Molecular Basis of Cysteine Biosynthesis in Plants

In plants, cysteine biosynthesis plays a central role in fixing inorganic sulfur from the environment and provides the only metabolic sulfide donor for the generation of methionine, glutathione, phytochelatins, iron-sulfur clusters, vitamin cofactors, and multiple secondary metabolites. O-Acetylserine sulfhydrylase (OASS) catalyzes the final step of cysteine biosynthesis, the pyridoxal 5′-phosphate (PLP)-dependent conversion of O-acetylserine into cysteine. Here we describe the 2.2 Å resolution crystal structure of OASS from Arabidopsis thaliana (AtOASS) and the 2.7 Å resolution structure of the AtOASS K46A mutant with PLP and methionine covalently linked as an external aldimine in the active site. Although the plant and bacterial OASS share a conserved set of amino acids for PLP binding, the structure of AtOASS reveals a difference from the bacterial enzyme in the positioning of an active site loop formed by residues 74-78 when methionine is bound. Site-directed mutagenesis, kinetic analysis, and ligand binding titrations probed the functional roles of active site residues. These experiments indicate that Asn77 and Gln147 are key amino acids for O-acetylserine binding and that Thr74 and Ser75 are involved in sulfur incorporation into cysteine. In addition, examination of the AtOASS structure and nearly 300 plant and bacterial OASS sequences suggest that the highly conserved β8A-β9A surface loop may be important for interaction with serine acetyltransferase, the other enzyme in cysteine biosynthesis. Initial protein-protein interaction experiments using AtOASS mutants targeted to this loop support this hypothesis.

Plants assimilate inorganic sulfur as sulfate from the soil or as sulfur dioxide and hydrogen sulfide from the atmosphere (1)(2)(3). Incorporation of sulfide into cysteine in chloroplasts, mitochondria, and the cytoplasm is the final metabolic step in environmental sulfur assimilation. Cysteine is the metabolic sulfide donor for all cellular components containing reduced sulfur. In addition to its role in protein structure, cysteine is a precursor of methionine, glutathione, phytochelatins, iron-sulfur clusters, vitamin cofactors, and multiple secondary metabolites.
Two enzymes catalyze the formation of cysteine in plants and bacte-ria ( Fig. 1A) (2). Serine acetyltransferase (SAT 4 8) uses pyridoxal 5Ј-phosphate (PLP) as a cofactor to catalyze the formation of cysteine from O-acetylserine and sulfide. Association of these two enzymes into an assembly called the cysteine synthase complex coordinates sulfate assimilation and modulates cysteine synthesis at the cellular level (4 -6). This macromolecular assembly contains one SAT hexamer and two OASS dimers (4), and when the two enzymes are complexed, SAT activity increases and OASS activity decreases (6). Environmental stresses alter the expression and enzymatic activity of OASS. Although OASS is constitutively expressed, sulfur, nitrogen, and carbon starvation conditions and abiotic stresses like salt and heavy metal exposure increase expression of OASS in Arabidopsis thaliana (AtOASS) (7,8). Overexpression of OASS in transgenic plants improves heavy metal tolerance (8), increases cysteine biosynthesis in response to sulfur stress (9), and provides protection against oxidative stress (10,11). In Arabidopsis, three tissue-specific OASS isoforms catalyze the formation of cysteine from sulfide and O-acetylserine (12)(13)(14)(15). Likewise, Spinacia oleracea (spinach) expresses a similar set of isoforms (16 -18). The OASS from different plant organelles share ϳ40% amino acid sequence identity with the bacterial enzymes (Fig. 1B).
The chemical reaction mechanism of OASS is well studied (19). The enzyme active site contains PLP linked to a lysine as an internal Schiff base (Fig. 1C, step 1). Binding of O-acetylserine displaces the lysine (Fig.  1C, step 2), initiating the first half-reaction yielding an ␣-aminoacrylate intermediate linked to PLP (Fig. 1C, step 3). The second half-reaction involves sulfide addition to the intermediate, thereby generating an external aldimine with the amino acid (Fig. 1C, step 4). The active site lysine reacts with this intermediate, releasing cysteine and regenerating the Schiff base (20,21). Site-directed mutagenesis of Salmonella typhimurium OASS (StOASS) confirmed the importance of the lysine (22) and the contribution of an active site serine to stabilizing the pyridoxal ring in the reaction (23). Determination of the three-dimensional structure of StOASS and subsequent elucidation of the crystal structure of the StOASS K41A mutant complexed with methionine linked to PLP as an external aldimine showed that an active site loop changes conformation to interact with the ligand (24,25). Resembling O-acetylserine, methionine is a substrate analog that reacts with PLP but does not complete the reaction cycle. To date, an extensive investigation of the functional roles of OASS active site residues involved in either PLP or substrate binding has not been undertaken.
