Recoding elements located adjacent to a subset of eukaryal selenocysteine-specifying UGA codons

RNA secondary structure predictions and comparative sequence analysis

Sequences located downstream of 36 selenocysteine-encoding UGA codons from 25 human selenoprotein genes, including the selenoprotein P gene, which contains 10 selenocysteine-encoding UGA codons, were examined for potential RNA secondary structures using mfold version 3. 1 (for specific folding parameters and accession numbers, see Materials and methods). Mfold predicts the minimum free energy, ΔG, for multiple RNA foldings of an input RNA sequence (Mathews et al, 1999; Zuker, 2003). The 60 nucleotides located just 3′ of the UGA codon were used as input and then ranked according to the most favorable predicted ΔG value for each RNA sequence. For the 36 selenocysteine insertion sites analyzed, ΔG values for the best RNA folding varied from −2.2 to −36.4 kcal/mol. RNA structures with ΔG values lower than −25 kcal/mol were predicted downstream of five UGA codons; SelO ΔG=−25.8, SEPN1 UGA1 ΔG=−30.5, SEPN1 UGA2 ΔG=−29.5, SelH ΔG=−32.4, and Sps2 ΔG=−36.4 (see Discussion). Analysis of sequences downstream of the 10 selenocysteine-encoding UGA codons in the human selenoprotein P gene reveals a range of ΔG values from −2.2 to −16.6. The identification of several selenocysteine-encoding UGA codons with little potential to form downstream RNA secondary structure suggests that, unlike the case in bacteria, a stem loop near the UGA codon is not required for selenocysteine insertion in eukarya. However, the possibility of structures formed by long-range RNA interactions cannot be ruled out.

Based on promising phylogenetic conservation and the identification of sequence variations that maintain base-pairing potential (see below), the sequence surrounding the selenocysteine-encoding codon located in exon 10 of the SEPN1 gene, SEPN1 UGA2, was selected for further examination in this study. SEPN1 genes were identified by BLAST analysis of publicly available protein and nucleotide databases using the human SEPN1 protein sequence as query. Genes with significant similarity to the human SEPN1 gene were identified in 14 vertebrates and two urochordates. Additional BLAST analysis using the distantly related sequence from Ciona intestinalis as query did not identify additional genes with significant similarity to SEPN1. Sequence alignments were performed using ClustalW (Thompson et al, 1994) (Figure 1), revealing a high degree of conservation surrounding the selenocysteine-encoding UGA2 codon. The first UGA codon, UGA1, in the human SEPN1 gene is not phylogenetically conserved and occurs only in an alternately processed transcript with uncertain biological significance. The codon preceding the UGA2 codon and five nucleotides downstream are universally conserved in all SEPN1 genes identified. The potential for a stem loop secondary structure starting seven nucleotides downstream of the UGA codon is predicted and supported by sequence variations that maintain base-pairing potential (Figure 2). The first five base pairs and the size of the predicted stems (17 nucleotides) are preserved. The length of each stem assumes that A:U and G:U pairs are formed at the apical end. These pairings may be nonexistent or only transiently formed in vivo. In addition, three nucleotides in the loop are identical for all predicted structures and a conserved C–A bulge is found after the ninth base pair in the stem of seven out of 13 predicted structures.

An important stem loop

The effect of the surrounding sequence context on decoding of the UGA2 codon located in the human SEPN1 gene was examined using a dual luciferase reporter assay in cultured human embryonic kidney cell line (HEK293). A total of 35 nucleotides located 5′ and 46 nucleotides 3′ of the UGA codon were cloned between the Renilla and firefly luciferase reporter genes in plasmid p2luc (Grentzmann et al, 1998) to create the reporter construct UGA2. The firefly luciferase gene lacks an initiation codon and can only be expressed as a fusion protein with Renilla luciferase if translational stop codon readthrough occurs. Readthrough efficiency is calculated as a ratio of the firefly to Renilla luciferase activities standardized to an in-frame control in which the intervening stop codon has been altered to a sense codon (see Materials and methods; Grentzmann et al, 1998).

