Sulfur Mobilization in Cyanobacteria

Sulfur mobilization represents one of the key steps in ubiquitous Fe-S clusters assembly and is performed by a recently characterized set of proteins encompassing cysteine desulfurases, assembly factors, and shuttle proteins. Despite the evolutionary conservation of these proteins, some degree of variability among organisms was observed, which might reflect functional specialization. l-Cyst(e)ine lyase (C-DES), a pyridoxal 5′-phosphatedependent enzyme identified in the cyanobacterium Synechocystis, was reported to use preferentially cystine over cysteine with production of cysteine persulfide, pyruvate, and ammonia. In this study, we demonstrate that C-DES sequences are present in all cyanobacterial genomes and constitute a new family of sulfur-mobilizing enzymes, distinct from cysteine desulfurases. The functional properties of C-DES from Synechocystis sp. PCC 6714 were investigated under pre-steady-state and steady-state conditions. Single wavelength and rapid scanning stopped-flow kinetic data indicate that the internal aldimine reacts with cystine forming an external aldimine that rapidly decays to a transient quinonoid species and stable tautomers of the α-aminoacrylate Schiff base. In the presence of cysteine, the transient formation of a dipolar species precedes the selective and stable accumulation of the enolimine tautomer of the external aldimine, with no formation of the α-aminoacrylate Schiff base under reducing conditions. Effective sulfur mobilization from cystine might represent a mechanism that allows adaptation of cyanobacteria to different environmental conditions and to light-dark cycles.

Sulfur is required for the synthesis and modification of a variety of biomolecules such as Fe-S clusters, thiamine, biotin, lipoic acid, and thiouridine (1,2). Persulfides constitute the biologically reactive form of sulfur mobilized from cysteine (2)(3)(4). The terminal sulfur in persulfide groups is labile, being rapidly released as elemental sulfur or sulfide (2, 4 -6), and requires carrier proteins for stabilization and transport to acceptor molecules (7,8). Among others, cysteine desulfurases are a well characterized family of enzymes that catalyze the removal of sulfur from cysteine with the formation of alanine. The first member of this group of enzymes to be studied was the NifS (cysteine desulfurase) protein from Azotobacter vinelandii (9), which now represents the prototype of a large family encompassing all three domains of life, the NifS-like proteins. Bacterial cysteine desulfurases are pyridoxal 5Ј-phosphate (PLP) 2 -dependent enzymes belonging to fold type I (1,10) and are classified into two groups, named I and II, based on sequence similarity (11). NifS-like proteins share a distinctive reaction mechanism with the nucleophilic attack of a conserved cysteine residue on the sulfhydryl group of the ketimine intermediate (12)(13)(14)(15). The cysteine persulfide is used as a building block for the synthesis of sulfur-containing biomolecules, likely through the intervention of intermediate acceptors that function as assembly scaffolds, such as NifU (16 -19) and IscU (20 -22), or shuttle proteins, such as SufE (23)(24)(25). It is presently well assessed that proteins involved in Fe-S cluster biosynthesis are organized in self-regulating macromolecular complexes (24,26). Mechanisms underlying sulfur mobilization and Fe-S cluster assembly are well conserved in all organisms, nonetheless showing some degree of variability (1). The cyanobacterium Synechocystis is an oxygenic photosynthetic microorganism and constitutes an interesting system for the investigation of enzymes involved in the response to oxidative stress (27). Its genome has recently been sequenced (28), and the differential gene expression in response to stress-inducing modifications of the environment has been studied (Ref. 29 and references therein). The number and function of proteins involved in the assembly of Fe-S clusters in Synechocystis and the gene organization of their coding sequences suggest the existence of distinct pathways for Fe-S cluster assembly in cyanobacteria with respect to those identified in other organisms (11,30). In the genome of Synechocystis, at least five open reading frames were identified coding for proteins involved in Fe-S cluster assembly (28,31) as follows: slr0387 (12,30,32), sll0704 (33), slr0077 (14,30,34), ssl2667 (18,27,30,35), and slr2143 (12,30). The first three open reading frames code for NifS-like proteins, whereas the fourth one codes for a small protein, Syn-nifU, that is considered an assembly factor for Fe-S clusters (18). The product of slr2143 has been isolated from Synechocystis sp. PCC6714 and has been given the name of cyst(e)ine C-S lyase (C-DES) (36). C-DES is a homodimeric PLP-dependent enzyme distantly related to the NifS family and belongs to fold type I (8). In the presence of L-cysteine and iron, C-DES is able to reconstitute ferredoxin (36) and biotin synthase (37) from the respective apo-forms. However, based on reactivity studies (36) and on investigations with substrate analogues (38), cystine is indicated as the natural substrate of the enzyme. Unlike cysteine desulfurases, C-DES does not have a conserved cysteine residue in the active site (11) and was proposed to catalyze a typical ␤-elimination reaction on both L-cysteine and L-cystine with formation of hydrogen sulfide and cysteine persulfide, respectively, pyruvate and ammonia (36,38). A detailed description of the reaction mechanism, including the observation of intermediates in the presence of either of the two substrates, is still lacking. A tentative catalytic mechanism was proposed on the basis of the enzyme three-dimensional structure (8) and single crystal polarized absorption microspectrophotometric studies (39). In this work, we have carried out spectroscopic measurements to determine the reaction mechanism of C-DES with L-cystine and L-cysteine. In addition, an analysis of sequences of C-DES homologues, retrieved in genomic data bases, has allowed the identification of conserved residues that might have a specific role in cyanobacterial enzymes.

