Regulation of protein translation through mRNA structure influences MHC class I loading and T cell recognition

Purine Bias Within the GAr Domain of EBNA1 Results in an Unstable mRNA Secondary Structure.

In delineating the mechanism by which the GAr domain exerts its cis-inhibitory effect on self-synthesis, an examination of the EBNA1 sequence revealed a dramatic overrepresentation of purine residues within the GAr domain (14). More than 99% of the glycine residues and 100% of the alanine residues within the GAr domain are comprised of purine codons (GGG, GGA, and GCA) compared with expected human averages for glycine and alanine purine codons of 49.3% and 33.3%, respectively (Fig. 1A). RNA secondary structure analysis (MFOLD) (15, 16) of 400 nt of native EBNA1 GAr sequence revealed a complete absence of secondary structure (Fig. 1B). MFOLD analysis reveals many potential structures. Fig. 1B is a representation of the most stable predicted structure. Similar analysis after modification of this same 400-nt GAr sequence to remove purine bias, while still maintaining the encoded protein sequence, demonstrated an enhanced and more stable mRNA secondary structure, reflected by the significantly more negative Gibbs free energy value, δG, of −218.7 kcal/mol compared with −50.8 kcal/mol for the native form (Fig. 1B). Based on this analysis, we hypothesized that an abnormal mRNA structure caused by unusual codon bias may play a crucial role in regulating the synthesis of EBNA1. To test this hypothesis, a number of glycine (GGN) and alanine (GCN) codons within the GAr domain were modified, by mutating the choice of third bases (N) to reflect the expected human average codon usage (Fig. 1A) (17). A series of sequentially designed GAr oligonucteotides were synthesized and inserted into the coding sequence of EBNA1 devoid of its normal GAr domain (E1-ΔGA) (Fig. 1C). These engineered GAr domains, GAr(N) referring to native EBNA1 GAr sequence or GAr(M) referring to codon-modified EBNA1 GAr sequence, encoded varying lengths of the GAr sequence (100–500 nt) (Fig. 1C).

A Codon-Modified GAr Influences EBNA1 Expression.

Analysis of in vitro-transcribed EBNA1 mRNA comprised of native or codon-modified GAr sequences by in vitro translation (IVT) assays demonstrated that the translational efficiency of EBNA1 sequences encoding a codon-modified GAr domain was significantly enhanced when compared with its matched pair of EBNA1 sequences encoding a native GAr domain (Fig. 2 A–C). In addition, there was a significant reduction in translational efficiency as the native or modified GAr domain was increased in length (Fig. 2 A and B). Furthermore, in agreement with the IVT assays, Western blot analysis of intracellular expression of EBNA1-GFP sequences in an epithelial cell line (HEK293) also showed a dramatic increase in protein expression for codon-modified EBNA1 constructs (Fig. 2 D and E). Differences in protein expression levels between native and codon-modified EBNA1-GFP sequences were not caused by differential transfection efficiencies because transfectants showed similar percentages (33–36%) of GFP-positive cells. The correlation between GAr size and level of translation varies slightly between the in vitro and in vivo model systems, suggesting that the overall translational efficiency may be related not only to predicted mRNA secondary structure but also to other factors such as tRNA abundance or other fortuitous binding proteins to Gly/Ala-rich proteins. Taken together, these observations demonstrate that purine bias within the EBNA1 GAr domain directly impacts EBNA1 protein expression.

Regulation of EBNA1 Synthesis Occurs at the Translational Level.

