Inhibition of Selenoprotein Synthesis by Selenocysteine tRNA [Ser]Sec Lacking Isopentenyladenosine*

A common posttranscriptional modification of tRNA is the isopentenylation of adenosine at position 37, creating isopentenyladenosine (i 6 A). The role of this modified nucleoside in protein synthesis of higher eu-karyotes is not well understood. Selenocysteyl (Sec) tRNA (tRNA [Ser]Sec ) decodes specific UGA codons and contains i 6 A. To address the role of the modified nucleoside in this tRNA, we constructed a site-specific mutation, which eliminates the site of isopentenylation, in the Xenopus tRNA [Ser]Sec gene. Transfection of the mutant tRNA [Ser]Sec gene resulted in 80% and 95% reduction in the expression of co-transfected selenoprotein genes encoding type I and II iodothyronine deiodinases, re-spectively. A similar decrease in type I deiodinase synthesis was observed when transfected cells were treated with lovastatin, an inhibitor of the biosynthesis of the isopentenyl moiety. Neither co-transfection with the mutant tRNA gene nor lovastatin

Nucleoside modifications involving tRNAs at positions in or near the anticodon have been shown to affect any of a number of steps in the translation process (1). A tRNA position commonly modified is position 37, the site immediately 3Ј of the anticodon. Eukaryotic and bacterial tRNAs reading codons starting with uracil often have a large hydrophobic moiety attached to position 37 (2). One common modification is isopentenylation of the adenosine at this position, creating isopente-nyladenosine (i 6 A). 1 Isopentenyl pyrophosphate:tRNA isopentenyltransferase is responsible for isopentenylation of this adenosine (3). The genes for isopentenyl pyrophosphate:tRNA isopentenyltransferase have been isolated in yeast and bacteria. The Escherichia coli gene is miaA (4), the Saccharomyces cerevisiae gene is MOD5 (5), and the Schizosaccharomyces pombe gene most likely is sin1 (6). Roles for the i 6 A modification in cellular metabolism have mainly been deduced from mutant strains of yeast and bacteria lacking isopentenyl pyrophosphate:tRNA isopentenyltransferase activity. E. coli miaA Ϫ mutants show decreased translation of the trp operon (7) and manifest a significant decrease in their ability to read through the stop codons, UAG and UGA, that is reduced by 75% and 60-fold, respectively, compared with wild-type E. coli (8). In MOD5 mutants of S. cerevisiae, the efficiency of a tyrosine inserting nonsense suppressor tRNA is reduced relative to that of the wild-type organism (9). In addition, lack of i 6 A leads to inactivation of a suppressor tRNA in the antisuppressor strain sin1 of S. pombe (6).
In E. coli, the i 6 A modification is not required for attachment of an amino acid to tRNA by aminoacyl tRNA synthetases (10). However, this modification does appear to be important for efficient binding of aminoacyl-tRNA to ribosomes (11). Isopentenylation of position 37 in tRNA appears to stabilize the first position of the codon-anticodon interaction and thereby assists in preventing first position misreading (10). However, both in vitro (12) and in vivo (13) evidence indicate that the i 6 A modification increases third position misreading due to decreased proofreading. Together these results show that, in yeast and bacteria, isopentenylation at position 37 in tRNAs reading codons beginning with uridine has important physiologic effects on translation efficiency, especially with regard to tRNA suppression of stop codons.
Mammalian selenocysteine (Sec) tRNA [Ser]Sec also contains i 6 A at position 37 (see Ref. 14 for review and Ref. 15 for subsequent work). Compared with other mammalian tRNAs, this tRNA is unique in that Sec biosynthesis occurs on the tRNA after it is aminoacylated with serine, it is 90 nucleotides long, which is the longest eukaryotic tRNA sequenced to date, and it contains only four other modified nucleosides in addition to i 6 A whereas other tRNAs contain many more modified bases (14). Sec tRNA [Ser]Sec decodes UGA and inserts Sec into nascent selenopolypeptides. Recognition of UGA as a Sec codon and not as a stop codon requires a stem-loop structure or structures in the 3Ј-untranslated region of selenoprotein mRNAs as well as other translational factors specific for Sec incorporation (16).
In this paper, we examined the role of i 6 A at position 37 of tRNA [Ser]Sec on selenoprotein biosynthesis in vivo by generating Sec tRNAs lacking this modified base. Our studies indicate that structural changes at position 37 affect the decoding properties of the altered tRNA [Ser]Sec and its ability to decode UGA and incorporate Sec into proteins. The effect of altered Sec tRNA [Ser]Sec on selenoprotein expression is not because of differences in aminoacylation of the tRNA, biosynthesis of Sec on the tRNA, transcriptional events, protein degradation, or selenoprotein translation initiation. Most likely, the presence of i 6 A is important for efficient translation of Sec codons by tRNA [Ser]Sec .