EXPERIMENTAL PROCEDURES
Chemicals-Chemicals were purchased from Sigma and were used as received. L-Cysteine was dissolved into a deoxygenated buffer just prior to use in order to minimize the oxidation to cystine. L-Cystine stock solutions were prepared by dissolving the powder in 1 N HCl to a final concentration of about 200 mM. The stock solution was diluted prior to use in the reaction buffer, and the pH was adjusted as needed.
Protein Expression and Purification-An N-terminal modified derivative of C-DES from the cyanobacterium Synechocystis sp. PCC 6714 (8) was expressed and purified as described previously (38). The enzyme was kept at Ϫ80°C as a stock solution in 10 mM MOPS buffer, pH 6.5. The absorption of a 1% C-DES solution at 280 nm is 14.5 (38), which corresponds to a molar extinction coefficient at 426 nm of 7756 M Ϫ1 cm Ϫ1 .
Sequence Analysis and Multiple Sequence Alignments-The amino acid sequence of C-DES from Synechocystis was used to search nonredundant sequence data bases using BLAST software (40) at the National Center for Biotechnology Information. Retrieved sequences were aligned using the ClustalW program (41). The sequence conservation pattern was visualized using the program ESPript (42).
Spectrophotometric Measurements-Absorbance spectra were collected in 50 mM Bicine buffer, pH 8. Temperature was maintained at 20 Ϯ 0.5°C with a circulating water bath. Spectra in the presence of substrates were taken either in deoxygenated conditions (i.e. under helium flux) or, when feasible, in the presence of DTT, to prevent cysteine oxidation and the development of turbidity as a consequence of S 0 precipitation. Spectra were corrected for buffer contribution.
The dependence of the absorbance at a defined wavelength on substrate concentration was fitted to a binding isotherm, as shown in Equation 1, where A and A 0 are absorbance values at a given ligand concen-tration and in the absence of ligand, respectively; B is the total amplitude; [L] is the concentration of the ligand, and K d is the dissociation constant. Rapid Scanning and Single Wavelength Stopped-flow Spectroscopy-Kinetic experiments were carried out in 50 mM Bicine buffer, pH 8. A 34 M C-DES solution was mixed with either 4 mM cysteine or 2 mM cystine solutions. The temperature of the loading syringes and the stopped-flow cell compartment was maintained at 20 Ϯ 0.5°C with a circulating water bath.
Rapid scanning stopped-flow kinetic measurements were carried out using an SX-18 MV apparatus (Applied Photophysics) equipped with a 150-watt Xenon lamp and coupled to an MS 125TM 1/8-m spectrograph and Instaspec II photodiode array (Lot-Oriel). Spectra in the presence of cystine were collected between 320 and 550 nm every 4 ms and while in the presence of cysteine between 336 and 500 nm every 1.2 ms and Single wavelength stopped-flow kinetic experiments were performed with the same apparatus, using a 75-watt Xenon lamp as a light source and a photomultiplier as a detector. Kinetic traces were collected at 345, 362, and 420 nm in the presence of cysteine and at 338, 362, 420, and 470 nm in the presence of cystine. 1200 -1400 data points were recorded on a logarithmic scale with a total acquisition time of about 500 ms and 2 s for cysteine and cystine, respectively. For both set ups the instrumental dead time was 1.56 ms.