As the first step toward understanding how the plant enzymes involved in cysteine biosynthesis function and interact with each other, we have determined the three-dimensional structures of AtOASS and the AtOASS K46A mutant. Using these structures as a guide, we examined the role of active site residues in catalysis and ligand binding by site-directed mutagenesis, kinetic analysis, and ligand titration assays.
In addition, we test the potential role of a highly conserved loop in AtOASS as a protein-protein interaction site with A. thaliana SAT (AtSAT). These experiments demonstrate the critical role of key amino acids in the OASS active site and provide insight into the molecular basis of cysteine biosynthesis in plants. Cloning and Mutagenesis-The cytosolic form of AtOASS (EMBL number CAA56593) was amplified by PCR from an A. thaliana cDNA library using 5Ј-dGCTTGACATATGGCCTCTCGTATTGCTAAAG-ATGTG-3Ј as the forward primer (NdeI site is underlined; the start codon is in boldface type; and two codons optimized for E. coli expression are in italic type) and 5Ј-dGGATCCTCAAGCCTCGAAGGT-CATGGCTTCCGC-3Ј as the reverse primer (BamHI site is underlined and the stop codon is in boldface type). The 1.1-kb PCR product was subcloned into the pPCR-Script Amp SK(ϩ) vector (Stratagene). Automated nucleotide sequencing confirmed the fidelity of the AtOASS PCR product. Digesting the pPCR-Script-AtOASS vector with NdeI and BamHI and then ligating the 1.1-kb DNA fragment into NdeI/BamHIdigested pET28a yielded the pET28a-AtOASS expression vector. Sitedirected mutants of AtOASS (K46A, T74A/S, S75A/T/N, N77A/D,  T78A/S, Q147A/E, H157Q, H157N, T182A/S, T185A/S, K217A, H221A, K222A, S269A/T) were generated using the QuikChange (Stratagene) PCR method. Where possible, degenerate oligonucleotides were used to introduce multiple mutations in a single PCR.

EXPERIMENTAL PROCEDURES
The AtSAT cytosolic isoform (Swiss-Prot number Z34888) was amplified by PCR from an A. thaliana cDNA library using 5Ј-dTTC-CATGGTATGCCACCGGCCGGAGAACTC-3Ј as the forward primer (NcoI site is underlined; the start codon is in boldface type) and 5Ј-dTTGCGGCCGCTTATATGATGTAATCTGACCATTCCGAG-ATG-3Ј as the reverse primer (NotI site is underlined, and the stop codon is in boldface type). The resulting PCR product was subcloned into pHIS8 (26) to generate the pHIS8-AtSAT expression vector.
Protein Overexpression and Purification-Wild-type and mutant AtOASS expression constructs were transformed into E. coli BL21(DE3). Transformed E. coli were grown at 37°C in Terrific broth containing 50 g ml Ϫ1 kanamycin until A 600 ϳ0.8. After induction with 1 mM isopropyl 1-thio-␤-D-galactopyranoside, the cultures were grown at 20°C for 6 h. Cells were pelleted by centrifugation and resuspended in 50 mM Tris (pH 8.0), 500 mM NaCl, 20 mM imidazole, 1 mM ␤-mercaptoethanol, 10% (v/v) glycerol, and 1% (v/v) Tween 20. After sonication and centrifugation, the supernatant was passed over a Ni 2ϩ -NTA column. The column was washed with lysis buffer minus Tween 20, and the His-tagged protein was eluted with elution buffer (wash buffer with 250 mM imidazole). Incubation with thrombin during overnight dialysis at 4°C against wash buffer removed the N-terminal His tag. Dialyzed pro-tein was reloaded on a mixed Ni 2ϩ -NTA/benzamidine-Sepharose column to deplete the mix of uncleaved protein and thrombin. The flowthrough of this step was dialyzed overnight against 30% (v/v) glycerol, 25 mM Hepes (pH 7.5), 100 mM NaCl, and 1 mM dithiothreitol and then loaded onto a Superdex-75 size exclusion FPLC column equilibrated in the same buffer without glycerol. Fractions containing AtOASS were concentrated and stored at Ϫ80°C.
