A nifS-like Gene, csdB, Encodes anEscherichia coli Counterpart of Mammalian Selenocysteine Lyase

Selenocysteine lyase is a pyridoxal 5′-phosphate (PLP)-dependent enzyme that catalyzes the exclusive decomposition of l-selenocysteine tol-alanine and elemental selenium. An open reading frame, named csdB, from Escherichia coli encodes a putative protein that is similar to selenocysteine lyase of pig liver and cysteine desulfurase (NifS) of Azotobacter vinelandii. In this study, the csdB gene was cloned and expressed inE. coli cells. The gene product was a homodimer with the subunit M r of 44,439, contained 1 mol of PLP as a cofactor per mol of subunit, and catalyzed the release of Se, SO2, and S from l-selenocysteine,l-cysteine sulfinic acid, and l-cysteine, respectively, to yield l-alanine; the reactivity of the substrates decreased in this order. Although the enzyme was not specific for l-selenocysteine, the high specific activity for l-selenocysteine (5.5 units/mg compared with 0.019 units/mg for l-cysteine) supports the view that the enzyme can be regarded as an E. coli counterpart of mammalian selenocysteine lyase. We crystallized CsdB, the csdB gene product, by the hanging drop vapor diffusion method. The crystals were of suitable quality for x-ray crystallography and belonged to the tetragonal space group P43212 with unit cell dimensions of a = b = 128.1 Å and c = 137.0 Å. Consideration of the Matthews parameter Vm (3.19 Å3/Da) accounts for the presence of a single dimer in the crystallographic asymmetric unit. A native diffraction dataset up to 2.8 Å resolution was collected. This is the first crystallographic analysis of a protein of NifS/selenocysteine lyase family.

Selenocysteine lyase (SCL) 1 (EC 4.4.1.16) and cysteine desulfurase (commonly referred to as NifS) are pyridoxal 5Ј-phosphate (PLP)-dependent enzymes that catalyze the same type of reaction, i.e. the removal of a sulfur or selenium atom from L-cysteine or L-selenocysteine to produce L-alanine. NifS acts on both L-cysteine and L-selenocysteine (1,2). Several enzymes participating in sulfur metabolism also act on the selenium analogs of the substrates (3). In contrast, SCL exclusively decomposes L-selenocysteine. Selenium is specifically metabolized by such enzymes as selenophosphate synthetase (4), selenocysteine synthase (5), and selenocysteine methyl transferase (6). Discrimination of selenium from sulfur is important for establishing the role of selenium as an essential trace element in mammals and other organisms (7). SCL of pig liver (8) was characterized as the first enzyme that specifically acts on a selenium-containing substrate. We have found that the peptide sequences obtained from the proteolysate of SCL are similar to those of NifS proteins. 2 However, a gene encoding SCL has not been cloned yet, and the physiological role of the enzyme remains to be investigated.
Selenoproteins such as formate dehydrogenase from Escherichia coli contain selenocysteine residues (9,10). Selenocysteyl-tRNA Sec is required for the biosynthesis of these selenoproteins (5,11,12). Selenophosphate is a highly reactive selenium compound, and serves as a selenium donor for the selenocysteyl-tRNA Sec production (4,7,12). Selenophosphate is synthesized from selenide and ATP, which is catalyzed by selenophosphate synthetase (13,14). Recently, Lacourciere and Stadtman (2) found that the replacement of selenide by NifS and L-selenocysteine in an in vitro selenophosphate synthetase assay resulted in an increased rate of formation of selenophosphate, indicating that selenium derived from L-selenocysteine by the action of NifS serves as a better substrate than selenide for selenophosphate synthetase.
In Azotobacter vinelandii, NifS functions in nitrogen fixation by supplying sulfur to stabilize or repair the Fe-S cluster of the nitrogenase component protein (15). NifS homologs also occur in many nondiazotrophic procaryotes, including E. coli (16,17) and Bacillus subtilis (18), and in eucaryotes, including Saccharomyces cerevisiae (19), Caenorhabditis elegans (20), mice (21), and humans. 2 These NifS homologs are proposed to play a general role in the mobilization of sulfur for Fe-S cluster synthesis (22). However, the exact roles of the nifS-like genes in these non-nitrogen-fixing organisms have not been clarified, and it is possible that some of these NifS homologs act physiologically as selenocysteine-specific enzymes (e.g. SCL) to facilitate the selenophosphate synthesis, as proposed by Lacourciere and Stadtman (2).
