Selective Restoration of the Selenoprotein Population in a Mouse Hepatocyte Selenoproteinless Background with Different Mutant Selenocysteine tRNAs Lacking Um34*

Novel mouse models were developed in which the hepatic selenoprotein population was targeted for removal by disrupting the selenocysteine (Sec) tRNA[Ser]Sec gene (trsp), and selenoprotein expression was then restored by introducing wild type or mutant trsp transgenes. The selenoprotein population was partially replaced in liver with mutant transgenes encoding mutations at either position 34 (34T→A) or 37 (37A→G) in tRNA[Ser]Sec. The A34 transgene product lacked the highly modified 5-methoxycarbonylmethyl-2′-O-methyluridine, and its mutant base A was converted to I34. The G37 transgene product lacked the highly modified N6-isopentenyladenosine. Both mutant tRNAs lacked the 2′-methylribose at position 34 (Um34), and both supported expression of housekeeping selenoproteins (e.g. thioredoxin reductase 1) in liver but not stress-related proteins (e.g. glutathione peroxidase 1). Thus, Um34 is responsible for synthesis of a select group of selenoproteins rather than the entire selenoprotein population. The ICA anticodon in the A34 mutant tRNA decoded Cys codons, UGU and UGC, as well as the Sec codon, UGA. However, metabolic labeling of A34 transgenic mice with 75Se revealed that selenoproteins incorporated the label from the A34 mutant tRNA, whereas other proteins did not. These results suggest that the A34 mutant tRNA did not randomly insert Sec in place of Cys, but specifically targeted selected selenoproteins. High copy numbers of A34 transgene, but not G37 transgene, were not tolerated in the absence of wild type trsp, further suggesting insertion of Sec in place of Cys in selenoproteins.

There are 24 known selenoproteins in rodents and 25 in humans (1). The targeted removal of specific selenoproteins has shown that some are essential in development, whereas others appear to be nonessential. For example, the loss of selenoproteins glutathione peroxidase 4 (GPx4) (2) or thioredoxin reductase 1 (TR1 or Txnrd1) (3) or 2 (TR3 or Txnrd2) (4) is embryonic lethal, whereas the loss of glutathione peroxidase 1 (GPx1) (5) or 2 (GPx2) (6) appears to be of little or no consequence. Other studies, however, suggest that those selenoproteins whose loss results in little or no phenotypic change may function in protective mechanisms against certain environmental stresses (see Ref. 6 and references therein). There are selenoproteins whose removal or mutation results in dramatic effects on health. For example, knock-out of selenoprotein P (SelP) 6 causes neurological problems (7,8), and knock-out of type 2 iodothyronine deiodinase results in a variety of defects, including an impaired adaptive thermogenesis and hypothermia in cold-exposed mice (see Ref. 9 and references therein), retarded cochlear development and hearing loss (10), and a pituitary resistance to thyroxine (11). Mutations affecting selenoprotein N (SelN) result in several muscle disorders (12,13).
LoxP-Cre technology, which allows the removal of embryonic lethal genes in specific tissues and organs (3,4,14), has been used to examine the roles of essential selenoprotein genes in development and health. Such studies have elucidated key roles of TR1 in embryogenesis of numerous tissues and organs, except heart (3), and of TR3 in hematopoiesis and in heart development and function (4). The targeted removal of the nuclear form of GPx4 (designated snGPx4) results in viable and completely fertile animals, although the overall structural stability of sperm chromatin is diminished (14). Loss of SelP in liver, achieved by targeted knock-out of the selenocysteine (Sec) tRNA [Ser]Sec gene (designated trsp), implicated SelP in transport functions in plasma and substantiated its essential role in brain (15).
Selenoprotein synthesis is dependent on the presence of Sec tRNA [Ser]Sec . Given this dependence, selenoprotein expression can be modulated by perturbing Sec tRNA [Ser]Sec expression, providing a means of elucidating the role of selenoproteins and selenium in development and health (16). The Sec tRNA [Ser]Sec population in higher vertebrates consists of two isoforms that differ by a single 2Ј-O-methyl group. One isoform contains 5-methoxycarbonylmethyluridine (mcm 5 U) at position 34, and the other is methylated on the ribosyl moiety at that position generating 5-methoxycarbonylmethyl-2Ј-O-methyluridine (mcm 5 Um; see Ref. 17). The presence of this 2Ј-methyl ribose modification (designated Um34) confers several unique properties on mcm 5 Um. For example, Um34 affects Sec tRNA [Ser]Sec mcmUm secondary and tertiary structure (18). Um34 addition is dependent on the prior synthesis of the four modified bases found in tRNA [Ser]Sec and on an intact tertiary structure (19). Synthesis of all other modified nucleosides of Sec tRNA [Ser]Sec , including mcm 5 Um, is less stringently associated with primary and tertiary structure. In addition, synthesis of Um34 is dependent on the selenium status of the organism, with increased dietary selenium increasing Um34 levels (17).