Pre-steady-state Kinetic Data Analysis-Single wavelength kinetic traces were fitted independently and simultaneously using a sum of exponential functions, as shown in Equation 2, where A and A 0 are the absorbance values at a given time and at time t3 ∞, respectively; A i values are pre-exponential factors, and i values are the relaxation times.
The time-resolved difference spectra, obtained by subtracting the spectrum of the free enzyme to the RSSF spectral set, were analyzed by singular value decomposition (SVD) (43,44) using the program MATLAB (The Mathworks, Inc. Natick, MA). Each experiment provided a (m ϫ n) data matrix (the A matrix), where m is the number of wavelengths, and n is the number of spectra collected. The data matrix was resolved into a product of three matrices, usually named U, S, and V, as shown in Equation 3.
The (m ϫ n) U matrix consists of n orthonormal eigenvectors; S is a square diagonal matrix, and V T is a (n ϫ n) matrix whose rows are also orthonormal. The columns of U are basis spectral components that combine to yield each observed spectrum with a contribution weighted by its singular value and a time evolution derived from the corresponding row in V T matrix. It is important to stress that each column of U does not represent the spectrum of a pure component (45). The first criterion for the selection of usable components is the magnitude of the singular values, the higher values being the meaningful ones. The selected components can be further screened by evaluating the autocorrelations of the corresponding columns of U and V and rejecting the component if either autocorrelation falls below 0.8 (43). The columns of V were fitted to sums of exponentials (Equation 2).

Identification of C-DES Homologues-A
Blast search of nonredundant sequence data bases with Synechocystis C-DES as a query gave low scoring matches with sequences from bacteria, fungi, and plants. Most of these sequences showed identities to C-DES lower than 30% and are annotated as SufS or NifS homologues. A low similarity of C-DES to NifS proteins has been reported previously (38), allowing the inclusion of C-DES into the NifS family following the relaxed definition of Ouzounis and Sander (46). The alignment of A. vinelandii NifS sequence with that of C-DES is reported in Fig. 1 (group 2). Interestingly, the search gave highly significant matches with entries from cyanobacteria. Sixteen sequences displayed a similarity higher than 50% and an identity higher than 40%, up to 55%. The 16 high scoring matches are annotated as cysteine/cystine lyases (including putative proteins and homologues; 12 entries), class V aminotransferases (two entries), selenocysteine lyase (see Ref. 46), or isopenicillin-N epimerase. A representative selection of sequences is reported in Fig. 1 (group 1).
Analysis of the C-DES three-dimensional structure ( Fig. 2) (8) with the CASTp program (47) revealed that many of the conserved residues belong to the active site ( Fig. 2) and are involved in cofactor and substrate/product binding (8). In particular, the active site lysine that binds the cofactor (Lys-223 in the Synechocystis sequence) is conserved among all the sequences, together with a string of surrounding residues. Asp-197 and Gln-200, which are responsible for the stabilization of the electronic distribution in the delocalized -system of the cofactor (8), are present not only in sequences of C-DES homologues but also in NifS-like proteins. Phe-25, His-114, Ala-199, and Thr-276, which are part of the cofactor-binding site (Fig. 2) and anchor the phosphate group to the protein matrix, and Arg-360 and Arg-369, which bind the two carboxylates of the substrate cystine (8), are all conserved among cyanobacterial sequences ( Fig. 1). Significantly, the multiple alignment reveals the absence of the catalytic cysteine residue that is conserved among NifS-like proteins (Cys-325 in NifS from A. vinelandii, red square below the sequence in Fig. 1) and suggests a distinct catalytic mechanism for cysteine/cystine lyases. Furthermore, the comparison between the two groups of sequences ( Fig. 1) allows the identification of residues that, being conserved only among C-DES homologues, may have a specific role in cyanobacterial enzymes. Among others, two cysteine residues (Cys-92 and Cys-112 in the sequence of C-DES from Synechocystis) are present only in C-DES-like proteins (Fig. 1), and one of the two arginines responsible for the binding of the two carboxylates of cystine (Arg-360) is absent in group I NifS-like proteins. Cys-112 is localized in an ancillary pocket in communication with the active site through a tunnel that allows the entry of a water molecule as well as of a persulfide group (Fig. 2, B-D). The presence of C-DES homologues in all sequenced cyanobacterial genomes indicates the evolutionary conservation of the distinctive properties of this enzyme, which might respond to specific metabolic requirements of photosynthetic organisms.
