Structural Insights into Catalysis and Inhibition of O-Acetylserine Sulfhydrylase from Mycobacterium tuberculosis

Cysteine biosynthetic genes are up-regulated in the persistent phase of Mycobacterium tuberculosis, and the corresponding enzymes are therefore of interest as potential targets for novel antibacterial agents. cysK1 is one of these genes and has been annotated as coding for an O-acetylserine sulfhydrylase. Recombinant CysK1 is a pyridoxal phosphate (PLP)-dependent enzyme that catalyzes the conversion of O-acetylserine to cysteine. The crystal structure of the enzyme was determined to 1.8Å resolution. CysK1 belongs to the family of fold type II PLP enzymes and is similar in structure to other O-acetylserine sulfhydrylases. We were able to trap the α-aminoacrylate reaction intermediate and determine its structure by cryocrystallography. Formation of the aminoacrylate complex is accompanied by a domain rotation resulting in active site closure. The aminoacrylate moiety is bound in the active site via the covalent linkage to the PLP cofactor and by hydrogen bonds of its carboxyl group to several enzyme residues. The catalytic lysine residue is positioned such that it can protonate the Cα-carbon atom of the aminoacrylate only from the si-face, resulting in the formation of l-cysteine. CysK1 is competitively inhibited by a four-residue peptide derived from the C-terminal of serine acetyl transferase. The crystallographic analysis reveals that the peptide binds to the enzyme active site, suggesting that CysK1 forms an bi-enzyme complex with serine acetyl transferase, in a similar manner to other bacterial and plant O-acetylserine sulfhydrylases. The structure of the enzyme-peptide complex provides a framework for the design of strong binding inhibitors.

enzymes to the same extent and is therefore designated as either CysK2 or CysM2.
The crystal structure analysis of OASS from Salmonella typhimurium (14) Haemophilus influenzae (15), Escherichia coli (16), and Arabidopsis thaliana (17) revealed that it belongs to the fold type II class of vitamin B 6 -dependent enzymes (18,19). The enzymatic mechanism of OASS follows a ping-pong mechanism typical for PLP-dependent enzymes (10,11,20). The reaction can be divided into two halves (Scheme 1). In the resting state, the cofactor forms an internal Schiff base with the invariant catalytic lysine residue (compound I in Scheme 1). The incoming substrate then forms an external Schiff base with PLP (II) followed by a ␤-elimination in which acetate is released and a proton is abstracted from the ␣ position, most likely by the lysine side chain (21). A similar role of the invariant lysine residue has been proven in the homologous tryptophan synthase ␤-subunit (22). The product of the first half-reaction is the ␣-aminoacrylate intermediate (23), covalently linked to PLP (III). The second half-reaction starts with a nucleophilic attack of the sulfide (or HS Ϫ ) on the ␤-carbon of the aminoacrylate intermediate, and the ␣-carbon is reprotonated (20,24), resulting in cysteine bound as external Schiff base (IV). Finally, the product is released via regeneration of the internal aldimine. Formation of the external aldimine appears to be coupled to a large conformational change that leads to the closure of the active site, as suggested by the structure of the S. typhimurium CysK K41A mutant in complex with methionine, a substrate analogue (25). However, in the corresponding structure of OASS from A. thaliana, the enzyme also remained in the open conformation after formation of the covalent PLP-methionine adduct (17). The latter is assumed to mimic the external aldimine (II).
In this work, we describe the cloning, purification, characterization, and 1.8 Å x-ray structure of CysK1 from M. tuberculosis. We were also able to trap the ␣-aminoacrylate in the crystal and report the structure of this reaction intermediate to 2.2 Å resolution, thus providing novel structural insights in the reaction cycle of CysK1. We also determined the crystal structure of CysK1 from M. tuberculosis bound to an inhibitory four-residue peptide derived from the C terminus of M. tuberculosis CysE (SAT, Rv2335). The structure of this inhibited form of CysK1 may provide the basis for the design of strong binding inhibitors of this enzyme.