Removal of trsp is embryonic lethal (20,21). Therefore, to alter the Sec tRNA [Ser]Sec population, techniques of influencing Sec tRNA [Ser]Sec levels other than the sole removal of trsp must be employed. We previously generated transgenic mice with extra copies of wild type or mutant Sec tRNA [Ser]Sec transgenes (22) and mice with a conditional knock-out of trsp (21), and we then rescued selenoprotein expression in trsp null mice with wild type or mutant Sec tRNA [Ser]Sec transgenes (23,24). Consistent with reports that the Sec tRNA [Ser]Sec population is not limiting in selenoprotein biosynthesis (20,22,25), we found little or no effect of extra copies of wild type transgenes on selenoprotein expression in the tissues and cells examined (22). In contrast, multiple copies of a mutant trsp transgene can lead to specific altercations in the selenoprotein population (22). For example, transgenes with a mutation at position 37 (37A3G) produce a tRNA gene product that not only lacks isopentenyladenosine (i 6 A) at this site but also lacks Um34 (19). Selenoprotein synthesis was affected in mice carrying the G37 Sec tRNA [Ser]Sec transgene in a protein-and tissue-specific manner (22). Rescue of selenoprotein expression in trsp null mice with the G37 Sec tRNA [Ser]Sec transgene results in the recovery of housekeeping selenoproteins, whereas numerous stress-related selenoproteins that are nonessential to survival are either not rescued or are poorly rescued (23,24).
Although the wild type and mutant trsp transgenic models and transgenic trsp rescue models have provided considerable insight into selenoprotein expression and the hierarchy of selenoprotein expression (21)(22)(23)(24), they have limitations. For example, when expressing mutant trsp transgenes in mice carrying the endogenous allele of trsp, expression from the wild type trsp can confound the studies. Studies with rescue models like that described above, with a germ line conditional trsp allele, focus on the selenoprotein population in the whole animal. The targeted removal of floxed trsp in defined cell types using transgenic mice with tissue-specific expression of Cre recombinase permitted some study of the effects of selenoprotein loss in specific tissues and organs in the absence of endogenous trsp. However, the resulting animals have a variety of defects, including embryonic mortality or early adult death (26), thus restricting the use of these models for studying the role of selenium and selenoproteins in health.
In this study, we generated a mouse model that targets the removal of trsp in liver for use in elucidating the role of selenium and selenoproteins and the contributions of housekeeping and stress-related selenoproteins in health. trsp loxP-albumin Cre mice (27) were crossed with i 6 A-Um34-deficient Sec tRNA [Ser]Sec transgenic mice (designated here as G37 transgenic mice (22)) or with another mutant trsp (34T3A) transgenic mouse described here that lacks mcm 5 U and consequently the Um34 modification (designated here as A34 transgenic mice). The resulting mouse lines lack trsp in liver and are dependent on the A34 mutant transgene or the G37 mutant transgene for selenoprotein expression. These new mouse models provide us with novel experimental systems for investigating the role of numerous stress-related selenoproteins in health in a specifically targeted organ.

EXPERIMENTAL PROCEDURES
Materials-[ 75 Se]Selenium (specific activity 1000 Ci/mmol) was obtained from the Research Reactor Facility, University of Missouri, Columbia; [ 3 H]serine (specific activity 29 Ci/mmol) was from GE Healthcare, and [␣-32 P]dCTP (specific activity ϳ6000 Ci/mmol) was from PerkinElmer Life Sciences. Hybond Nylon Nϩ membranes were purchased from GE Healthcare; NuPAGE 10% polyacrylamide gels, polyvinylidene difluoride membranes, Superscript II reverse transcriptase, and SeeBlue Plus2 protein markers were from Invitrogen; SuperSignal West Dura extended duration substrate was from Pierce; Universal Reference RNA was from Stratagene; ADP-Sepharose 4B resin was from GE Healthcare; and anti-rabbit horseradish peroxidase-conjugated and anti-chicken horseradish peroxidase-conjugated secondary antibodies were from Sigma. Reagents for the TR1 assay (22) were purchased from Sigma. Antibodies against GPx1 were obtained from Abcam, and antibodies against GPx4, TR1, TR3, SelR, and SelT were from our laboratories (1,24,27). All other reagents were obtained commercially and were products of the highest grade available.
Animals and Genotyping of Mice-Homozygous floxed trsp C57BL/6 mice (trsp fl ) that were also homozygous for albumin Cre (AlbCre) were designated ⌬trsp after trsp fl was removed by the Cre recombinase (23,24,27). ⌬trsp C57BL/6-FVB/N transgenic mice were homozygous for one of three types of trsp transgene (trsp t ) alleles as follows: 1) wild type transgene encoding 10 copies of wild type trsp t /allele (22); 2) G37 low or high copy transgene encoding either 1 (low) or 8 (high) copies of the 37A3G mutant trsp t /allele; or 3) A34 transgene encoding one copy of the 34T3A mutant trsp t /allele. The product of the G37 transgene lacks the highly modified base, i 6 A, at position 37 and also Um34. The single copy G37 transgenic mouse was generated specifically for this study (22) to compare with the effects of the single copy A34 transgene. The product of the A34 transgene lacked the highly modified base, mcm 5 U, and also lacked Um34. 34T3A transgenic mice were generated exactly as described (22) except that the transgene construct contained an A at position 34 instead of a T, and the base at position 37 was the wild type A base; and three founders that were heterozygous for 1, 4, and 6 transgene copies were obtained. A34 transgenic mice were in strain FVB/N, and founders were bred to obtain the corresponding homozygous mice (designated A34-2, A34-8, and A34-12, respectively).