Steady-state Spectra of C-DES with Substrates and Substrate Analogues-The absorption spectrum of native C-DES (Fig. 3) shows a major peak centered at 426 nm, attributed to the ketoenamine tautomer of the internal aldimine (Scheme 1) (39). The peak is asymmetric, with a shoulder around 400 nm. A minor peak, centered at around 500 nm, is because of a catalytically inert species generated during the heterologous overproduction in Escherichia coli. 3 Reaction with cystine ( Fig. 3A) leads to the accumulation of a species absorbing at around 470 nm, attributed to the ketoenamine tautomer of the ␣-aminoacrylate Schiff base (Scheme 1A) (39). A broad increase in the absorbance at wavelengths shorter than 400 nm hampers the accurate identification of specific bands and is likely because of the accumulation of reaction products, i.e. cysteine persulfide and pyruvate, which absorb below 350 nm (7,48). Indeed, it was observed that spectra recorded on C-DES crystals suspended in a solution containing cystine exhibited a well defined peak at 338 nm, tentatively assigned to the enolimine tautomer of the external aldimine, with a possible contribution from the enolimine tautomer of the ␣-aminoacrylate (39).
C-DES reacts with cysteine leading to the accumulation of a species absorbing at 345 nm (Fig. 3B). Minor peaks and shoulders are centered at around 400, 430, and 470 nm. Previous studies on C-DES crystals (39) attributed the peak at 345 nm to the enolimine tautomer of the external aldimine. In the presence of the reducing agents dithiothreitol (DTT) or tris(2-carboxyethyl)phosphine (TCEP), the spectrum of the cysteinebound enzyme is stable for at least 1 h, and in its absence the band at 345 nm slowly increases and blue-shifts (data not shown). Furthermore, negligible amounts of the product pyruvate were detected in coupled assays under reducing conditions (data not shown). The dependence of the spectral changes on L-cysteine concentration ( Fig. 3B) was fitted to a binding isotherm (Equation 1) with a dissociation constant of 110 Ϯ 8 M (Fig. 3B, inset). Even if the dissociation constant for cystine could not be measured because of the fast accumulation of reaction products, steady-state measurements have provided for K m an estimate of 1 mM (38). Interestingly, a higher affinity of C-DES for cystine with respect to L-cysteine is suggested by the direct competition for binding to the active site. In fact, a solution of C-DES, saturated with 2 mM cysteine, immediately acquires the spectral features of cystine-bound C-DES when 1 mM cystine is added to the reaction mixture (data not shown).
The spectral changes accompanying the formation of the external aldimine were investigated using the substrate analogues L-serine and L-glycine, the latter being known from crystallographic data to form an external aldimine adduct in an inactive mutant of C-DES (39). Both serine and glycine (Fig. 3C) induce a red shift of the visible absorbance band from 426 to ϳ434 nm that indicates the accumulation of the ketoenamine tautomer of the external aldimine. Even though serine induces a small increase in the absorbance at around 330 nm, the accumulation of the enolimine tautomer of the external aldimine is not appreciable with either of the two amino acids.
Single Wavelength Stopped-flow Studies of the Reaction of L-Cystine and L-Cysteine with C-DES-The pre-steady-state kinetics of the reaction of C-DES with 1 mM cystine was monitored at 338, 420, and 470 nm by stopped-flow (Fig. 4, left panels). The absorption band of the internal aldimine at 420 nm disappears with a concomitant increase of absorbance at 338 nm. At 470 nm there is a fast accumulation of a species, followed first by a slow decrease and, at times longer than 2 s, by a further absorbance increase (data not shown). The fitting of the time courses with a biexponential function gave similar relaxation times independent of the analyzed wavelength. Relaxation times for the fast and slow phase, calculated from the global analysis of the traces, are 86 Ϯ 8 and 527 Ϯ 50 ms, respectively ( Table 1).
The reaction in the presence of 2 mM cysteine was monitored at 345, 362, and 420 nm (Fig. 4, right panels). The decrease in the absorbance of the internal aldimine at 420 nm is accompanied by an increase in the absorbance at 345 nm. At 362 nm (Fig. 4), a very fast absorbance increase is followed by a small, slow decrease. This process is barely detectable at this wavelength and becomes even less pronounced at shorter wavelengths. In the presence of cystine no accumulation of species absorbing at 360 nm was detected (data not shown). The time courses can be fitted, both individually and globally, to a biexponential function, with relaxation times of 6.9 Ϯ 0.6 and 41 Ϯ 1 ms (Table 1).