EXPERIMENTAL PROCEDURES
Gene Cloning and Protein Production-The DNA sequence coding for CysK1 from M. tuberculosis (Rv2334) was amplified by PCR and cloned using upstream NdeI and downstream HindIII restriction sites. Subsequently, the sequence was verified, and the fragment was cloned into the expression vector pET28a (Novagen), resulting in a cleavable six-histidine tag at the N terminus of the polypeptide chain. In this construct, thrombin cleavage results in a recombinant protein with three additional amino acids (Gly-Ser-His) at the N terminus.
E. coli BL21(DE3) carrying the expression construct pET-His6CysK1 was cultivated in 1.5 liters of LB medium supplemented with kanamycin (30 g/ml) at 21°C. At an A 600 value of 0.5-0.6, gene expression was induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. After ϳ24 h, the cells were harvested and then resuspended in a buffer consisting of 10 mM Tris-HCl, pH 8.0, 300 mM NaCl, and 10 mM imidazole. Cells were disrupted by freeze-thaw cycles, lysozyme, and DNase I treatment and sonication. The clarified lysates were loaded on a nickel-nitrilotriacetic acid column (Qiagen) with a column volume of 1.8 ml. After extensive washing with 10 mM imidazole, remaining proteins were eluted with an imidazole step gradient. Pure His 6 -CysK1 eluted in fractions containing 100 -200 mM imidazole. His 6 -CysK1-containing fractions were pooled, loaded on a desalting PD10 column (GE-Healthcare), and eluted in a buffer of 25 mM Tris-HCl, pH 8.0, and 150 mM NaCl. The N-terminal His 6 tag was removed by thrombin cleavage. Residual non-cleaved His-CysK1 was removed by passage through a nickel-nitrilotriacetic acid column. The flowthrough containing the cleaved CysK1 protein was further purified using an ion-exchange step on a Q-Sepharose (Amersham Biosciences) column and NaCl gradient elution. The CysK1containing fractions (200 -250 mM NaCl) were pooled and loaded on a Superdex 200 column equilibrated with 25 mM Tris-HCl, pH 8.0, and 150 mM NaCl. This procedure resulted in homogeneous CysK1 as verified by SDS chromatography. CysK1 protein was concentrated to 10 mg/ml using an Amicon centrifugation device with a 3-kDa molecular mass cut-off. Aliquots of the protein preparations were flash-frozen in liquid nitrogen and stored at Ϫ80°C until further use.

Structure of the ␣-Aminoacrylate Intermediate in CysK1
Spectrophotometry-UV-visible absorbance spectra of CysK1 were recorded at a concentration of 3.3 mg/ml using a Jasco-V65 spectrophotometer. Spectra of the covalent aminoacrylate complex were obtained after the addition of 300 M O-acetylserine to the enzyme solution.
Enzyme Activity-Recombinant CysK1 was assayed for O-acetylserine sulfhydrylase activity spectrophotometrically at 560 nm by monitoring the formation of cysteine using the acid-ninhydrin method (26). The reactions were carried out in 1500 l of 100 mM MOPS buffer at pH 7.0. O-Acetyl-Lserine and sodium sulfide (dissolved in 1 mM NaOH) were added to final concentrations of 10 and 0.25 mM, respectively, and the reactions were started by the addition of the enzyme (0.025 g) to the assay mixtures. Aliquots of 250 l were taken at various time points, and the reaction was stopped by the addition of trichloroacetic acid (final concentration 16.6%). Following centrifugation, the amount of product was determined spectrophotometrically using a standard curve for cysteine in the range 0.04 -0.8 mM. Specific activities were derived from the linear part of the time curves, and each measurement was carried out in triplicate at 30 and 37°C, respectively.