Genotype designations and definitions are given in the legend of Table 1. All mice used in this study were males. Matings to obtain mouse lines carrying wild type, A34, and G37 transgenes and ⌬trsp in their liver are summarized in Table 1.
Primers used for detecting trsp, trsp fl , ⌬trsp, trsp t , G37, or A34 by PCR are designated CKNO2 (forward primer) and RES1 and VP1 (reverse primers) (see Fig. 1) (27). Primers used for detecting AlbCre are designated LIV1 (forward primer) and LIV2 (reverse primer) (see Fig. 1 Tissues and organs were excised, immediately frozen in liquid nitrogen, and stored at Ϫ80°C. Tissues were homogenized, and 40 g of protein were electrophoresed on NuPAGE 10% polyacrylamide gels. Gels were stained with Coomassie Blue, dried, and exposed to a PhosphorImager as described (see Refs. 23 and 27 and references therein). To further assess 75 Se labeling of TR1, the labeled protein was purified from crude extracts of tissue using ADP-Sepharose 4B prior to gel electrophoresis as described (22). TR3 is also enriched by the ADP-Sepharose procedure, but the amounts of TR3 relative to TR1 are only about 10% and likely not to influence overall levels of the TR population (e.g. in Fig. 3B).
GPx and TR1 Activities and Selenium Assays-Total GPx activity was measured using a standard assay with hydrogen peroxide as substrate (see Refs. 23 and 27 and references therein). TR1 activity was determined in cytosol-enriched protein extracts using the insulin reduction method (8). The amount of selenium in extracts of liver, kidney, testes, brain and plasma was determined by Oscar E. Olsen Biochemistry Laboratories at South Dakota State University as described (23,27).
Isolation, Aminoacylation, Fractionation, and Sequencing of tRNA and Coding Studies-Total tRNA was isolated from liver of each mouse line. The tRNA was aminoacylated with [ 3 H]serine (29), and the resulting aminoacylated tRNA was fractionated on an RPC-5 column (30) as described (22,23,27). Sec tRNAs that were synthesized intracellularly from the A34 mutant transgene encoding a base change were sequenced using an RT-PCR technique (31) in which individual fractions from the RPC-5 columns were used as designated in Fig. 6. Codon recognition studies were carried out on [ 3 H]seryl-tRNA [Ser]Sec fractions from the RPC-5 column using the ribosomal binding technique of Nirenberg and Leder (32) as described (29). Trinucleoside diphosphates AGA, GGA, CGA, UGA, UGU, and UGC were the gift of Dr. M. Nirenberg or were prepared as described (29).

Characterization of Transgenic Mice Carrying the A34
Mutant trsp Transgene-Founder mice containing 1, 4, and 6 copies (heterozygous animals) of the A34 mutant transgene were generated as described under "Experimental Procedures." Each mouse line was bred to yield homozygous trsp transgenic animals, and the resulting lines were characterized in parallel with transgenic mice carrying low and high copy numbers of wild type or G37 trsp t transgenes (22). To determine the effects of A34 transgenes on selenoprotein biosynthesis, wild type and A34 transgenic mice carrying 4 or 12 copies of mutant transgenes were labeled with 75 Se, and the resulting labeled selenoprotein population was analyzed (Fig. 1). The highest A34 transgene copy number inhibited GPx1 expression, whereas TR1 expression showed little change in liver and kidney of both A34 transgenic mice. Thus, the low copy A34 transgene number appeared to have only a minor effect on the overall synthesis of selenoproteins. Similar observations were reported previously for the G37 transgene carrying similar trsp t copy numbers (22). These data suggested that the mutations at position 34 and 37 in Sec tRNA [Ser]Sec , which both result in loss of Um34, had similar effects on the expression of selected selenoproteins.
Replacement of Housekeeping Selenoproteins in Liver-Mouse lines carrying the targeted removal of trsp in liver (⌬trsp) and carrying either A34, low or high copy G37 transgenes, or the wild type (trsp t ) transgenes were generated (see Table 1). Importantly, trsp is not expressed in liver, whereas it, along with each transgene, is expressed in all other organs and tissues. Thus, the effects of the transgene products on selenoprotein expression and function occur independently of the wild type gene only in the trsp-targeted organ. Because the trsp and trsp t mouse lines express wild type Sec tRNA [Ser]Sec and overexpression of trsp has little or no effect on selenoprotein synthesis (17,20,21,22,37), they both were considered as controls. Genotypes were determined in liver and kidney, as the affected and control organs, respectively, and the expected genotypes were found as shown in Fig. 2.