To determine whether the enhanced synthesis of codon-modified EBNA1 was being influenced by transcriptional and/or translational factors, we first assessed mRNA synthesis levels of EBNA1 sequences encoding either native or codon-modified GAr domains. pcDNA3 expression constructs encoding E1-GAr(400N), E1-GAr(400M), E1-GAr(500N), or E1-GAr(500M) were linearised with XbaI and transcribed with T7 RNA polymerase by using a Riboprobe in vitro transcription system supplemented with 50 μCi [α-³²P]UTP. Data presented in Fig. 3A demonstrate that modification of the GAr coding sequences had minimal impact on mRNA synthesis levels, consistent with previous studies comparing EBNA1 and EBNA1-ΔGA steady-state mRNA levels (11). To discount the possibility that modification of codons within the GAr domain may have altered EBNA1 mRNA turnover, we next performed mRNA degradation analyses by using quantitative real-time RT-PCR (qRT-PCR). The results presented in Fig. 3B demonstrate that mRNA from EBNA1 sequences encoding a 400-nt modified repeat, GAr(400M), displayed a half-life of 231 ± 142 min, compared with a value of 366 ± 30 min for mRNA from EBNA1 encoding a 400-nt native repeat, GAr(400N). However, after statistical analysis there was no significant difference in the two half-lives (P > 0.05) (Fig. 3B), indicating that the overall rates of decay of mRNAs of the two representative EBNA1 constructs encoding either native or codon-modified GAr sequences are comparable. As expected, protein degradation rates of EBNA1-GFP sequences encoding either GAr(400N) or GAr(400M) domains displayed similar intracellular kinetics after cycloheximide treatment over a 30-h time course (Fig. 3C).

As the observed differences in protein expression of EBNA1-GFP encoded by either native or modified repeat sequences were not caused by significant differences in mRNA transcription or turnover, we assessed absolute rates of EBNA1 synthesis relative to a common protein standard by using a double labeling protocol (18). Although the rate of synthesis for tubulin remained unchanged in HEK293 cells transfected with the different EBNA1 expression vectors, the translation rates for two representative EBNA1 constructs encoding modified GAr sequence, E1-GAr(400M)-GFP and E1-GAr(300M)-GFP, were 60–70% higher, respectively, than their corresponding matched pair of native GAr sequence (Fig. 3 D and E), confirming earlier results that a codon-modified GAr domain of EBNA1 critically modulates translational efficiency (Fig. 2A). Transfection efficiencies for native and codon-modified EBNA1-GFP sequences in the above assays were similar as assessed by the percentage of GFP-positive cells.

A GAr Antisense Oligonucleotide Enhances EBNA1 Synthesis in Vitro.

Having observed an increase in translation efficiency as the mRNA of the codon-modified GAr domain becomes more structured because of a greater propensity to form stem loops after modification of the codons, we reasoned that if we could stabilize the mRNA of native GAr sequences, this modification may override the GAr cis-inhibitory affect on translation. To test this hypothesis, we performed IVT assays of EBNA1 encoded by native or codon-modified GAr sequences in the presence of a 21-mer GAr(N) antisense oligonucleotide. Fig. 3F demonstrates that addition of the antisense oligonucleotide results in an increase in translation product for E1-GAr(300N) and full-length EBNA1, with the EBNA1 expression construct encoding 300 nt of native GAr sequence, E1-GAr(300N), reaching 50% of the translational level observed for the matching codon-modified E1-GAr(300M) construct. An increase in translation product was also observed for E1-GAr(300M) because stretches of the antisense oligo also matched regions of the codon-modied GAr sequence. As the E1-ΔGA expression level was not enhanced after the same antisense oligonucleotide treatment, the data clearly implicate the EBNA1 GAr in translational regulation (Fig. 3F).

Increased Translation Efficiency of EBNA1 Encoded by a Modified GAr Results in Enhanced Antigen Processing.

