Altered RNA Binding Activity Underlies Abnormal Thyroid Hormone Metabolism Linked to a Mutation in Selenocysteine Insertion Sequence-binding Protein 2*

The expression of selenoproteins requires the translational recoding of the UGA stop codon to selenocysteine. In eukaryotes, this requires an RNA stem loop structure in the 3′-untranslated region, termed a selenocysteine insertion sequence (SECIS), and SECIS-binding protein 2 (SBP2). This study implicates SBP2 in dictating the hierarchy of selenoprotein expression, because it is the first to show that SBP2 distinguishes between SECIS elements in vitro. Using RNA electrophoretic mobility shift assays, we demonstrate that a naturally occurring mutation in SBP2, which correlates with abnormal thyroid hormone function in humans, lies within a novel, bipartite RNA-binding domain. This mutation alters the RNA binding affinity of SBP2 such that it no longer stably interacts with a subset of SECIS elements. Assays performed under competitive conditions to mimic intracellular conditions suggest that the differential affinity of SBP2 for various SECIS elements will determine the expression pattern of the selenoproteome. We hypothesize that the selective loss of a subset of selenoproteins, including some involved in thyroid hormone homeostasis, is responsible for the abnormal thyroid hormone metabolism previously observed in the affected individuals.

The expression of selenoproteins requires the translational recoding of the UGA stop codon to selenocysteine. In eukaryotes, this requires an RNA stem loop structure in the 3-untranslated region, termed a selenocysteine insertion sequence (SECIS), and SECIS-binding protein 2 (SBP2). This study implicates SBP2 in dictating the hierarchy of selenoprotein expression, because it is the first to show that SBP2 distinguishes between SECIS elements in vitro. Using RNA electrophoretic mobility shift assays, we demonstrate that a naturally occurring mutation in SBP2, which correlates with abnormal thyroid hormone function in humans, lies within a novel, bipartite RNA-binding domain. This mutation alters the RNA binding affinity of SBP2 such that it no longer stably interacts with a subset of SECIS elements. Assays performed under competitive conditions to mimic intracellular conditions suggest that the differential affinity of SBP2 for various SECIS elements will determine the expression pattern of the selenoproteome. We hypothesize that the selective loss of a subset of selenoproteins, including some involved in thyroid hormone homeostasis, is responsible for the abnormal thyroid hormone metabolism previously observed in the affected individuals.
Selenium is an essential micronutrient for human health. Most of the selenium in the body is found as selenocysteine, which is incorporated into proteins termed selenoproteins. Selenoproteins are believed to be involved in such diverse aspects of human health as immunity, fertility, cancer prevention, and atherosclerosis (1). The human selenoproteome is composed of ϳ25 members (2). Those of known function include the glutathione peroxidases, the thioredoxin reductases, and the deiodinases, all of which are involved in oxidation-reduction reactions within cells (3,4).
Selenoprotein synthesis is unusual, because it relies on the translational recoding of a UGA stop codon to selenocysteine within the coding region of selenoprotein mRNAs. The dual nature of this codon implies that chain termination and selenocysteine insertion are competing processes; however, this area remains poorly understood at this time. In recent years, much progress has been made in delineating the necessary components for selenocysteine incorporation. This process requires an RNA stem-loop structure termed the selenocysteine insertion sequence (SECIS), 2 which is located in the 3Ј-untranslated region of eukaryotic selenoprotein mRNAs (5). In addition, a selenocysteine-charged tRNA (Sec-tRNA Sec ) (6) and an elongation factor dedicated to selenocysteine insertion (EFSec) (7,8) are necessary. Several other protein factors have roles in Sec-tRNA Sec synthesis and delivery, and interactions between the selenoprotein mRNA, Sec-tRNA Sec , and the ribosome have also been identified. Currently, the exact order of events and series of interactions remains unclear. For a comprehensive review of selenocysteine insertion see reviews by Hatfield and Gladyshev (9) and Hoffman and Berry (10).