Attempts to restore the housekeeping selenoprotein population in liver of ⌬trsp mice by mating with A34 transgenic mice carrying a higher copy number of A34 transgenes than the two copies used in replacing selenoprotein expression in these mice were unsuccessful. In addition, we were not able to rescue trsp null mice with any A34 transgenic mice regardless of the transgene copy number. Furthermore, the number of matings to Matings to obtain each mouse line used in this study are shown in the  obtain restored selenoprotein expression in ⌬trsp liver with the A34 transgene exceeded, by more than 3-fold, those required to obtain the A37 and trsp t selenoprotein liver replacement mice. Se Labeling-The expression of selenoproteins in liver, kidney, testis, brain, heart, and plasma of the six mouse lines was examined by labeling animals with 75 Se and analyzing labeled proteins in tissue extracts following gel electrophoresis. Coomassie Blue-stained gels of total proteins (Fig. 3A, lower panels) from the different tissues served as loading controls, and similar patterns and amounts of total proteins from the same tissues were observed with the possible exception of an enriched band indicated with an arrow in the liver ⌬trsp extract. Identification of this band as glutathione S-transferase and its   (27) and an unidentified protein (indicated with a ?) that varied inconsistently in amounts in liver of the mutant trsp t transgenic mice (see text). During preparation of plasma from trsp and ⌬trsp mice, greater hemolysis of the red blood cells occurred, which accounted for the globin observed in the Coomassie Blue-stained gels (see lanes 1 and 6 in the Plasma panel). B, 75 Se-labeled TR1 was purified from crude extracts of liver and kidney using ADP-Sepharose 4B prior to gel electrophoresis as described under "Experimental Procedures." Partially purified 75 Se-labeled TR1 from the two organs is shown. NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45 JOURNAL OF BIOLOGICAL CHEMISTRY 32595 possible significance were reported elsewhere (27). Another protein band that migrated near 20 kDa was observed to vary in amounts in liver of mutant Sec tRNA [Ser]Sec mouse lines (Fig. 3A, lower Liver panel, see arrow with a question mark). This band did not vary reproducibly in livers from mutant trsp mice, and therefore we have not identified it.

Partial Restoration of Selenoproteins in Liver
Variations were observed in the selenoprotein labeling patterns within organs and tissues of mice with the six different genotypes, particularly in liver between the two control mice containing wild type trsp and trsp t and the four mouse lines containing the defective tRNAs [Ser]Sec (Fig. 3). Some 75 Se-labeled bands have been identified previously, including TR1, GPx1, GPx4, and Sep15, which are indicated in Fig. 3, Liver panel (23,27,33). SelW, indicated with a question mark in Fig.  3A, has been tentatively identified (see Ref. 27 and references therein). The band that migrates just below TR1 is likely selenophosphate synthetase 2 (SPS2), although SelP also migrates at this position (1). GPx3 and SelP, which are indicated in the Plasma panel of Fig. 3A, have been characterized in plasma (see Ref. 27 and references therein).
The selenoprotein labeling patterns from trsp and trsp t control mice (1st and 2nd lanes, respectively, in each panel of Fig. 3) were similar with some minor differences. For example, GPx1 and GPx4 appeared to be more enriched in kidney (control tissue) and liver of trsp t mice than in the corresponding tissues in trsp mice. Comparison of mice encoding wild type trsp with those lacking trsp in liver (⌬trsp) showed that, as expected, most of the selenoprotein population is absent in ⌬trsp liver. The minor selenoprotein bands observed in ⌬trsp mice are likely proteins from liver cell types other than hepatocytes (27), which is the only cell type in which trsp deletion is targeted (28).
The presence of the A34 or G37 mutant transgenes in ⌬trsp mice resulted in TR1 and GPx4, and possibly SelP and/or selenophosphate synthetase 2 (SPS2), being restored in liver. The selenoprotein-labeled population in kidney appeared to be similar in the four tRNA [Ser]Sec -defective mouse lines with the possible exceptions of a reduced level of GPx1 in the high copy G37 transgene line (Fig. 6) and the elevated levels of GPx1 and -4 in ⌬trsp. The higher number of transgenes in high copy G37 than A34 would also seem to account for the reduced amounts of GPx1 observed in the other tissues. The G37 low copy number and A34 transgenes resulted in similar effects in labeling in the tissues examined except plasma. SelP, which is synthesized largely in the liver and transported to other tissues (see Ref. 15 and references therein), is reduced in plasma of A34 and G37 mice and possibly in testes of these mouse lines compared with control mice. SelP is also reduced in plasma in ⌬trsp mice compared with the two control mice, trsp and trsp t . These observations are further considered under "Consequences of A34 Sec tRNA [Ser]Sec in Selenoprotein Synthesis." GPx4 also appeared to be slightly enriched in testes of several mouse lines compared with trsp mice. GPx levels were further examined by measuring GPx activity in each tissue of the six mouse lines (see GPx assays below and in Table 1).
