction was tested for BCZ103 T cell-stimulating activity with Kb+B7.2+ L cells as APCs. After overnight incubation, the β-galactosidase induced in activated T cells was measured using the substrate chlorophenol red-β-pyranoside, which yields a colored product with absorbance at 595 nm. The arrows indicate the reproducible peak elution times for the MYL8 and the LYL8 peptides. Injections of buffer alone (Buffer) were carried out under identical conditions, and the fractions were assayed in parallel to ensure absence of cross-contamination between runs.
(B–D) Extracts from COS-7 cells transfected with cDNA encoding Kb and the indicated constructs were separated by RP-HPLC. Fractions were tested for BCZ103 T cell-stimulating activity as in (A).
(E) The indicated synthetic peptides mixed with extracts from untransfected COS-7 cells were separated by RP-HPLC. Each fraction was tested for the specific DBFZ T cell-stimulating activity with Db+B7.2+ L cells as APCs as in (A). The arrows indicate the reproducible peak elution times for the MM9 and the LM9 peptides.
(F and G) Extracts from COS-7 cells transfected with cDNA encoding Db and the indicated constructs were separated by RP-HPLC. Fractions were tested for DBFZ T cell-stimulating activity as in (E).
(H) A range of concentrations of the indicated synthetic peptides was tested for BCZ103 T cell-stimulating activity with Kb+B7.2+ L cells as APCs.
(I) A range of concentrations of the indicated synthetic peptides was tested for DBFZ T cell-stimulating activity with Db+B7.2+ L cells as APCs.
We next tested whether the LYL8 coding sequence itself enabled the leucine start. We examined the initiating amino acid used for a different peptide presented by the Db MHC class I molecule that satisfied the conditions required for our assay: that an MHC molecule present the peptide, that HPLC allow distinction between the leucine- and methionine-initiated forms, and that a T cell cross-react with the peptides with leucine or methionine residues at the first position. When COS-7 cells were transfected with cDNA constructs encoding the Db MHC molecule and the *(CTG)-SNEN-METM peptide derived from the influenza nucleoprotein, only the leucine-initiated LM9 peptide was detected in the HPLC-fractionated cell extracts (
At present it is difficult to quantify the fraction of the total translated material that is initiated with leucine, because the different peptides may have different stabilities in the cell. Furthermore, the T cell hybridomas could respond to the methionine- and leucine-initiated peptides with differing sensitivities. Indeed, the BCZ103 T cell hybridoma responds to LYL8 approximately 30-fold better than to MYL8 (
In previous studies, we explored the possibility that leucine was used as the first amino acid because the ribosome may have begun at an upstream alternate initiation codon (there are no upstream ATGs in these constructs) and read through the stop codon before the CTG initiation codon in the LYL8 coding sequence. We showed that increasing the number of stop codons upstream of CTG from one to six had no effect on LYL8 expression, and that substituting the CTG codon with other leucine-encoding triplets essentially ablated peptide expression (
(A) The nucleotide sequences of R0 and R1 constructs encode the SVL9 peptide followed by a termination codon and the LYL8 peptide. In the R0 construct, the SVL9 peptide is in frame with an ATG initiation codon. In the R1 construct, a single nucleotide is inserted after the ATG codon and causes the SVL9*LYL8 coding sequence to be out-of-frame with the ATG. The in-frame translation products are underlined and arrows indicate the potential initiation codons.
(B and C) The R0 and R1 constructs were transfected into Lmtk– cells along with the appropriate MHC molecule. They were tested for SVL9/Db expression using the 30NX/B10Z hybridoma (B) and LYL8/Kb expression using the BCZ103 T cell hybridoma (C).
(D) Extracts from COS-7 cells transfected with cDNA encoding Kb and the indicated constructs were separated by RP-HPLC. Fractions were tested for BCZ103 T cell stimulating activity as in
We also considered the possibility that leucine was used as the first amino acid because of an RNA modification that introduced an AUG immediately upstream of the peptide-coding sequence. However, the experiments below show that the 5′ UTR influences the leucine start because it is affected by the Kozak context, by the presence of an upstream hairpin, and by the presence of upstream initiation codons. Together, these findings demonstrate that the mRNA remained intact.
