A Novel Cysteine-rich Domain of Sep15 Mediates the Interaction with UDP-glucose:Glycoprotein Glucosyltransferase*

Selenium is an essential trace element with potent cancer prevention activity in mammals. The 15-kDa selenoprotein (Sep15) has been implicated in the chemopreventive effect of dietary selenium. Although the precise function of Sep15 remains elusive, Sep15 co-purifies with UDP-glucose:glycoprotein glucosyltransferase (GT), an essential regulator of quality control mechanisms within the endoplasmic reticulum. Recent studies identified two GT and two Sep15 homologues in mammals. We characterize interactions between these protein families in this report. Sep15 and GT form a tight 1:1 complex, and these interactions are conserved between mammals and fruit flies. In mammalian cells, Sep15 co-immunoprecipitates with both GT isozymes. In contrast, a Sep15 homologue, designated selenoprotein M (SelM), does not form a complex with GT. Sequence analysis of members of the Sep15 family identified a novel N-terminal cysteine-rich domain in Sep15 that is absent in SelM. This domain contains six conserved cysteine residues that form two CxxC motifs that do not coordinate metal ions. If this domain is deleted or the cysteines are mutated, Sep15 no longer forms a complex with GT. Conversely, if the cysteine-rich domain of Sep15 is fused to the N-terminus of SelM, the resulting chimera is capable of binding GT. These data indicate that the cysteine-rich domain of Sep15 exclusively mediates protein-protein interactions with GT.

Selenium is present in 25 human proteins in the form of selenocysteine, which is the 21st natural amino acid in the genetic code and is encoded by the UGA codon (1,2). Recent studies provide strong evidence that dietary selenium supplements reduce the incidence of cancer in animal models and human clinical trials (3)(4)(5). Low molecular weight selenium compounds and selenium-containing proteins have been implicated in this chemopreventive effect. However, the mechanism by which selenium suppresses tumor development remains to be established.
One candidate protein that may mediate the chemopreventive effects of selenium is the 15-kDa selenoprotein (Sep15) 3 (6,7). The Sep15 gene is located at position p31 on human chromosome 1. This region has been shown to undergo loss of heterozygosity, which is frequently associated with tumor progression, suggesting a potential tumor suppressor role for this selenoprotein (8,9). In addition, two polymorphisms found at positions 811 (C/T) and 1125 (G/A), the latter of which is located within the selenocysteine insertion sequence element, have been identified within the 3Ј-untranslated region of the human Sep15 gene. The nucleotide identity at these polymorphic sites has been shown to influence the efficiency of selenocysteine incorporation in a selenium-dependent manner (10).
The precise biological function of Sep15 has not been firmly established; however, this selenoprotein co-purifies with the essential regulator of the calnexin cycle designated UDP-glucose:glycoprotein glucosyltransferase (GT) (11). The calnexin cycle is a quality control pathway localized to the endoplasmic reticulum that specifically assists in the folding of N-linked glycoproteins (12). Once a newly synthesized protein enters the endoplasmic reticulum, a glycan core of 14-oligosaccharides (Glc 3 Man 9 GlcNAc 2 ) is covalently linked to appropriate asparagine residues. N-Linked glycoproteins are immediately recognized by glucosidase I and II, the first components of the calnexin cycle, which sequentially trim the two outermost glucose moieties from the glycan core (13). Once processed, membrane-bound calnexin and its lumenal homologue calreticulin bind to the glycan (14 -17). This triggers the binding of ERp57 (a lumenal protein disulfide isomerase) to calnexin and calreticulin and accelerates folding by catalyzing disulfide bond exchange (18 -22). Regardless of the folded status of the bound glycoprotein, glucosidase II cleaves the remaining glucose residue and the complex disassociates, releasing the glycoprotein. If the glycoprotein is properly folded, it is rapidly exported from the endoplasmic reticulum by vesicular transport to the Golgi complex. Misfolded proteins are re-glucosylated by GT, a 175-kDa lumenal enzyme that catalyzes the transfer of a glucose moiety from UDP-glucose to the distal mannose of the glycan core (23). This event creates a retention signal that re-initiates the calnexin cycle.