We next investigated the impact of the enhanced translational efficiency of codon-modified EBNA1 on the endogenous presentation of MHC class I-restricted epitopes derived from EBNA1. In the first set of these experiments we assessed the loading of MHC class I molecules with a H-2K^b-restricted epitope from ovalbumin, SIINFEKL (Ser–Ile–Ile–Asn–Phe–Glu–Lys–Leu, residues 257–264), which was inserted into the EBNA1 sequence (13). H-2K^b-expressing HEK293 cells were transiently transfected with either E1-GFP, E1-ΔGA-GFP, E1-GAr(300N)-GFP, E1-GAr(300M)-GFP, E1-GAr(400N)-GFP, or E1-GAr(400M)-GFP expression constructs, with average transfection efficiencies between 75% and 80% (data not shown). For each SIINFEKL containing construct, a separate transfection of the parent construct without SIINFEKL was done to provide a negative control. After an overnight transfection, cells were assessed by flow cytometry for GFP expression and surface expression of H-2K^b-SIIN complexes by using a mAb (25-D1.16) that recognizes the SIIN epitope bound to H-2K^b molecules (19). For each SIINFEKL-containing construct a separate transfection of the parent plasmid without SIINFEKL was done to provide a baseline above which an increase in fluorescence would indicate H-2K^b-SIIN expression. Representative data from one experiment (Fig. 4A) show that the percentage of cells expressing H-2K^b-SIINFEKL complexes was 4- to 11-fold higher when constructs with EBNA1 encoded by codon-modified GAr sequences were transfected compared with the EBNA1 sequences encoding a native GAr domain. We next assessed the endogenous processing and surface presentation of two different CD8⁺ T cell epitopes. One of these was a HLA B8-restricted epitope derived from the EBNA3 protein, FLRGRAYGL (Phe–Leu–Arg–Gly–Arg–Ala–Tyr–Gly–Leu, residues 345–361), which was inserted into the C terminus of the EBNA1 sequence (13), while the second epitope was encoded within the EBNA1 sequence and restricted through HLA B35, HPVGEADYFEY (His–Pro–Val–Gly–Glu–Ala–Asp–Tyr–Phe–Glu–Tyr, residues 407–417). HLA B8 and HLA B35-positive human cells expressing EBNA1 encoded by either native or codon-modified GAr domains were incubated with the CD8⁺ T cells specific for FLRGRAYGL or HPVGEADYFEY epitopes and stimulation was assessed by using intracellular cytokine assays. Data presented in Fig. 4 B and C show that human cells expressing EBNA1-GFP protein encoded by a codon-modified GAr domain [E1-GAr(300M)-GFP, E1-GAr(400M)-GFP and/or E1-GAr(500M)-GFP] stimulated 2- to 7-fold more IFN-γ producing EBV-specific T cells compared with cells expressing EBNA1-GFP encoded by a native GAr domain [E1-GAr(300N)-GFP, E1-GAr(400N)-GFP and/or E1-GAr(500N)-GFP]. Interestingly, these experiments demonstrated that as the GAr domain is increased in length within the EBNA1 sequence the number of IFN-γ producing EBV-specific T cells are decreased, closely mimicking the IVT data (Fig. 2 A and B).

Cell Lines.

Cell lines (HeLa, DG75, and SVMR6) were routinely maintained in RPMI medium 1640 supplemented with 2 mM l-glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin plus 10% FCS (referred to as growth medium). HEK293 cells were grown in DMEM/10% FCS.

Generation of EBNA1 Expression Constructs.

Full-length EBV-encoded EBNA1, (E1) and EBNA1-ΔGA, (E1-ΔGA) were cloned into the expression vector pcDNA3 (Invitrogen). To generate EBNA1 expression constructs with varying lengths of native or modified GAr sequences, 105-mer oligonucleotides were synthesized, representing sequential 102-nt increments of the EBNA1 GAr sequence [supporting information (SI) Table S1]. Mutagenesis of the E1-ΔGA expression construct allowed the insertion of PacI and ClaI restriction sites at nucleotide position 250. The first 105-mer sense and antisense oligonucleotides of native or modified GAr sequence were annealed and cloned into the PacI and ClaI restriction sites to generate E1-GAr(100N) or E1-GAr(100M), where N represents the native GAr sequence and M represents the codon-modified GAr sequence. Sequential 105-mer GAr oligonucleotides were cloned into the 3′ ClaI restriction site of each preceding GAr insertion to generate the following EBNA1 expression constructs: E1-GAr(200N), E1-GAr(200M), E1-GAr(300N), E1-GAr(300M), E1-GAr(400N), E1-GAr(400M), E1-GAr(500N), and E1-GAr(500M). The above EBNA1 expression vectors were also subcloned in-frame with a sequence coding for GFP (pEGFP-N1; Clontech). To assess EBNA1 endogenous processing, a sequence encoding the HLA-B8-restricted, EBNA3 epitope FLRGRAYGL (23) was inserted into EBNA1 constructs as described (13). For the assessment of endogenous loading of MHC class I molecules, a H-2K^b-restricted epitope from ovalbumin, SIINFEKL (24), was inserted in-frame between EBNA1 and GFP.