Peptide Binding Assay and Inhibition Studies-The chemically synthesized oligopeptide DFSI was obtained from the GenScript Corp., Scotch Plains, NJ. The binding interaction was probed in a ThermoFluor experiment (27) in the buffer consisting of 25 mM Tris-HCl, pH 8.0, and 150 mM NaCl in the presence of the dye Sypro Orange (Sigma). CysK1 at a concentration of 0.4 g/l (12 pmol/l) in a total volume of 25 l was mixed with the DFSI peptide at concentrations of 0.0016, 0.016, 0.16, and 1.6 mM. Fluorescence at the wavelength 575 nm was recorded in the temperature interval 20 -90°C.
The initial rate of cysteine formation by CysK1 was determined at 22°C in the absence and presence of various DFSI peptide concentrations, and varying the OAS concentration in the range 1.0, 2.0, 5.0, 8.0, and 10 mM at a constant sodium sulfide concentration of 1.0 mM in 100 mM MOPS buffer, pH 7.0. At various time intervals, aliquots of the reaction mixture were removed and assayed for L-cysteine formation as described above. The mode of inhibition was determined from double reciprocal plots, and the K I value for the peptide was derived using the formalism described in (28).
Crystallization and Data Collection-Crystallization screening was carried out using the vapor diffusion method and a Phoenix crystallization robot. After extensive screening and optimization, two conditions were established that resulted in single crystals of CysK1. Both contained unusually high 2-methyl-2,4-pentane-diol (MPD) concentrations. Crystal form I was obtained by mixing 3 l of protein solution (10 mg/ml) with 2 l of reservoir solution (Bis-Tris-propane, pH 7.5, and 60 -65% MPD) in sitting drops. The rod-like crystals diffracted, however poorly (ϳ6 Å), and were not characterized further. Crystal form II was obtained by mixing 2 l of protein solution (10 mg/ml) with 2 l of reservoir solution (sodium HEPES, pH 7.2-8.2, and 80% MPD) in either sitting or hanging drops. The diamond-shaped single crystals reproducibly diffracted to 1.8 -2.0 Å resolution.
The CysK1 ␣-aminoacrylate complex was prepared by a combination of soaking and freeze-trapping. Sitting drops con-taining CysK1 crystals obtained at pH 8.2 were layered cautiously with a droplet of mother liquor containing the substrate O-acetylserine at concentrations of 1 and 2 mM, respectively. At various time intervals (1 mM OAS concentration after 10, 15, 20, 30, 35, and 30 min; at 2 mM OAS concentration after 10 and 20 min), crystals were removed, flash-frozen, and stored in liquid nitrogen until x-ray analysis. Due to the high MPD concentration in the mother liquor, no cryoprotectant was necessary.
Co-crystallization of CysK1 with the DFSI-peptide was achieved using the crystallization conditions established for the holoenzyme. CysK1 at a concentration of 10 mg/ml was mixed with the peptide at various peptide concentrations (0.1, 1.0, 4.0, 5.0, 10.0 mM) and preincubated at 22°for 30 -60 min before setting up the crystallization drops using the conditions described above.
Diffraction data for crystals of the holoenzyme were collected at beamline ID14:3 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). Diffraction data for the peptide complex were collected at the beamline ID14:1 of the ESRF. Diffraction data sets for the CysK1-aminoacrylate complexes were collected at beamline 9/11 at MAX-lab (Lund, Sweden). All data sets were collected in a nitrogen gas stream at 110 K.
The x-ray data were processed and scaled with the programs MOSFLM and SCALA from the CCP4 suite (29). The native crystals belong to the tetragonal space group P4 1 Table 1.
Molecular Replacement and Crystallographic Refinement-The structure was determined by molecular replacement using the program MOLREP (30). The coordinates for OASS from A. thaliana (17) (Protein Data Bank accession code 1Z7W) with the cofactor PLP and the bound sulfate ion omitted were used as the search model. The best solution had a score of 0.435 and an R-factor of 48.3%, with a single monomer in the asymmetric unit. To monitor the behavior of the refinement process, 5% of the x-ray data were removed for the calculation of R-free. Initial cycles of restrained refinement using Refmac5 (31) resulted in a drop of the R-factor by 12.8%. The correctness of the molecular replacement solution was confirmed by electron density for the cofactor, including the covalent linkage to the active site lysine residue appearing at the expected positions.