To examine the 75 Se labeling of TR1 in liver and kidney in more detail, TR1 was enriched from these tissues by passing tissue extracts over an ADP-Sepharose 4B affinity column (22). As shown in Fig. 4B, similar amounts of TR1 were present in both tissues of each mouse line with the exception of liver from ⌬trsp mice which expressed TR1 poorly. In addition, TR1 appeared to be slightly enriched in the kidney of ⌬trsp mice and possibly in both tissues of trsp t mice.
Northern Blot Analysis-Analysis of mRNA levels is an alternative means of examining the status of selenoprotein expression. The presence of UGA in the coding region targets some selenoprotein mRNAs for nonsense-mediated decay (NMD) (34,35). As shown previously (34,35), the NMD of some selenoprotein mRNAs is dependent on selenium status (reviewed in Refs. 17,36). GPx1 mRNA was present in low but detectable levels in liver of the four tRNA [Ser]Sec -defective mice compared with those of the two control mouse lines (Fig. 4). The levels of SelW mRNA were reduced substantially in liver of the four tRNA [Ser]Sec -defective mice and in reduced levels in kidney of A34 and high copy G37 transgenic mice. SelT mRNA levels were reduced in liver of the four tRNA [Ser]Sec -defective mice but present in similar levels as control mice in kidney. SelR mRNA was reduced in liver of high copy G37 transgenic mice but remained at similar levels in liver and kidney of the other . Northern blot analysis. RNA was extracted from liver and kidney of the six mouse strains and electrophoresed. RNA was then transblotted onto the appropriate membrane and hybridized with the indicated probes. Relative labeling was assessed using a PhosphorImager as described under "Experimental Procedures." Staining of developed gels with ethidium bromide showed that identical amounts of 18 S and 28 S rRNA were present in all tissue extracts (loading control, data not shown). Each of the Northern blots shown was carried out separately on two occasions with tissues from different mice with similar results. SelK mRNA was not examined in liver and kidney of low copy G37 mice as the levels of SelK mRNA from both A34 or high copy G37 mice were very similar. mice, with only slightly lower amounts in liver of the low copy G37 transgenic and ⌬trsp mice. The level of SelK mRNA was similar in the mouse lines examined with the exception of a slightly reduced level in liver of ⌬trsp mice. The mRNA levels of the other selenoproteins examined appeared to be present in similar amounts in both tissues of the six mouse lines or to vary only slightly in the four tRNA [Ser]Sec -defective mice as compared with those of the controls.
Selenium Status-The selenium levels were determined in liver, kidney, testes, brain, and plasma ( Table 2). The amounts of selenium were similar in each tissue of the trsp and trsp t control mice with the exception of liver and kidney that appeared to have a somewhat higher selenium level in mice carrying the trsp t transgene. The four defective tRNA [Ser]Sec mice had lower but similar levels in liver, kidney, and plasma compared with the two control mice, whereas A34 and the low copy G37 transgenic mice had similar levels as the two controls in testes and brain. The high copy G37 transgenic and ⌬trsp mice had similar but slightly lower selenium levels than the two other transgenic defective mice in testes.
Western Blot Analysis-The expression levels of several selenoproteins were further examined by Western blot analyses. Because GPx1, SelR, and SelT were not rescued in an earlier study involving the G37 transgene (23,24), as assessed by Western blot analysis, we focused on these selenoproteins to determine whether they might be restored in liver by the A34 transgene. GPx4 expression was also examined because its 75 Se-labeling patterns varied in liver and kidney within the different mouse lines (Fig. 3).
GPx1 was not detected in liver of ⌬trsp mice or in liver of mice carrying either the A34 or G37 transgenes (Fig. 5). GPx1 was present in lower levels in kidney of ⌬trsp, A34, and low copy G37 transgenic mice than in the other mice and was not detected in kidney from high copy G37 transgenic mice. The relative amounts of GPx1 were similar in kidney of ⌬trsp and A34 mice but less than observed in the low copy G37 transgenic mouse. Possible reasons for these differing levels of GPx1 expression are further considered below.
GPx4 was present in liver of low copy G37 transgenic mice and was virtually absent in liver from the other three tRNA [Ser]Secdefective mice. This selenoenzyme appeared to be reduced but present in kidney of ⌬trsp, A34, and low copy G37 transgenic mice and virtually absent in high copy G37 transgenic mice. SelR was poorly expressed in liver of both G37 transgenic mouse lines relative to control mice but slightly better expressed in A34 mice. SelR was also not expressed in kidney of the high copy G37 mice and was weakly expressed in the other tRNA [Ser]Sec -defective mice (Fig. 5), although SelR mRNA was expressed in this tissue (Fig. 4). Interestingly, SelT was not expressed in kidney of high copy G37 mice and partially or poorly expressed in kidney or liver of the tRNA [Ser]Sec -defective mice (Fig. 5) even though its mRNA appeared to be synthesized in sufficient levels within these tissues for adequate translation (Fig. 4).