His-tagged AtSAT was expressed and purified using a similar procedure, except that E. coli Rosetta(DE3) cells were used for protein expression, and a Superdex-200 size exclusion FPLC column replaced the Superdex-75 column.
Enzyme Assays-OASS activity was assayed by measuring the formation of cysteine (27). Standard assay conditions were 100 mM Mopso (pH 7.0), 10 mM O-acetylserine, and 0.25 mM Na 2 S in a 0.25-ml reaction volume. Reactions were initiated by the addition of 0.01 g of AtOASS and incubated for 0, 5, 10, and 15 min at 25°C. The addition of 50 l of tricholoroacetic acid (20% (v/v) final concentration) quenched the reactions. Following centrifugation, 0.25 ml of supernatant was mixed with an equal volume of acid-ninhydrin reagent. The mixture was heated (95°C) for 5 min and then cooled on ice for 5 min. Following the addition of 0.5 ml of cold 100% ethanol, the A 546 was measured. The amount of cysteine in each reaction was determined using a standard curve (0.01-0.5 mol). Steady-state kinetic parameters were determined by initial velocity experiments, in which product formation was linear over the time monitored (2-20 min) using the above assay with varied concentrations of O-acetylserine (0 -10 mM) at 1 mM Na 2 S or varied concentrations of Na 2 S (0 -1.5 mM) at 10 mM O-acetylserine. Substrate concentrations were chosen to avoid substrate inhibition. Mutations that increased K m values for a given substrate were reassayed with concentrations up to 10-fold higher than the K m value. Data for both substrates were fitted to the Michaelis-Menten equation, v ϭ (V max [S])/(K m ϩ [S]), using Kaleidagraph (Synergy Software, Reading, PA).
Ligand Binding-Binding of O-acetylserine, cysteine, and methionine was measured by monitoring the change in absorbance of the PLP using a Cary Bio300 UV-visible spectrophotometer. Crystallography-Crystals of AtOASS were grown by the vapor diffusion method in 2-l hanging drops of a 1:1 mixture of protein (10 mg ml Ϫ1 ) and crystallization buffer (1.8 M ammonium sulfate; 0.1 M Tris, pH 8.0) at 4°C over a 0.5-ml reservoir. Macroseeding was used to improve crystal size by transferring individual crystals to hanging drops containing a 1:1 mixture of protein (7.5 mg ml Ϫ1 ) and crystallization buffer (1.8 M ammonium sulfate and 0.1 M Tris, pH 8.0). Before freezing at 100 K, crystals were washed in 30% glycerol, 1.8 M ammonium sulfate, and 0.1 M Tris (pH 8.0) as a cryoprotectant. All data were collected at 100 K using a Proteum-R Smart 6000 CCD detector connected to a Bruker-Nonius FR591 rotating anode generator. Diffraction intensities were integrated, merged, and scaled using the Bruker Proteum software suite (TABLE ONE). The AtOASS structure was solved by molecular replacement with CNS (29) using a homology model of AtOASS (residues 3-308) generated from the structure of StOASS (Protein Data Bank number 1OAS) as the search model. Cross-rotation and translation searches yielded a single solution ( 1 ϭ 8.08°; 2 ϭ 46.55°; 3 ϭ 6.79°; x ϭ 97.30; y ϭ 48.13; z ϭ Ϫ52.96) with monitor and packing values of 0.394 and 0.345, respectively, consistent with the presence of a single monomer in the asymmetric unit. After rigid body refinement using CNS (R cryst ϭ 48.7%; R free ϭ 49.6%), difference electron density indicated the presence of PLP at the active site. Following an initial round of simulated annealing, positional, and B-factor refinement (R cryst ϭ 32.3%; R free ϭ 35.6%), PLP was added and the additional C-terminal residues (positions 309 -322) were built into the model. After iterative rounds of manual rebuilding in O (30) and refinement in CNS, the R-factors converged (TABLE ONE). The final model includes residues 3-322, the PLP cofactor, 172 water molecules, and a sulfate ion.