Our lab has a long standing interest in identifying proteins that interact with the SECIS. We previously demonstrated that SECIS-binding protein 2 (SBP2) (11,12) and ribosomal protein L30 (13) bind to the SECIS and promote selenocysteine insertion. The interactions of these proteins with the SECIS are mutually exclusive because they have overlapping, possibly identical binding sites and compete for binding to the conserved core element of the SECIS (13). Both of these proteins are members of the L7Ae family of RNA-binding proteins, although it should be noted that SECIS binding activity is not a general property of this family. Two models are currently proposed for the interaction of SBP2 and the SECIS. In one model, free SBP2 binds to the SECIS before encountering the translation apparatus (13), whereas in the other model ribosomebound SBP2 interacts with the SECIS element (14,15). Based on previous studies in our lab, we proposed that the SECIS acts as a molecular switch (13). Short term interaction of SBP2 with the SECIS would serve to prime the system for selenocysteine insertion, and then the complex would be transferred to L30 on the ribosome for multiple rounds of translation. Because SBP2 is a limiting factor in selenocysteine insertion, this model would allow for greater selenoprotein synthesis than one that requires constant association between SBP2 and the SECIS. All known functions of SBP2 are performed by the C-terminal half of the protein, SBP2 CT (residues 399 -846 in rat) (12). Previous dele-* This work was supported by Public Health Services Grant HL29582 from the tion analysis in our lab identified two domains in rat SBP2 CT (16). Residues 399 -516 of rat SBP2 constitute the functional domain, which is required for selenocysteine insertion but not SECIS binding. The region bounded by residues 517 and 777 contains the RNA binding activity of the protein (see Fig. 1A) and includes the L7Ae motif, spanning amino acids 650 -752 using the Pfam alignment (accession number PF01248) (17,18).
Although the exact details of the mechanism of selenocysteine insertion remain elusive, the requirement for the SBP2-SECIS interaction is well established, because mutations that abolish the RNA binding activity of SBP2 also abolish selenocysteine insertion in vitro (16). Several recent studies also suggest the importance of this interaction in vivo. The SePN-1 related myopathies are a group of early onset congenital muscular dystrophies in humans that result from different mutations in the selenoprotein N (SelN) gene (19 -23). Rigid spine muscular dystrophy, one member of this group, can be caused by a point mutation in the SECIS of SelN mRNA that abolishes SBP2 binding, resulting in the loss of SelN expression (24). The deiodinases are required for human thyroid hormone homeostasis, and mutations in SBP2 that are linked to thyroid dysfunction in humans were recently reported in two different families (25). The activities of two selenoproteins, GPx1 and Dio2, were impaired despite the maintenance of mRNA levels in affected individuals. In one case, two different mutations on each SBP2 allele led to nonfunctional protein, either by truncation or frameshifting. This combination of mutations is easy to reconcile with a loss of SBP2 function, because the C-terminal half of the protein, which is largely missing in these affected individuals, performs all known functions. More intriguingly, the second family carried a simple homozygous point mutation, causing a single amino acid change from arginine to glutamine at position 540 of human SBP2. The location of this mutation led us to hypothesize that it could disrupt the RNA binding activity of SBP2. Although the activities of GPx1 and Dio2 were decreased, these patients had only mild symptoms. Because some selenoproteins are essential in mouse knock-out models, more severe symptoms might be expected if this mutation had a global impact on selenoprotein synthesis.
Given the importance of the interaction between the SBP2 and the SECIS for successful selenoprotein synthesis, this study identifies the region(s) of SBP2 required for RNA binding and determines the effect of the medically relevant R 3 Q mutation on SBP2-SECIS interactions.

EXPERIMENTAL PROCEDURES
DNA Constructs-The constructs used to examine the RNAbinding domain boundaries were created using primers that introduced a start codon into the 5Ј primer and a stop codon into the 3Ј primer. The resulting PCR products were cloned into the KpnI/XbaI sites of pcDNA 3.1 (Invitrogen). SBP2 CT and SBP2[517-777] were previously described (16). All point mutations (Arg 3 Glu and Met 3 Ala) were introduced using DpnI-mediated mutagenesis. Briefly, complementary primers with the mutation of interest in the center were used to amplify the plasmid. The resulting product was incubated with DpnI, which digests the template DNA. The remaining mutagenized DNA was transformed into Top10 cells and plated onto selec-tive medium, and colonies were isolated and analyzed. All of the mutations were confirmed by DNA sequencing. For the internal deletions in SBP2[517-777], DpnI mutagenesis was used to introduce AfeI sites at various positions (545/546, 566/567, 593/594, and 611/612). Then AfeI digests were performed to excise the intervening sequence. The human deiodinase 2 (Dio2) SECIS probe was made by PCR from whole brain cDNA (Clontech) followed by cloning into pCRIIblunt TOPO (Invitrogen). The thioredoxin reductase 1 (TR1) SECIS element was cloned using the mouse TR1 3Ј-untranslated region as a template for PCR. The selenoprotein 15 SECIS element was obtained by PCR from a rat whole brain library and cloning into the EcoRI/HindIII sites of pGEM-3Zf(ϩ). The phospholipid hydroperoxide glutathione peroxidase (PhGPx), glutathione peroxidase 1 (GPx1) (26), and deiodinase 1 (Dio1) SECIS probe sequences were previously described (11).