The efficiency of initiation at a given AUG codon depends on the identity of the surrounding nucleotides. These nucleotides, commonly referred to as the “Kozak context,” have a substantial influence on protein synthesis. Kozak found that
We inserted synthetic oligonucleotides *(
(A and B) The indicated degenerate oligonucleotides were cloned into the pcDNA1 vector. “
(C) Three sets of 18 representative plasmids, each yielding high, intermediate, and low responses (as shown) were selected for nucleotide sequencing.
(D) Summary of the nucleotide sequences of plasmids yielding high, intermediate, and low responses. The left, middle, and right panels, respectively, correspond to the plasmids shown in (C). Each panel shows the percent of each nucleotide found at the –6, –3, and +4 degenerate positions indicated by the “
In the above model, because the initiation codon was not included within the final SEL8 antigenic peptide product protected by the MHC molecule, the identity of the amino acid residue specified by the CUG initiation codon could not be determined. Thus, we could not distinguish whether the Kozak context affected the leucine start, the methionine start, or both. To resolve this question, we turned to the (CTG)YL8 model, in which the predominant T cell-stimulating activity is the leucine-initiated LYL8 peptide (see
We first transfected cells with the two constructs as well as the appropriate Kb MHC cDNA. After 2 d the transfected cells were assayed for their ability to stimulate the BCZ103 T cell. Cells expressing the LYL8 peptide with its CTG initiation codon in the “Excellent Kozak” context (T at –6, A at –3) were superior to those with the CTG codon in a “Poor Kozak” context (G at –6, T at –3) in stimulating the T cell response (
(A) Lmtk– cells were transfected with Kb cDNA and the indicated “Excellent” and “Poor” constructs encoding the (CTG)YL8 peptide. They were tested for LYL8/Kb expression using the BCZ103 T cell hybridoma.
(B) Extracts from COS-7 cells transfected with cDNA encoding Kb and the indicated constructs were separated by RP-HPLC. Fractions were tested for BCZ103 T cell-stimulating activity with Kb+B7.2+ L cells as APCs. The arrows indicate the peak elution positions for the MYL8 and the LYL8 peptides. T cell responses to fractions collected after injecting sample buffer alone (Buffer) are also shown to indicate absence of cross-contamination between runs.
In most cases, ribosomes bind mRNA at the 5′ cap and scan in the 3′ direction for the first AUG in an appropriate Kozak context. Thus, for approximately 90% of mRNA transcripts, the 5′-most AUG initiates protein synthesis (
We first transfected COS-7 cells with constructs encoding the ATG-initiated MYL8 peptide and another encoding MYL8 downstream of the heat-stable hairpin. We titrated the transfected cells and assayed the T cell response to the peptides presented on the cell surface. As expected, the presence of the hairpin inhibited MYL8 expression (
COS-7 cells were transfected with cDNA encoding Kb and the indicated constructs. (A and C) The cells were titrated and peptide expression was tested with BCZ103 T cells. (B and D) GFP expression in the transfected cells was assayed by fluorescence-activated cell sorting. GFP fluorescence (shaded histograms) is not observed in untransfected cells (or in cells transfected with a vector not encoding GFP [unpublished data]).
We next asked whether ribosomes responsible for the CUG/leucine start were scanning specifically for CUG initiation codons, or whether they were able to start at conventional AUG initiation codons as well. To address this question, we placed “decoy” ATG and CTG codons upstream of and out of frame with the CTG codon initiating expression of the peptide, and asked whether their presence affected peptide translation.
As a positive control, we transfected cells with a construct encoding the ATG-initiated MYL8 and another encoding the same MYL8 peptide but with three ATGs upstream of and out of frame with the peptide (ATG)3ATG. The control constructs had CAGs instead of ATGs, because the CAG codon does not possess initiation activity. As expected, the presence of upstream out-of-frame ATGs dramatically reduced ATG-initiated MYL8 peptide expression. The reduction in MYL8 peptide was seen both when the transfected cells were used directly to stimulate T cells and when peptides from the transfected cells were extracted, separated by HPLC, and then assayed with T cells (
(A, C, and E) Lmtk– cells were transfected with the indicated constructs and Kb cDNA. After 2 d they were tested for MYL8/Kb or LYL8/Kb expression using the BCZ103 T cell hybridoma. Error bars represent the standard deviation of three replicate wells. (ATG)3ATG (solid circles, A) denotes the ATG-initiated peptide preceded by three ATGs upstream of and out of frame with the peptide; (CAG)3ATG (open circles, A) is the identical DNA construct but the upstream ATGs were replaced with CAG. (ATG)3CTG (solid circles, C) denotes the CTG-initiated peptide preceded by three ATGs upstream of and out of frame with the peptide; (CAG)3CTG (open circles, C) is the identical DNA construct but the upstream ATGs were replaced with CAG. (CTG)3CTG (solid circles, E) denotes the CTG-initiated peptide preceded by three CTGs upstream of and out of frame with the peptide; (CAG)3CTG (open circles, E) is the identical DNA construct but the upstream CTGs were replaced with CAG.