The E. coli genome contains three genes with sequence homology to nifS. Two enzymes, IscS (17) and cysteine sulfinate desulfinase (CSD) ( can deliver the sulfur from L-cysteine for the in vitro synthesis of the Fe-S cluster of dihydroxyacid dehydratase from E. coli (17). CSD exhibits both selenocysteine lyase and cysteine desulfurase activities in addition to cysteine sulfinate desulfinase activity, and the enzyme is distinct from A. vinelandii NifS in its amino acid sequence, absorption spectrum, and lack of cysteine residues catalytically essential for the decomposition of L-selenocysteine (16). Neither enzyme shows strict specificity for L-selenocysteine, and both act on L-cysteine. Thus, we have explored the possibility that the last nifS homolog (csdB) 3 mapped at 37.9 min (23) in the chromosome encodes SCL, which plays a crucial role in selenophosphate synthesis. We have isolated the gene product (CsdB), studied its enzymatic properties, and carried out preliminary x-ray crystallographic studies.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and other DNA modifying enzymes were purchased from New England Biolabs (Beverly, MA) and Takara Shuzo (Kyoto, Japan); molecular weight markers for SDS-PAGE and gel filtration were from Amersham Pharmacia Biotech (Uppsala, Sweden) and Oriental Yeast (Tokyo, Japan); oligonucleotides were from Biologica (Nagoya, Japan); Gigapite was from Seikagaku Corporation (Tokyo, Japan); DEAE-Toyopearl, Phenyl-Toyopearl and Butyl-Toyopearl were from Tosoh (Tokyo, Japan). L-Selenocystine was synthesized as described previously (24). L-Selenocysteine was prepared from L-selenocystine according to the previous method (8). The Kohara/Isono miniset clone No. 430 (25) was a kind gift from Dr. Yuji Kohara, National Institute of Genetics, Japan. All other chemicals were of analytical grade.
Cloning of the csdB Gene-The DNA fragment containing csdB was cloned from the chromosomal DNA of E. coli K-12 ICR130 by polymerase chain reaction in a manner identical to that used for cloning of csdA (16). Oligonucleotide primers used were 5Ј-GGAATTCAGGAGGTGC-CATATGATTTTTTCCGTCGAC-3Ј (upstream) and 5Ј-CCCAAGCTTA-TCCCAGCAAACGGTG-3Ј (downstream); underlining indicates EcoRI and HindIII sites, respectively, and bold face indicates a putative ribosome binding sequence. The polymerase chain reaction product was ligated into pUC118, and then the resultant expression plasmid, pCSDB, was introduced into E. coli JM109 competent cells.
Enzyme Assays-The enzyme was assayed in 0.12 M Tricine-NaOH buffer at pH 7.5. The enzymatic activities toward L-selenocysteine and L-cysteine were measured with lead acetate as described previously (16). The previously reported value (8) for a molar turbidity coefficient of PbSe at 400 nm was corrected as 1.18 ϫ 10 4 M Ϫ1 ⅐cm Ϫ1 , and this value was used in this study. Sulfite produced from L-cysteine sulfinic acid was determined with fuchsin (26). Production of alanine from substrates was determined with a Beckman 7300 high performance amino acid analyzer (Beckman Coulter, Fullerton, CA). Specific activity was expressed as units/mg of protein, with 1 unit of enzyme defined as the amount that catalyzed the formation of 1 mol of the product in 1 min.