The combination of reasonable mRNA levels with low protein levels strongly suggests that the defect in stress-related selenoprotein synthesis in the mutant trsp t mice is in translation rather than via effects on mRNA stability. More specifically, these findings raise the possibility that NMD of selenoprotein mRNAs (34,35) results from a block to translation because of the absence of the Um34 isoform. The corresponding mRNA with the encoded NMD criteria are then degraded. The site of regulation of mRNA decay may therefore be the Um34 methylation step that is known to be sensitive to selenium status (37). In fact, the Um34 isoform is dramatically reduced during selenium deficiency leading us to speculate that phenotypes displayed by the A34 and G37 transgene lines (Northern data shown in Fig. 4 and Western data shown in Fig.  5) likely mimic those provoked by selenium-deficient conditions, thus pinpointing this phenomenon as a failure of mRNA translation by the Um34 isoform.
Glutathione Peroxidase and TR1 Activities-Because the labeling of GPx1 and GPx4 with 75 Se appeared to be enhanced in kidney of trsp t , ⌬trsp, A34, and low copy G37 mice (Fig. 3), but their levels were diminished in this tissue as assessed by Western blotting (Fig. 5), we examined the cytosolic GPx activities in kidney and several other tissues from the six mouse lines ( Table 3, Experiment I). The assay did not distinguish between the different peroxidases but reveals whether total GPx activity was increased or decreased. In liver, where most of the GPx activity is because of GPx1, the activities were similar in the two control mice, trsp and trsp t , but extremely low in ⌬trsp mice and ⌬trsp mice carrying either the A34 or G37 transgene. In the other tissues examined, GPx levels were also similar in the two control mice. However, non-liver tissues of the four tRNA [Ser]Sec -   NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45 defective mouse lines had variable amounts of GPx activity. For example, in kidney, GPx activities were reduced in the ⌬trsp mice or mice carrying A34 and low copy G37 transgenes but were even lower in mice carrying the high copy G37 transgene. Testes and brain had normal GPx activities in the ⌬trsp mice, whereas mice encoding the A34 and low copy G37 transgenes had reduced activities, and those encoding the high copy G37 transgene had even lower levels. Plasma had low and similar activity levels in the four defective tRNA [Ser]Sec mouse lines, although the levels in low copy G37 mice appeared to be slightly higher. Thus, the mutations in trsp did not support full GPx1 activities. In particular, the high copy G37 transgene apparently exerted dominant negative effects in kidney and brain even in the presence of wild type trsp alleles, because GPx activities in high copy G37 mutants were below the levels in ⌬trsp mutants. These observations appear to exclude simple effects because of impaired selenium transport to these tissues. Interestingly, GPx activities in kidney of ⌬trsp mice were lower than in control mice even though kidney was not the targeted tissue (Table 3). We have proposed that this is a result, at least in part, of reduced SelP expression and thus impaired selenium transport of SelP from ⌬trsp liver to kidney (15). The selenium levels observed in kidney in this study were lower in ⌬trsp mice than control mice (Table 2). This result seems to conflict with a previous study with the same mouse lines (27), but gender differences likely explain this apparent conflict. In this study, male mice were analyzed, whereas in the earlier study, female mice were studied (27). We have recently described lower selenium levels and GPx activities in mutant kidneys from male ⌬trsp mice (15). Because gender differences were observed in these experiments and female ⌬trsp mice were less affected than male mutants, 7 the gender-specific differences in expression of prototype selenoenzymes in major organs of the mouse were systematically explored (38,39). GPx and deiodinase type I activities were higher in female kidney compared with male kidney, whereas hepatic deiodinase type I activity was higher in male compared with female mice. Because GPx activity is a good indicator of tissue selenium content, it would seem that the reduced selenium level detected in kidney of ⌬trsp mice compared with that of control mice (Table 2) may account in part for the lower GPx activities found in ⌬trsp mice, whereas an additional dominant negative effect of high copy G37 further reduced kidney GPx activity. It should also be noted that the observation of gender differences with respect to selenium metabolism in rodents is not a recent one. For example, Burk et al. (40) found that male rats were more affected by selenium deficiency than female rats, and the nutritional requirement for selenium is higher in males than females.

Partial Restoration of Selenoproteins in Liver
The Western blot analysis of GPx1 and GPx4 in liver and kidney (Fig. 5) was in closer agreement with the direct GPx enzyme assays (Table 3, Experiment I) than with the 75 Se-labeling data (Fig. 3). This discrepancy almost certainly reflects variations in both selenoprotein pool sizes and turnover rates that would influence patterns of 75 Se labeling. In addition, the A34 tRNA [Ser]Sec isoform may also insert Sec in place of Cys in response to certain Cys codons, UGU and UGC, that would also affect labeling patterns as well as the activity and Western blot analysis (see under "Consequences of A34 Sec tRNA [Ser]Sec on Selenoprotein Synthesis").
