Escherichia coli NifS-like Proteins Provide Selenium in the Pathway for the Biosynthesis of Selenophosphate*

Selenophosphate synthetase (SPS), theselD gene product from Escherichia coli, catalyzes the biosynthesis of monoselenophosphate, AMP, and orthophosphate in a 1:1:1 ratio from selenide and ATP. Kinetic characterization revealed the K m value for selenide approached levels that are toxic to the cell. Our previous demonstration that a Se0-generating system consisting ofl-selenocysteine and the Azotobacter vinelandiiNifS protein can replace selenide for selenophosphate biosynthesisin vitro suggested a mechanism whereby cells can overcome selenide toxicity. Recently, three E. coli NifS-like proteins, CsdB, CSD, and IscS, have been overexpressed and characterized. All three enzymes act on selenocysteine and cysteine to produce Se0 and S0, respectively. In the present study, we demonstrate the ability of each E. coliNifS-like protein to function as a selenium delivery protein for thein vitro biosynthesis of selenophosphate by E. coli wild-type SPS. Significantly, the SPS (C17S) mutant, which is inactive in the standard in vitro assay with selenide as substrate, was found to exhibit detectable activity in the presence of CsdB, CSD, or IscS and l-selenocysteine. Taken together the ability of the NifS-like proteins to generate a selenium substrate for SPS and the activation of the SPS (C17S) mutant suggest a selenium delivery function for the proteins in vivo.

The highly reactive, reduced selenium compound monoselenophosphate is required for the insertion of selenium into selenium-dependent enzymes (1) and seleno-tRNAs (2). The selD gene product from Escherichia coli catalyzes the synthesis of monoselenophosphate, AMP, and orthophosphate in a 1:1:1 ratio from selenide and ATP (Reaction 1).
Ϫ ϩ AMP ϩ P i REACTION 1 Selenophosphate synthetase (SPS) 1 from E. coli (3) and the closely related enzyme from Haemophilus influenzae (4) have been characterized. Both enzymes are active in the presence of high levels of free selenide (1-5 mM) and dithiothreitol (8 -10 mM) that are included in the in vitro assay system employed routinely. Under these conditions, the apparent K m values for ATP and selenide are 1 mM and 20 M, respectively. Although the reactions are carried out under argon, any traces of contaminating oxygen tend to reduce the effective concentration of selenide. The standard use of high selenide levels in vitro is a partial solution of this problem. The K m value determined for selenide in the in vitro system is far above the optimal concentration (0.1-1 M) of selenium needed for the growth of various bacterial species and cultured mammalian cells. In fact, levels above 10 M are toxic for many bacterial species. The fact that the high reactivity and thus the toxicity of selenide in biological systems is even greater than that of sulfide is apparent from the greater extent of ionization at physiological pH of H 2 Se with a pK 1 ϭ 3.89 as compared with that of H 2 S with a pK 1 ϭ 7.5. Rate acceleration by ionized selenols of disulfide reduction and disruption of protein structures or destabilization of coordinated metal ions from metallothionines (5) can occur if selenide concentrations are not maintained at low and optimal levels. Potential candidates for control of selenium levels are the specific L-selenocysteinelyase enzymes that decompose selenocysteine to elemental selenium (Se 0 ) and alanine (6, 7) and a closely related enzyme D-selenocystine-lyase (8). The attractive possibility that these lyases also may serve normally to deliver Se 0 to SPS was suggested by previous studies in which it was shown that L-selenocysteine and the Azotobacter vinelandii NifS protein effectively replaced the high level of free selenide in the in vitro SPS assay (9). In fact, even though the normal substrate for A. vinelandii NifS is cysteine (10,11), the extent of utilization of selenocysteine in vitro was sufficient to promote a higher rate of selenophosphate formation by SPS than was observed with free selenide alone.