Site-directed Mutagenesis-The construct (A37C-XTSS) was created by site-directed mutagenesis using a 1.04-kilobase EcoRI/BamHI fragment of the Xenopus tRNA [Ser]Sec gene (18) as a template. Site-directed mutagenesis was carried out by polymerase chain reaction amplification of the 1.04-kilobase EcoRI/BamHI fragment in 2 reaction steps. Initially, a fragment of A37C-XTSS was generated using the KpnItailed 29-mer (5Ј-CGGGGTACCCCACATCCACTAACAAACAG-3Ј) and the 19-mer (5Ј-GCTACAGGTGTGAAGCCTG-3Ј) to yield a 348-bp product. Simultaneously, a partially overlapping fragment was generated using the 20-mer (5Ј-GGTATGTAAGCGGCGATACG-3Ј) and a PstItailed 29-mer (5Ј-AAACTGCAGCCTGGTTGTTATGGATACTG-3Ј) to yield a 365-bp product. A portion of each reaction was combined along with the KpnI-tailed 29-mer and PstI-tailed 29-mer and amplified by polymerase chain reaction using temperatures of 94°C for denaturation, 50°C for annealing, and 72°C for extension. The polymerase chain reaction product was purified on a 1% agarose gel, cut with KpnI and PstI, and cloned into pGEM-3zf(ϩ). The mutation was verified by DNA sequencing (19) performed by the Tufts University Department of Physiology Core Facility. A clone (A37A-XTSS) lacking the A37C mutation in the tRNA [Ser]Sec gene sequence was isolated, sequenced, and used as a control.
Transfecting CHO Cells-For isolation of tRNA [Ser]Sec fractions used in the codon binding studies, CHO cells were transfected by electroporation using 50 g of plasmid DNA from either A37C-XTSS or the polymerase chain reaction-derived nonmutant A37A-XTSS. Cells were harvested 4 days post-transfection and stored at Ϫ80°C until tRNA isolation.
For [ 75 Se]selenoprotein synthesis and experiments involving Western and Northern blotting, CHO cells were transfected, unless indicated otherwise, using LipofectAMINE Plus (Life Technologies, Inc.) reagent according to the manufacturer. Briefly, CHO cells were seeded on day 0 in 6-well plates at a density of 8 ϫ 10 5 cells/well in 1.5 ml of H-NCS. On day 1, the cells were transfected using 6 l each of LipofectAMINE and Plus reagent with the indicated plasmids in serum-and antibiotic-free Ham's F12. After a 3-h incubation period, an equivalent volume of Ham's F12 containing 10% (v/v) NCS, 200 units/ml penicillin, 200 mg/ml streptomycin, 4 mM glutamine was added. On day 2, the cells were refed with H-NCS. Further processing of cells is described in the respective figure legend.
Extraction and Isolation of tRNA [Ser]Sec -Transfer RNA was isolated by phenol extraction of CHO cells transiently transfected with A37A-or A37C-XTSS. The extract was loaded onto a DE-52 column and the tRNA eluted in 1 M NaCl, 0.1 M Tris-HCl, pH 7.4. Total tRNA was deacylated by incubation in 1.8 M Tris-HCl, pH 8.0, for 1 h at 37°C. Total tRNA was aminoacylated with [ 3 H]serine under conditions of limiting tRNA and chromatographed in the absence of Mg 2ϩ (20) on a RPC-5 column (21). Aminoacylated tRNA was eluted and chromatographed on a RPC-5 column in a linear gradient at pH 4.5 containing 10 mM sodium acetate, 1 mM EDTA, and 0.575-0.8 M NaCl for A37C-XTSS or 0.6 -0.8 M NaCl for A37A-XTSS. CHO cells transfected with A37A-XTSS and treated with 20 M lovastatin were fractionated in a gradient of 0.575-0.825 M NaCl. Fractionated [ 3 H]seryl-tRNA [Ser]Sec isoacceptors were pooled and used in ribosome binding studies according to the procedure of Nirenberg and Leder (22) as described (23).
To examine the intracellular [ 75 Se]selenocystenyl-tRNA [Ser]Sec population, CHO cells were transfected with either A37A-XTSS or A37C-XTSS DNA by electroporation and allowed to grow for 4 days in H-NCS supplemented with 50 ng/ml Na 2 SeO 3 . The cells were harvested, radiolabeled for 4 h with [ 75 Se]selenious acid as described previously (24) at a concentration of 44 Ci/ml of (A37C-XTSS) or 56 Ci/ml (A37A-XTSS), washed, and then frozen until used for RNA extraction and RPC-5 chromatography as described above.
[ 75 Se]Selenoprotein Synthesis-CHO cells were seeded and transfected as described above. On day 3, the cells were radiolabeled for 3 h in H-NCS containing 2 Ci/ml [ 75 Se]selenious acid unless otherwise indicated. Cells were washed and harvested for protein determination (25). Laemmli (26) or Schagger (27) SDS-PAGE was performed as indicated. Polyacrylamide gels were processed for fluorography and quantification of radiolabel incorporation was accomplished using a PhosphorImager (Molecular Dynamics) and ImageQuant software (Molecular Dynamics).
Northern Hybridization-Total cellular RNA was isolated (28), electrophoresed on 1% agarose/formaldehyde gels (29), transferred to Ge-neScreen Plus, and probed for type I deiodinase using a 770-bp SmaI/ NcoI fragment of the G21-D10 plasmid (30) in a standard hybridization buffer (1 M NaCl, 1% (w/v) SDS, 10% (w/v) dextran sulfate, and 25 mg/ml sheared salmon sperm DNA). The blot was stripped and reprobed for GAPDH using the full-length rat cDNA (31). Alternatively, for Northern blot analyses of Sec tRNA, total RNA, or RNA subjected to immunoprecipitation were separated by electrophoresis on 15% acrylamide, 4 M urea gels, transferred to GeneScreen Plus and probed for Sec tRNA (15) using a 193-bp AvaI-HindIII fragment of the human tRNA [Ser]Sec gene (17) as probe.