In Vitro Transcription/Translation Assays.

EBNA1/pcDNA3 expression constructs were linearized with XbaI and 1 μg of template transcribed with T7 RNA polymerase by using a Riboprobe in vitro transcription system (Promega) supplemented with 50 μCi [α-³²P]UTP (Amersham Biosciences). For translation assays EBNA1/pcDNA3 vectors were transcribed and translated in vitro with T7 RNA polymerase by using a coupled transcription/translation reticulocyte lysate system (Promega) supplemented with 250 μCi ³⁵[S]methionine (Amersham Biosciences). Lysates were subjected to SDS/PAGE and autoradiography as described (13). IVT assays were performed in the presence or absence of a 21-mer antisense oligonucleotide, (5′-TCCCGCTCCTGCTCCTGCTCC-3′) to the native GAr sequence at a molar ratio of 10:1 to the GAr.

Transfection of EBNA1 Constructs and Detection by Immunoblotting.

DG75 cells (5 × 10⁶) were transfected with 10 μg of expression constructs by using the BioRad Gene Pulser (960 μF, 250 V, 0.4-cm gap electrode, 300-μl assay volume, 25°C). For adherent cell lines HEK293 or SVMR6 cells (2 × 10⁵) were transfected with 0.4 μg of EBNA1 expression constructs by using Effectene (Qiagen) according to the manufacturer's instructions. Twenty-four hours posttransfection, cells were harvested and samples were subjected to SDS/PAGE and immuno-blotted with either anti-GFP (1:2,000) or an actin mAb (1:1,000) as described (25). Half-life analysis of EBNA1-GFP fusion proteins were performed as described (25).

Isolation of RNA and RT-PCR.

Total RNA was extracted from HEK293 cells transiently transfected with EBNA1 expression constructs after treatment with 5 μg/ml Actinomycin D (Sigma) over a time course of 2 h. The RNA was purified with an RNeasy Plus Mini Kit (Qiagen) and quantified by spectrophotometric measurements at 260 and 280 nm, and its integrity was verified by the OD₂₆₀/OD₂₈₀ absorption ratio (>1.8) and visualization on an agarose gel.

qRT-PCR.

cDNA synthesis of selected EBNA1 sequences was undertaken with 1 μg of isolated RNA per sample by using MMLV SuperScript III reverse transcriptase (Invitrogen) and an anchored oligo(T)₁₈ primer combined with random hexamers. qRT-PCR using the Sybr Green-based fluorescent detection system and the ABI Prism 7900 Sequence Detection System (Applied Biosystems) was used to measure mRNA abundance. Primers were designed by using DS Gene (version 1.5) (Accelrys) software (Table S2). A constant amount of cDNA, corresponding to 10 ng of reverse-transcribed RNA derived from each sample was used for qRT-PCR measurements. Four technical replicates were performed for each gene investigated. This process allowed quantification of the target gene relative to a constant reference gene in each sample by using threshold cycle (Ct) data. Ribosomal protein P0 (RPLP0; GenBank accession no. NM_053275) was used as the reference gene for all samples.

Each qRT-PCR (5 ml total volume) contained 2.5 ml of 2× Sybr Green Master Mix (Applied Biosystems), 0.25 ml of each primer giving a final concentration of 500 nM each, 1.0 ml water, and 1.0 ml of a 1/10 dilution of the stock cDNA template. The cycling conditions consisted of 40 cycles of 95°C for 15 s and 60°C for 1 min. At the completion of each run, a dissociation melt curve analysis was performed. All melt curves for detectable amplicons showed a single peak and were consistent with the presence of a single specific amplicon. Data analyses used Q-gene qRT-PCR analysis software (Gene Quantification; www.gene-quantification.info) and gene expression was normalized relative to RPLP0 expression. A linear least-squares statistical analysis was applied for half-life calculations.