The A. vinelandii NifS protein and the selenocysteine-lyases are pyridoxal 5Ј-phosphate-dependent enzymes that catalyze the elimination of sulfur from L-cysteine and selenium from L-selenocysteine (10,11). Because in vivo concentrations of sulfur-containing compounds are a thousand-fold or more greater than their selenium analogs, proteins such as NifS presumably bind and metabolize L-cysteine preferentially. Thus, lyase proteins that are highly specific for L-selenocysteine as substrate may function as the actual selenium delivery proteins in vivo. In E. coli three NifS-like proteins were identified recently (12). All three proteins resemble A. vinelandii NifS in amino acid sequences and catalytic properties. Sequence alignments also revealed similarities between the Nterminal region amino acid sequence of the pig liver selenocysteine-lyase (6). Characterization of the three E. coli NifS-like proteins CsdB, CSD, and IscS revealed considerable differences in the degree of discrimination between L-cysteine and Lselenocysteine as substrates (12). The CsdB protein, although not entirely specific for the seleno-amino acid, substrate was 290 times more active on L-selenocysteine than on L-cysteine. This protein thus appears to be the E. coli counterpart of the mammalian selenocysteine-lyase.
In an attempt to identify residues that are essential for catalysis in the SPS enzyme, mutagenesis experiments have been performed (13,14). Certain amino acids in a glycine-rich potential ATP-binding site in the N-terminal region of the protein were selected as targets. Mutagenesis of cysteine 17, located in the potential ATP-binding site, to serine resulted in the complete loss of detectable SPS activity in both the in vitro selenide-dependent assay and the in vivo complementation of a selD lesion in E. coli strain MBO8 (13). The inability to detect SPS catalytic activity of the cysteine 17 mutant enzyme and the failure of this mutated enzyme to catalyze a positional isotope exchange reaction between the ␥-phosphoryl group and the ␤-phosphoryl group of ATP 2 supported a catalytic role for cysteine 17. However, a catalytic role was contradicted by the finding that substitution of selenocysteine for cysteine at position 17, which occurs normally in H. influenzae SPS, fails to elicit an increase in catalytic activity (4). In the present work, we report the ability of the CsdB, CSD, and IscS proteins to function both in vitro and in vivo as selenium delivery proteins to the E. coli wild-type SPS and also to SPS (C17S) mutant.

Methods
Complementation of the selD Mutant MBO8 -E. coli strain MB08 (16,17) transformed with plasmids containing either the SPS gene or the SPS (C17S) mutant gene was grown anaerobically at 37°C overnight in Luria broth containing 0.5% glucose. Cultures were observed after 24 -48 h for the production of H 2 , a product of an active formate dehydrogenase (FDH H )-hydrogenase H complex (also known as formate hydrogen-lyase) (18). Synthesis of the selenocysteine-containing FDH H depends on the availability of selenophosphate.
In parallel experiments, MBO8 cultures containing both plasmids were streaked out on Luria medium plus glucose agar plates and incubated anaerobically for 48 h at 37°C. After growth, all individual colonies were colorless. The plates were then overlaid with agar containing benzyl viologen, an electron acceptor for active FDH H that turns blue when reduced. Thus, any colonies appearing blue in color provide a test for the presence of FDH H (19).
[ 75 Se]Selenocysteine-labeled FDH H -Single colonies of MBO8, MBO8/SPS, and MBO8/SPS (C17S) were grown anaerobically in Luria broth ϩ 20 Ci 75 SeO 3 2Ϫ at 37°C for 24 h. Following growth, cells were harvested, resuspended in 100 mM potassium phosphate buffer, pH 7.2, and sonicated. Supernatants were analyzed for the presence of the selenocysteine containing FDH H by SDS-polyacrylamide gel electrophoresis and PhosphorImager detection of radioactivity.
Enzyme Assays-All assays were performed anaerobically three to six times and averaged for each reported value. The coupled SPS-lyase assay reaction mixtures contained 50 mM Tricine⅐KOH, pH 8.0, 8 mM MgCl 2, 20 mM KCl, 50 mM dithiothreitol, 200 M pyridoxal 5Ј-phosphate, 2 mM ATP, 10 mM L-selenocysteine, 10 M SPS, 0.2 Ci [8-14 C] ATP, and the indicated concentration of added lyase. Reaction mixtures were incubated at 37°C for 30 min, terminated by the addition of 1.2 N HClO 4 , and neutralized with KOH. A standard assay, in which selenide replaced L-selenocysteine and lyase protein, was performed at the same time. The nucleotides in the supernatant solutions were separated by chromatography on cellulose-polyethyleneimine thin layer sheets developed in 1.0 M formic acid and 0.5 M LiCl. Nucleotide spots detected by UV quenching were cut out, and radioactivity was measured by liquid scintillation spectroscopy.
