Cloning and functional characterization of human selenophosphate synthetase, an essential component of selenoprotein synthesis.

Selenocysteine is co-translationally incorporated into prokaryotic and eukaryotic selenoproteins at in-frame UGA codons. However, the only component of the eukaryotic selenocysteine incorporation machinery identified to date is the selenocysteine-specific tRNA. In prokaryotes, selenocysteine is synthesized from seryl-tRNA and the active selenium donor, selenophosphate. Selenophosphate is synthesized from selenide and ATP by the selD gene product, selenophosphate synthetase, and is required for selenocysteine synthesis and incorporation into bacterial selenoproteins. We have now cloned human selD and shown that transfection of the human selD cDNA into mammalian cells results in increased selenium labeling of a mammalian selenoprotein, type 1 iodothyronine deiodinase. Despite significant differences between the mechanisms of selenoprotein synthesis in prokaryotes and eukaryotes, human selD weakly complements a bacterial selD mutation, partially restoring selenium incorporation into bacterial selenoproteins. Human selenophosphate synthetase has only 32% homology with the bacterial protein, although a highly homologous region that has similarity to a consensus ATP/GTP binding domain has been identified. Point mutations within this region result in decreased incorporation of selenium into type 1 iodothyronine deiodinase in all but one case. Further analysis revealed that reduced selenium labeling was due to altered ATP binding properties of the mutant selenophosphate synthetases.

Several eukaryotic selenoproteins have also been identified, including glutathione peroxidases (16), selenoprotein P (17,18), and the type I iodothyronine 5Ј-deiodinase (19), all of which contain selenocysteine encoded by UGA codons. However, the mechanisms of recognition and translation of selenocysteine codons differ between prokaryotes and eukaryotes. Although stable stem-loop structures are required for eukaryotic selenoproteins synthesis, these elements are situated within the 3Јuntranslated region of the mRNAs (20). In addition, the sequences of prokaryotic and eukaryotic stem-loop structures are not conserved (2,20). Mutagenesis studies have also indicated that the position of the stem-loop in bacterial formate dehydrogenase is essential to function (2), whereas in 5Ј-deiodinase, the position of the stem-loop is quite flexible (21). Finally, cloning of the rat and human selenoprotein P cDNAs revealed the presence of 10 in-frame UGA codons (17,18) and two 3Ј-untranslated stem-loops (21). As yet, no prokaryotic seleno protein containing multiple selenocysteine residues has been identified.
Apart from the tRNA Sec (22,23), none of the components of the eukaryotic selenocysteine incorporation machinery have been identified. This may be explained by poor conservation of the selenoprotein translational machinery. For example, eukaryotic selenoprotein mRNAs are unable to direct selenocysteine incorporation when expression is attempted in Escherichia coli. 1 In addition, cloning of the tRNA Sec from several organisms has revealed evolutionary differences in the sequences for this tRNA within both the eukaryotic (23) and the prokaryotic (24) kingdoms.
We have now cloned the human homologue of bacterial selD and shown that it is functional in mammalian cells and also in bacteria. Comparison of the human and bacterial selenophosphate synthetase peptide sequences reveals a high homology glycine-rich sequence that is similar to conserved sequences found in many ATP/GTP binding proteins and protein kinases (25,26,27,28). Mutational analysis has revealed that conserved amino acids within this region are important for selenophosphate synthetase activity and for the ATP binding properties of the enzyme.

75
Se was a gift from Dr. Dolph Hatfield (National Institutes of Health, Bethesda, MD) and was originally obtained from Dr. Kurt Zinn (University of Missouri Research Reactor, Columbia, MO). The Xenopus tRNA Sec gene in pGEM3 was kindly provided by Dr. Dolph Hatfield (22,23). E. coli strain WL400, containing an inactivated selD gene, and pMN330, which contains the bacterial selD gene in pBR322 (5), were gifts from the laboratory of Dr. August Bock.

Isolation of a Human selD Clone
Partial Clone-A partial human selD cDNA clone was identified using a yeast two-hybrid system as described previously (29, 30, 31). The bait protein used to clone selD was the mammalian spliceosomeassociated protein SAP62.
Full-length Clone-A human liver cDNA library (a gift of Dr. Brian Seed) in the plasmid CDM8 (33) was transformed into MC1061-P3 host cells and plated onto 150-mm agar plates. Colonies were lifted onto nylon filters (GeneScreen Plus, DuPont NEN) and denatured in 0.5 M NaOH, neutralized in 0.5 M Tris (pH 7.5), 1.5 M NaCl, and cross-linked using a UV-Stratalinker 1800 (Stratagene, La Jolla, CA). Filters were prehybridized as described previously (34) and then hybridized overnight (34) with a random primed (Prime It, Stratagene, La Jolla, CA) 32 P-labeled PstI fragment (nucleotides 1023-1623) from the partial human selD clone. Membranes were washed to a final stringency of 0.1 ϫ SSC (1 ϫ SSC ϭ 0.5 M NaCl and 15 mM sodium citrate), 0.1% SDS at 65°C. Four positive clones were identified and colony-purified. Two of these, selD2 and selD4, were selected for further characterization. Dideoxynucleotide sequencing was carried out using a Sequenase kit (Version 2, U. S. Biochemical Corp).