Purification of the csdB Gene Product (CsdB)-Purification was carried out at 0 -4°C, and potassium phosphate buffer (KPB) (pH 7.4) was used as the buffer throughout the purification. E. coli JM109 cells harboring pCSDB were grown in 9 liters of LB medium containing 200 g/ml ampicillin and 1 mM isopropyl-1-thio-␤-D-galactopyranoside at 37°C for 16 h. The cells were harvested by centrifugation, suspended in 10 mM KPB, and disrupted by sonication. The cell debris was removed by centrifugation, and the supernatant solution was fractionated by ammonium sulfate precipitation (25-50% saturation). The enzyme was dissolved in 10 mM KPB and dialyzed against the same buffer. The enzyme was applied to a DEAE-Toyopearl column (3 ϫ 15 cm) equilibrated with the same buffer. After the column was washed with the same buffer, the enzyme was eluted with a 0.8-liter linear gradient of 0 -0.25 M NaCl in the buffer. The active fractions were pooled (110 ml) and concentrated by ultrafiltration through an Advantec UP-20 membrane (Advantec, Naha, Japan). The enzyme was dialyzed against 10 mM buffer containing 0.65 M ammonium sulfate and applied to a Phenyl-Toyopearl column (3 ϫ 15 cm) equilibrated with the same buffer. The enzyme was eluted with a 0.7-liter linear gradient of 0.65-0.3 M ammonium sulfate in the buffer, and the active fractions were pooled and concentrated as above. The enzyme was dialyzed against 10 mM buffer and applied to a Gigapite column (3 ϫ 10 cm) equilibrated with the same buffer. The enzyme was eluted with a 1-liter linear gradient of 10 -150 mM KPB, and the active fractions were collected and concentrated. The final preparation was further concentrated with Centriprep-10 (Millipore, Bedford, MA) to a volume of 2.7 ml.
Purification of CSD and IscS-Purification of CSD from E. coli JM109 transformed with a plasmid pCSD1 containing the csdA gene was performed as described previously (16). Expression and purification of recombinant IscS will be described elsewhere. 4 Briefly, the iscS gene was amplified by polymerase chain reaction with the Kohara miniset clone No. 430 (25) as a template and inserted into the NdeI and HindIII sites in pET21a (Novagen, Madison, WI) to yield pEF1. IscS was isolated from BL21 (DE3) pLysS cells harboring pEF1 by sonication, ammonium sulfate fractionation, and chromatography with Phenyl-Toyopearl, DEAE-Toyopearl, Butyl-Toyopearl, Gigapite, and Superose 12 (Amersham Pharmacia Biotech, Uppsala, Sweden) columns.
Analytical Methods-Protein was quantified by the Bradford method (27) using Protein Assay CBB solution (Nacalai Tesque, Kyoto, Japan) with bovine serum albumin as a standard. The concentration of the purified enzyme was determined with the value ⑀ M ϭ 4.8 ϫ 10 4 M Ϫ1 ⅐cm Ϫ1 at 280 nm, which was calculated from the content of tyrosine, tryptophan, and cysteine (28). The subunit and the native M r of CsdB were determined by SDS-PAGE (29) and gel filtration with Superdex 200 (Amersham Pharmacia Biotech, Uppsala, Sweden), respectively. The PLP content of the enzyme was determined fluorometrically with KCN according to the method of Adams (30).
Crystallography-Crystals of CsdB were grown by the hanging drop vapor diffusion method. Each droplet was prepared by mixing 5 l of 20 mg/ml enzyme in 10 mM KPB (pH 7.4) with an equal volume of each reservoir solution of the Crystal Screen™ (Hampton Research, CA) initially and of a modified reservoir solution subsequently. The yellow crystals of CsdB were mounted in glass capillaries with the crystallographic c* axis along the rotation axis of the spindle and subjected to x-ray experiments. Native data for structure determination were collected at 20°C with a Rigaku R-AXIS IIC imaging plate detector using double focusing mirror-monochromated CuK ␣ radiation that was generated with a 0.3-mm focal cup of an x-ray generator RU300 (Rigaku, Tokyo, Japan) operated at 40 kV and 100 mA. The crystal-to-detector distance was set to 130.0 mm. Data reduction was carried out using the R-AXIS IIC software package.