(B, D, and F) Extracts from COS-7 cells transfected with cDNA encoding Kb and the indicated constructs were separated by RP-HPLC. Fractions were tested for BCZ103 T cell-stimulating activity with Kb+B7.2+ L cells as APCs. Arrows indicate the peak elution positions of the MYL8 and the LYL8 peptides. Points on graphs correspond to those in (A), (C), and (E).
We then transfected cells with a construct encoding (CTG)YL8 and another encoding (CTG)YL8 with three ATGs upstream of and out of frame with the peptide (ATG)3CTG. When the transfected cells were used directly to stimulate T cells, the upstream ATGs had little effect (
Finally, we transfected COS-7 cells with constructs encoding (CTG)YL8 with three CTGs upstream of and out of frame with the peptide (CTG)3CTG. When the transfected cells were used directly to stimulate T cells, we saw a small but consistent inhibition (
Note, however, that the upstream CUGs, despite an “Excellent Kozak” context, inhibited the leucine start weakly. This effect contrasts with the inhibition caused by the upstream AUGs on the AUG/methionine or the CUG/methionine starts and suggests that other features are required for an efficient CUG/leucine start. Interestingly, one form of the ASCT2 amino acid transporter is initiated with multiple CUG and GUG codons in close proximity (
Finally, we were interested to know whether the leucine start requires eukaryotic translation initiation factor 2 (eIF2), which is responsible for loading the RNAiMet onto the 40S ribosome. Cells target eIF2 by phosphorylating its α subunit (eIF2α) to inhibit protein synthesis in response to a number of stress signals, including viral infection, starvation, and the accumulation of unfolded proteins. Ribosomes release eIF2-guanosine diphosphate (GDP) after the AUG initiation codon is reached, and GDP is exchanged for guanosine triphosphate (GTP) with the assistance of another protein, eIF2B, before eIF2 can be used for another round of translation initiation. When eIF2α is phosphorylated, it binds eIF2B with unusually high affinity and thus prevents subsequent nucleotide exchange. Because eIF2B is limiting in the cell, phosphorylation of only a fraction of eIF2α can substantially inhibit translation globally (
To approach this question, we transfected HeLa cells with (CTG)YL8 or (ATG)YL8 constructs. We then assayed peptide expression in cells that had or had not been treated to induce phosphorylation of eIF2α. It was a challenge to induce phosphorylation of eIF2α for long enough to see an effect on peptide expression without causing substantial toxicity. Furthermore, we could not disrupt peptide/MHC assembly in the endoplasmic reticulum, a requirement that ruled out standard reagents such as dithiothreitol, thapsigargin, and tunicamycin. We also did not want to inhibit peptide elongation, which ruled out amino acid starvation.
The optimal treatment for our purposes was sodium arsenite (NaAs). Arsenite reacts with sulfhydryl groups and causes phosphorylation of eIF2α presumably by inducing an unfolded protein response, although the precise mechanism remains unknown (
HeLa cells transfected with cDNA encoding Kb together with cDNA encoding either the ATG- or CTG- initiated peptides were treated for 4 h with 50 μM NaAs, with brefeldin A (BfA), or left untreated (UT).
(A) Transfected cells treated with NaAs or without (UT) were lysed and tested for phosphorylation of eIF2α by Western blot and for tubulin as a loading control.
(B) The transfected cells were titrated and tested for their ability to stimulate BCZ103 T cells.