Manual rebuilding of the model was carried out with the program Coot (32) based on -A weighted 2F o Ϫ F o and F o Ϫ F o electron density maps (33). Manual adjustment of the model was interspersed with rounds of refinement by Refmac5 (31). Water molecules were added based on peak heights, shape of the electron density, temperature factor, and capability to form hydrogen bonds to surrounding protein residues and/or other water molecules. The final model contains amino acid residues 1-300 from CysK1, one PLP molecule, four MPD molecules, one chloride ion, and 205 water molecules. The last 10 residues at the C terminus did not have defined electron density.
Due to the change in cell dimensions upon soaking with OAS, molecular replacement was used to solve the structure of the CysK1 ␣-aminoacrylate complex. An initial CysK1 model was used as search model. The electron density maps calculated from the data set collected from a crystal soaked in 1 mM OAS for 30 min clearly indicated formation of the ␣-aminoacrylate intermediate. Refinement of the complex was carried out using a similar protocol as outlined above. At the final stage, TLS refinement (34) was used with 20 TLS groups as defined by the TLSMD server (35), which resulted in a 2% drop in the free R-factor. The final model contains residues 1-306 from CysK1, the amino-acrylate PLP covalent complex, one MPD, and 47 crystallographic water molecules.
The structure of the CysK1-peptide complex was solved by molecular replacement using the initially refined CysK1 structure as a search model. The electron density calculated after the initial refinement rounds indicated the presence of the peptide bound at the enzyme active site. Refinement of the complex was carried out using a similar protocol as outlined above. The final model comprises CysK1 residues 1-300, the four-residue-long peptide, one MPD molecule, and 163 water molecules.
The protein models were analyzed with PROCHECK (36) to monitor the stereochemistry. Details of the refinement and protein models are given in Table 1. Figures were made using the program PyMOL. The sequence alignment was made by ClustalW (37).

RESULTS AND DISCUSSION
Characterization of CysK1 from M. tuberculosis-CysK1 was produced with a hexahistidine tag at the N terminus. The tag was removed by thrombin leaving three extra residues Gly-Ser-His on the N terminus of the purified protein. Recombinant CysK1 has a yellow color and shows the typical absorption spectrum of PLP-dependent enzymes (38). The absorbance peak at 410 nm is characteristic of the internal Schiff base formed between the cofactor and the active site lysine residue (Fig. 1).
Gel filtration chromatography suggested that CysK1 is a dimer in solution and that the enzyme migrated in a native gel as a single well defined band (data not shown). Purified recombinant CysK1 is catalytically active as an O-acetylserine sulfhydrylase. The specific activity was determined with O-acetyl-Lserine and sodium sulfide as substrates. The turnover number was 211 s Ϫ1 at 30°C and 352 s Ϫ1 at 37°C. These values compare well with those determined for the related enzyme from S. typhimurium, which showed 280 s Ϫ1 at 25°C as turnover rate for the reaction (39).

Structure of the ␣-Aminoacrylate Intermediate in CysK1
Overall Structure of M. tuberculosis CysK1-The asymmetric unit contains a single copy of the CysK1 subunit, corresponding to a solvent content of 65%. The subunit consists of two domains, each having a ␣/␤-fold ( Fig. 2A and  supplemental Fig. S1), typical of the fold type II PLP enzymes (18,19). The N-terminal domain (residues 43-145) is built up by a central four-stranded parallel ␤-sheet, which is flanked by three ␣-helices on one side and one ␣-helix on the other side. The C-terminal domain consists of residues 11-42 and 151-304 that fold into a six-stranded mixed ␤-sheet sandwiched by four ␣-helices. In the crystal, the enzyme is a dimer, with the molecular axis coinciding with the 2-fold crystallographic axis. The buried surface at the dimer interface is 2130 Å 2 , which corresponds to 15.7% of the total surface of one monomer.