TR1 assays were also carried out in liver and kidney of the six mouse lines (Table 3, Experiment II). TR1 activities were similar in kidney of each mouse line except in high copy G37, which was about half that of controls. In liver, ⌬trsp had low detectable activity, which for the most part was likely due to liver cell types other than hepatocytes (27). TR1 activity was recovered with the three mutant transgenes, but low copy G37 was slightly less than controls; high copy G37 was about half that of controls, and A34 was slightly less than that of low copy G37.
Fractionation, Sequencing, and Codon Recognition of 34T3 A Sec tRNA [Ser]Sec -The Sec tRNA [Ser]Sec population was examined in liver from five of the six mouse lines. The low copy G37 transgenic mouse was excluded because it had been examined previously (27). The endogenous wild type Sec tRNA [Ser]Sec was absent in liver of ⌬trsp mice (data not shown), which allowed us to examine the A34 mutant tRNA population without any influence of host wild type tRNA. Total tRNA was isolated from liver of the five mouse lines, aminoacylated with [ 3 H]serine and the resulting [ 3 H]seryl-tRNA isoforms chromatographed over an RPC-5 column. The elution profile of the tRNA [Ser]Sec population from the A34 replacement mouse is shown in Fig. 6. The mutant tRNA eluted from the column as two major peaks. A small aliquot of two fractions from each peak was taken for sequencing, whereas the remainder of each peak was pooled for coding studies. The codon recognition 7 U. Schweizer, unpublished data. properties of peak I demonstrated that it decoded UGU, UGC, and UGA (Fig. 6) suggesting that the anticodon was ICA. Peak II decoded UGU suggesting that its anticodon was ACA. Sequences of two separate fractions of peak I showed that the base in the wobble position was G, which corresponds to I in the actual sequence (31). The anticodon was therefore ICA. Sequences of the two fractions of peak II demonstrated that the base in the wobble position was A and the anticodon was ACA.  Fig. 6. Um34 Is Important for Stress-related Selenoprotein Expression-The A34 and G37 mutations result in the loss of highly modified but very different bases, mcm 5 U in A34 and i 6 A in G37. The common feature of these mutant tRNAs is that they lack Um34. The two tRNAs are clearly capable of decoding Sec UGA codons as both support selenoprotein synthesis in mouse liver lacking wild type Sec tRNA [Ser]Sec (see Figs. 3 and 5), but neither is able to restore stress-related selenoprotein synthesis in liver following trsp knock-out (Fig. 5). Although minor differences were observed in mRNA stability of certain selenoprotein mRNAs and in levels of certain selenoproteins, the overall effects of both mutant tRNAs were similar providing strong evidence that the Um34 Sec tRNA [Ser]Sec isoform is responsible for stress-related selenoprotein synthesis. The minor differences in the effects of the two tRNAs on selenoprotein synthesis are likely because of the loss of the large, highly modified base in each mutant tRNA. Furthermore, the fact that selenium deficiency in rodents mimics the effects of the two Um34 lacking tRNAs in that the level of the Um34 Sec tRNA [Ser]Sec isoform and the expression of stress-related selenoproteins are reduced (reviewed in Ref. 17) also provides strong support that this isoform is indeed responsible for stress-related selenoprotein synthesis. from the A34 mouse line, and the arrows indicate the fractions from which small aliquots of the two peaks were taken for sequencing, and ICA and ACA show the anticodon sequences determined from sequencing these samples. The hatched areas show the fractions pooled for coding studies that were carried out using 10,000 total cpm of peak I/assay wherein the cpm bound to ribosomes in the absence of codon were 1,170, and using 10,000 total cpm of peak II/assay wherein the cpm bound to ribosomes in the absence of codon were 1,452. Counts/min bound to ribosomes in the absence of codon were subtracted from the counts/min bound in the presence of codon and given as ⌬CPM Bound. The inset shows the corresponding [ 3 H]seryl-tRNA [Ser]Sec isoforms from the other four mouse lines and the relative counts/min at the highest point of each peak. The fraction numbers of each highest point are as follows: G37, 14,010 cpm (19); trsp, 5,065 cpm (49); and trsp t , 37,805 cpm (44). The distributions of the two isoforms were determined as described previously (22): ICA and ACA were present at 66.6 and 33.4%, respectively.
How does the Um34 Sec tRNA [Ser]Sec isoform regulate stressrelated selenoprotein synthesis? We examined many of the more likely features that might be expected to play a role in the selective expression of selenoproteins by the Um34 isoform (37). These included nucleotide context of the UGA Sec codon, the total length of the cDNA coding region, the number of exons within the gene, the exon within the gene wherein the UGA resides, the number of nucleotides between the UGA Sec codon and the stop codon, the number of nucleotides between the stop codon and the highly conserved AUGA sequence within the SECIS element, and the number of nucleotides between the highly conserved AUGA sequence within the SECIS element and the downstream poly(A) signal. We concluded that none of these components are likely involved. Remaining candidates for mediating the effects of Um34 include the uncharacterized Um34 methylase and/or different SECIS-binding proteins. Interestingly, a new SECIS-binding protein that preferentially binds to different selenoprotein mRNAs has been detected. 8 We are currently working to identify and characterize the Um34 methylase.