Transient Transfections
pSL21 was created by subcloning the entire E. coli selD cDNA from pMN330 (5) into the HindIII/NotI sites of CDM8. 10 g of selD4, pSL21, or mutants of selD4 were transfected into HtTA-1 cells (35) in combination with two other plasmids, G21-D10 (5 g), encoding rat 5Ј-deiodinase (36), and XtRNA (5 g), encoding the Xenopus tRNA Sec gene in pGEM3 (23,24). 3 days prior to transfection, cells were plated onto 60-mm dishes in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. 1 day prior to transfection, the medium was changed to Dulbecco's modified Eagle's medium containing 1% fetal calf serum supplemented with 100 pM 3,5,3Ј-triiodothyronine, 10 g/ml transferrin, 20 mg/ml insulin, and 50 nM hydrocortisone to decrease the endogenous serum and selenium concentrations. Plasmids were transfected by CaPO 4 co-precipitation as described previously (37) and maintained in the supplemented low serum media. 1 day following transfection, cells were shocked with 10% Me 2 SO, and fresh medium containing 5 nM unlabeled selenium plus 6 Ci of 75 Se was added. In vivo labeling was carried out for 18 h. Cells were harvested, washed, and resuspended in 0.1 M potassium phosphate (pH 6.9), 1 mM EDTA containing 0.25 M sucrose, and 10 mM dithiothreitol. Aliquots of labeled cell extracts were sonicated and diluted in protein electrophoresis buffer (0.125 mM Tris-HCl (pH 6.8), 2.5% SDS, and 0.5 M mercaptoethanol). Samples were boiled for 2 min and electrophoresed on 10% acrylamide (acrylamide to bis, 37.5:1) SDS gels, followed by autoradiography.

Bacterial Transformations
Since E. coli contain endogenous prokaryotic selD, expression of exogenous human and E. coli selD in bacteria was carried out in strain WL400, in which the endogenous selD gene has been interrupted by insertion of the chloramphenicol acetyl transferase gene, producing a selD Ϫ phenotype (5). All of the selD constructs are in CDM8 vector, a supF suppressor tRNA plasmid that carries no antibiotic resistance. CDM8 therefore must be amplified in nonsuppressing hosts containing the P3 plasmid, which encodes amber-mutated ampicillin and tetracycline drug resistance elements, and a wild-type kanamycin resistance gene (38). In order to express the selD-CDM8 constructs in E. coli, a new host strain, WL400-P3, was created by transfer of the P3 plasmid from MC1061-P3 to WL400, followed by selection for P3 with kanamycin (50 g/ml) and for WL400 with chloramphenicol (20 g/ml). selD4 and pSL21 were then transformed into WL400-P3. As a positive control, E. coli strain BL21(DE3), which contains endogenous selD and the T7 RNA polymerase gene under control of the lacUV5 promoter in the chromosome (39), was also used. The human selD2 coding region plus 23 nucleotides of the 3Ј-untranslated region (nucleotides 254-1476) was amplified by polymerase chain reaction using oligos designed to create 5Ј NdeI and 3Ј BamHI restriction sites respectively. This polymerase chain reaction fragment was subcloned into the NdeI and BamHI sites in pET-3a (Novagen, Madison, WI) and transformed into BL21(DE3).
Single colonies of BL21(DE3), BL21(DE3)/pET-selD2, WL400-P3, WL400-P3/selD4, and WL400-P3/pSL21 were grown in LB with appropriate antibiotics until the cells were in log phase growth. Cultures were then diluted 1:4 in LB and 6 Ci of 75 Se was added to each. 4 mM isopropyl-␤-D-thiogalactopyranosidase was added to both BL21(DE3) cultures to induce the T7 RNA polymerase gene and consequently transcription of the selD plasmid. 75 Se labeling was continued for 3 h before bacterial pellets were recovered, resuspended in distilled water, and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography.