In this quality control pathway, GT functions as the folding sensor that assesses the folded status of glycoproteins. Although we do not understand how GT senses structural fidelity, several important clues have been revealed. (i) Sensing involves the innermost (and normally buried) N-GlcNAc glycosidic bond of the glycoprotein (24,25). (ii) GT preferentially monoglucosylates partially folded late folding intermediates as compared with their fully denatured counterparts (26 -28). (iii) GT is capable of distinguishing folded from misfolded domains within a single polypeptide (29,30). (iv) Localized folding defects in otherwise correctly folded glycoproteins or individual domains are recognized by GT (30,31). (v) GT monitors the tertiary and quaternary structures of glycoproteins (32).
Given the sequence diversity of glycoproteins, it is unlikely that the recognition mechanism used by GT involves a specific primary structure motif. A more robust mechanism, involving common biophysical properties of misfolded proteins, seems more plausible. Accordingly, the distribution of glycans and their spatial relationship to normally surface-inaccessible hydrophobic residues influence the order in which polypeptide segments achieve their native conformation (12,27,28). GT preferentially recognizes hydrophobic motifs proximal to glycosylation sites in glycopeptides (33). The efficiency of glucosylation can be positively correlated with the spatial proximity of site-directed mutations to each glycan, lending support to the idea that normally buried hydrophobic residues found adjacent to glycans mediate interactions with GT (31).
Our initial findings of the interaction between mammalian Sep15 and GT (11) raised the possibility that Sep15 might be a misfolded glycoprotein. Although Sep15 does not contain an endoplasmic reticulum retention sequence, it is neither secreted or glycosylated, suggesting that this selenoprotein might function as a co-factor that assists GT in assessing structural fidelity. Sep15 deficiencies observed in tumors may also impair GT-mediated regulation of the calnexin cycle.
In this report, we characterize interactions between members of Sep15 and GT protein families. Isothermal titration calorimetry was used to demonstrate that Sep15 forms a 1:1 complex with GT with an apparent K d value of 20 nM. Moreover, co-immunoprecipitation revealed that Sep15 binds to both human GT isozymes. Although selenoprotein M (SelM) shares significant sequence homology with Sep15 and has also been localized to the endoplasmic reticulum, it does not form a complex with GT. The interaction between Sep15 and GT is exclusively mediated by a novel cysteine-rich domain located at the N-terminus of Sep15. This conserved domain is present in other functional Sep15 homologues and contains six conserved cysteine residues that do not coordinate metal ions. When the cysteine-rich domain of Sep15 is removed, it no longer forms a complex with GT. Site-directed mutagenesis was used to demonstrate that the structural integrity of the cysteine-rich domain is an essential factor in the ability of Sep15 to bind GT.
A fusion protein of Sep15 and SelM from mouse (designated Sep15::SelM) was generated by fusing the cysteine-rich domain of Sep15 to the N-terminus of SelM. The cysteine-rich domain of Sep15, together with its native signal peptide, was PCR amplified using the following primers: 5Ј-TATAGACTAGACTCGAGGGCAGACCGCAG-3Ј and 5Ј-GCAGACTTCAAGGATGAATTCTGCATACAGCTTTTTGG-3Ј. This PCR product was digested with XhoI and EcoRI and ligated into the pCI-neo vector. SelM-U48C was PCR amplified without its signal peptide using the following primers: 5Ј-CTTGTGGCTCCAGC-CGAATTCACCACCAACTACCGACC-3Ј and 5Ј-ATGCGGCCGC-TACAGGTCGTCGTGTTCTGAAGCTTCCTCG-3Ј. The PCR product was digested with EcoRI and NotI and ligated into the pCI-neo vector containing the cysteine-rich domain of Sep15.
Mammalian expression vectors encoding protein C (PC)-tagged human GT1 and GT2 have been described previously (34). The gene encoding Sep15 from Drosophila melanogaster was PCR amplified from an expressed sequence tag clone using the following primers: 5Ј-TGC-TCGAGCAAATTGGCGCCGAATTGACGGCCGCCGACTG-3Ј and 5Ј-CAGTTTTGGGTATGGGTACCTTAGATCCTGTTGGTT-CGCAGG-3Ј. The PCR product was digested with EheI and KpnI and ligated into Baculovirus expression vector pFastBac-HTa (Invitrogen). A previously described cDNA clone (35) served as the template to introduce a C-terminal hexahistidine-tag into Drosophila GT. To improve secretion from insect cells the native signal sequence was replaced with the honeybee melittin signal peptide. Moreover, the C-terminal endoplasmic reticulum retention sequence was removed. Primer sequences are available upon request.