RESULTS AND DISCUSSION
Cloning and Expression of the csdB Gene and Purification of the Product-For the production of a large amount of CsdB, expression plasmids were constructed as described under "Experimental Procedures" with chromosomal DNA isolated from E. coli K-12. The nucleotide sequence of csdB in the expression vector (pCSDB) was confirmed to be identical with that registered in GenBank ™ accession number D90811 (open reading frame o320#17). The clone provided overexpression of the cloned gene: about 10% of the total protein in the extract of E. coli JM109 recombinant cells. In the representative purification (Table I), about 8 mg of homogeneous preparation of CsdB was obtained per liter of culture.
Physical Characterization-CsdB provided a single band corresponding to the M r of 43,000 on SDS-PAGE (Fig. 1). The N-terminal sequence of the purified enzyme, MIFSVDKVRA, agreed with that deduced from the nucleotide sequence of csdB.
The M r of the native enzyme was determined to be 88,000 by gel filtration. Consequently, the enzyme is a dimer composed of two identical subunits. The spectrophotometric properties of the enzyme were very similar to those of CSD with an absorption maximum at 420 nm ( Fig. 2) at pH 7.4. No significant changes in the absorption spectrum were observed in the range of pH 6 -8. This absorption peak is characteristic of PLP enzymes, which contain the cofactor bound to the ⑀-amino group of a lysine residue at the active site. However, CsdB is distinct from either of two A. vinelandii proteins, NifS and IscS, and also from IscS of E. coli, all of which have an absorption maximum around 390 nm (15,17,22). Reduction with sodium borohydride resulted in a decrease in the absorption peak at 420 nm with a concomitant increase in the absorbance at 335 nm (Fig. 2). This result is consistent with that this is a PLP enzyme. The PLP content of CsdB was determined to be 1.0 mol per mol of subunit by the fluorometric method (30). Catalytic Activity and Substrate Specificity-CsdB catalyzed the removal of a substituent at the ␤-carbon of L-selenocysteine, L-cysteine, and L-cysteine sulfinic acid to yield L-alanine. The production of elemental selenium and elemental sulfur from L-selenocysteine and L-cysteine, respectively, in the reaction was confirmed in the same manner as reported previously (31). The optimal pH value for the removal of selenium from L-selenocysteine was between 6.5 and 7.5 in Tricine-NaOH or Mes buffer. The substrate specificity of the enzyme is summarized in Table II; L-selenocysteine was the best substrate followed by L-cysteine sulfinic acid and L-cysteine, in that order. The specific activity of CsdB on L-selenocysteine (5.5 units/mg) was comparable with that of CSD and IscS (Table III) but was about 7 times lower than that of SCL (37 units/mg) (8). The cysteine desulfurase activity of CsdB was about 2 and 5% of that of CSD and IscS, respectively, at a substrate concentration of 12 mM (Table III). The specific activity of CsdB for L-cysteine was about 1/290 of the activity with L-selenocysteine (Table  III). This value is much lower than those of CSD and IscS (Table III). In contrast with CsdB, A. vinelandii NifS favors L-cysteine as a substrate over its selenium analog (2). CsdB acted on L-cystine, L-selenocystine, and L-aspartic acid, although at extremely low rates (Ͻ0.08% of the rate for L-selenocysteine) (Table II).
Crystallization and Preliminary X-ray Characterization-CsdB was crystallized at 25°C within 2 days by hanging drop vapor diffusion against a 100 mM cacodylate solution (pH 6.8) containing 1.4 M sodium acetate, which corresponds to the solution No.7 in the Crystal Screen ™ . The crystals were also obtained in 100 mM KPB (pH 6.8) containing 1.4 M sodium acetate and 10 M PLP, and these conditions were further used for the crystallization of the enzyme. The yellow crystals (0.5 ϫ 0.5 ϫ 0.4 mm 3 ) had tetragonal-bipyramidal shapes (Fig. 3). They were grown in amorphous debris, which was removed  c Discrimination factor was calculated from the specific activity of the enzymes for L-selenocysteine divided by that for L-cysteine. Activity was measured in the reaction mixture containing 120 mM Tricine-NaOH (pH 7.5), 50 mM dithiothreitol, 0.2 mM PLP, and 12 mM substrate. from the crystals before they were sealed in thin-walled glass capillaries.