The selenocysteine-lyase activity of each protein was measured in the absence of selenophosphate synthetase by quantitation of liberated H 2 Se after reaction with lead acetate as described previously (6). Reaction mixtures contained 50 mM Tricine⅐KOH, pH 8.0, 8 mM MgCl 2 , 20 mM KCl, 50 mM dithiothreitol, 200 M pyridoxal 5Ј-phosphate, 10 mM L-selenocysteine, and 2.5 M lyase. Reaction mixtures were incubated at 37°C for the indicated amounts of time, and reactions were quenched by the addition of lead acetate. The amount of H 2 Se produced was quantitated using a molar turbidity coefficient of colloidal PbSe at 400 nm of 1.18 ϫ 10 4 cm Ϫ1 M Ϫ1 .
The competition experiments with the CsdB (R379A) mutant were performed as described above for the coupled assay. A total of five assays was performed, each in the presence of a different concentration of CsdB (R379A) (0.25, 0.5, 1, 5, and 10 M). The assays performed with SPS (C17S) mutant were performed as described above.

RESULTS AND DISCUSSION
Selenophosphate Synthetase-and Lyase-coupled Assays-As previously reported, the high levels of free selenide that are used routinely for SPS can be replaced by the A. vinelandii NifS protein and Lselenocysteine in the in vitro SPS assay (9). Even though the normal substrate for NifS is cysteine, the ability to act also on the selenium analog makes it an effective selenium donor in vitro. In this coupled assay, an increase in selenophosphate formation was observed, indicating that Se 0 generated by the NifS protein is a more effective substrate than free selenide (9). Recently, three E. coli genes encoding the NifSlike proteins CsdB, CSD, and IscS were cloned, and the gene products were characterized (12). All three proteins exhibit lyase activity on L-cysteine and L-selenocysteine as substrates and produce sulfane sulfur, S 0 , or Se 0 respectively. One of the proteins, CsdB, utilizes selenocysteine more effectively than cysteine, indicating a more specific role for Se 0 generation. To test the relative abilities of the three NifS-like proteins from E. coli to provide Se 0 to SPS, assays were performed with L-selenocysteine and CsdB, CSD, or IscS in place of high levels of free selenide. SPS catalytic activity was supported in the presence of each NifS-like protein (Fig. 1A). The determined activity of SPS was directly related to the concentration of lyase used in each assay. When lyase concentrations were increased above 0.025 M, the amount of AMP generated by SPS was slightly higher than AMP formed in assays using 1.5 mM free selenide as substrate (data not shown). However, at lyase concentrations above 0.05 M, the amount of selenide generated was much higher (Fig. 1B) as compared with the amount of AMP generated by SPS. (In the absence of an acceptor (SPS) and in free solution the product of the lyase reaction is referred to as selenide.) The inability to detect increased SPS activity in the presence of higher concentrations of lyase indicates SPS is saturated with selenium. SPS may have an inherently low catalytic rate because of a possible rate-limiting step in the reaction mechanism. The rate-limiting step may occur after selenium binding and could include bound ADP hydrolysis (20), product release (AMP) or an enzyme conformational change.