We assayed the HeLa cells for peptide expression using the BCZ103 T cell hybridoma (
The effect of eIF2α phosphorylation on the leucine start strikingly mirrors the effect of eIF2α phosphorylation on proteins whose synthesis is directed by the CPV-IRES, which does not require eIF2 (
The CPV-IRESs are to date the only known sequences that allow eIF2-independent initiation in eukaryotic cells. Viruses employ a host of creative strategies to prevent phosphorylation of eIF2α (
In summary, we found that when CUG acts as an alternate initiation codon, it can be decoded as leucine as well as methionine. The leucine start does not depend on mRNA structure or sequence, but its efficiency can be enhanced by the Kozak context. A set of ribosomes is scanning 5′ to 3′ specifically for the CUG initiation codon. While the methionine start is inhibited when cells are treated with NaAs, the leucine start is enhanced, suggesting that leucine initiation is independent of eIF2. This novel translation initiation mechanism provides cells not only antigenic peptides but also a potential tool for translational control.
Lmtk–, COS-7, Kb+B7.2+ L, Db+B7.2+ L, BCZ103, 30NX/B10Z, and DBFZ cells have been described (
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
Sequences for constructs depicted in
The DEAE-dextran transfection method was used in
The T cell assay has been described previously (
The HPLC assay has been described previously (
Cells were transfected with GeneJuice (Novagen) as described above, but the transfection medium was left on for only 4 h before the cells were lifted and split into multiple dishes with fresh medium. Cells were allowed to rest for 12 h before sodium arsenite treatment. They were treated with 50 μM NaAs (Sigma, St. Louis, Missouri, United States), 1× GolgiPlug containing brefeldin A (PharMingen, San Diego, California, United States), or left untreated for 4 h. The cells were then lifted and counted. For the T cell assay, they were titrated in a 96-well plate. The assay was as described above, except that 1× GolgiPlug was added in order to “freeze” the cells in their state at the end of treatment. For the Western blot, they were incubated on ice for 5–10 min in lysis buffer (20 mM HEPES [pH 7.5], 150 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM EDTA, 10 mM tetrasodium pyrophosphate, 100 mM NaF, 17.5 mM β-glycerophosphate, 0.4 U/ml aprotinin, 10 μg/ml leupeptin, 1 mM PMSF, 0.1 mM pepstatin A, and complete protease inhibitor cocktail [Roche, Basel, Switzerland]). The lysate was spun for 15 min at 4 °C, and the supernatant was transferred to a tube containing an equal volume of 2× SDS-PAGE sample buffer (100 mM Tris-HCl [pH 6.8], 20% glycerol, 4% SDS, bromophenol blue, and 5% β-mercaptoethanol). The sample was then heated in water just off the boil for 5 min, separated on a 10% SDS-PAGE gel, and transferred to a nitrocellulose membrane. The membrane was blocked for 1 h at room temperature in TBS–0.1% Tween 20–5% bovine serum albumin (TBS, 0.02 M Tris-HCl [pH 7.6] with 0.137 M NaCl), incubated for 1 h with primary antibody (#44–728, at a 1:2000 dilution; Biosource, Camarillo, California, United States) in TBS–0.1% Tween 20–5% bovine serum albumin, washed four times for 5 min with TBS–0.1% Tween 20, incubated for 40 min with secondary antibody (anti-rabbit-HRP #NA934V, at a 1:30,000 dilution; Amersham, Little Chalfont, United Kingdom), washed four times for 5 min with TBS–0.1% Tween 20, incubated for 5 min in substrate (SuperSignal West Femto Maximum Sensitivity Substrate, #34095; Pierce Biotechnology, Rockford, Illinois, United States), and exposed to film. An antibody to α-tubulin (#sc-5546; Santa Cruz Biotechnology, Santa Cruz, California, United States) was used as a loading control.
We thank P. Sarnow and E. Jan (Stanford University) for discussions and advice, D. King for peptide synthesis, S. Bakkour, M. Hutchinson, and Y. Ow for excellent technical assistance, and T. Serwold and G. Hammer for thoughtful suggestions. This research was supported by grants to N. Shastri from the NIH.
antigen-presenting cell
cricket paralysis virus-like internal ribosome entry site
cytotoxic T cell
eukaryotic translation initiation factor 2
eIF2 α subunit
guanosine diphosphate
green fluorescent protein
guanosine triphosphate
high performance liquid chromatography
internal ribosome entry site
major histocompatibility complex
sodium arsenite
methionyl initiator RNA
reverse-phase HPLC
untranslated region