In Vitro Transcription and Translation-Plasmid DNAs were linearized and used as templates for in vitro transcription using T7 RNA polymerase (Ribomax T7; Promega). An aliquot of RNA was then used in an in vitro translation reaction in the presence of [ 35 S]Met, using rabbit reticulocyte lysate (RRL), according to the manufacturer's suggestion (Promega). A 2-l aliquot of each translation reaction was resolved on an SDS-PAGE gel and quantitated by PhosphorImager analysis. The concentrations were determined by comparison with known standards. Calculations were performed using a concentration of 5 M cold methionine in the RRL lysate as indicated by the manufacturer (Promega).
RNA Probes, Electrophoretic Mobility Shift Assays, and Competition-SECIS probes were synthesized from linearized templates with T7 RNA polymerase using 10 mM GTP, 10 mM ATP, 10 mM CTP, 0.05 mM UTP, and 25 Ci of 32 P-labeled UTP for 3 h at 37°C. The transcription reactions were treated with DNase I for 20 min and then phenol:chloroform-extracted. The aqueous phase was passed through a Micro Bio-Spin P30 col-umn according to the manufacturer's instructions (Bio-Rad). The probes were heated to 95°C for 2 min and slow cooled to room temperature. The final RNA electrophoretic mobility shift assay (REMSA) binding buffer concentrations were 140 mM KCl, 10 mM HEPES, pH 7.9, 5% glycerol, 1 mM dithiothreitol, and 0.33 mg/ml tRNA. The reaction was further supplemented with 15 g of salmon sperm DNA to reduce nonspecific interactions from the RRL lysate. For the competition reac-tions, increasing concentrations of unlabeled SECIS RNA was added in excess, as indicated. The reactions were incubated at 37°C for 30 min, and complexes were resolved on 6% nondenaturing polyacrylamide gels. The gels were dried, and the appearance of complexes was analyzed on a PhosphorImager.
In Vitro Selenocysteine Insertion Assay-The luciferase reporter construct luc/UGA 258 /PhGPx was previously described. To generate luc/ UGA 258 /GPx1 and luc/UGA 258 / Dio1, the PhGPx SECIS in the 3Ј-untranslated region was replaced with the GPx1or Dio1 SECIS elements. In vitro translation reactions were assembled for a total volume of 12.5 l, including 100 ng of luc/ UGA 258 mRNA in the presence or absence of SBP2 CT or mutants thereof. The reactions were incubated at 37°C for 30 min. Each reaction was tested in triplicate by adding 2.5 of the translation mixture to 50 l of luciferase substrate (Promega) and reading the sample for 10 s in a PerkinElmer Victor 3 . To examine the effect of multiple SECIS elements, increasing amounts of SECIS RNA was added to the translation reaction as indicated.

RESULTS
Identification of the SBP2 RNAbinding Domain-To investigate our hypothesis that the R 3 Q mutation interferes with SECIS binding, we first needed to define the RNA-binding domain of SBP2. Using rat SBP2 as our model, previous work in our lab determined that the RNA binding activity of SBP2 was located within the area bounded by amino acids 517 and 777 (16). This region also contains the L7Ae motif, which is the only region of protein homology within SBP2 (Fig. 1A). Another L7Ae family member, ribosomal protein L30, was also shown by our lab to bind to the SECIS element and promote selenocysteine insertion (13). Given this common binding activity, we first wanted to determine whether the L7Ae motif in SBP2 was sufficient for SECIS binding. UV cross-linking assays were not used because specific binding of the in vitro translated proteins could not be detected. The experimental method selected was REMSAs using the SECIS element from PhGPx in combination with various in vitro translated truncation mutants of SBP2.