Native Gel Electrophoresis-Recombinant Drosophila Sep15 and GT proteins were expressed in Sf9 cells and purified by immobilized metalaffinity chromatography. Complexes of Sep15 and GT were prepared by incubating 5 g of GT dissolved in 10 mM Tris-HCl (pH 7.5) and 10 mM CaCl 2 with the indicated amounts of Sep15 for 10 min at 37°C. Samples were subjected to non-denaturing PAGE using Tris-glycine gels (Invitrogen) and transferred to polyvinylidene difluoride membranes. The Sep15-GT complex was visualized by Western blotting with Sep15specific antibodies.
Isothermal Titration Calorimetry-The apparent association constant for Sep15 binding to GT was determined using isothermal titration calorimetry. Protein solutions of Drosophila Sep15 and GT were dialyzed for 2 days against 25 mM Tris-HCl (pH 7.5), 200 mM NaCl, and 1 mM CaCl 2 . The concentrations of both proteins were determined using Beer's law. This experiment consisted of a series of 8-l injections of Sep15 into a cell containing 1.4 ml of GT equilibrated at 20°C. Heats associated with Sep15 binding to GT were measured by an isothermal titration calorimeter (Microcal). The apparent association constant derived from these data were determined using software included with the calorimeter.
Metal Ion Analysis-The concentrations of twenty biologically relevant metal ions were determined by inductively coupled argon plasma emission spectrometry at the Chemical Analysis Laboratory, University of Georgia. Protein samples dissolved in 50 mM sodium phosphate (pH 7.0), 300 mM NaCl, and 250 mM imidazole were analyzed. Metal ion concentrations were calculated by subtracting the ion content of the buffer from that of the protein sample.

Sep15, but Not SelM, Interacts with Human GT Homologues-Sep15
is a resident protein of the endoplasmic reticulum that has been previously shown to form a complex with GT when purified from rat and mouse liver (11). Two human GT homologues (GT1 and GT2) have been identified (36). These enzymes are highly homologous and display 83% sequence identity in their proposed C-terminal catalytic domains. However, their N-terminal non-catalytic domains are more diverse (49% sequence identity), and only GT1 displays measurable catalytic activity. To assess whether Sep15 functions as a protein co-factor that binds both GT isozymes, mouse Sep15 was co-expressed with PC-tagged GT1 or GT2 in NIH 3T3 cells. Sep15 co-immunoprecipitated with GT1 and GT2 (Fig. 1A), indicating that both human GT isozymes are capable of binding Sep15 in vivo. The reduced intensity of the Sep15 signal observed with GT2 is consistent with its lower level of expression as compared with that of GT1.
The recently described selenoprotein SelM (37) shares 31% sequence identity with Sep15 (Fig. 2, A and B). Both proteins encode a single selenocysteine residue. Primary sequence alignments illustrate that SelM has an elongated C terminus, whereas Sep15 possesses an extended cysteine-rich N terminus. The N-terminal sequence of Sep15 is conserved among functional Sep15 homologues and is conversely absent in sequence homologues of SelM (Fig. 2C). As observed with Sep15, selenoprotein SelM is also localized to the endoplasmic reticulum. Given the common intracellular location of Sep15 and SelM and their respective sequence homology, we speculated that SelM might also bind to GT. However, co-immunoprecipitation experiments indicate that mouse SelM does not form a complex with either human GT1 or GT2 (Fig. 1B).

Cysteine-rich Domain of Sep15 Mediates the Interaction with GT-
The absence of an N-terminal extension in SelM suggested that this region of Sep15 may potentially mediate protein-protein interactions with GT. To test this hypothesis, the cysteine-rich N-terminal extension of mouse Sep15 was fused to the N terminus of mouse SelM. When expressed in NIH 3T3 cells, this mouse protein chimera co-immunoprecipitates with human GT1 (Fig. 3). These data suggest a functional role for the N-terminal extension of Sep15 in the formation of a complex with GT.