Readthrough efficiency was calculated to be 6% for UGA2 (Figure 3B). An equivalent sequence context was tested for all 10 of the human selenoprotein P UGA codons. Translational readthrough levels were measured at 1% or less for all stop codons with the exception of the first UGA codon from the selenoprotein P gene, which revealed translational readthrough levels of approximately 1.7% (data not shown). The importance of the predicted stem loop in SEPN1 UGA2 for stimulation of translational readthrough was addressed by systematic mutagenesis of the stem loop region. Three nucleotides at a time were changed beginning at the bottom of the stem to interrupt base pairing (Figure 3A). Each disruption mutation reduced stop codon readthrough levels to less than 1% (Figure 3B). Pairing potential was restored at each position by making compensatory mutations such that for each consecutive block of three base pairs in the stem, G:C, A:U, and G:U pairs were altered to C:G, U:A, and U:G, respectively (Figure 3A, SC1-3, SC2-3, SC3-3, SC4-3, SC5-3). Altering the first three base pairs from G:C to C:G (SC1-3) failed to restore readthrough levels. However, restoring base–pairing potential at the second (SC2-3), third (SC3-3), fourth (SC4-3), and fifth (SC5-3) set of three base pairs resulted in partial restoration of stop codon readthrough levels to 5, 1, 2.5, and 4.5%, respectively (Figure 3B).

The stem loop structure predicted for the human SEPN1 gene contains a single C:A mismatch following the ninth base pair of the stem. Converting the C:A mismatch to a C:G base pair increased readthrough levels to 10% (Figure 3B, PS), whereas inverting the C:A mismatch to an A:C mismatch reduced readthrough levels to less than 1% (Figure 3B, IMM). Finally, changing the sequence of the loop reduced readthrough levels to background (Figure 3B, LCa), implying that, in addition to the stem, sequences contained in the loop are important for readthrough stimulation.

To determine the ability of this stem loop readthrough element to induce stop codon readthrough in the presence of the natural SEPN1 3′UTR SECIS element, the human SEPN1 SECIS element was cloned into the 3′UTR of the dual luciferase vectors containing the UGA2 stem loop element, or the stem sequence variants SC2-1, SC2-2, and SC2-3 (see Materials and methods). UGA2 readthrough levels increased to approximately 12% with the addition of the SECIS element to the 3′UTR (Figure 3C, UGA2 SECIS). Mutations that disrupt the stem, thereby inactivating the readthrough activity of the stem loop structure (SC2-1 and SC2-2), reduced frameshifting levels to approximately 6% when tested in the presence of the SECIS (Figure 3C, SC2-1 SECIS, SC2-2 SECIS). Restoring the base-pairing potential (SC-3) revealed levels of readthrough (11%) approximating those observed for the wild-type UGA2 in the presence of the SECIS structure (Figure 3C, SC-3 SECIS). These results demonstrate that, in this reporter system, the SEPN1 SECIS element and the stem loop structure each contribute approximately equally to readthrough efficiency at the UGA2 selenocysteine-encoding codon.

Stop codon and upstream sequences

The ability of the SEPN1 sequences to induce stop codon readthrough of UAA and UAG stop codons was examined in HEK293 cells as described above. When the UGA2 stop codon was replaced with UAG, only a slight reduction in readthrough to 4.5% was observed (Figure 4B, UAG). However, low-level readthrough (<1%) was obtained when UGA was replaced with the UAA stop codon (Figure 4B, UAA).

The contribution of SEPN1 upstream RNA sequence to stop codon readthrough efficiency was examined by altering the third position of each codon (with the exception of the single Trp codon, which is encoded by a single codon, UGG) such that the amino-acid sequence encoded was maintained (Figure 4A, 5′ 3POS). Stop codon efficiency was reduced approximately two-fold by these changes (Figure 4B, 5′ 3POS). In a second construct, a single base was deleted from the 5′ end of the readthrough cassette and a single base was inserted three nucleotides prior to the stop codon (Figure 4A, 5′ FS). In this case, the RNA sequence is maintained essentially intact but the sequence of the nascent peptide is changed due to the shift in reading frame. Changing the composition of the nascent peptide did not significantly alter readthrough efficiency.