Crystals of the AtOASS K46A mutant grew using the vapor diffusion method in 2-l hanging drops of a 1:1 mixture of protein (15 mg ml Ϫ1 ) and crystallization solution (4.5 M sodium formate) at 4°C over a 0.5-ml reservoir. Individual crystals were used to macroseed hanging drops containing a 1:1 mixture of protein (10 mg ml Ϫ1 ) and 4.0 M sodium formate. Crystals were briefly washed in 30% glycerol and 4.0 M sodium formate as a cryoprotectant. All data were collected at 100 K as described above. Diffraction intensities were integrated, merged, and scaled using the Proteum software suite (TABLE ONE). The structure was obtained by difference Fourier methods and was refined using CNS. The initial model included the K46A mutation and lacked the PLP molecule. CNS parameter and topology files for PLP⅐methionine external aldimine were generated using the HIC-Up server (available on the World Wide Web at x-ray.bmc.uu.se/cgi-bin/gerard/hicup_server.pl). After iterative rounds of manual rebuilding in O and refinement in CNS, the R-factors converged to those reported in TABLE ONE. The final model of the AtOASS K46A mutant includes residues 3-322, the PLP⅐methionine external aldimine molecule, and 72 water molecules.

Three-dimensional Structure
Recombinant AtOASS was expressed in E. coli as a hexahistidinetagged protein and purified by nickel affinity and size exclusion chromatography. Digestion with thrombin removed the His tag from the expressed protein. The purified protein was yellow with an absorbance maximum at 412 nm, indicating the presence of PLP as an internal Schiff base in the active site, and displayed a specific activity of 1400 mol of cysteine produced/min/mg of protein.
The overall structure of AtOASS resembles that of the Salmonella enzyme (24) with an r.m.s. deviation for the aligned 315 C␣ atoms of 1.2 Å 2 (Fig. 2B). Structural differences occur in the ␤1A-␤2A and ␤8A-␤9A loops and the C terminus of the two proteins. There is a three-amino acid insertion in the ␤1A-␤2A loop and a six-residue deletion in the ␤8A-␤9A loop of AtOASS compared with StOASS (Fig. 1B). The C terminus of AtOASS extends along the surface of the adjacent monomer, whereas the last 20 amino acids of the StOASS structure were disordered (24).

AtOASS K46A
Space group P4 3 where summation is over the data used for refinement. d R free is defined the same as R cryst but was calculated using 5% of data excluded from refinement.
is related by 14 -21% amino acid sequence identity to these proteins with r.m.s. deviations for the C␣ atoms of 2.3-3.0 Å 2 . Since these enzymes require PLP as a cofactor, conservation of the dual ␣/␤-domain structure in each protein maintains the cofactor binding site with structural and sequence variations tailoring substrate recognition and reaction chemistry.

Active Site Architecture: PLP and Sulfate Binding
The location of PLP in the AtOASS structure defines the location of the active site (Figs. 2, C and D) with clear electron density for the internal Schiff base between PLP and Lys 46 observed (Fig. 2E). In addition to the covalent linkage, Asn 77 and Ser 269 form hydrogen bonds with the oxygen and nitrogen, respectively, of the pyridine ring. PLP is anchored in the active site by multiple interactions. Hydrogen bonds are formed between Gly 181 , Thr 182 , Gly 183 , Thr 185 , and the phosphate group. His 157 and Thr 185 interact with two water molecules, which in turn form hydrogen bonds with the phosphate moiety. The amino acids involved in PLP binding are conserved in both plant and bacterial OASS structures and between the Arabidopsis isoforms (Fig. 1B).
The ␤3B-␣2 loop of AtOASS, corresponding to residues 74 -78, forms one side of the active site. This loop, which was previously suggested to be involved in substrate binding, adopts the same conformation as observed in the StOASS structure (24). The sequence of this loop (TSGNT) is highly conserved in all plant and bacterial OASS sequences with the exception of the StOASS used for previous crystallization experiments, which replaces the serine with an asparagine (TNGNT) (Fig. 1B). In the AtOASS active site, electron density for a sulfate ion, presumably from the crystallization solution, was observed near the substrate-binding loop. The side-chain hydroxyl group of Thr 74 (2.79 Å), the backbone nitrogen of Ser 75 (2.77 Å), and the side-chain nitrogen of Gln 147 (2.90 Å) interact with the sulfate ion (Fig. 2C).

Three-dimensional Structure of an External Aldimine Complex
The sequence difference in the substrate-binding loop of AtOASS and StOASS suggests that the protein-substrate interactions may differ in these enzymes. Efforts to obtain a substrate and/or product analog complex with OASS are complicated by the presence of the active site lysine, which reacts with any external aldimine formed in the active site.
To obtain a structure of AtOASS complexed with a ligand in the active site, the AtOASS K46A mutant was generated. The purified mutant protein was inactive for cysteine formation, which was consistent with the catalytic role of the active site lysine (22). An absorbance maximum at 418 nm indicated that PLP had, however, formed a stable external aldimine in the active site (21).