Rapid Scanning Stopped-flow Studies of the Reaction of L-Cystine and L-Cysteine with C-DES-To detect transient and poorly populated species, the time dependence of spectral changes associated with the reaction of cystine with C-DES was further investigated via rapid scanning stopped-flow experiments (Fig. 5A). The first spectra of the kinetic series show a broadening of the internal aldimine band with a small decrease in its intensity because of the progressive formation of the ketoenamine tautomer of the ␣-aminoacrylate Schiff base absorbing at 472 nm. The reaction proceeds with a marked red shift of the visible band and a further decrease in its intensity. The disappearance of the internal aldimine peak is also accompanied by the accumulation of a species with an absorption maximum at around 335 nm, likely the enolimine tautomer of the ␣-aminoacrylate. There is an almost perfect isosbestic point at 375 nm, whereas there are no isosbestic points on the red side of the 426 nm band, indicating that in this spectral region more than one species is simultaneously contributing to the observed absorbance changes. The time evolution of the spectral changes was analyzed by singular value decomposition (SVD) (43,44). On the basis of the first-order autocorrelations, two spectral compo-  vinelandii, gi 128312), is reported for comparison, and for clarity, the sequence of C-DES was omitted from display. Identical residues across the two groups have a red background; residues with similar physicochemical properties within groups are shown in red; and similar residues across the groups are boxed. Similarity scores were calculated by the ESPript program (42) using the Blosum62 matrix set at global score of 0.3 and difference score of 0.6. Secondary structure elements indicated above the alignment are from Synechocystis three-dimensional structure (Protein Data Bank code 1ELQ). The first nine amino acids of the Synechocystis sequence are not shown because they are absent in the three-dimensional structure. Sequence is numbered according to Ref. 8, where the first residue corresponds to residue number 13 in the wild type sequence. The active site Lys-223 is indicated by a yellow arrowhead. The catalytic cysteine residue of NifS is indicated by a red square. Some of the residues conserved only among C-DES sequences mentioned in the text are indicated by a blue star. The protein surface is colored blue, and cavities and clefts are colored green, yellow, and orange as a function of depth. One of the two symmetry-related active sites is shown on the right side of the structure with the cofactor and the product cysteine persulfide shown in ball and stick mode colored red and by atom type, respectively. B, close up of the active site where a tunnel connecting the active site to an ancillary pocket is visible, with Cys-112 shown in ball and stick mode at the bottom. C, front view of the active site with the ancillary pocket on the right. The tunnel connecting the two sites is covered by residues belonging to ␤-strands ␤9-␤10 and ␤14-␤15 (see Fig. 1). PLP is shown in ball and stick mode colored red. D, some of the external residues belonging to ␤-strands ␤9-␤10 and ␤14-␤15 have been removed to display the tunnel connecting the active site to the ancillary pocket. Cysteine persulfide in the active site and Cys-112 in the ancillary site are shown in ball and stick mode colored by atom type. For a better visualization of Cys-112 the oxygen of the backbone carbonyl group of Asp-111 has been removed. DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 38773
nents were judged to be significant and retained for subsequent analysis (Fig. 5). The first component U 1 (S 1 ϭ 23) exhibits two maxima at 334 and 488 nm and a minimum at 422 nm (Fig. 5B). A second component U 2 , accounting for only a small fraction (S 2 ϭ 2.4) of the spectral changes, shows a maximum at 489 nm (Fig. 5D). The time courses V 1 (Fig. 5C) and V 2 (Fig. 5E) can be simultaneously fitted using a sum of two exponentials with relaxation times of 76 Ϯ 7 and 357 Ϯ 40 ms ( Table 1).