Site-specific Mutagenesis
Oligonucleotide mutagenesis by polymerase chain reaction amplification using Taq DNA polymerase (Boehringer Mannheim) was performed as described previously (40) to create a conserved amino acid change, His 275 3 Asn, and a nonconserved amino acid change, His 275 3 Tyr. Polymerase chain reaction products were subcloned into T-vector (Promega), and individual colonies were sequenced to confirm the presence of the mutations. Gly 269 3 Cys, Gly 271 3 Arg, Gly 274 3 Asp, Gly 274 3 Ala, and Gly 274 3 Val mutations were created using the Altered Sites in vitro mutagenesis system (Promega) as described previously (41), and the mutations were confirmed by sequencing. Each of the mutated sequences was then subcloned into the XmaI/EcoRV sites of selD4 in CDM8 vector. The entire subcloned regions were sequenced to confirm that no other mutations were present.

Photoaffinity Labeling with 8-Azido-ATP
HtTA cells (35), were transiently transfected with 10 g of selD4 or 10 g each of the mutants of selD4 as described above. Co-transfection of 3 g of a constitutive human growth hormone-expressing plasmid, pTKGH, was used to monitor transfection efficiencies as described previously (37). Cells were harvested 2 days following transfection, washed, and resuspended in 0.1 M potassium phosphate (pH 6.9), 1 mM EDTA containing 0.25 M sucrose. Aliquots of transfected cells were stored at Ϫ70°C until use. ATP-binding assays were performed using 8-azido-[␥-32 P]ATP (specific activity 12.8 Ci/mmol, ICN Flow), an analogue of ATP that can be cross-linked under UV irradiation. Reaction mixtures (30 l) consisted of an aliquot of cell extracts corrected for the transfection efficiency, 100 mM Tricine 2 (pH 7.2 with KOH) and 7.2 M 8-azido-[␥-32 P]ATP. Reactions were irradiated at 254 nm for 3 min on ice in the absence of visible light. The reaction was stopped by the addition of excess dithiothreitol. Labeled proteins were analyzed by 12% polyacrylamide SDS gel electrophoresis followed by autoradiography.
Time course experiments with 8-azido-[␥-32 P]ATP were also carried out to examine potential differences between selD4 and the selD mutants. Reaction mixtures were assembled as described above and allowed to incubate on ice for varying lengths of time before UV crosslinking, then reactions were stopped by the addition of excess dithiothreitol. Labeled proteins were analyzed by polyacrylamide SDS gel electrophoresis and autoradiography.