Crystals of the AtOASS K46A mutant were obtained, and the 2.7 Å resolution structure of the protein was determined (TABLE ONE). Initial electron density maps unambiguously showed the alanine substitution at position 46 and indicated that PLP formed an external aldimine with an unknown ligand in the active site (Fig. 3A). Given the reaction catalyzed by OASS, glycine and cysteine where modeled in the external aldimine linkage, but electron density extending beyond the terminal side-chain atom of either amino acid was observed. For model building and refinement, the external aldimine was modeled as a PLP-methionine linkage, since the crystal structure of the StOASS K41A mutant revealed this molecular bond in the active site (25). The overall structures of the wild-type and mutant AtOASS are nearly identical, with an r.m.s. deviation of 0.37 Å 2 . Comparison of the AtOASS K46A mutant, the StOASS, and the StOASS K41A mutant structures shows that the external aldimine molecule binds in a similar conformation in the active sites of the plant and bacterial enzymes; however, the substrate-binding loop of AtOASS does not reposition as observed in StOASS (24, 25) (Fig. 3B).
Within the active site of the AtOASS K46A mutant, the side chain of the methionine is oriented toward the active site entrance, and the car-  NOVEMBER (Fig.  3, C and D). These interactions are similar to those observed with the corresponding residues in the StOASS K41A mutant structure (25). In the AtOASS structure, Ser 75 does not interact with the external aldimine, whereas the active site loop of StOASS shifts to place the side chain of an asparagine residue, which corresponds to the serine, within 3 Å of the methionine carboxylate group (25).

Functional Analysis
PLP Binding Site-An extensive set of conserved hydrogen bond interactions anchor the cofactor in the active site of AtOASS (Fig. 2, C  and D). To probe the importance of specific interactions in positioning PLP, amino acid substitutions (H157Q, H157N, T182A, T182S, T185A, T185S, S269A, and S269T) were introduced into AtOASS using sitedirected mutagenesis. The mutant proteins were overexpressed, purified, and assayed for cysteine formation (TABLE TWO). Compared with wild-type AtOASS, substitutions of His 157 and Thr 182 altered the catalytic efficiency (k cat /K m ) of the enzyme less than 2-fold for either substrate. The decreased reaction rates observed for the S269A and S269T mutants probably result from changes in the position of the pyridine ring, as observed in StOASS (23). Mutation of Thr 185 to either an alanine or a serine resulted in 400 -1600-fold decreases in turnover rates (k cat ). The absorbance spectra of the T185A and T185S mutants showed a ϳ10-fold reduction in PLP signal (A 412 ) compared with the wild-type enzyme, indicating that this residue provides a crucial binding interaction with the cofactor.
Substrate Binding Site-Although the substrate-binding loop of OASS appears important for cysteine formation, the functional roles of amino acids in this loop remain untested. To examine the contributions of Thr 74 , Ser 75 , Asn 77 , Thr 78 , and Gln 147 to catalysis and ligand binding, we generated the following mutants: T74A, T74S, S75A, S75T, S75N, N77A, N77D, T78A, T78S, Q147A, and Q147E. Using purified recombinant proteins, the kinetic parameters of each AtOASS mutant for O-acetylserine and sulfide were examined for comparison with the wildtype enzyme (TABLE TWO).
Mutation of Gln 147 , the only active site residue not on the substratebinding loop that contacts methionine in the external aldimine structure, to either an alanine or glutamic acid reduced the catalytic efficiency of AtOASS by 1900 -25,400-fold. Incorporation of a negatively charged side chain in the active site (Q147E) decreased the turnover rates more than elimination of the glutamine amide group (Q147A). Although the effects of the Q147E and Q147A mutations were primarily on k cat , the K m value of each mutant for O-acetylserine also increased.
Of the mutations in the substrate-binding loop, replacement of Thr 78 (T78A and T78S) had a modest effect on catalytic efficiency (k cat /K m ), resulting from a slower turnover rate with either substrate. Both the T74A and T74S mutants displayed roughly 15-fold higher K m values for sulfide and slower reaction rates with both O-acetylserine and sulfide. Mutation of Ser 75 to either alanine or threonine mainly changed k cat with each substrate, suggesting that the amino acid side chain at this position influences cysteine formation. Surprisingly, the S75N mutation, which places the asparagine side chain of StOASS into the AtOASS active site, yielded an enzyme with a 10 5 -fold lower catalytic efficiency than wild-type enzyme. Substitution of Asn 77 (N77A or N77D) reduces k cat /K m by 10 5 -fold.