As shown in Fig. 1B, in vitro translation of a synthetic RNA encoding SBP2[517-777] results in two protein bands. We previously postulated that the lower band is caused by internal initiation at Met 535 (16). Initiation at this site was confirmed by mutation of this methionine to alanine, SBP2[517-777,M535A], which resulted in only one protein band when in vitro translated (Fig. 1B). When SBP2[517-777] was tested in a REMSA using PhGPx SECIS as probe, a doublet was observed (Fig. 1C). In contrast, SBP2[517-777,M535A] only results in a single shift, with the same electrophoretic mobility shift as the lower band of the wild type doublet. This indicates that the upper band of the doublet is due to interactions of the internally initiated protein SBP2[535-777] with the PhGPx SECIS. However, previous work in this lab demonstrated that SBP2[535-777] does not support selenocysteine insertion in vitro (16). Thus, the lower band of the doublet represents the functional interaction between SBP2 and the PhGPx SECIS RNA. All of the subsequent experiments using the PhGPx SECIS will focus on the presence or absence of the lower band. Additionally, when using the PhGPx SECIS as a probe, a background band is consistently observed (Fig. 1, B-D) in the RRL lanes. This is due to a low affinity interaction between an unidentified protein in the RRL and the PhGPx SECIS, which is displaced in the presence of a specific SBP2-SECIS interaction. The SBP2[517-777] translation product was tested across a range of protein concentrations in a REMSA assay (Fig. 1C), with volume matched RRL as a control. With increasing amounts of protein added, the nonspecific RRL band decreases, the lower band of the doublet intensifies in signal, and the upper band is displaced. Given that the L7Ae motif within SBP2 is not adequate for SECIS binding activity, we wanted to define the boundaries of the RNA-binding domain. Previous work with deletion mutants in our lab had localized the N-terminal boundary of the domain from residues 517 to 535 and the C-terminal boundary to between residues 728 and 777 (Fig. 1E) (16). The gene structure of SBP2 was used as a guide to choose a starting point from which to investigate the location of these boundaries. Sites of exon-exon junctions were used as initial boundaries; amino acid 526 for the N-terminal end and amino acid 749 for the C-terminal end. Subsequently, a series of mutants were constructed across these boundary regions by serially bisecting the regions between residues 517 and 526 or between residues 749 and 777. When we examined the C-terminal end truncation mutants, we found that three of the mutants retained their binding activity (Fig. 1E, lanes 1-3). However, when a further truncation was made from residues 756 to 749, the binding activity was lost (Fig. 1E, compare lanes 3 and 4). This indicates that the C-terminal boundary of the RNA-binding domain lies between residues 749 and 756. When the N-terminal end truncation mutants were tested, SBP2[520 -777] was able to bind, whereas SBP2[526 -777] was not (Fig. 1E, compare lanes 5  and 6), indicating that the N-terminal boundary of the domain is between residues 520 and 526. Interestingly, the C-terminal boundary corresponds to the end of the L7Ae domain. In contrast, the N-terminal boundary is more than 120 amino acids upstream from the L7Ae domain. This analysis clearly demonstrates that sequences outside of the L7Ae domain are required for SBP2 RNA binding.