Probes were radiolabeled for 2 h at 37°C with [␣-32 P]dCTP using DECAprime II (Ambion Inc.), added to a final concentration of 0.1 nM, and incubated at 62°C for a minimum of 16 h. Following hybridization, the membranes were washed for 2 h in 1% (w/v) SDS, 0.1ϫ SSC at 62°C with multiple changes of the wash buffer.
Dot blot analysis of tRNA [Ser]Sec was accomplished as described previously (15). Immunoprecipitation of total RNA for Northern blot analysis by anti-i 6 A serum (32) was performed essentially as described (15).
Hybridizations were monitored by autoradiography at Ϫ80°C in the presence of intensifying screens, and quantification was accomplished using a PhosphorImager (Molecular Dynamics) and ImageQuant software (Molecular Dynamics).
Western Blotting-CHO cells were seeded and transfected as described above. Cells were harvested for protein determination (25) and SDS-PAGE (26) as described. Protein gels were transferred to Protran (Schleicher and Schuell) as described previously (30). Membranes were blocked with 5% (w/v) nonfat milk in TBS-Tween (20 mM Tris, pH 7.4, 150 mM NaCl, and 0.1% Tween 20). Rabbit antisera directed against a peptide in the N-terminal region of type I deiodinase (30) was used at a 1:1000 dilution in TBS-Tween containing 5% (w/v) nonfat milk. Alternatively, rabbit antisera directed against ␤-galactosidase was used at a 1:1000 dilution. Goat anti-rabbit horseradish peroxidase was used as the secondary antibody at a 1:4000 dilution in TBS-Tween containing 5% (w/v) nonfat milk. Enhanced chemiluminescence was performed according to the manufacturer (NEN Life Science Products).

RESULTS
One means of defining the physiological role of i 6 A on tRNA [Ser]Sec is to eliminate the adenosine at position 37, the site of isopentenylation. As described under "Experimental Procedures," we created an A37C mutation in the Xenopus tRNA [Ser]Sec gene (A37C-XTSS) by site-directed mutagenesis. Next, we transfected A37C-XTSS into CHO cells, isolated total RNA, and analyzed for Sec tRNA expression by northern hybridization with tRNA [Ser]Sec probe (Fig. 1A, lanes 1 and 2). This experiment shows that A37C-XTSS transfected cells express a new Sec tRNA with faster mobility than the endogenous CHO Sec tRNA. Upon quantification, we observed that the expression levels of A37C-Sec tRNA exceeded the level of the endogenous CHO tRNA by almost 4-fold.
To determine if A37C-Sec tRNA was isopentenylated, we immunoprecipitated total RNA from cells transfected with either empty vector or A37C-XTSS with antiserum prepared in rabbits immunized with isopentenyladenosine (32) and performed Northern blots on the immunoprectipates (lanes 3 and 4). Previous work showed that these antibodies do not recognize unmodified adenosine (32). Whereas the endogenous, CHO Sec tRNA was precipitated by our anti-i 6 A antibodies in both samples, the A37C-Sec tRNA (lane 4) previously observed (lane 2) was not. These data indicated that transfection of A37C-XTTS results in overexpression of a Sec tRNA lacking isopentenyladenosine.
We also determined whether lovastatin caused reduced levels of isopentenyladenosine in Sec tRNAs (Fig. 1B). Lovastatin is a potent, specific inhibitor of mevalonic acid biosynthesis (35), the precursor for the isopentenyl moiety at position 37 of Sec tRNAs. Lovastatin treatment of A37A-XTSS transfected cells did not seem to affect either endogenous CHO or exogenous A37A-Sec tRNA expression (lanes 1 and 2). However, the amount of i 6 A-containing Sec tRNA was reduced ϳ50% in total RNA from lovastatin-treated cells (compare lanes 3 and 4). The decreased anti-i 6 A immunoprecipitation in lovastatin-treated cells is consistent with reduced isopentenyl modification of the tRNAs.
To address the effect that A37C-Sec tRNA has on cells, we examined the capability of A37C-XTSS transfected cells to decode UGA Sec codons. Fig. 2 depicts the synthesis of selenoproteins in CHO cells as measured by 75 Se incorporation. As seen in untransfected cells (lane 1), there are several endogenously expressed proteins that contain 75 Se. This pattern is consistent with what others have found in similar labeling experiments (33,34). Furthermore, we analyzed selenoprotein synthesis in CHO cells transfected with plasmids containing cDNAs for type I deiodinase (G21-D10) and type II deiodinase (hD2/SelP). Transfection with G21-D10 alone resulted in the expression of deiodinase (compare lanes 1 and 2), and co-transfection with the A37A-XTSS construct appeared to have little or no effect on the level of type I deiodinase expression (compare lanes 2 and 3). However, co-transfection with the mutant A37C-XTSS construct resulted in a dramatic reduction in type I deiodinase expression (lane 4). Although type II deiodinase expression was not as strongly expressed (lanes 5 and 6) as the type I form (lanes 2 and 3), its expression was affected similarly by co-transfection with the wild-type (lane 6) and the mutant  2). Alternatively, 20 (lane 3) or 30 g (lane 4) of total cellular RNA from electroporated cells were immunoprecipitated with anti-i 6 A serum, and the immunoprecipitates analyzed by Northern blotting for isopententylated Sec tRNA expression as described previously (15). In B, all cells were Lipo-fectAMINE-transfected with A37A-XTSS and incubated in the absence (lanes 1 and 3) or presence (lanes 2 and 4) of 10 M lovastatin for 18 h. Total RNA was prepared, and 20 g were blotted for Northern blot analysis directly (lanes 1 and 2). Also, 40 g of total RNA was subjected to anti-i 6 A immunoprecipitation (15) prior to Northern blot analysis (lanes 3 and 4). Northern blot analysis and quantification of Sec tRNA expression were performed as described under "Experimental Procedures."