The reaction of C-DES with cysteine leads, during the instrument dead time, to a marked decrease in the 426 nm band and to the formation of a shoulder around 360 nm (Fig. 6A). The band at 360 nm increases up to 8.8 ms, in agreement with single wavelength kinetics (Fig. 4, right panels). The peak at 360 nm is likely the result of the deprotonation of the phenolic oxygen of the cofactor with formation of a dipolar species, as already observed for the PLP-dependent enzymes glycogen phosphorylase (49,50) and aspartate aminotransferase (51). The reaction proceeds with the accumulation of a species absorbing at 347 nm (Fig. 6A). The SVD analysis of the time-resolved difference spectra for the reaction with cysteine indicates the presence of three significant spectral components, as judged by their singular values and autocorrelations (Fig. 6, B, D, and F). The first component U 1 (S 1 ϭ 24.22) exhibits a maximum at 346 nm and a minimum at 420 nm (Fig. 6B). The second component U 2 (S 2 ϭ 0.65) shows two minima at 338 and 400 nm (Fig. 6D), and the third component U 3 , with a relative weight comparable with that of the second component (S ϭ 0.36), shows essentially only a maximum at 367 nm (Fig. 6F). The global fitting of V 1 (Fig. 6C), V 2 (Fig. 6E), and V 3 (Fig. 6G) gives a fast relaxation time of 7.3 Ϯ 0.4 ms and a slow relaxation time of 42 Ϯ 2 ms (Table 1), in excellent agreement with single wavelength stopped-flow data (Table 1).

DISCUSSION
The catalysis of PLP-dependent enzymes proceeds via a sequence of chemical transformations among different PLPbound derivatives. Each species is characterized by defined spectral properties and is associated with a conformation of the active site stabilizing a specific catalytic intermediate. Versatility and plasticity are the key features that allow the family of PLP-dependent enzymes to carry out an extraordinary variety of chemical reactions with high specificity. These general properties are well evidenced in many enzymes involved in sulfur mobilization and Fe-S cluster formation. Three major cysteine desulfurases classes have been identified in bacteria, based on the genetic context of their coding sequences (4,14). NifS and the housekeeping enzyme IscS belong to group I, following the definition of Mihara et al. (11), whereas SufS, which provides a backup system for iron-sulfur cluster repairs during oxidative stress, belongs to group II. A common reaction mechanism is shared among cysteine desulfurases, with the abstraction of sulfur from the substrate cysteine and formation of an enzyme-bound cysteine persulfide and alanine via a ketimine intermediate.
In Synechocystis, an enzyme distinct from cysteine desulfurases, C-DES, exploits a typical ␤-elimination reaction on cystine to catalyze the formation of free cysteine persulfide, pyruvate and ammonia (38). The recent burst in whole cyanobacterial genome sequencing (29) now allows us to report on rather close homologues to Synechocystis C-DES (31,34). Homologies are also found with NifS group II proteins, with an identity of about 25% (38). Our search of nonredundant sequence data bases retrieved 16 sequences belonging to eight different cyanobacterial genera. Most of the sequences were annotated as cysteine/cystine lyases or selenocysteine lyases, although, not surprisingly, some sequences were annotated as aminotransferases of class V or isopenicillin-N epimerase. As a matter of fact, C-DES has a high sequence and structural homology to class V aminotransferases (8), and it was proposed that NifS and isopenicillin-N epimerase evolved from the common ancestor for the aminotransferases (see Ref. 11 and references therein). Furthermore, the incorrect annotation of C-S lyases as aminotransferases has already been pointed out for plant cystine C-S lyases (52).
The high homology among the sequences retrieved and the widespread occurrence in most sequenced cyanobacterial genomes strongly indicate that C-DES constitutes a new family of sulfur-mobilizing enzymes, distinct from the desulfurases known so far. Furthermore, the pattern of conserved residues might allow the identification of amino acids with functional and/or structural roles. One striking feature is the conservation among cystine lyases of two arginines, namely Arg-360 and Arg-369, both also present in group II NifS proteins (39). The two residues are involved in substrate binding through hydrogen bonding to cystine carboxylates (8) and are a hallmark for the preferential binding of bipolar substrates (see also Refs. 8 and 38). Arg-369 is present in all NifS homologues (11), and in aminotransferases as well (53), where it fulfills the common function of ␣-carboxylate binding.