Pyridoxal Phosphate Binding Site-The PLP binding site is located in a cleft formed between the C-terminal ends of the ␤-sheets of the two domains within one subunit ( Fig.  2A). The cofactor is bound via a covalent linkage to the side chain of Lys 44 (Fig. 2B). The phosphate group forms hydrogen bonds to the highly conserved glycine-and threonine-rich loop ( 178 GTGGT 182 ) at the N terminus of helix 6 and three water molecules (Fig. 2, C and D). The aromatic ring of PLP is positioned at a crossover connection in the ␤-sheet of the C-terminal domain. The N-1 atom of the pyridine ring is hydrogen-bonded to the side chain of Ser 266 , located in another highly conserved region ( 264 GISSGAA 270 ) of the polypeptide chain. The 3Ј-hydroxyl group of PLP forms a hydrogen bond to the side chain of Asn 74 from the highly conserved sequence stretch 69 EPTSGNTG 76 and to one water molecule. Related OASS enzymes (14 -17) show basically identical interaction networks with the PLP cofactor.
The Active Site in the CysK1 Holoenzyme-The active site of the CysK1 holoenzyme contains several ligands from the crystallization solution. An MPD molecule is bound in the active site pocket and, with the exception of a hydrogen bond to carbonyl oxygen of Gly 222 , mostly forms hydrophobic interactions with residues lining the pocket. The electron density map further contained a strong spherical electron density close to the internal aldimine linkage and the conserved loop ( 71 TSGNT 75 ) at the N terminus of helix ␣2, which was modeled as a chloride ion, mostly likely originat-ing from the crystallization buffer (Fig. 2, C and D). A similar anion binding site has also been found in other O-acetylserine sulfhydrylases (14,17).
Structure of the CysK1-␣-Aminoacrylate Complex-The formation of the aminoacrylate Schiff base intermediate, the product of the first half-reaction, was observed in solution by spectrophotometry. The UV-visible spectrum of CysK1 in the presence of 300 M OAS (Fig. 1) showed the absorbance maxima at 321 and 469 nm, which are characteristic for the formation of the ␣-aminoacrylate intermediate (23,38). The subse-quent addition of 0.25 mM sodium sulfide resulted in the fast disappearance of the band at 469 nm, indicating decomposition of the intermediate.
The ϳ60 nm shift of the absorbance maximum in the spectrum corresponds to a visible color change, which was also observed in the crystals during soaking experiments with OAS. Diffraction data collected from such a crystal after arresting the reaction by flash-freezing were used to solve the structure of CysK1 with the trapped ␣-aminoacrylate intermediate. Most notably, parts of the N-terminal domain had moved significantly upon formation of the reaction intermediate and had to be rebuilt completely. The largest conformational changes (maximal atomic displacements of 6 Å) involve the loop region consisting of the conserved sequence stretch 71 TSGNT 75 , the loop between strand ␤5 and helix ␣3 (residues 94 -101) and between strand ␤6 and helix ␣4 (residues 117-124) that fold over the active site (Fig. 3A). The changes in cell dimensions in the crystals containing the reaction intermediate most likely reflect the conformational changes of this part of the polypeptide chain. The domain rotation (Fig. 3A) results in the closure of the active site, which becomes less accessible and thus protects the reaction intermediate from solvent.
The electron density map clearly showed that the internal Schiff base linkage between PLP and the catalytic lysine residue was broken and that the external Schiff base with the ␣-aminoacrylate had formed in the crystal (Fig. 3B). We did not observe any electron density beyond the C␤-carbon atom of the external Schiff base, indicating that the acetate moiety had been expelled from the substrate, OAS. The carbon atoms of the aminoacrylate moiety and the nitrogen involved in the Schiff base linkage are planar, as expected for the aminoacrylate intermediate. One of the oxygen atoms of the carboxylate group is within hydrogen-bonding distance to the side chains of Gln 144 and Thr 71 and the main chain nitrogen atom of Ser 72 . The second oxygen atom is tethered to the backbone amide groups of Asn 74 and Thr 75 (Fig. 3, C and D). Most of these residues are part of the anion binding site in the holoenzyme, but as a result of the loop movement, they are closer to the active site and interact with the carboxyl group of the aminoacrylate.