tRNA [Ser]Sec gene constructs (lane 7).
Quantification of exogenously derived deiodinases and endogenous selenoprotein synthesis is shown in Table I. Expression of wild type A37A-XTSS gene had variable, but marginal affects on the expression of the deiodinases when compared with cells transfected with either G21-D10 or hD2/SelP alone. However, A37C-XTSS co-transfection with either deiodinase gene resulted in 78 and 94% decreases in the syntheses of type I and type II deiodinases, respectively. Whereas, expression of the wild type, Xenopus A37A-XTSS construct produced a slight increase in endogenous selenoprotein biosynthesis, the mutant A37C-XTSS gene led to a slight decrease. Synthesis of thioredoxin reductase (ϳ54 kDa) and the 24-and 18-kDa selenoproteins were reduced between 9 and 26% by A37C-Sec tRNA expression. The difference in the effect of A37C-Sec tRNA expression seen between the deiodinase proteins and other endogenous selenoproteins may be related to transfection or other translation phenomena and is addressed under "Discussion." Neither result could be explained by decreased cellular uptake of 75 Se. Uptake levels of 75 Se, when normalized for cell protein, were not affected by experimental treatment (data not shown).
Although the effect was not as dramatic as in CHO cells, co-transfection of A37C-XTSS along with the deiodinase constructs into HEK-293 cells led to decreased syntheses of both type I and type II deiodinases (data not shown). It should be noted that [ 75 Se]selenious acid incorporation into either type I or II deiodinase was not detected in either CHO (Fig. 2) or HEK-293 (data not shown) cells transfected with pGEM-T vector alone.
To determine if the effect of A37C-XTSS was because of a general inhibition of exogenous gene expression, we transfected CHO cells with either G21-D10 or the ␤-galactosidase encoding plasmid DNA, pNASS, in the presence or absence of either A37A-XTSS or A37C-XTSS and performed Western blot analyses. As seen in Fig. 3, transfection of A37C-XTSS reduced type I deiodinase levels (compare lanes 2 and 4) to a similar extent as that seen in Fig. 2. However, ␤-galactosidase expression was unaffected by co-expression of A37C-Sec tRNA (compare lanes   5 and 7). These data indicate that A37C-Sec tRNA does not act as a general inhibitor of protein synthesis, but inhibits selenoprotein expression (note that the immunoreactive protein seen at ϳ21 kDa represents nonspecific binding of our antiserum).
In the above experiments, the function of i 6 A was addressed by creating a mutation in the tRNA gene sequence, transfecting this mutant construct into CHO cells, and examining the ability of cells transfected with the mutant Sec tRNA gene to support selenoprotein synthesis. This method addressed the function of the intact, modified nucleoside at position 37 of Sec tRNA. As a second approach to discern the function of the isopentenyl moiety, we treated cells with 10 M lovastatin. Lovastatin is a potent, specific inhibitor of mevalonic acid biosynthesis (35), and mevalonic acid is a precursor for the iso-  CHO cells were transfected with type I (G21-D10) or type II (hD2/ SelP) deiodinase plasmid DNA alone or in combination with A37A-XTSS or the mutant A37C-XTSS gene as described under "Experimental Procedures." On day 3, CHO cells were pulsed for 3 h with 2 Ci/ml 75 Se in H-NCS. The cells were harvested, an extract was prepared, and 100 g of cell protein were electrophoresed on 13% SDS-PAGE gels. Gels were processed for fluorography, and quantification of [ 75 Se]selenoprotein synthesis was performed.
a Quantitative data of transfected proteins from Fig. 2 are represented and expressed as the percentage of control in comparison to the respective deiodinase construct alone.
b Quantitative data of endogenous selenoproteins is an average of expression levels from transfection with either G21-D10 or hD2/SelP and expressed as percentage of control relative to the average of lanes 2 and 5.
pentenyl moiety at position 37 of tRNA. As observed above (Fig.  1C), lovastatin treatment of cells expressing A37A-Sec tRNA resulted in an ϳ50% decrease in the amount of i 6 A-containing Sec tRNA without affecting total Sec tRNA expression. This experimental protocol allowed us to further address the function of the isopentenyl modification on selenoprotein biosynthesis.