Sequence analyses of the N-terminal domain of Sep15 proteins identified six conserved cysteine residues (Fig. 2C). Four of these six cysteine residues form a pair of CxxC motifs. To establish a structural and/or functional role for these cysteine residues and evaluate their potential involvement in constituting the GT-binding site, all cysteine residues were individually replaced with serine residues. Co-immunoprecipitation experiments demonstrate that formation of the Sep15-GT complex was abolished for each mutant (Fig. 4). These data indicate that the N-terminal extension of Sep15 forms a distinct cysteine-rich domain and that the structural integrity of this domain is required for Sep15 to form a complex with GT. Sep15 and GT Form a 1:1 Complex-One technical obstacle to investigating protein-protein interaction between Sep15 and GT is the naturally occurring selenocysteine residue encoded within the Sep15 gene. Selenocysteine-containing proteins are frequently difficult to express and purify. However, the Sep15 protein of Drosophila is not a selenoprotein and encodes two cysteine residues in place of the CxU motif (Fig. 2B). To examine the binding between Drosophila Sep15 and GT, both recombinant proteins were expressed and purified from Sf9 cells.
Formation of the complex between Sep15 and GT was analyzed by polyacrylamide gel electrophoresis under non-denaturing conditions (Fig. 5). Western blotting using antibodies specific for Sep15 detected a low molecular weight band corresponding to unbound Sep15 in addition to a higher molecular weight band that co-migrates with GT. Several minor low molecular weight bands were also detected, indicating that Sep15 may form homomultimers and/or is degraded under these experimental conditions. The band intensity of the Sep15-GT complex  continued to increase as the molar equivalents of Sep15 and GT were increased from 0.2 to 2.0, at which point the amount of unbound and GT-bound Sep15 became equivalent.
The stoichiometry of the Sep15-GT complex was precisely determined by isothermal titration calorimetry (Fig. 6). Sep15 forms a 1:1 complex with GT with an apparent K d value of 20 nM. The mixed species Sep15::SelM chimera also bound tightly to GT (supplemental Fig. S1). In contrast, no interactions were observed for SelM and Sep15 without its N-terminal cysteine-rich domain. These isothermal titration calorimetry results were confirmed by measuring the intensity of the strongest methyl resonance by one-dimensional 15 N-edited proton NMR spectroscopy (supplemental Fig. S2). A decrease in the strongest methyl resonance intensity in the proton NMR spectrum of the protein mixture in comparison to the added strongest methyl resonance intensities of the individual proteins indicates specific protein-protein interactions (38). Consistent with co-immunoprecipitation and isothermal titration calorimetry data, these NMR data indicate that both Sep15 and the Sep15::SelM chimera form complexes with GT. Conversely, SelM and Sep15 without its corresponding cysteine-rich domain do not bind to GT. Taken together these complementary techniques demonstrate that the cysteine-rich domain of Sep15 mediates that formation of a tight 1:1 complex with GT.
Formation of the Sep15-GT Complex Is Not Mediated by Coordinated Metal Ions-Four of the six conserved cysteine residues in the cysteinerich domain of Sep15 form two CxxC motifs, which are typically involved in metal ion coordination. To establish whether the cysteinerich domain of Sep15 does coordinate metal ions, we further analyzed Drosophila Sep15. When expressed in bacterial or insect cells, purified Sep15 protein was colorless. Moreover, UV-visible spectrophotometric analysis revealed only a single peak centered at 280 nm, indicating the absence of certain co-factors and iron-sulfur clusters. Sep15 protein samples were also analyzed for the presence of twenty biologically relevant metal ions by inductively coupled argon plasma emission spectrometry. Only Co 2ϩ was detected (0.24 molar equivalents), which is most likely non-specifically bound during protein purification by immobilized metal-affinity chromatography using cobalt-based resin. These data suggest that the cysteine-rich domain itself does not coordinate metal ions.
It has been previously reported that the catalytic activity of GT is dependent upon the presence of divalent cations (preferably Ca 2ϩ ) (39). Although metal ions were not detected in Sep15 protein preparations, it is possible that metal ions mediate the interaction between Sep15 and GT. To address this possibility, complex formation between Sep15 and GT was analyzed by polyacrylamide gel electrophoresis under non-denaturing conditions in the absence and presence of 1 mM Ca 2ϩ , Cu 2ϩ , Fe 2ϩ , and Zn 2ϩ . However, the Sep15-GT complex was observed independently of the presence of metal ions (data not shown). These results indicate that formation of the Sep15-GT complex is not mediated by coordinated metal ions.