RESULTS
Isolation of a Human selD cDNA-Screening of a human liver cDNA library with a fragment of the partial selD clone resulted in the isolation of four positive clones with inserts varying from 1225 to 1672 nucleotides in length. All four clones showed significant sequence similarity to E. coli selD; two clones, selD2 and selD4, were further investigated. Two alternative polyadenylation sites appear to be used in the human selD mRNA. The partial clone contains a consensus polyadenylation signal at nucleotides 1588 -1593, followed by a poly(A) tail beginning at 1613. selD2 and selD4 are polyadenylated at an alternative site approximately 80 nucleotides 5Ј of the position in the partial clone. This site is also preceded by a polyadenylation signal at 1507-1512. A third clone extended 64 base pairs 3Ј of the partial clone and contained no poly(A) tail, while the fourth clone terminated at nucleotide 1374. No consensus polyadenylation sequence is present in this region. Of the four clones isolated, selD2 extended the furthest 5Ј, approximately 240 base pairs upstream of the ATG start codon. Included in this region were approximately 80 base pairs of sequence of Ͼ90% GC content, which was highly compressed and therefore difficult to sequence accurately.
The deduced amino acid sequence of human selD is 381 amino acids with a calculated molecular mass of approximately 45 kDa and shows overall 32% identity and 55% similarity with E. coli selD. Alignment of the human and E. coli peptide sequences (Fig. 1) revealed a conserved glycine rich region with the sequence Gly-Phe-Gly-Ile-Leu-Gly-His at 269 -275 in the human peptide, and at 232-238 in the E. coli sequence (Fig. 1). This sequence is similar to the conserved ATP/GTP binding consensus sequences, Gly-X-X-X-X-Gly-Lys-Ser/Thr and Gly-X-Gly-X-X-Gly found in many ATP/GTP binding proteins and protein kinases, respectively (26,27,28,29).
Expression of Human and Prokaryotic selD in Mammalian Cells-To examine the functional activity of selD in mammalian cells, human (selD4) and E. coli (pSL21) selD cDNAs were transfected into HtTA cells in combination with G21-D10, which expresses the mammalian selenoprotein 5Ј-deiodinase, and XtRNA, which encodes the Xenopus tRNA Sec . Transfected G21-D10 alone is poorly labeled with 75 Se (Fig. 2). However in the presence of co-transfected selD4, 75 Se labeling of 5Ј-deiodinase is greatly enhanced, indicating that the human selD cDNA is functional in mammalian cells (Fig. 2). In addition, labeling of 5Ј-deiodinase is increased by co-transfection of XtRNA, either in the absence or presence of selD. E. coli selD (pSL21) also leads to increased 75 Se labeling of 5Ј-deiodinase compared with G21-D10 alone, indicating that the action of selD is probably very similar in bacteria and mammalian cells (Fig. 2).
Complementation of a Bacterial selD Mutation-E. coli strain BL21(DE3) contains endogenous selD, thus in vivo labeling with 75 Se results in the incorporation of selenium into formate dehydrogenase and 5-methylaminomethyl-2-selenouridine residues in tRNAs (Fig. 3). Interestingly, when BL21(DE3) was transformed with pET-selD2, selenophosphate synthetase itself became labeled with 75 Se. In contrast, in WL400-P3, which contains an interrupted and therefore inactive selD gene (5), neither the tRNAs nor formate dehydrogenase are labeled with 75 Se, confirming that an active selD gene is required for selenium incorporation into bacterial selenoproteins and tRNAs (Fig. 3). 75 Se labeling of formate dehydrogenase and tRNAs is restored when WL400-P3 are transformed with an intact E. coli selD gene (pSL21) or the intact human selD cDNA (selD4). However, the levels of incorporation of 75 Se into formate dehydrogenase and tRNAs in the presence of Upper sequence is human selD, and lower sequence is E. coli selD. Conserved amino acids between the human and bacterial enzymes are indicated by vertical lines. The glycine-rich, putative ATP binding consensus sequence is indicated by an asterisk; methionine residues in both sequences are indicated by ϩ.
selD4 are much lower than in bacteria transformed with E. coli selD (Fig. 3).
Expression of selD Mutants in Mammalian Cells-Point mutations within the putative ATP-binding domain of human selD were created, and the effects of these amino acid substitutions on selenium incorporation into 5Ј-deiodinase determined by transfection into HtTA-1 cells. Compared with wild type selD4, all but one of the selD mutants resulted in less efficient incorporation of 75 Se into 5Ј-deiodinase (Fig. 4, a and b). Mutation of the histidine at position 275 to tyrosine resulted in increased selenium labeling of 5Ј-deiodinase, in contrast with the decreased 75 Se incorporation observed when the same residue is mutated to aspartate (Fig. 4a). Mutational analysis of each of the conserved glycine residues indicates that all three are important for normal functioning of human selD, since selenium labeling of 5Ј-deiodinase is decreased in the presence of each of these mutants compared with wild-type human selD (Fig. 4b).
ATP Binding Properties of Mutant Selenophosphate Synthetases-Since the amino acid substitutions made were to the putative ATP-binding domain of human selD, we next tested the possibility that these mutants may have lost their ability to bind ATP, thus explaining the decreased selenium labeling of 5Ј-deiodinase. As shown in Fig. 5, several of the mutant selD proteins bind ATP less efficiently than the wild-type human selD. In particular, mutation of the glycine residue at position 274 to either asparagine, valine, or alanine resulted in an almost complete inability of these mutants to bind ATP. In addition, mutation of histidine 275 to aspartate also produced decreased ATP binding compared with wild-type selD (Fig. 5). In contrast, mutation of the same histidine to tyrosine resulted in increased ATP binding, indicating that the amino acid at position 275 is involved in ATP binding in wild-type selD. Mutation of the glycine residues at positions 269 and 271 to cysteine and arginine respectively, had no effect on ATP binding, although both of these mutations resulted in decreased selenium labeling of 5Ј-deiodinase ( Fig. 4b and data not shown), suggesting that the glycine residues at these positions have roles in normal selenophosphate synthetase activity other than ATP binding.
Following ATP binding to selenophosphate synthetase, ATP is hydrolyzed to produce AMP and a pyrophosphate-enzyme intermediate (9,10). In order to examine ATP hydrolysis by wild-type human and mutant selenophosphate synthetases, cell extracts were incubated with ATP, and the reaction products were analyzed by thin-layer chromatography. Data from this study indicated that the wild-type and mutant enzymes do hydrolyze ATP (data not shown). However cell extracts alone also extensively degrade ATP, and thus rates of hydrolysis by the selenophosphate synthetases were difficult to measure accurately. Therefore we have examined the time course of ATP binding. Wild-type selD4 initially exhibits efficient ATP binding, followed by a rapid loss of bound label (Fig. 6). In contrast, the amount of ATP binding by the His 275 3 Tyr mutant decreases gradually over the same time period, suggesting less ATP hydrolysis by this mutant (Fig. 6). Gly 269 3 Cys and Gly 271 3 Arg also bind more ATP for a longer period of time than wild-type selD (Table I, and data not shown) suggesting that glycine residues at position 269 and 271 may be involved in ATP hydrolysis (Table I). However, due to degradation of ATP by cell extracts, definitive information on ATP hydrolysis by wild-type human selD and the mutants will require purification of the enzymes.