The RNA-binding Domain of SBP2 Is Bipartite-The boundaries of the RNA-binding domain as defined above (residues 520 -756) would give rise to a domain of ϳ235 amino acids, which is extraordinarily large for an RNA-binding domain. In comparison, the RNA recognition motif, one of the most common eukaryotic motifs, averages 85-90 amino acids (27), whereas the double stranded RNA-binding domain averages ϳ70 -75 amino acids (28). When the RNA-binding domain of rat SBP2 is aligned to the comparable region of other vertebrate SBP2 sequences, two separate blocks of conservation are immediately apparent (amino acids 520 -541 and 615-756 within rat SBP2; Fig. 2). Notably, the Arg 3 Glu mutation, R531Q in rat SBP2, falls within the first conserved stretch of amino acids. The divergence of the intervening sequence (amino acids 542-614) suggested that it was not required for RNA binding per se, although the region could potentially be required as a hinge or as a spacer between the conserved sequences. To investigate this possibility, four internal deletions were made across this region in rat SBP2: ⌬567-593, ⌬567-611, ⌬546 -593, and ⌬546 -611. Each deletion was made in the context of SBP2[517-777] and constructed so that no extraneous amino acids were added. The mutants were in vitro translated and analyzed on an SDS-PAGE gel (Fig. 3A). When tested for their ability to bind the PhGPx SECIS in gel shift assays, all four internal deletion mutants retained the ability to bind the RNA and appeared to bind slightly better than wild type, based on the amount of free probe remaining (Fig. 3B). This binding was specific, because the internal deletion constructs did not bind to a mutant PhGPx SECIS in which the core motif was deleted (data not shown).
Although this nonconserved region was not required for RNA binding, it may be necessary for selenocysteine insertion. To examine whether these internal deletions would impair SBP2 function, the deletions were also made in the context of SBP2 CT, and the proteins were expressed by in vitro transcription and translation. Equimolar amounts of wild type and mutant proteins were tested using an in vitro translation assay employing a luciferase reporter whose expression is dependent on selenocysteine insertion. In the luc/UGA 258 /PhGPx reporter, a cysteine codon at position 258 has been mutated to UGA and the PhGPx SECIS placed in the 3Ј-untranslated region. This reporter has been previously validated to be spe-cific for assaying selenocysteine insertion because luciferase expression is dependent on both the presence of a functional SECIS and an in-frame UGA codon (13,29). Although a small amount of endogenous SBP2 is found in RRL, it is the limiting factor, and the addition of exogenous SBP2 CT results in strong stimulation of in vitro selenocysteine insertion. As shown in Fig. 3C, none of the internal deletion mutants were impaired in their ability to promote selenocysteine insertion. In fact, all four mutants were more active than the wild type protein (p Ͻ 0.001) in this assay, which correlates with the enhanced SECIS binding activity observed in the REMSA assays.
The R531Q Mutation Impairs the SECIS Binding Activity of SBP2-Based on the boundaries that we have identified, the R531Q mutation does fall within the RNA-binding domain of SBP2 and suggests that this protein could be defective in SECIS binding. We made the equivalent mutation in rat SBP2 (R531Q), and examined its ability to bind to various SECIS elements. The original study reported defects in GPx1 and Dio2 activity (25). When the SECIS elements of these selenoproteins were tested in a REMSA assay, we found that the wild type protein bound to both the GPx1 and Dio2 SECIS RNAs, whereas the R531Q mutant did not (Fig. 4A). Nonspecific interactions with proteins in the RRL were also observed with the Dio2 SECIS and are marked with an asterisk. In addition, the R531Q mutant did not bind the SECIS element of Dio1 (data not shown), which can catalyze the same reactions as Dio2 in the thyroid hormone metabolism pathway (30). Because SBP2 is required for selenoprotein synthesis, the previous report assumed that this defect would have a generalized effect on all selenoproteins. However, given the mild phenotype of the affected individuals, Berry (31) proposed that the expression of other selenoproteins might be preserved in these individuals. In particular, PhGPx and TR1 are essential selenoproteins, because knock-out of these proteins in mouse models results in embryonic lethality (32,33). Thus, the loss of expression of these proteins in humans might be expected to have more severe consequences. When we used the SECIS elements from these two selenoprotein mRNAs as probes in a REMSA assay, the R531Q mutant was able to bind both the PhGPx SECIS and TR1 SECIS (Fig. 4B).