CHO cells were transfected with G21-D10 plasmid DNA and subsequently cultured for the final 21 h under a variety of conditions in the presence or absence of lovastatin. [ 75 Se]Selenious acid was added during the final 3 h to monitor selenoprotein synthesis. Experiments were carried out in duplicate (Fig. 4), and the resulting data were quantified (Table II). The pattern of selenoprotein labeling in the absence of lovastatin is shown in lanes 1 and 2. In the presence of 10 M lovastatin, selenoprotein labeling was dramatically reduced (lanes 3 and  4), whereas the addition of mevalonic acid to the culture medium reversed the effect of lovastatin (lanes 5 and 6). Quantifying this data showed that all cellular selenoproteins were reduced from 55 to 95% with the exception of an 8.5-kDa protein. The presence of the 8.5-kDa protein is enhanced, and its identity is unknown.
It has previously been shown that 10 M lovastatin results in cell-cycle arrest at G 0 /G 1 phase (32). To ascertain how much of the lovastatin effect was because of cell cycle arrest, we cultured cells under serum-free conditions. CHO cells cultured for 21 h under serum-free conditions (lanes 7 and 8) showed reduced 75 Se incorporation into several selenoproteins compared with CHO cells cultured in H-NCS (lanes 1 and 2). The latter growth conditions prevent cell cycle arrest. The magnitude of inhibition, however, was much less than that observed with lovastatin treatment. Note that 75 Se incorporation into type I deiodinase (ϳ27 kDa) was unaffected by serum removal. CHO cells cultured under serum-free conditions in the presence of 10 M lovastatin showed a further reduction in 75 Se incorporation into all proteins, except the 8.5-kDa protein (lanes 9 and 10). Most notably affected was type I deiodinase, whose synthesis was unaffected by serum-free conditions.
Together these data point to a crucial role of the isopentenyl group at position 37 of the Sec tRNA in selenoprotein expression. However, the decreased synthesis of type I deiodinase may be explained by lower levels of deiodinase mRNA, a decrease in protein stability, or a reduction in mRNA translation.
First, we used Northern blot analysis to determine whether the type I deiodinase mRNA levels were reduced in A37C-XTSS-transfected or lovastatin-treated CHO cells. As shown in  (lanes 2-5).
These results indicate that the manipulations that cause a decreased synthesis of type I deiodinase do not alter rates of  9 and 10). CHO cells were pulsed for 3 h with 2 Ci/ml of 75 Se in serum-free Ham's F12 containing 5 ng/ml Na 2 SeO 3 in the presence or absence of the respective drug treatments. The cells were harvested for PAGE, and 100 g of cell protein were electrophoresed on 10% acrylamide spacer, 12% acrylamide resolving Schagger gels. Gels were processed for fluorography, and [ 75 Se]selenoprotein synthesis direct quantification was accomplished by Phosphorimage analysis using ImageQuant software. gene transcription or mRNA stability. This implicates a posttranscriptional event as the cause for the observed reduction in selenoprotein synthesis by the mutant A37C-Sec tRNA and the lovastatin-induced Sec tRNA lacking i 6 A.
Next, we performed a pulse-chase experiment to determine if co-transfection with the mutant A37C-XTSS construct affected the degradation rate of type I deiodinase protein (Fig. 6). CHO cells were co-transfected with G21-D10 and either A37A-XTSS (lanes 1-4) or A37C-XTSS (lanes 5-8). After one day, the cells were pulse-labeled with 75 Se for 1.5 h and chased for various times, and the amount of radiolabeled type I deiodinase was determined after SDS-PAGE and fluorography. Densitometric scans showed that A37C-XTSS-transfected cells synthesized ϳ35% less type I deiodinase during the pulse interval compared with A37A-XTSS-transfected cells. However, the rate of disappearance of type I deiodinase was similar in both sets of cells. These data indicate the expression of the mutant tRNA does not affect the stability of type I deiodinase.
Lastly, we sought to determine if A37C-XTSS inhibited type I deiodinase expression by decreasing translation. Inhibition of translation could occur at two possible sites: initiation of synthesis or, given the unique nature of selenoprotein biosynthe-sis, decoding of the internal Sec UGA. Impaired read through of the internal UGA would cause termination of translation and release of a truncated protein. Fig. 7 is a Western blot analysis showing the level of both the full-length and truncated type I deiodinase in the absence (lanes 1 and 3) or presence (lanes 2 and 4) of co-transfected A37C-XTSS. We also analyzed the content in cells treated with ALLN, an inhibitor of cellular proteolysis (lanes 3 and 4). The approximate relative expression of the full-length protein in lanes 1-4 is 1.0, 0.35, 0.15, and 0.06, respectively.
If A37C-Sec tRNA caused decreased initiation of translation, we would expect to observe similar reductions in both the full-length and truncated type I deiodinase proteins. On the other hand, if the mutant Sec tRNA inhibited expression of the full-length type I deiodinase by blocking decoding of the internal UGA, we would observe a reciprocal increase in the amount of truncated protein. In comparing lanes 1 and 2, we see an ϳ3-fold reduction in the amount of full-length protein, yet no change in the expression of the truncated translate. This observation does not support either mechanism for A37C-Sec tRNA inhibition of translation.
We then looked at the effect of ALLN on the expression of type I deiodinase proteins (Fig. 7, lanes 3 and 4). ALLN caused a dramatic increase in the amount of the truncated protein. This observation indicates the truncated protein is subject to extremely rapid degradation, probably as a result of the proteosome-dependent protease system (36). Given the abundance of truncated versus full-length protein in the presence of ALLN, we conclude that the overwhelming majority (ϳ95%) of type I deiodinase mRNA-directed translation is terminated at the internal UGA. Similarly, synthesis of the full-length protein is extremely inefficient (ϳ5% of total mRNA-directed translation).