The space group of the CsdB was P4 3 2 1 2 with the cell dimensions of a ϭ b ϭ 128.1 Å, and c ϭ 137.0 Å. The assumption that a single dimer (89 kDa) exists in the asymmetric unit of the crystal gives a V m value of 3.19 Å 3 /Da, which is equivalent to a solvent content of 62%. These values lie within the range of values commonly found for proteins (32). A set of native data was collected to 2.8 Å resolution on a Rigaku R-AXIS IIC using 1.5°oscillation over a range of 45°(94.2% complete with 23,770 independent reflections). The R merge value for the intensity data was 7.22%. The data collection statistics obtained for the native CsdB crystals are given in Table IV. The x-ray crystal structure determination of the enzyme is now under way by the multiple isomorphous replacement method.
We also obtained crystals of CSD at 25°C by hanging drop vapor diffusion against a 100 mM sodium acetate solution (pH 4.6) containing 200 mM ammonium acetate and 30% (w/v) polyethylene glycol 4000. However, these crystals were small and not suitable for x-ray analysis. Further optimization of crystallization conditions by changing pH, polyethylene glycol concentration, and salt has resulted in little improvement.
Comparison with Other PLP-dependent Enzymes-Grishin et al. (33) classified PLP enzymes into seven distinct fold types on the basis of primary structure, secondary structure prediction, and biochemical function. NifS proteins have been classified as "aminotransferases class V" in the fold type I together with serine-pyruvate aminotransferase (EC 2.6.1.51), phosphoserine aminotransferase (EC 2.6.1.52), isopenicillin N epimerase, and the small subunit of the soluble hydrogenase. Recently, three-dimensional structures of phosphoserine aminotransferases from Bacillus circulans sbsp. Alkalophilus 5 and E. coli 6 were solved and deposited in the Protein Data Bank, Brookhaven National Laboratory, with the codes 1BT4 and 1BJN, respectively. Comparison of the structures of phosphoserine aminotransferases with that of CsdB will contribute to the understanding of how the related proteins confer separate reaction specificities on the same coenzyme.
The reaction of CsdB shares some common features with that of other PLP-dependent enzymes such as aspartate ␤-decarboxylase (EC 4.1.1.12) (34), kynureninase (EC 3.7.1.3) (35), and SCL. These enzymes catalyze removal of ␤-substituent from the substrate to form alanine. None of their structures have been solved. The solution of the three-dimensional structure of CsdB would contribute to the understanding of the mechanisms of these PLP-dependent enzymes.
A Possible Role of CsdB in Vivo-Genome sequencing projects have revealed that homologs of A. vinelandii nifS are widespread throughout nature and that some organisms contain more than one copy of a nifS homolog (16,22). Some of the "NifS-like proteins" characterized so far prefer L-cysteine to L-selenocysteine, and some of them show the opposite preference. Further experiments will need to be done to determine whether putative NifS-like proteins can play a role in Fe-S cluster assembly.
Lacourciere and Stadtman (2) have pointed out that in vivo concentrations of sulfur-containing compounds are on the order of a thousand times greater than those of their selenium analogs (36). Thus, enzymes showing higher activity toward Lcysteine, such as A. vinelandii NifS, will preferentially utilize L-cysteine over L-selenocysteine in vivo (2). Therefore, it may be reasonable to assume that enzymes which are specific toward L-selenocysteine probably function as a physiological selenide delivery system in E. coli. Although CsdB is not strictly specific to selenocysteine, its discrimination factor (290 times over the activity on cysteine) is much higher than those of other NifS homologs of E. coli. Accordingly, the enzyme can be regarded as an E. coli counterpart of mammalian selenocysteine lyase. It would be particularly intriguing to determine whether CsdB is more effective than CSD and IscS as a selenide delivery protein in the formation of selenophosphate catalyzed by E. coli selenophosphate synthetase. The in vivo function of CsdB is now being studied by disrupting its gene.