A significant spacer sequence

The RNA sequences located between the stop codon and the stem loop were changed to examine the effect of this spacer region on stop codon readthrough efficiency (Figure 5A). Mutagenesis of the six-nucleotide spacer, GGUUCA, was tested using the dual luciferase reporter system in HEK293 cells as described above. Comparison of the six-nucleotide SEPN1 spacer region GGUUCA to the previously identified CARYYA readthrough motif revealed a match for the last three nucleotides. Altering the last three nucleotides to AGU or ACA resulted in a modest, approximately two-fold, decrease in readthrough efficiency (Figure 5B, Spa, Spd). However, alterations that retain the YYA composition at this site, CUA or UUA, maintained readthrough levels near wild-type efficiency (Figure 5B, Spb, Spc). When the first three nucleotides of the spacer were changed to CAG and CAA to match the CARYYA motif, readthrough efficiency increased to 12 and 17%, respectively (Figure 5B, Spe, Spf). The contribution of the first three nucleotides of the spacer to readthrough efficiency was further examined by changing each nucleotide independently to a C (Figure 5A, Spg, Sph, Spi). Changing either of the first two nucleotides to a C reduced readthrough levels to 0.2 and 3%, respectively (Figure 5B, Spg, Sph), whereas changing the third nucleotide had no appreciable effect on readthrough (Figure 5B, Spi). The role of spacer length was examined by changing the length to eight nucleotides (Figure 5A, Spj, Spk, Spl) or four nucleotides (Figure 5A, Spm). Each resulted in a reduction in readthrough efficiency to approximately 2% (Figure 5B).

The RNA element described here contains a key sequence separating the UGA codon from the essential downstream stem loop structure. Mutations designed to conform the spacer to the previously identified CARYYA motif (Skuzeski et al, 1991; Beier and Grimm, 2001; Harrell et al, 2002) resulted in higher readthrough levels, with the exception of changing the first G of the SEPN1 spacer to C, which eliminated readthrough. Those changes that were designed to reduce similarity to the known readthrough motif lowered readthrough efficiency. The spacer length is also critical and likely serves to position the structure at the predicted distance to be near the mRNA entrance site (Yusupova et al, 2001) and the ribosome mRNA unwinding center (‘helicase') (Takyar et al, 2005).

MuLV gag stop codon readthrough stimulation by the SEPN1 stem loop

Comparison of the spacer sequence of SEPN1 to that of the spacer region known to be important for MuLV readthrough reveals that the first two and last three nucleotides are identical; MuLV spacer=GGAGGUCA (Figure 5). A dual luciferase reporter was constructed that contains the MuLV stop codon (UAG) and downstream pseudoknot (Figure 6A, MuLVWT). The ability of the SEPN1 stem loop to stimulate readthrough of the MuLV stop codon was tested by replacing the MuLV pseudoknot with the SEPN1 stem loop (Figure 6A, MuLV1). Readthrough efficiency promoted by the wild-type MuLV pseudoknot was 7% and replacing the pseudoknot with the SEPN1 stem loop reduced readthrough to 1% (Figure 6B, MuLVwt, MuLV1). Reducing the spacer length to six nucleotides by replacing it with the SEPN1 spacer or deleting the first and fourth Gs of the MuLV spacer partially restored readthrough efficiency to approximately 3% (Figure 6B, MuLV2, MuLV3). Adding two nucleotides to the end or between the GGU and UCA of the SEPN1 spacer reduced readthrough of the UAG to 1 and 0.5% respectively (Figure 6B, MuLV4, MuLV7), and 2% readthrough was observed with an eight-nucleotide spacer, GGGUGUCA (Figure 6B, MuLV8). Finally, the MuLV and SEPN1 spacer regions do not induce high-level readthrough in the absence of downstream structures, as readthrough efficiency dropped to approximately 0.3% when the MuLV or SEPN1 structures were deleted entirely (Figure 6B, MuLV5, MuLV6).

The strong sequence bias for the six nucleotides located downstream of redefined stop codons, experimental evidence that the spacer region in MuLV contains key sequence stimulators of gag-pol readthrough (Wills et al, 1994), and mutagenesis of the SEPN1 spacer region further illustrate that sequences located adjacent to stop codons can be utilized as a means to facilitate stop codon readthrough for gene expression purposes. In addition, structure ‘swapping' experiments here demonstrate that the SEPN1 stem loop can induce stop codon readthrough of the MuLV gag stop codon, albeit at a lower level than the wild-type MuLV pseudoknot structure.