Unfortunately, this grossly disproportional pattern of expression for the two forms of type I deiodinase prevents us from determining if A37C-Sec tRNA directly inhibits decoding of the internal UGA. Co-transfection of the mutant tRNA gene reduces full-length deiodinase expression by ϳ60% in ALLNtreated cells. However, (as noted above) only ϳ5% of mRNA-directed translation produces the full-length protein. Therefore, if the mutant tRNA does inhibit decoding, we would expect to observe only an ϳ3% (60% of 5%) increase in the amount of truncated protein. Such a small difference would not be detected in the immunoblots from ALLN-treated cells.
If the initiation of translation is inhibited, we would predict a 60% decline in truncated deiodinase expression in the ALLN-  Fig. 4 are represented and expressed as percentage of control relative to incubation of CHO cells in H-NCS. CHO cells were transfected with type I deiodinase plasmid DNA (0.4 g of G21-D10) as described under "Experimental Procedures," except that on day 2, cells received 10 M lovastatin or 10 M lovastatin ϩ 2 mM mevalonic acid. In addition, CHO cells were incubated in serum-free conditions (2% w/v) bovine serum albumin in Ham's F12 containing 100 units/ml penicillin, 100 mg/ml streptomycin, 2 mM glutamine, and 25 mM Hepes, pH 7.4) for 18 h in the absence or presence of 10 M lovastatin. CHO cells were pulsed for 3 h with 2 Ci/ml of 75 Se in serum-free Ham's F12 containing 5 ng/ml Na 2 SeO 3 in the presence or absence of the respective drug treatments. The cells were harvested for PAGE, and 100 g of cell protein were electrophoresed on 10% acrylamide spacer, 12% acrylamide resolving Schagger gels. Gels were processed for fluorography, and  Total RNA (5 g) was resolved on a 1% agarose/formaldehyde gel and transferred to GeneScreen Plus. The membrane was first probed for type I deiodinase using a 770-bp SmaI/NcoI fragment of the G21-D10 plasmid (A), stripped, and subsequently probed for GAPDH (B) as described under "Experimental Procedures." treated cells. However, densitometric scans of films with reduced exposures show no diminution in the amount of the truncated protein in ALLN-treated cells (data not shown). These data solidify our conclusion (from lanes 1 and 2) that the mutant tRNA does not inhibit initiation of translation.
Another possible explanation for the reduction in 75 Se incorporation into protein seen with lovastatin treatment and A37C-XTSS expression could be because of the lack of or reduced selenocystenyl-tRNA [Ser]Sec biosynthesis because of changes at position 37. Fig. 8 shows the RPC-5 column elution profiles of [ 75 Se]selenocystenyl-tRNA [Ser]Sec from CHO cells transfected with either A37A-XTSS (Fig. 8A) or A37C-XTSS (Fig. 8B). The elution profiles of both in vivo 75 Se-labeled tRNA (Fig. 8A) and in vitro labeled [ 3 H]seryl-tRNA (Fig. 8C) from CHO cells overexpressing A37A-XTSS show two peaks; the position 34, 5-methylcarboxymethyluridine (mcm 5 U, peak II) and 5-meth-ylcarboxymethyluridine-2Ј-O-methylribose tRNAs [Ser]Sec (mcm 5 U m , peak III) as previously reported (37). Peak I c observed in Fig. 8B is most likely the A37C-Sec-tRNA given its relatively early elution from the RPC-5 column as was seen for tRNA [Ser]Sec lacking i 6 A (38). Similarly peak I a in Fig. 8D isolated from cells overexpressing A37A-XTSS and subsequently treated with 20 M lovastatin most certainly lacks i 6 A. Therefore it can be concluded that i 6 A is not necessary for aminoacylation with serine or in the biosynthesis of Sec.
The efficiency of aminoacylation of tRNA [Ser]Sec and biosynthesis of Sec were still unknown. Measurement of these parameters was accomplished by quantifying the relative amount of tRNA [Ser]Sec in each of these peaks by dot blotting and the amount of 75 Se associated with the purified Sec tRNA by liquid scintillation counting. The calculated specific radioactivity (Table III) is a measure of the ability of the tRNA to become aminoacylated with Sec. Experiment 1 shows the specific activities of the two wild type forms of Sec-tRNA (derived from peaks II and III, Fig. 8A) from cells transfected with A37A-XTSS. Experiment 2 compares the purified mutant, A37C-Sec-tRNA (peak I c , Fig. 8B), with the two forms of A37A-Sec-tRNA (peaks II and III, Fig. 8B). Note, the specific radioactivity is approximately the same for all three Sec-tRNA species. These data indicate that the mutant Sec tRNA is aminoacylated with [ 75 Se]Sec to a similar extent as A37A-Sec tRNA. Thus, a lack of or a reduction in aminoacylation of A37C-Sec tRNA or in the biosynthesis of Sec cannot explain the observed decreases in selenoprotein synthesis seen in Figs. 2 and 3.