To determine the effect of the point mutations on ligand binding, titration of wild-type and mutant AtOASS with O-acetylserine, cys-teine, and methionine was performed. In the absence of sulfide, O-acetylserine reacts with PLP to yield the ␣-aminoacrylate intermediate (absorbance maximum at 470 nm), resulting from elimination of acetate from the substrate (Fig. 4A) (19). Titration using O-acetylserine yields a binding constant (K d ) for external Schiff base formation (Fig. 4B). Similarly, the addition of either cysteine (Fig. 4, C and D) or methionine (Fig.  4, E and F) to the enzyme results in formation of an external aldimine, maximum absorbance at 418 nm (21). Binding constants for wild-type and mutant AtOASS for O-acetylserine, cysteine, and methionine are summarized in TABLE THREE. Titrations of the T74S, S75T, T78A, and T78S mutants showed that these mutations do not significantly change the binding constant for O-acetylserine, whereas substitution of Asn 77 with an alanine increased the K d by 28-fold (TABLE THREE). Titrations using cysteine and methionine showed a similar trend of less than 3-fold differences between wild-type AtOASS and the T74S, S75T, T78A, and T78S mutants. The K d of the N77A mutant for methionine was 10-fold higher than wild-type enzyme, and cysteine binding was not saturated at 20 mM ligand. Changes in the PLP-signal at 412 nm were not observed in the T74A, S75A, S75N, N77D, Q147A, and Q147E mutants, although these enzymes catalyzed cysteine formation with turnover rates a 1000-fold lower than wild-type AtOASS.

AtOASS-AtSAT Interaction Site
Comparison of ϳ300 plant and bacterial OASS sequences reveals that amino acids in the ␤8A-␤9A loop are highly conserved. In the Arabidopsis and Salmonella enzymes (Fig. 1B), this loop contains a block of amino acids (Lys 217 to Phe 230 in AtOASS) in common. The ␤8A-␤9A loop forms part of a surface cleft ϳ15-20 Å away from the active site (Fig. 5A). Interestingly, evolutionary conservation of surface loops implies a functional or structural role (38). Since OASS and SAT physically associate to form the cysteine synthase complex (4 -6), in which OASS is inactive and SAT is activated, the location of the ␤8A-

Steady-state kinetic parameters
All reactions were performed as described under "Experimental Procedures." WT, wild-type. All k cat and K m values are expressed as a mean Ϯ S.E. for n ϭ 3. ␤9A loop at the active site entrance suggested this as a potential interaction site with SAT.

O-Acetylserine
Within the ␤8A-␤9A loop of AtOASS, Lys 217 , His 221 , and Lys 222 were mutated to alanine. These amino acids were chosen for mutagenesis, because polar amino acids are often found at protein interaction hot spots (38 -40). Moreover, His 221 is highly conserved in the plant and bacterial OASS. The K217A, H221A, and K222A AtOASS mutants were expressed and purified as described for wild-type enzyme to yield proteins lacking the His tag. Each mutant protein catalyzed the formation of   cysteine with kinetic parameters similar to the wild-type enzyme, indicating that the role of these conserved residues is not the direction of substrates to the active site.
To test whether the K217A, H221A, and K222A mutant proteins interacted with AtSAT, a protein pull-down assay was used. Purified octahistidine-tagged AtSAT (monomer molecular mass ϭ 36.3 kDa) with a specific activity for O-acetylserine formation of 60 mol of product formed/min/mg of protein was bound to an Ni 2ϩ -NTA column. Following incubation of His-tagged AtSAT with non-His-tagged wildtype, K217A, H221A, or K222A AtOASS (monomer mass of 33.9 kDa), the column was washed to remove unbound protein. Next, bound protein was eluted from the column with imidazole. The wash and elution fractions were analyzed by SDS-PAGE, and SAT activity was assayed. Fig. 5B shows a comparison of the SDS-PAGE results obtained with the H221A mutant (lanes A and B) and wild-type AtOASS (lanes C and D). After incubation of AtSAT with AtOASS, excess AtOASS was collected from the wash step. The elution fraction contained AtSAT and AtOASS, indicating formation of the cysteine synthase complex. The identities of the two bands in the elution fraction were confirmed by matrix-assisted laser desorption ionization mass spectrometry peptide fingerprinting performed by the Danforth Center Proteomics facility (not shown). The specific activity of AtSAT in this fraction was 400 mol of product formed/min/mg of protein with low OASS activity (100 mol of product formed/min/mg of protein) detected. The increase in SAT activity and decrease in OASS activity are consistent with formation of the cysteine synthase complex (6,41). In contrast, incubation of AtSAT with the H221A AtOASS mutant yielded only AtSAT in the elution fraction with a specific activity of 55 mol of product formed/min/mg of protein, indicating that the cysteine synthase complex was not formed. Results similar to those obtained for the H221A mutant were also observed with the K217A and K222A AtOASS mutants.