However, although the R531Q mutant was capable of binding to the PhGPx SECIS, it did not appear to bind as well as wild type (Fig. 4B). Because this suggested that the R531Q mutant has a weaker RNA binding affinity than wild type, we wanted to examine a range of protein concentrations in these experiments. Increasing amounts of protein corresponded to a 24ϫ range of concentration. Equimolar amounts of protein and RNA probes were used across experiments, but it should be noted that the GPx1 probe has a higher specific activity, resulting in stronger probe signals. When PhGPx was used as the SECIS element (Fig. 4C, upper panel), increasing the concentration of wild type protein results in an increase of the intensity of the lower functional band and the loss of the upper band as previously mentioned (compare with Fig. 1C). Although the R531Q protein can interact with the PhGPx SECIS, the binding does not exactly recapitulate that of the wild type protein because the doublet remains even at higher concentrations. If GPx1 SECIS is used as probe (Fig. 4C, lower panel), only the binding of the wild type protein is detected. Increasing the amount of R531Q protein does not result in complex formation. The wild type protein bound to all SECIS elements tested, but SBP2 appears to bind better to the PhGPx SECIS than the GPx1 SECIS (Fig.  4C, compare upper and lower panels). This is the first evidence that SBP2 can distinguish among SECIS elements in vitro.
To examine the relative affinities of the wild type and R531Q mutant proteins for different SECIS elements, REMSAs were performed with increasing amounts of unlabeled GPx1 or PhGPx SECIS competitor RNA, and the loss of the labeled SECIS-protein complex was observed. As shown in Fig. 5A, SBP2[517-777] binding to the GPx1 probe can be competed away by both unlabeled PhGPx (left panel) or GPx1 (right panel). However, although 2.5ϫ molar excess of cold PhGPx SECIS was able to completely compete away the binding activity, 250ϫ molar excess of GPx1 SECIS was needed to achieve the same result. Similar discrepancies in competition effectiveness are observed when the converse experiment is performed using PhGPx SECIS as the labeled probe. Although 20ϫ molar excess of cold PhGPx SECIS results in a loss of signal (Fig. 5B,  left panel), 500ϫ molar excess of GPx1 was not sufficient to completely deplete the signal (Fig. 5B, right panel). When the R531Q mutant protein is challenged in these competition assays (Fig. 5C), the binding to the PhGPx probe is competed off with the unlabeled PhGPx SECIS in a manner comparable with that of the wild type protein because 20ϫ molar excess efficiently decoys both proteins (Fig. 5, compare B and C, right  panels). In contrast, GPx1 SECIS RNA does not decoy the R531Q protein away from the PhGPx probe, because the signal with 500ϫ molar excess is the same as in the absence of cold competitor. This indicates that the affinity of the mutant protein for the GPx1 SECIS is further reduced compared with wild type.
The R531Q Mutant Can Support Selenocysteine Insertion in Vitro-In vitro translation assays were performed using either luc/UGA 258 /PhGPx, luc/UGA 258 /GPx1, or luc/UGA 258 /Dio1 reporter constructs. Similar levels of luciferase activity were  observed whether SBP2 CT or SBP2 CT[R531Q] were used in the assay, and surprisingly, the R531Q mutant was able to support selenocysteine insertion from all three reporters. This suggests that although the R531Q mutation alters SBP2 RNA binding affinity so that an interaction with GPx1, Dio1, or Dio2 SECIS elements is not detected by REMSA, a transient interaction may be sufficient to support selenocysteine insertion in vitro.
At first glance, these findings seem to be in contrast with the findings in the affected patients; however, these translation reactions were primed with a single type of SECIS-containing mRNA. In a cellular environment, multiple SECIS elements are present. Because SBP2 is a limiting factor for selenocysteine insertion, competition for SBP2 binding would result. The luciferase experiments were repeated with the luc/UGA 258 / Dio1 reporter, adding increasing amounts of SECIS elements as decoys to mimic a competitive environment. As shown in Fig.  6A, when PhGPx SECIS is used as the decoy RNA, similar inhibition profiles are obtained for both the wild type (left panel) and R531Q mutant proteins (middle panel). An 80% reduction in luciferase activity is obtained with a 0.8ϫ molar ratio of reporter to PhGPx SECIS RNA, regardless of whether wild type or mutant SBP2 CT was used. There was no general effect on translation because the activity of a luciferase control (right panel) was unaffected. When Dio2 SECIS RNA was used as competitor, a 0.8ϫ molar ratio still resulted in an 80% reduction in activity when wild type protein was used (Fig. 6B, left  panel). In contrast, a 0.8ϫ molar ratio only results in a 15% inhibition of activity when the R531Q mutant protein is used. Because PhGPx SECIS RNA competes four times better than Dio2 SECIS RNA when tested at equimolar amounts, this reflects a difference in affinity of the R531Q mutant for these elements. Increasing the amount of Dio2 competitor to a 4ϫ molar ratio results in a 60% inhibition of R531Q-dependent luciferase activity and still does not reach the level of inhibition obtained in the wild type reactions (Fig. 6B, compare left and middle  panels). Again, no inhibitory effects were observed on the activity of a luciferase control (Fig. 6B, right  panel). Similar results were found when GPx1 SECIS RNA was used as competitor. A 70% inhibition was achieved with a 0.8ϫ molar ratio of decoy in the presence of SBP2 CT, whereas only a 15% reduction occurs with a 0.8ϫ molar ratio when the R531Q mutant protein is used. Increasing the amount of GPx1 decoy added improved the inhibition but did not equal that observed with the wild type protein. One difference among the SECIS elements is that at higher molar ratios, GPx1 SECIS RNA has a nonspecific effect, because the activity of the luciferase control reactions was decreased by the addition of the competitor (Fig. 6C, right  panel).