In Vivo Selenophosphate Synthetase Assay-An in vivo test to detect activity of the formate hydrogen-lyase complex in E. coli depends on the evolution of gas (H 2 ) from formate generated during anaerobic growth on glucose (18). This test was later extended to detect in vivo activity of the selD gene (13,14). The selD gene product, SPS, forms the reactive selenium donor compound, selenophosphate, which is required for the production of selenocysteyl-tRNA Sec (21). The synthesis of an active selenocysteine containing formate dehydrogenase depends on the availability of selenocysteyl-tRNA Sec . A more sensitive in vivo test, devised for the detection of redox-active enzymes such as formate dehydrogenase, involves the ability of the enzyme to reduce benzyl viologen to a blue color when agar containing the dye was overlaid on bacterial colonies grown anaerobically on glucose agar plates. E. coli MBO8, a selD mutant strain, cannot make selenophosphate and lacks an active selenocysteine containing formate dehydrogenase. Therefore, MBO8 fails to produce H 2 when grown in glucose media (16,17) and is unable to reduce benzyl viologen, resulting in colonies that do not turn blue. Introduction of a plasmid bearing a wild-type selD gene into MBO8 complements the selD lesion and allows the synthesis of an active selenocysteine containing formate dehydrogenase that can be detected in both in vivo assays (Data not shown) (22).
The SPS (C17S) mutation failed to complement the selD lesion in MBO8 as judged by lack of visible H 2 production from glucose during anaerobic growth (13,14). The purified mutant enzyme was also unable to catalyze the formation of selenophosphate from 1.5 mM selenide and ATP in the usual in vitro assay (13,14). To re-evaluate the activity of the SPS (C17S) mutant, MBO8 cells were transformed with a plasmid bearing the mutant gene. In the benzyl viologen overlay assay the SPS (C17S) mutant gene was able to complement the selD lesion, and transformed MBO8 colonies turned blue in color (data not shown). Although reduction of benzyl viologen is a more sensitive test for FDH H activity than production of H 2 gas in glucose media, the extent of blue color development is more qualitative than quantitative. Nevertheless, the benzyl viologen assay is a rapid and selective assay for determining formate dehydrogenase activity (19). MBO8 colonies bearing either the wild-type SPS or the mutant SPS (C17S) gene turned blue in the assay.
To compare the in vivo activities of wild-type SPS and SPS (C17S), MBO8 cells containing plasmids bearing either gene were grown anaerobically in the presence of 75 SeO 3 2Ϫ . After growth, cell extracts were prepared and examined for the presence of [ 75 Se]selenocysteine containing FDH H (Fig. 2). MBO8 cells alone are unable to incorporate selenocysteine into FDH H . However, the selD lesion in MBO8 cells can be complemented either by a plasmid bearing the SPS gene or a plasmid bearing the SPS (C17S) mutant gene, thus restoring incorporation of selenocysteine into FDH H . Quantitation of the amount of radioactivity in FDH H in response to the two gene products (Fig. 2) revealed that the level of selenophosphate synthesis by the SPS (C17S) gene product was sufficient to allow incorporation of 44% as much 75 Se into FDH H as observed with the wild-type SPS. The fact that the activity of the SPS (C17S) mutant could be detected in vivo is indicative of an essential component present in E. coli that is absent in our in vitro assay containing purified SPS and free selenide. Competition Experiments-The low amount of lyase required for detectable SPS activity suggests a complex may form between a NifSlike protein and SPS. This complex would allow the direct delivery of selenium as a putative Se 0 to the active site of SPS. As a means of detecting complex formation with a NifS-like protein and SPS, competition assays were performed. As a potential competitor, a catalytically inactive mutant CsdB (R379A), 3 which is unable to bind L-selenocysteine, was added to assay mixtures containing CsdB(1 M), L-selenocysteine (10 mM), and SPS (10 M). If SPS and CsdB actually bind to form a complex, then the addition of CsdB (R379A) would displace CsdB. To observe the possible competitive capability of CsdB (R379A), reaction mixtures were prepared that contained 0.25, 0.5, 1, 5, and 10 M CsdB (R379A). The SPS selenium-dependent hydrolysis of ATP to AMP was measured in each sample. Over the range of added CsdB (R379A) examined, the observed rate of AMP formation was unaffected (data not shown), indicating that a 5-10-fold molar excess of this mutant is ineffective as a competitor of putative complex formation between SPS and the active delivery protein. Although no decrease in SPS activity was observed in the presence of excess CsdB (R379A), the substitution of alanine for arginine, which is known to affect selenocysteine binding, might also affect binding to SPS (23,24). Because wild-type CsdB has both L-cysteine desulfurase and L-selenocysteine-lyase activities, it is possible that binding of L-selenocysteine induces a conformational change that allows complex formation with SPS, and this change cannot occur in the mutant.