DISCUSSION
Compared with the mechanisms of nitrogen and phosphorus uptake and assimilation, plant sulfur metabolism is less well understood at the molecular level, although plants play a central role in the environmental sulfur cycle (2). Within the pathway of sulfur assimilation, OASS catalyzes the production of cysteine and is part of the regulatory system that responds to changes in cellular sulfur supply (5,6). The crystal structures and functional analysis of AtOASS presented here provide molecular insight into cysteine biosynthesis in plants.
Detailed studies of StOASS provide a structural model for the reaction catalyzed by the enzyme (45). Crystal structures of uncomplexed (open) and complexed (closed) StOASS formed the basis of this model (24,25). The reaction mechanism begins with O-acetylserine binding to the resting enzyme, which is in the open conformation. Interaction between the substrate binding loop and the ␣-carboxylate of bound O-acetylserine (or methionine) was proposed to trigger closure of the active site. In particular, the side chain of an asparagine (Asn 69 ) in StOASS moves 7 Å to interact with the substrate's carboxylate, locking the substrate in the active site by shifting the position of the loop (25).
The first half-reaction occurs, resulting in the release of acetate. Next, sulfide enters the active site to complete the second half-reaction, yielding cysteine.
Experiments presented here and recent work described by others question the proposed role of the asparagine-substrate interaction as the primary trigger in conformational switching between open and closed active site conformations in OASS. First, the amino acid sequence of the AtOASS active site loop (TSGNT) is highly conserved in all plant and bacterial OASS sequences, whereas the active site loop sequence of StOASS (TNGNT) only occurs in that enzyme (24). Moreover, the S75N AtOASS mutant catalyzes cysteine formation with a 10 5 -fold decrease in catalytic efficiency, indicating that AtOASS and the other OASS enzymes with a serine at this position are functionally different from StOASS. Second, crystal-packing forces probably determine the protein conformation observed in the various three-dimensional structures of OASS. Both wild-type AtOASS and the AtOASS K46A mutant with PLP and methionine linked as an external aldimine crystallized in the open conformation. Likewise, structures of E. coli OASS and a mutant version of the same protein adopted open and closed conformations, respectively, in the absence of substrates or analogues bound in the active site (44). Although the asparagine-substrate interaction in StOASS may not act as a general triggering mechanism, the open and closed conformations observed in different OASS structures strongly suggest a functional role for domain movement in the catalytic cycle of the enzyme, as originally proposed.
Extensive studies have examined the chemical reaction of OASS (19), but only limited information is available on the functional roles of residues in the OASS active site. In AtOASS, Lys 46 forms a Schiff base linkage to PLP, and multiple interactions anchor the cofactor in the active site (Fig. 2). As observed in StOASS (22), the active lysine of the Arabidopsis enzyme is crucial for the catalysis. In addition, Thr 185 and Ser 269 maintain PLP in a catalytically productive orientation (23), since mutation of either residue affects cysteine synthesis (TABLE TWO).
Based on the AtOASS crystal structures, kinetic analysis, and ligand binding experiments, Asn 77 and Gln 147 are key residues for O-acetylserine binding. Of the AtOASS mutants assayed, the N77A, N77D, Q147A, and Q147E mutants increased the K m values for O-acetylserine (TABLE TWO). Likewise, mutation of Asn 77 to an alanine raised the K d for O-acetylserine by 28-fold (TABLE THREE). Moreover, the N77A mutant displayed the largest decreases in binding affinity for cysteine and methionine (TABLE THREE). Although titrations of the N77D, Q147A, and Q147E mutants with O-acetylserine yielded no detectable ␣-aminoacrylate intermediate, these mutants still catalyze cysteine formation with 1000-fold lower turnover rates compared with wild-type enzyme. In the OASS reaction, formation of the ␣-aminoacrylate intermediate is the rate-limiting step, and the intermediate decays with time in the absence of sulfur (20,46,47). The lack of detectable intermediate in the N77D, Q147A, and Q147E mutants indicates that the rate of intermediate formation is now slower than the dissociation rate, consistent with an increased K d value for O-acetylserine. These mutants retain the ability to form cysteine at reduced turnover rates, because the addition of sulfide is diffusion-limited (45)(46)(47), thus rapidly transforming any ␣-aminoacrylate intermediate formed into product.