Consequences of A34 Sec tRNA [ (41) and EFsec (42,43)), which are required for incorporation of Sec into protein. These mutant tRNAs probably insert Sec at some Cys codons in selenoprotein mRNAs with insertion governed by the same criteria that control Sec insertion at UGA (location of the Sec UGA codon relative to the SECIS element, for example; see Refs. 44 and 45). These tRNAs would likely compete with tRNA Cys for decoding specific Cys codons.
Replacement of Cys with Sec in selenoproteins would most likely result in lower enzymatic activity as is found in the GPxs in the mutant tRNA lines (Table 3). That the level of SelP was severely reduced in the presence of the A34 transgene (see Plasma panel in Fig. 3) supports the idea of competition, but it also may indicate that a protein with multiple amino acid replacements is more rapidly degraded and/or poorly transported. Clearly, SelP has multiple Sec and Cys residues, and the repeated use of a Sec tRNA [ [Ser]Sec may explain the inability to rescue trsp null mice with any A34 transgenic mouse regardless of the transgene copy number even though the same breeding scheme successfully rescued trsp null mice using wild type or G37 transgenic mice (23,24). We were able to restore selenoprotein expression in liver of mice targeted for removal of ⌬trsp with A34 transgenic mice carrying two copies of the mutant transgene, but not with higher copy numbers employing the same breeding scheme as that which replaced selenoprotein expression with trsp t and G37 transgenes. Sec tRNA [Ser]Sec isoforms with anticodons ACA and ICA may compete more effectively with Cys tRNA in decoding selenoprotein mRNAs and insert Sec in place of Cys disrupting function and resulting in lethality. Although high copy numbers were tolerated by A34 transgenic mice, these animals also had wild type trsp.
Significance of Developing Novel Mouse Models-Selenium is reported to have many health benefits, including roles in preventing cancer, heart disease, and other cardiovascular diseases, in delaying the aging process and the onset of AIDS in human immunodeficiency virus-positive patients, male fertility, immune function, mammalian development, and viral inhibition (46). Numerous large scale, human clinical trials have been undertaken to examine the effect of selenium in preventing the onset of disease with most focusing on the effects of selenium in cancer prevention. These trials are very costly and were designed with little understanding of how selenium acts at the molecular level. Development of animal models to elucidate the metabolic roles of selenium, selenoproteins, and low molecular weight selenocompounds is essential to understanding roles of selenium in health and development and in designing better human clinical trials. For example, recent evidence suggests that selenium may be ambivalent in its metabolic action in cancer in that it has cancer chemopreventive activity through some selenoproteins, but once a malignancy begins, selenium also promotes growth through selenoprotein TR1 (47,48). Furthermore, it is possible that the outcome of selenium supplementation at the doses used in human clinical trials may depend on individual genotypes, disease states, and other factors that can be elucidated through animal models.
We have therefore generated several mouse models to provide a better understanding of the role of selenium, selenoproteins, and low molecular weight selenocompounds in health and development. Our transgenic mouse model employing high copy G37 transgenic mice has been used to show that both selenoproteins and low molecular weight selenocompounds have a role in preventing colon cancer (49) and selenoproteins have a role in preventing prostate cancer (50). This model has also been used to examine other aspects of the role of selenoproteins in health (51,52). Our floxed trsp model using loxP/ Cre technology allowing targeted removal of trsp in specific tissues or organs has been used to show that selenoproteins play a role in endothelial development and heart disease prevention (26), proper liver function (27), neuronal function, 9 and in skin function and development. 10 The above useful models are surpassed by the one presented here, which allows alteration of the selenoprotein population in liver with wild type or mutant trsp transgenes. These mice are phenotypically normal, allowing study of the role of selenium, housekeeping selenoproteins, and stressrelated selenoproteins as well as the entire selenoprotein population in resistance to various factors, such as toxic metabolites, hepatocarcinogens, and liver cancer driver genes. The approach of targeting specific tissues for trsp removal can be used to generate other model systems for studying the role of selenium and selenoproteins in tissues of interest. The possibility that the expression of G37 and A34 mutant tRNAs [Ser]Sec in all tissues and organs may hinder experimentation seems not to be an issue. The high copy G37 transgenic mouse, which expresses G37 tRNA in all tissues and organs, has been used to show that selenoproteins and low molecular weight selenocompounds have a role in colon cancer prevention (49) and selenoproteins in prostate cancer prevention (50). The major advantage of the current model is that we can target the removal of trsp and then replace or partially replace the selenoprotein population. The targeted mouse models presented in this study are among the most sophisticated mouse models developed to date for studying the role of selenium, selenoproteins, and low molecular weight selenocompounds in health and development.