Re-evaluation of the SPS (C17S) Mutant-Previously, it was shown that the SPS (C17S) mutant enzyme is unable to catalyze the selenide dependent conversion of ATP to AMP (13). However, the ability of the SPS (C17S) mutant to complement MBO8 in the in vivo benzyl viologen overlay assay and incorporate selenocysteine into FDH H suggests either free selenide in the in vitro assay does not mimic the actual in vivo substrate or, alternatively, an additional factor is needed for selenophosphate formation and is supplied by MBO8. To determine whether a selenium delivery protein might be the factor required, the SPS (C17S) mutant enzyme was assayed in the presence of a selenocysteinelyase protein. Fig. 3A shows that replacement of free selenide with 10 M CsdB, CSD, or IscS and L-selenocysteine results in detectable activity in the SPS (C17S) in vitro assay. Fig. 3B shows that in the presence of a lyase (IscS), SPS (C17S) is saturated with Se 0 and activity is detected. Maximum activity is observed when concentrations of IscS approach one-half or greater the concentration of SPS (C17S). In contrast, maximum activity of wild-type SPS is observed in the presence of a selenium delivery protein at concentrations as low as 0.25 M. The ability to detect catalytic activity of the SPS (C17S) mutant further supports the involvement of a lyase protein in selenophosphate biosynthesis and provides additional evidence that cysteine-17 of SPS does not have a direct catalytic role.
Biological roles of NifS proteins include the mobilization of sulfur [S 0 ] from cysteine to a specific protein acceptor. The mobilized sulfur can be distributed throughout the cell for iron-sulfur cluster assembly such as those found in nitrogenase (10,11), FNR, SoxR (25,26), and apodihydroxy acid dehydratase (27). NifS-like proteins have been identified in several organisms (28), and in some sources such as E. coli (29,30), H. influenzae (31), and Synechocystis (32,33) there are three nifS-like genes. Additional in vivo functions of NifS-like proteins including tRNA sufurtransferase activity in E. coli (34), NAD biosynthesis in Bacillus subtilis (35) and E. coli (36), tRNA splicing in Saccharomyces cerevisiae (37), and the biosynthesis of sulfur-containing cofactors such as biotin and thiamin have been determined (36,38). These additional functions, together with the identification of multiple NifS-like proteins support roles for NifS-like proteins that are not limited to iron-sulfur cluster assembly.
The generation of Se 0 by the NifS-like proteins reflects an additional in vivo function in which these enzymes participate as components of a selenium delivery system for the biosynthesis of selenoproteins and selenium modification of tRNA nucleosides (Fig. 4). The chemical similarity between selenium and sulfur allows selenium to enter bacterial metabolism via the cysteine biosynthetic pathway where it can be incorporated into free selenocysteine (39). Free selenocysteine can be inserted into proteins nonspecifically, in place of cysteine, or a NifS-like protein can utilize it to generate selenium for SPS. The biological relevance of the NifS-like proteins in selenophosphate biosynthesis is particularly evident in the case of the SPS (C17S) mutant. Catalytic activity of this mutated SPS enzyme could be detected in the presence of CsdB, CSD, or IscS (Fig. 3). In the in vitro coupled assays, all three E. coli NifS-like proteins were able to function well as selenium delivery proteins with SPS. However, in vivo concentrations of sulfur-containing compounds are much higher compared with their selenium analogs. For the E. coli NifS-like proteins to be effective selenium delivery proteins under normal conditions additional proteins may be involved to assist in the discrimination between cysteine and selenocysteine. Future work will be focused on identifying additional cellular components and proteins that participate in selenium metabolism and selenophosphate biosynthesis. Clearly, generation of Se 0 at or in proximity to SPS may be the mechanism whereby the essential substrate of the enzyme is made available and the obstacle of free selenide toxicity is avoided.