Functional analysis of the AtOASS mutants also demonstrates that Thr 74 and Ser 75 are important active site residues. With regard to sulfide binding, substitution of either amino acid yields mutant enzymes (T74A, T74S, S75A, and S75N) that display more than 10-fold increases in the K m value for sulfide (TABLE TWO). A defined sulfide binding in OASS has not been identified (47); however, the effects of mutating Thr 74 and Ser 75 suggest that these residues play a role in stabilizing the NOVEMBER 18, 2005 • VOLUME 280 • NUMBER 46 transition state of the second half-reaction. The T74A, S75A, and S75N mutations also affect ligand binding, since titrations of these mutant enzymes showed no detectable signal for the ␣-aminoacrylate intermediate or the external aldimine with cysteine or methionine. As described above, these mutations probably decrease the rate of the first half-reaction, leading to ␣-aminoacrylate intermediate formation. The T74A mutation eliminates a potential hydrogen bond donor in the active site (Fig. 3, C and D), unlike the T74S mutation, which retains a hydroxyl group at this position and causes minimal changes in binding affinity with O-acetylserine, cysteine, or methionine (TABLE THREE). Although Ser 75 does not interact with methionine in the AtOASS K46A crystal structure, in the recently determined structure of E. coli OASS (44), the side-chain hydroxyl group of the corresponding serine contacts the modeled ␣-aminoacrylate intermediate and sulfur. Thus, the S75A and S75N mutations may disrupt the local structure of the substrate binding loop to affect catalysis when the active site adopts the closed conformation.

Structure of Arabidopsis O-Acetylserine Sulfhydrylase
The formation of multienzyme complexes provides cells with a means of controlling metabolism by organizing the location of key enzymes (48,49). In bacteria and plants, SAT and OASS associate to form the cysteine synthase complex, which acts as a molecular sensor in the regulatory circuit that coordinates sulfate assimilation and modulates intracellular cysteine levels (4 -6, 50). When the two enzymes are complexed, SAT activity increases and OASS activity decreases. This results in production of O-acetylserine. When intracellular sulfur levels are low, O-acetylserine accumulates, because free OASS is unable to generate cysteine due to a lack of sulfide. Elevated O-acetylserine levels cause the complex to dissociate, thereby down-regulating SAT. Meanwhile, the increased O-acetylserine concentration also activates expression of genes encoding sulfate transporters, OASS, and SAT. This leads to increased sulfur uptake and reduction. As sulfur levels elevate, free OASS begins to catalyze cysteine formation and reduce O-acetylserine levels. This allows association of SAT and OASS, activation of SAT, and resumption of cysteine biosynthesis. Yeast two-hybrid (51,52), gel filtration chromatography (6,53), and surface plasmon resonance methods (54) demonstrate assembly of this macromolecular complex. Based on these studies, the C terminus of SAT is crucial for interaction with OASS; however, until recently, the region of OASS involved in formation of the cysteine synthase complex has been unclear.
Determination of the three-dimensional structure of the OASS from Haemophilus influenzae complexed with a peptide corresponding to the 10 C-terminal residues of the Haemophilus SAT showed the peptide bound in the OASS active site (55). This structure explains the reduced OASS activity as part of the cysteine synthase complex and the dissociation of the assembly in the presence of O-acetylserine, which competes for the binding site. Earlier studies using E. coli SAT showed that the full-length enzyme is a 250-fold better inhibitor of OASS than a peptide corresponding to the C-terminal of SAT, suggesting that additional structural features are required for efficient formation of the cysteine synthase complex (53). The protein-protein interaction experiments described here demonstrate the importance of structural features adjacent to the OASS active site (i.e. the ␤8A-␤9A loop) for providing an interaction site with SAT. Mutation of Lys 217 , His 221 , and Lys 222 in AtOASS disrupts formation of the cysteine synthase complex with AtSAT without altering OASS activity. Further structural and protein interaction studies will better define the molecular and energetic details of how OASS and SAT form the cysteine synthase complex.