Taken together, these results suggest that the R531Q mutation alters the RNA binding activity of SBP2. The addition of this subtle binding deficit into a situation where differential binding affinities between SECIS elements inherently exist results in the loss of expression of a subset of selenoproteins.

DISCUSSION
The interaction of SBP2 with the SECIS is critical for selenoprotein expression. The importance of this interaction is underscored by the diseases that can arise when this interaction fails, either through mutation of the SECIS RNA as in rigid spine muscular dystrophy (24) or mutation of the SBP2 protein, which can lead to thyroid hormone dysfunction (25). In the study, we show that the R531Q mutant associated with abnormal thyroid function lies within a novel bipartite RNA-binding domain within SBP2. This mutation selectively interferes with the ability of SBP2 to bind to the SECIS and promote selenocysteine insertion in a subset of selenoprotein mRNAs.
Previous studies in our lab demonstrated that the RNA-binding domain of SBP2 fell in the region between residues 517 and 777, which also contains the L7Ae motif (16). In this study, we show that the L7Ae motif (amino acids 650 -752) is not sufficient for SECIS binding and that another region upstream (amino acids 520 -545) is also required. This suggests that the RNA-binding domain of SBP2 is actually bipartite and represents a unique binding motif. Secondary structure predictions for this region were generated using two web-based prediction methods, the self-optimized prediction method from Alignment (34) and sequence alignment and modeling system (SAM-T02) (35). The regions encompassed by residues 520 -545 and 622-756 are predicted to contain secondary structure elements (Fig. 7A), whereas the intervening unconserved region is not. We propose a model of a tertiary structure of SBP2 where the two regions combine to form the interface for the interaction of SBP2 with the SECIS RNA (Fig. 7B). The model depicted is using the rat SBP2 amino acids numbers; however, this predicted structure is conserved across species. Although the crystal structure of SBP2 has not been solved, using the PHYRE Protein Fold Recognition Server (36), the region of SBP2 from residues 659 to 729 can be directly modeled onto the structure of L7, with the same spacing and placements of helices and ␤-sheets. However, in addition to these shared secondary structure elements, two additional regions of secondary structure are predicted in SBP2. Within our defined RNA-binding region of residues 520 -756, one helix is predicted from residues 524 -542 and another from residues 623-632. Our experimentally determined requirements for SECIS binding merge exactly with these secondary structure predictions, because all regions of predicted structure were preserved in our internal deletion mutants, and their RNA binding activity was not disrupted. This suggests that the remaining unstructured region from residues 612 to 622 is sufficient in length and flexibility to allow the proper relative alignment of the structural units of the RNAbinding domains. Because this unstructured region is not required as a spacer for RNA binding, it is not clear what purpose this intervening sequence may serve. Intriguingly, each of the internal deletion mutants was more active than the wild type protein for in vitro selenocysteine insertion. This deleted region may be involved in the negative regulation of SBP2 in some manner, but more detailed studies will be required to determine the validity of this idea.