FIG. 5. 8-azido-[␥-32 P]ATP binding properties of selD.
HtTA-1 cells were transfected with wild-type human selD (selD4; 10 g/dish) or selD mutants (10 g/dish) as indicated. Photoaffinity labeling of cell extracts with 8-azido-[␥-32 P]ATP was carried out as described under "Experimental Procedures." ATP binding was analyzed on 12% polyacrylamide SDS gels followed by autoradiography. DISCUSSION The role of prokaryotic selenophosphate synthetase in the production of the active selenium donor, selenophosphate, has been clearly demonstrated (3,4,5,13,14,15). Mammalian selenoproteins have also been identified (16,17,18,19,20), although the mechanism whereby selenocysteine is synthesized and incorporated into these proteins is not yet clear. We have now cloned the human homologue of bacterial selD and shown that it is functional in both mammalian cells and bacteria. A partial clone of human selD was obtained using a yeast two-hybrid protein interaction system. The bait protein for this interaction was the mammalian spliceosome-associated protein SAP62, which is involved in mRNA splicing (42). Since appropriate control experiments were performed to minimize nonspecific interactions with the bait protein, the interaction between selD and SAP62 may be specific, although the relevance of this and the implication that selD may be involved in RNA splicing is not known.
Transfection of E. coli and human selD into mammalian cells showed that both prokaryotic and eukaryotic selenophosphate synthetases are functional in higher eukaryotes. However, when human selD was transformed into bacteria containing an inactive selD gene, it only partially complemented the bacterial system and produced low levels of 75 Se labeled tRNAs and formate dehydrogenase. One possible explanation for this may be that the human selD gene product is improperly folded in bacteria. Alternatively, a factor(s) found in mammalian cells but not bacteria may be required in addition to the mammalian selD gene product to maintain normal selenophosphate synthetase activity. Nevertheless, these data do indicate that although human and bacterial selenophosphate synthetase sequences are poorly conserved (32% homology), and there is a great evolutionary distance separating the two proteins, the action of selD is probably very similar in bacteria and higher organisms. When the E. coli strain BL21(DE3) was transformed with human selD and grown in the presence of 75 Se, in addition to the expected bacterial selenoproteins becoming labeled, selenophosphate synthetase itself was also labeled by 75 Se (Fig. 3). This was unexpected, since the sequence of human selD does not indicate the presence of selenocysteine within the protein itself. However, when the selenium to sulfur ratio is higher than normal, selenium may be nonspecifically misincorporated into proteins in the form of selenomethionine (43,44), or, if the methionine content of a protein is high, misincorporation of selenium into selenomethionine may also occur (45,46). It was therefore interesting to note that human selenophosphate synthetase contains 18 methionine residues, and the bacterial protein contains 13, both of which are much higher than the average protein methionine content and may therefore explain the observed labeling of selD.
Mutational analysis of the putative ATP binding consensus sequence in the human selD cDNA indicated that all of the amino acids considered are important in normal selenophosphate synthetase activity. We have also shown that the residues within this region are important for normal ATP binding properties of selenophosphate synthetase, although direct measurement of ATP hydrolysis will require purification of wild-type and mutated enzymes. Recent work on bacterial selD involved mutational analysis of the sequence encoding the amino terminus of the protein, which also shared some similarity with the consensus ATP binding sequence (47,48). It was shown that the cysteine residue at position 17 (Cys 17 ) and the lysine residue at position 20 (Lys 20 ) are both essential for the formation of selenophosphate from selenide and ATP (47,48), while Cys 17 was also shown to be required for ATP binding to selenophosphate synthetase. In contrast, mutation of Cys 19 had no effect on selenophosphate activity. The significance of these results in relation to human selenophosphate synthetase activity is not clear since the only residue in this region that is conserved between the two sequences is Cys 19 , which was not essential for catalytic activity in E. coli selD. However it is possible that there are differences in the three-dimensional structure of the bacterial versus human selD gene products, which could explain the discrepancies between the two studies. Further work is clearly required to determine the importance of individual amino acids in both selD homologues for normal selenophosphate synthetase function, ATP binding, and ATP hydrolysis.