RNA folding and comparative sequence analysis of readthrough regions

Accession numbers for the 25 Homo sapien selenoprotein genes are as follows: Selenoprotein P, NM_005410; Sps2, NM_012248; SelW, NM_003009; SelV, NM_182704; 15Kda, NM_004261, SelM, NM_080430; TR1, NM_003330; TR2, XM_051264; TR3, NM_006440; Gpx1, NM_000581; Gpx2, NM_002083, Gpx3, NM_002084; Gpx4, NM_002085; Gpx6, NM_182701; DI1, NM_000792; DI2, NM_013989; DI3, NM_001362; SelR, NM_016332; SelT NM_016275; SelN, NM_020451; SelH, NM_170746; SelK, NM_021237; SelS, NM_203472; SelI, NM_033505; SelO, NM_031454. RNA folding parameters for mfold v3.1 are as follows: 37°C, 1 M NaCl, percent suboptimality=5, window=2, maximum interior bulge loop size=30, maximum asymmetry of an interior/bulge loop=30.

Sequences with significant similarity to SEPN1 genes were identified by BLAST analysis. Accession numbers used for multiple alignments are as follows: H. sapiens, NM_206926; Danio rerio, NM_001004294; Mus musculus, BI555369; Gallus gallus, XM_417734; Xenopus tropicalis, BC063905; Ciona intestinalis, AK112355; Rattus norvegicus, CK600139; Xenopus laevis, CF288737; Salmo salar, CK887829; Oryzias latipes, BJ734513; Bos taurus, BF073924; Strongylocentrotus purpuratus, CD330282; Pan troglodytes, AACZ01296920; Fugu rubripes, CAAB01000005; Monodelphis domesticus, AAFR01073613; Tetraodon nigroviridis, CAAE01015019; Ciona savignyi, AACT01005174. Multiple alignments of sequences were generated using ClustalW (Thompson et al, 1994) and refined manually.

Luciferase reporter constructs

Complimentary oligonucleotides, to construct the sequences utilized in this manuscript (Supplementary data), were synthesized at the University of Utah DNA/Peptide Core Facility so that when annealed they would have appropriate ends to ligate into the SalI/BamHI restriction sites of the dual luciferase vector, p2luc (Grentzmann et al, 1998). Control constructs were designed such that the UGA codon was changed to CGA. Constructs containing the SECIS element were produced by insertion of the SEPN1 SECIS element into the NotI site of the dual luciferase vector. SEPN1 SECIS insert=GGCCGCAGTGGCTTCCCCGGCAGCAGCCCCATGAT GGCTGAATCCGAAATCCTCGATGGGTCCAGCTTGA TGTCTTTGCAGCTGCACCTATGGGGCGGCC. All dual luciferase constructs were sequence verified.

Cell culture and transfections

HEK293 cell line was obtained from ATCC and maintained in DMEM supplemented with 5% FBS in the absence of antibiotics. Cells used in these studies were subcultured at 70% confluence and used between passages 7 and 15. Cells were transfected using Lipofectamine 2000 reagent (Invitrogen), using the 1-day protocol in which suspension cells are added directly to the DNA complexes in 96-well plates. In all, 25 ng DNA and 0.2 μl Lipofectamine 2000 per well in 25 μl Opti-Mem (Gibco) were incubated and plated in opaque 96-well half-area plates (Costar). Cells were trypsinized, washed, and added at a concentration of 4 × 10⁴ cells/well in 50 μl DMEM and 10% FBS. Transfected cells were incubated overnight at 37°C in 5% CO₂, then 75 μl DMEM+10% FBS were added to each well, and the plates were incubated for an additional 48 h.

Dual luciferase assay of ribosomal readthrough efficiency

Luciferase activities were determined using the Dual Luciferase Reporter Assay System (Promega). Relative light units were measured on an MLX microplate luminometer (Dynex). Transfected cells were lysed in 12.5 μl lysis buffer and light emission was measured following injection of 25 μl of luminescence reagent. Percent readthrough was calculated by comparing firefly:Renilla luciferase ratios of experimental constructs with those of control constructs: (firefly experimental RLUs/Renilla experimental RLUs)/(firefly control RLUs/Renilla control RLUs) × 100. The total number of independent experiments for each construct varies between three and 11, and for each experiment, 3–12 independent data points were obtained for each construct. For each construct, all data points were then averaged and the standard deviation calculated. Data points that fell greater than 2 standard deviations from the mean were discarded as outliers.