To determine if tRNA [Ser]Sec minus i 6 A may have an effect at the level of translation, we examined the codon recognition properties of the mutant isoacceptor and the i 6 A lacking isoacceptor induced by lovastatin inhibition. As shown in experiments 1 and 2 in Table IV, changes at position 37 from i 6 A37 to either A37C or unmodified A37A did not alter the specificity of the seryl-tRNA [Ser]Sec for recognizing UGA in a ribosome binding assay. The cysteine codons, UGU and UGC, as well as the tryptophan codon, UGG, were poorly recognized by all seryl-tRNA [Ser]Sec isoacceptors tested in this assay. However, when Ci/ml of [ 75 Se]selenious acid for 1.5 h in selenium-free growth medium. Following the pulse, the monolayers were washed and chased for the indicated times in growth medium containing 50 ng/ml sodium selenite. Cells were processed for PAGE, and 100 g of cell protein were electrophoresed on 10% acrylamide spacer, 12% acrylamide resolving Schagger gels. Gels were processed for fluorography and 75 Se type I deiodinase quantification was accomplished by Phosphorimage analysis using ImageQuant software. The position of full-length type I deiodinase protein is indicated by the arrow.  ALLN (lanes 3 and 4) or Me 2 SO vehicle (lanes 1 and 2) for 24 h. On day 3, cells were harvested, extracts were prepared, and 60 g of cell protein were electrophoresed on Schagger-PAGE gels. Western blotting of type I deiodinase was performed as described under "Experimental Procedures." Arrows indicate the full-length (28 kDa) and truncated (13 kDa) in vivo translates.
we examined specific binding as a percentage of added seryl-tRNA [Ser]Sec for A37C-and i 6 -lacking seryl-tRNA [Ser]Sec to UGA in this assay, both tRNAs exhibited reduced binding to UGA in comparison to the wild-type isoacceptor. Specific binding of the A37C-and i 6 -lacking Sec tRNA to UGA was decreased by 62 and 69%, respectively.
Although E. coli ribosomes were used in the ribosomal binding studies, it should be noted that the genetic language of mammals and bacteria were shown to be the same in this assay using E. coli ribosomes (39) and that strong responses relative to their assigned codons of mammalian Cys-, Arg-, and Trp-tRNAs to UGA were observed in this same assay using E. coli ribosomes (40). The later observation suggests that mammalian Cys-, Arg-, and Trp-tRNAs are capable of misreading UGA and, indeed, these tRNAs serve as authentic suppressors of UGA stop codons in mammals (41,42). We conclude, therefore, that the data shown in Table IV demonstrate a defect in the decoding properties of tRNA [Ser]Sec lacking the i 6 A modification in mammalian protein synthesis.

DISCUSSION
In this study we have provided evidence that changes at position 37 of Sec tRNA result in a tRNA with reduced abilities to bind to UGA in an in vitro ribosomal binding assay and to support selenoprotein synthesis in vivo. These results elucidate the role of i 6 A in mammalian cells that seems to be similar to that observed in bacteria and yeast. In E. coli (8) and S. cerevisiae (9), the inability to synthesize i 6 A results in a decreased efficiency to suppress stop codons. Similarly, the lack of i 6 A on a nonsense serine suppressor, sup3-i, in an antisuppressor mutant of S. pombe, sin1, leads to its inactivation (6). In addition to the decreased stop codon suppression seen in bacteria and yeast, we have shown that removal of the i 6 A from Sec tRNA results in reduced selenoprotein synthesis in mammalian cells.
Our observations of reduced selenoprotein synthesis are quite intriguing. First, the effect of A37C-XTSS transfection on type I deiodinase synthesis (Table I) does not mimic the effect on endogenous selenoprotein synthesis. Whereas a definitive effect of A37C-Sec tRNA expression on endogenous selenoproteins was observed, the reduction in synthetic rates did not approach the 80 -90% decreases observed with type I and II deiodinases. Attempts to differentiate between a unique effect on deiodinase and a general effect on transfected selenoprotein cDNAs was attempted using HepG2 cells. This cell line reportedly expresses functional type I deiodinase activity (43). However, we were unable to observe detectable 75 Se-radiolabeling of type I deiodinase under our culture conditions indicating that the synthetic rate is too low for determining the effect of A37C-Sec tRNA on endogenous type I deiodinase synthesis.
Why does expression of the mutant, A37C-tRNA [Ser]Sec appear to preferentially inhibit deiodinase synthesis? A plausible explanation is that replacing the highly modified i 6 A nucleoside with an unmodified base results in an isoacceptor that, by analogy to the corresponding tRNAs in bacteria and yeast,

Inhibition of Selenoprotein Synthesis
decodes UGA less efficiently than the fully modified form. Our co-transfections were performed with the aid of liposomes, and the mutant tRNA and cDNA plasmids were mixed prior to the addition of the liposomes. Under these conditions, the mixture of exogenous DNA forms complexes with the liposomes. It is highly likely both types of DNA are present in every complex. Therefore, those cells that take up a liposome-DNA complex are receiving both types of plasmid, and it is unlikely a significant amount of cells take up one type of DNA without the other. Also, knowing that transfection efficiency is not 100%, some percentage of the population is not receiving any mutant tRNA. However, the entire population is exposed to a radiolabeled 75 Se precursor. Consequently, if the mutant tRNA were inhibiting endogenous selenoprotein synthesis, we would only observe a decline in radiolabeled incorporation equivalent to the percentage of cells expressing the mutant tRNA. As seen in Table I, we observed a small, but significant decline (9 -26%) in the incorporation of 75 Se into several endogenous selenoproteins when transfected with the mutant tRNA. On the other hand, all cells on the monolayer receive lovastatin and are inhibited by the drug. This may explain why endogenous selenoproteins and transfected type I deiodinase syntheses were inhibited by the drug.