The formation of the reaction intermediate results in a number of subtle changes in the active site. The pyridine ring of PLP is tilted by ϳ19°along an axis passing through the 2Ј-5Ј-carbon atoms. A small shift in the position of the phosphate group was found (0.5 Å), which leads to slightly different interactions when compared with the structure of the holoenzyme. One of the phosphate oxygen atoms forms a hydrogen bond with the ⑀-amino group of the catalytic residue Lys 44 , and it is likely that the protonated state of the Lys 44 , crucial in the second halfreaction, is ensured by this interaction. Another phosphate oxygen atom acquires a water molecule as a new hydrogenbonding partner. In total, the number of hydrogen bonds of the phosphate group increases from seven to nine and is thus firmly anchored in the active site.
In the structure of the CysK1-aminoacrylate complex, the C␤-carbon atom points into the active site pocket and is the only atom of the intermediate accessible for a nucleophilic attack by hydrogen sulfide through a narrow tunnel extending from the enzyme surface (Fig. 4). The ⑀-amino group of Lys 44 is within 4.3 Å distance of the C␣-atom of the aminoacrylate but can easily adopt a position suitable for the proposed role in protonation of the C␣-carbon atom (20, 24) by a simple rotation along the C␦-C⑀ carbon bond. Lys 44 is positioned such that protonation of the C␣-carbon can only occur from the si-face of the intermediate. The channel from the surface leading to the C␤-carbon atom (Fig. 4) directs the nucleophilic attack of the sulfide toward the re-face of the intermediate, leading to an anti-addition, resulting in the formation of L-cysteine.
In previous attempts to obtain structures of O-acetylserine sulfhydrylases with bound substrate and intermediates, active site mutants where the catalytic lysine residue had been replaced by alanine were employed. The structure analysis of such mutants from S. typhimirium (25) and A. thaliana (17) revealed that an external aldimine had formed via a reaction with a methionine residue during protein production (21). This structure represented a mimic of the first intermediate, the species preceding the formation of the ␣-aminoacrylate in the reaction cycle of OASS (compound II in Scheme 1). Although the methionine complex of OASS from A. thaliana (crystallized in 4.5 M sodium formate) did not show any major structural rearrangements upon ligand binding, the S. typhimirium (crystallized at pH 7.0, Tris buffer) enzyme showed a similar conformational change as seen in CysK1 from M. tuberculosis. The external methionine Schiff base adduct in S. typhimurium OASS and the aminoacrylate Schiff base in M. tuberculosis CysK1 overlap well upon superposition, in particular the C␣, C␤-carbon atoms and the carboxyl group (supplemental Fig.  S2A). Consistent with the key role of Thr 71 , Asn 74 , Ser 72 , and Gln 144 in recognizing and binding the carboxyl group of the aminoacrylate intermediate as suggested by the structure analysis, replacement of these amino acids in OASS from A. thaliana leads to mutants with a 10 3 -10 5 -fold lower catalytic efficiency than wild-type enzyme (17).
It is of interest to note that in tryptophan synthase from S. typhimurium, the only other member of the fold type II PLP enzymes with a known structure of an aminoacrylate external Schiff base (42), the interactions of the intermediate with the enzyme are rather different from that in CysK1 and are not conserved (supplemental Fig. S2B). Interactions are made to the side chain of Thr 110 and backbone nitrogens from Ala 112 , Gly 113 , and Gln 114 . These residues are structurally equivalent to the substrate binding loop (residues 71-75) of OASS enzymes but are not conserved. A notable difference in the recognition and binding of the aminoacrylate moiety in CysK1 and tryptophan synthase is the absence of Gln 144 in the latter. The opening of the hydrophobic channel that connects the two active sites in the heterodimeric tryptophan synthase is located at approximately this position and allows access of indole, formed in the ␣-subunit, to the PLP cofactor.