Measurement of EBNA1 Protein Synthesis.

HEK293 cells (2 × 10⁵) were transfected with selected EBNA1-GFP expression constructs. Twenty-four hours posttransfection the cells were metabolically labeled at 37°C for 12–14 h in growth medium containing 20 μCi/ml ³[H]methionine (Amersham Biosciences). Cells were washed in PBS and incubated in methionine-free growth medium for 30 min at 37°C preceding a 30-min pulse with 100 μCi ³⁵[S]methionine. Following the pulse, cells were lysed in Tris-buffered saline with 1% Triton X-100 and protease inhibitors and precleared with Protein A Sepharose, and lysates were immunoprecipitated with anti-GFP or a mAb to β-tubulin (Sigma). Immunoprecipitated samples were added to 10 ml of scintillant fluid, Ultima Gold (PerkinElmer Life and Analytical Sciences), and counted on a Packard liquid scintillation analyzer, Tri-carb 2100TR.

Detection of Cell Surface K^b-SIINFEKL.

293KbC2 cells, which stably express H-2K^b (26), were transfected with E1-SIINFEKL-GFP constructs by using Lipofectamine 2000 (Invitrogen). For each SIINFEKL-containing construct, a separate transfection of the parent construct without SIINFEKL was done to provide a negative control. Cells were harvested after an overnight transfection and stained with mAb 25D1.16 (19) conjugated to Alexa Fluor 647 (Molecular Probes/Invitrogen), a kind gift of M. Princiotta (Upstate Medical University, Syracuse, NY) and J. Yewdell (National Institute of Allergy and Infectious Diseases, Bethesda, MD) (27), on ice for 30 min. Cells were washed and analyzed by flow cytometry (LSR II; BD Biosciences) for GFP expression and 25D1.16 binding. GFP⁺ events were determined from a population gated tightly on forward scatter (FSC) × side scatter (SSC) (to eliminate cells adversely effected by the transfection), and average transfection efficiencies were 75% or better. As the majority of 25D1.16⁺ events were contained in the GFP^hi population of all SIINFEKL-containing constructs, we restricted further analysis to the brightest half of transfected cells. To determine the level of surface K^b-SIINFEKL for each version of construct, the 25D1.16-Alexa 647 intensity of the brightest 50% of GFP⁺ events was compared between non-SIINFEKL bearing parent constructs and the relevant test SIINFEKL-containing construct by using the Overton Subtraction method (28) (Flowjo software; Tree Star).

Intracellular Cytokine Staining.

SVMR6 or DG75 cells transfected with EBNA1-GFP constructs were fixed with 2% paraformaldehyde, washed, and resuspended in growth medium. T cells, either an FLRGRAYGL-specific clone (LC13) or a HPVGEADYFEY-specific T cell line, were incubated overnight (FLR) or for 6 h (HPV) at 37°C with fixed EBNA1-GFP transfectants at responder-to-stimulator ratios of 5:1, 10:1, and 20:1 in growth medium supplemented with Brefeldin A (BD Pharmingen). Cells were washed and incubated with either phycoerythrin (PE)-labeled HLA B3508 HPVGEADYFEY pentamer (ProImmune; Oxford) followed by peridinin chlorophyll protein (perCP)-conjugated anti-CD8 for HPV-specific T cells or allophycocyanin-conjugated anti-CD3 and perCP-conjugated anti-CD8 for LC13, rewashed, then fixed and permeabilized with cytofix/cytoperm (BD Pharmingen) at 4°C for 20 min. Cells were washed in perm/wash, incubated with allophycocyanin-conjugated or PE-conjugated anti-IFNγ (BD Pharmingen) at 4°C for 30 min, washed again with perm/wash, resuspended in PBS, and analyzed on a FACSCanto (BD Biosciences).