Of course other possibilities exist to explain the lack of effect of A37C-Sec tRNA on endogenous selenoprotein synthesis. For example, several laboratories have provided evidence for the existence of supramolecular translation complexes containing mRNAs, ribosomes, EF-1␣, tRNAs, and tRNA synthetases (44 -46). According to the channeling model of Stapolionis and Deutscher (46), tRNAs are directly transferred from aminoacyl-tRNA synthetases to elongation factor to ribosomes without dissociation into the cellular fluid. Upon leaving the ribosome, uncharged tRNAs are directly transferred back to their cognate synthetases, without being released into the surrounding cytosol. There, they are recharged with amino acid and then bound by EF-1␣ for another elongation cycle in the same supramolecular complex. This model may explain the effect of transfected Sec-specific tRNA on transfected but not endogenous selenoprotein synthesis. Selenoprotein mRNAs expressed from endogenous genes would be expected to be continuously transcribed and exported to the cytoplasm at a low, constitutive level. Newly synthesized selenoprotein mRNA and Sec tRNA from the transfected genes would likely undergo a sharp, steep burst of transcription and export. Thus, most of the endogenous selenoprotein mRNAs would already be in the cytoplasm assembled into translation complexes at the time of transfection. The newly transcribed mRNAs, following export, would predominantly be free to assemble together into new supramolecular complexes, producing an apparent transfection-specific effect of the introduced wildtype or mutant tRNA [Ser]Sec gene.
As was shown in Fig. 5, no effect of A37C-XTSS co-expression or lovastatin treatment on type I deiodinase mRNA levels was observed. Whereas increased selenoprotein degradation could account for some of the decreased 75 Se-labeled proteins in Fig.  2 (lanes 4 and 7) and Fig. 4 (lanes 3 and 4), we hypothesized that A37C-Sec tRNA and i 6 A-lacking Sec tRNA affect translation. To address this, we examined the ability of these tRNAs to recognize UGA in a well established ribosomal binding assay (22). The results in Table IV provide evidence that changes at position 37 of the Sec tRNA affect translation, specifically the step involved in codon binding. Specific binding to UGA was reduced ϳ70% for both the A37C-and i 6 A-lacking tRNA [Ser]Sec populations compared with the corresponding wild-type tRNA. These results are in close agreement to those of Ohama et al. (47). They observed that an early eluting 75 Se-labeled tRNA showed a 50% decrease in specific binding to UGA compared with the full modified Sec tRNA. They presumed this to be a Sec-tRNA lacking i 6 A based on previous observations in Xenopus oocytes (38).
Our hope was to test the ability of these individual Sec tRNA populations to support UGA translation in vitro. Rabbit reticulocyte translation systems programmed with specific mRNAs are widely used for in vitro synthesis of proteins. However, to our knowledge an in vitro translation system to faithfully measure the efficiency of selenoprotein synthesis does not exist. Previously, Berry et al. (30,48) demonstrated in vitro synthesis of type I deiodinase using a rabbit reticulocyte system. However, these investigators determined that the commercially available in vitro translation systems are, in addition to being quite inefficient, a measure of UGA suppression and not Sec incorporation into protein. Given that the A37C-and i 6 A-lacking tRNA [Ser]Sec both exhibit binding to UGA, we would expect them to show some level of UGA translation. However, without a reliable system to accurately measure translation efficiency, we are unable to definitively demonstrate a reduced translational ability of either the A37C-or i 6 A-lacking tRNA [Ser]Sec .
We are not aware of prior studies addressing the specific role of adenosine adjacent to the anticodon in any tRNA. In addition to preventing isopentenylation, inserting a cytosine for adenosine at position 37 in the Sec tRNA may have other effects on its function, e.g. inhibiting other modifications, affecting tRNA stability, altering tertiary structure, and inhibiting general protein synthesis. However, our data show the lack of adenosine at this position does not affect the ability of A37C-Sec tRNA to bind Sec, as judged by normal amounts of 75 Se incorporation into the mutant tRNA ( Fig. 8B and Table III). This observation indicates adenosine 37 is not required for serinespecific aminoacylation and tRNA-dependent conversion to Sec, mediated by aminoacyl-tRNA synthetase and selenocysteine synthase, respectively. Furthermore exchanging guanosine for adenosine at position 37 of Sec tRNA does not affect the tertiary interaction of D-T-arms in the tRNA. 2 Lastly, the expression of transfected ␤-galactosidase is not reduced by co-transfection with the A37C-Sec tRNA gene (Fig. 3), showing lack of adenosine at position 37 does not grossly affect general protein synthesis. These data provide evidence of a crucial role for the i 6 A modification on tRNA [Ser]Sec for decoding UGA Sec codons. Future directions are aimed at determining the precise mechanism by which changes at position 37 of Sec tRNA result in decreased selenoprotein synthesis.