It has been previously reported that SBP2 can distinguish among SECIS elements in cells (37). Competition studies in transfected HEK293 cells have shown that different SECIS elements varied in their effectiveness to block expression of a SECIS-dependent reporter construct and that competition could be rescued by overexpression of SBP2. To date, in vitro studies have not detected large differences in the binding of SBP2 CT to various SECIS elements, with respect to both affinity and binding site (38). This current study is the first demonstration that SBP2 CT can differentiate among SECIS elements in vitro. Although the previous data were attained using SBP2 CT in REMSA assays, the protein used in those studies was later found to aggregate, and the SECIS elements used were shorter. Here we show that the smaller region of SBP2[517-777] is capable of differentiating between SECIS elements in the REMSA assays and that SBP2 CT can distinguish the elements in the selenocysteine insertion assays. The R531Q mutation appears to impair the binding of SBP2 sufficiently so that high affinity interactions with a subset of SECIS elements, including Dio2 and GPx1, is no longer possible. While this manuscript was under review, Squires et al. (39) reported that SBP2 showed selective association with different SECIS elements in vivo.
The R531Q mutant protein shows differences in its interaction with various SECIS elements, but the basis for the difference is unclear. There are no obvious differences in primary sequence between those SECIS elements that were bound by the R531Q mutant in the REMSA assays and those that were not. However, SECIS elements can be subclassed into Type I and Type II elements, based on the presence of either a terminal apical loop or a terminal bulge. All three elements that were bound by the R531Q mutant are Type II SECIS elements, as determined by biochemical probing for PhGPx and selenoprotein 15 (38) and by NMR for TR1 (40). In contrast, GPx1 and Dio1 are both Type I SECIS elements based on structure probing (38). The only potential outlier is human Dio2, which is predicted to be a Type II element by SECISearch (2), but this has not been directly tested. Perhaps these subtle differences in structure that do not appear to correlate with the apparent binding affinity of the wild type protein are sufficient to destabilize the interaction with the R531Q mutant.
The R531Q mutant promotes Sec insertion, which may indicate that a stable interaction with the SECIS RNA is not a strict prerequisite for recoding functions in vitro. However, the patients with this mutation do not have active Dio2 or GPx1 protein, despite wild type levels of these selenoprotein mRNAs, suggesting that the translation of these mRNAs is severely impaired in vivo. The in vitro selenocysteine insertion assays with the decoy RNAs are a model for the cellular situation where multiple selenoprotein messages are present and competing for the protein factors required for selenocysteine insertion. Because SBP2 is a limiting factor in cells (37), the relative ability of the R531Q mutant to bind with higher affinity to some SECIS elements than others may be reflected in which members of the selenoprotein family are expressed. In the competitive cellular environment, if the R531Q mutant binds to GPx1 or Dio2 and then dissociates and binds to PhGPx or TR1, where it is stably bound, an imbalance in relative expression of selenoproteins is easily visualized, even in selenium-adequate conditions. A recent study that surveyed the relative abundance of selenoprotein mRNA levels in several mouse tissues found that PhGPx mRNA is highly expressed and always significantly more abundant than Dio1, Dio2, or GPx1 mRNA (41). In our competition assays, adding a 2ϫ molar excess of PhGPx SECIS RNA dropped the luc/UGA 258 /Dio1 reporter activity to background for both the wild type and R531Q mutant proteins. The  NOVEMBER 30, 2007 • VOLUME 282 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 34661
PhGPx:Dio1 ratio used in our assay is far lower than would be predicted to exist within a cell based on the recent study. The relatively high abundance of PhGPx SECIS within a cell, coupled with the higher affinity of SBP2 for PhGPx would result in a preference for PhGPx production over many other selenoproteins. Although our REMSA data show that the wild type SBP2[517-777] shows varying affinities for different SECIS elements, the selenocysteine insertion assays revealed that the R531Q mutation further decreases the affinity of SBP2 CT for a subset of SECIS elements. In patients whose selenoprotein expression depends on the SBP2[R531Q] mutant protein, this additional decrease in affinity results in the loss of a subset of selenoproteins including GPx1 and Dio2.
In conclusion, this study supports the model that the intracellular competition among SECIS elements for SBP2 binding will determine which selenoproteins are expressed and that mutations that affect SBP2 RNA binding affinity will affect the expression of the selenoproteome. Our results are in keeping with the hypothesis that the expression of a specific subset of selenoproteins is affected by the R531Q mutation in SBP2, which leads to the relatively mild phenotype of the affected individuals.