M. tuberculosis CysK1 Interacts with a SAT-derived C-terminal Peptide-OASS associates with SAT to form a bi-enzyme complex, in which the OASS activity is down-regulated (43,44). Crystallographic studies of OASS from H. influenzae and A. thaliana (15,45) have shown that an essential part of the interactions between the two enzymes most likely involves binding of the C-terminal four residues of SAT to the active site of OASS. We therefore addressed the question whether CysK1 from M. tuberculosis is able to form similar interactions with SAT and whether binding of a peptide derived from the C terminus of M. tuberculosis SAT (CysE) would lead to inhibition of CysK1. The addition of a peptide comprising the last four residues (DFSI) of CysE to CysK1 gave a shift in the thermal denaturation curve of the enzyme by 5°C. Inhibition studies showed the competitive nature of the inhibition with a K I value of 5.0 M (Fig. 5). These observations indicated formation of an enzyme-peptide complex, and the DFSI peptide was therefore used for co-crystallization.
The electron density maps calculated from x-ray data collected on crystals of CysK1 obtained in the presence of 1.0 and 4.0 mM DFSI peptide clearly showed that the peptide was indeed bound to CysK1 (Fig. 6, A and B). The peptide is bound in the active site cleft between the two domains and extends from the enzyme surface into the interior of the active site. It blocks the active site cleft completely and also acts like a wedge, preventing complete domain rotation seen in the enzyme-intermediate complex. CysK1 is thus trapped in an open conformation similar to the holoenzyme.
The peptide interacts with CysK1 through direct hydrogen bonds as well as hydrophobic interactions (Fig. 6A). Two main chain atoms of the peptide are engaged in hydrogen bonds. The carbonyl oxygen of Ser 3 * (the asterisk denotes residues from the DFSI peptide) interacts with the main chain nitrogen atom from Gly 222 , and the amide of Ser 3 * forms an indirect hydrogen bond to the carbonyl oxygen atom of Ile 223 via a water molecule. Residue Asp 4 * is firmly held in place through interactions with the side chain of Lys 215 and the main chain amide of Met 122 . The side chain of Phe 2 * binds to a hydrophobic pocket lined by Met 122 , Phe 145 , Ala 225 , and Phe 227 . Finally, Ile 4 * is buried in the interior of the active site and packs against Phe 145 , Gly 178 , Gly 222 , and Ala 225 . The carboxylate group of Ile 4 * occupies the position of the carboxylate of the aminoacrylate moiety in the CysK1-intermediate complex. Hydrogen bonds are formed to main chain nitrogen atoms from residues of the conserved sequence stretch 71 TSGNT 75 , the serine loop. Additional hydrogen bonds are made to the side chain of Gln 144 .
A comparison of the CysK1-DFSI complex with similar peptide complexes of OASS from H. influenzae and A. thaliana (15,45) reveals that peptide recognition is mediated via side chain interactions in a sequence-specific manner. The only enzyme-peptide interaction common to the three enzymes is the binding mode of the conserved C-terminal isoleucine residue via specific hydrogen bonds to its carboxyl group and hydrophobic interactions with the side chain. The residues of the remaining peptide residues are either not conserved, or in cases where they are, their interactions with the enzyme are different. For instance, Ser 3 * corresponds to asparagine in H. influenzae or valine in A. thaliana, and the hydrogen bond formed between the side chains of Ser 3 * and Ser 72 is specific for the M. tuberculosis Cysk1-peptide complex. The N-terminal peptide residue Asp 1 *, although conserved in A. thaliana, has a different conformation and interacts with different enzyme residues.
In conclusion, the strong inhibition and the unique interactions of the peptide with CysK1 suggest this complex to be an attractive scaffold for the rational design of CysK1 inhibitors  specific for M. tuberculosis. Furthermore, we have shown that CysK1 is able to interact with the C-terminal peptide from SAT in a similar manner as seen in OASSs, where formation of a bi-enzyme complex has been shown. This suggests that the M. tuberculosis enzyme may form such a complex in vivo and be subject to a similar mechanism of regulation.