Sep15 Does Not Affect Glucosylation of Thyroglobulin-The association of Sep15 with GT suggests that this selenoprotein may modulate the enzymatic activity of GT and/or be involved in assessing structural fidelity. To test this hypothesis, GT activity in the presence of Sep15 was analyzed by monitoring the amount of incorporated [ 14 C]glucose into denatured thyroglobulin. The presence of Sep15 did not affect the GT activity toward this protein (supplemental Fig. S3). These results indicate that, at least in the case of thyroglobulin, Sep15 does not modulate GT activity and is not required for glucosylation under these experimental conditions.

DISCUSSION
Selenoproteins have not been previously shown to participate in the quality control networks of the endoplasmic reticulum. The observation that Sep15 co-purifies with the essential regulator of the calnexin cycle FIGURE 4. Binding of Sep15 to GT is dependent upon the structural integrity of the cysteine-rich domain. NIH 3T3 cells were co-transfected with plasmid-encoded PCtagged human GT1 (GT1-PC) and the empty control vector pCI-neo or plasmid-encoded His-tagged mouse Sep15-U93C or Sep15-C38SU93C, Sep15-C49SU93C, Sep15-C52SU93C, Sep15-C67SU93C, Sep15-C70SU93C, and Sep15-C71SU93C mutants. Samples were immunoprecipitated (IP) using anti-PC antibodies, resolved by SDS-PAGE, and detected by Western blotting (WB) with anti-PC (first panel) and anti-His (second panel) antibodies. Cell lysates were also directly probed using anti-PC (third panel) and anti-His antibodies (fourth panel). suggests a potential mechanism for the chemopreventive effects of dietary selenium (11).
In this study, we characterized interactions between members of Sep15 and GT families using a series of complementary genetic and biophysical techniques and found that this selenoprotein interacts with both human GT isozymes. The 1:1 stoichiometry of this complex has an apparent K d value of 20 nM as determined by isothermal titration calorimetry using Sep15 and GT from Drosophila. In contrast, SelM (a sequence homologue of Sep15) does not interact with GT. Multiple sequence analyses identified a novel cysteine-rich domain that is conserved among the N termini of Sep15 proteins, which is not found in the sequences of SelM proteins. If the cysteine-rich domain of Sep15 is deleted, the truncated Sep15 protein no longer interacts with GT. However, when the cysteine-rich domain of Sep15 is fused to the N terminus of SelM, the resulting Sep15::SelM protein chimera forms a complex with GT.
The sequence of the cysteine-rich domain is highly conserved among Sep15 homologues from plants to humans (40) (Fig. 2C), suggesting that interactions between Sep15 and GT have been conserved throughout evolution. Mutating any one of these six conserved cysteine residues disrupts the interaction between Sep15 and GT. As four of these six cysteine residues form a pair of CxxC motifs, we speculated that these motifs might coordinate metal ions, such as iron or zinc. However, metal ion analysis of purified Sep15 protein samples did not identify specifically bound metal ions. It is more likely that these conserved cysteine residues form intramolecular disulfide bonds. Site-directed mutagenesis of cysteine residues may disrupt the proper arrangement of these disulfide bonds and compromise the structural integrity of the cysteine-rich domain, which in turn inhibits protein-protein interaction between Sep15 and GT. Taken together, these data unequivocally demonstrate that the cysteine-rich domain mediates the interaction between Sep15 and GT. However, the corresponding regions of GT that are involved in forming the binding surface for Sep15 remain to be established.
The fact that Sep15 does not possess the endoplasmic reticulum retention signal and its N-terminal part is responsible for maintaining Sep15 in the endoplasmic reticulum (41) suggests that its location in this cellular compartment is the result of the interaction with GT.
The biological significance of protein-protein interactions between Sep15 and GT within the endoplasmic reticulum is poorly understood. One possibility is that Sep15 modulates the enzymatic activity of GT by functioning as a protein co-factor that assists this folding sensor in assessing the structural fidelity of misfolded glycoproteins. However, our data show that GT-mediated labeling of thyroglobulin is not directly affected by Sep15. Sep15 appears to occur exclusively in the complex with GT, whereas GT is present in both selenoprotein-bound and selenoprotein-free form (11). Thus, it is possible that Sep15 is not required for the GT function per se but may be required for the recognition of glycoproteins with incorrectly formed disulfide bonds. It is an attractive possibility that the unusual CxU motif of Sep15 may participate in the reduction, isomerization, or oxidation of disulfide bonds of glycoprotein substrates of GT.