Cys-92 and Cys-112 are also conserved residues among C-DES-like enzymes and are distinct from the catalytic cysteine of NifS (red square in Fig. 1). Inspection of the C-DES threedimensional structure (Fig. 2) has led to the identification of an ancillary pocket close to the active site where Cys-112 lies with the sulfhydryl group exposed to the pocket surface. The tunnel connecting the active site to the ancillary pocket is narrow, allowing the diffusion of molecules not bigger than a sulfide. On the other hand, the external wall of the tunnel is formed by two SCHEME 1. Mechanism for the ␤-elimination reaction catalyzed by C-DES on L-cystine (A) and L-cysteine (B). Catalytic intermediates identified in the pre-steady-state regime are in boldface. Species that might be present or that are known to be obligatory intermediates, but were not detected in the present study, are depicted in gray. DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 38775
␤-hairpins, one of which constitutes the dimerization clamp (8), whose flexibility could allow a widening of the cavity. We speculate that Cys-112 might be involved in accepting a sulfur from the product cysteine persulfide (Fig. 2, B and C), that has been crystallographically localized in the vicinity of the proposed exit tunnel (Fig. 2) (8). Transfer of the product to a secondary site might avoid steric interference with an incoming cystine substrate molecule and favor the flow of sulfur to acceptor proteins that may bind at the entrance of the ancillary pocket. Moreover, the labile cysteine persulfide would be protected until the acceptor protein docks at the exit tunnel. Stud-ies via site-directed mutagenesis at Cys-112 will test the validity of this hypothesis.
The conservation of cystine lyases among cyanobacteria suggests a role for this protein in photosynthetic organisms, which should lie on the ability of using cystine in place of cysteine for Fe-S cluster assembly or, in general, for sulfur mobilization. It has already been proposed that cystine lyase could provide for sulfur mobilization in oxidized environments associated with the dark-light cycles of photosynthesis (2,32). If this was the case, cystine lyase homologues should be found in plant genomes, with a preferential plastidic localization. There are some reports on plant cystine lyases (52, 54 -56) restricted to Arabidopsis thaliana and Brassica oleracea and some more on the family of C-S lyases, which also comprise aliin lyases and cystathionine lyases (54,57,58). Although recent data demonstrate the conservation of this enzyme in A. thaliana and strongly support the utilization of cystine as the preferential substrate, no evidence of its presence in other plants or of its subcellular localization has been reported so far. Therefore, functional roles of cystine lyases strictly connected to cyanobacterial biology should be considered, e.g. in the versatile adaptation to changes in the environment or in cell differentiation. Even though the growing number of differential gene expression studies on Synechocystis could provide additional information on the role of C-DES, no effects on the expression of slr2143 have been reported in response to salt stress (59), light-dark cycles (60), high light (61,62), redox stress (27,63), and iron deprivation (64).
Catalytic Mechanism of C-DES in the Presence of Cystine-The formation of an external aldimine via a gem-diamine is the first step of the reaction of PLP-dependent enzymes with amino acid substrates. The ketoenamine and enolimine tautomers of the external aldimine absorb at 420 -440 and 330 -340 nm, respectively. Depending on the enzyme investigated, the selected substrate, and the experimental conditions, the external aldimine might stably or transiently accumulate. In the case of C-DES, the external aldimine does not accumulate in the reaction with cystine. Spectra collected under steady-state and pre-steady-state conditions show no evidence of formation of  Table 1.

TABLE 1 Relaxation times () and relative amplitudes (A) for the reaction of C-DES with either cysteine or cystine, obtained from the fitting of single wavelength kinetics and the significant columns of V matrix using Equation 2
Errors on relaxation times are the standard deviations from the mean of at least three experiments.

Cystine Cysteine
Single wavelength kinetics 1 (ms) 2 (ms) 1 63 37 either the gem-diamine or the external aldimine. In fact, the internal aldimine species disappears with the concomitant formation of bands at 338 and 470 nm (Fig. 3A), attributed to the enolimine and ketoenamine tautomers of the ␣-aminoacrylate Schiff base, by analogy to the spectral properties of this intermediate in other PLP-dependent enzymes catalyzing ␤-elimination/replacement reactions (65)(66)(67). Therefore, the formation of the gem-diamine and the external aldimine is followed by a faster conversion to subsequent reaction intermediates. In the first phase of the reaction, the appearance of the 338 and 470 nm bands and the disappearance of the 426 nm band take place with the formation of a single isosbestic point at 375 nm, but no isosbestic point at higher wavelengths. This finding indicates that catalysis proceeds via the chemical transformation of the internal aldimine absorbing at 426 nm to the species absorbing at 338 nm, and to more than a single absorbing species on the red side of the 426 nm band. Indeed, a quinonoid absorbing at 500 nm is unequivocally present in the U 2 component (Fig. 5D), partially mixed with a 470 nm species. This component accounts for only 10% of the spectral changes. Because quinonoid species are endowed with extinction coefficients at least 5-fold higher than other PLP intermediates (see Ref. 68 and references therein), very small changes in their concentration can significantly affect the evolving spectra. Furthermore, the components obtained from the SVD analysis may not be pure spectral species but a mixture of the species evolving with the same time courses. The V 2 time course is consistent with that at 470 nm in single wavelength measurements, suggesting that the observed kinetics is not due to the formation and disappearance of ␣-aminoacrylate but to the transient accumulation of quinonoid, because their spectral signatures almost overlap. The presence, in the first component U 1 , of positive bands absorbing at 338 and 470 nm and a negative peak at 420 nm supports the view that the internal aldimine is converted into distinct tautomers of the ␣-aminoacrylate intermediate. The biphasic nature of the process, detected at single wavelength, as well as for each SVD components, can originate from very different reaction schemes, as pointed out by Fisher and co-workers (see Ref. 69 and references therein) and also formally treated by Bernasconi (70). Assuming that only the ketoenamine tautomer of the internal aldimine reacts with L-cystine with formation of the external aldimine, quinonoid, and the two tautomers of the ␣-aminoacrylate, and that these species are in rapid equilibrium, the observed rate constants include both the forward and the reverse rate constants. Consequently, rates, as well as amplitudes, cannot be attributed to a specific catalytic step. This holds also under the assumptions that C-DES ␤elimination reaction on cystine is essentially irreversible, and the formation of the undetected external aldimine from the corresponding tautomeric mixture of the internal aldimine is at equilibrium. Despite these uncertainties, because ␣-aminoacrylate accumulates in the reaction of C-DES with cystine, the results clearly indicate that the rate of cysteine persulfide formation is faster than the rate of pyruvate and ammonia release.
Catalytic Mechanism of C-DES in the Presence of Cysteine-The experimental data reported in the literature on the reactivity of C-DES with cysteine are somewhat conflicting. In fact, on the basis of apoferredoxin reconstruction assays, it was originally proposed that cysteine could be used by the enzyme for the incorporation of sulfur in Fe-S clusters (36), even though it was immediately clear that cystine was a better substrate (36). Later, analysis of substrate preferences and identification of the reaction product cysteine persulfide (38) suggested a possible role for cystine as the natural substrate of C-DES. Nevertheless, a slow conversion of cysteine to sulfide, pyruvate, and ammonia was inferred from measurements of pyruvate production detected by a coupled optical assay (36). Enzyme assays, under either oxidizing or reducing conditions, e.g. absence of oxygen or presence of DTT, strongly indicate that cysteine is an extremely poor substrate of C-DES and that the observed catalytic activity is at least partly because of the slow oxidation of cysteine to cystine. In the simplest scenario, the almost complete loss of activity in the presence of DTT or TCEP could be a direct effect on the inhibition of cysteine oxidation. Nevertheless, as already demonstrated for some other PLP-dependent enzymes (67,71), the inhibition could be the consequence of the attack of nucleophilic agents, like DTT, on the ␣-aminoacrylate intermediate with formation of amino acids as those reported previously (71). The last hypothesis seems the more  Table 1.
feasible, considering the superimposing catalytic activities obtained in the presence of oxygen or under helium atmosphere. Furthermore, a band centered at 345 nm forms in the reaction of C-DES with cysteine both in the presence and absence of DTT. Taken together these results strongly suggest that ␣-aminoacrylate does not accumulate in the reaction with cysteine and that the band at 345 nm is attributable to the external aldimine. The external aldimine formed upon reaction with cysteine is spectrally different from those formed with substrate analogues like glycine and serine (Fig. 3C). The selective accumulation of the enolimine tautomer of the external aldimine in the presence of cysteine is likely a consequence of the stabilization of a less polar active site conformation (39).
Experimental evidence obtained under pre-steady-state conditions is fully consistent with steady-state measurements. The SVD analysis shows the presence of three significant compo-nents with the second and the third ones accounting for only about 3 and 1.5%, respectively. The first component, which predominantly contributes to the overall observed spectral changes, exhibits a positive band at 345 nm and a rather broad negative band at 420 nm. The second component accounts for a very small increase in absorbance at wavelengths lower than 340 nm, and the third component accounts for the transient species detected at 360 nm by single wavelength kinetics and originating from the 420 nm species. The transiently accumulated 360 nm band is likely attributable to a dipolar species of the cofactor directly formed from the 420 nm band (49,50,(72)(73)(74). Significantly, PLP-dependent enzymes where the dipolar species is predominant have evolved a network of ionic interactions aimed at stabilizing the positive charge on the pyridinium ring and the deprotonated phenolic oxygen (50,75). Two conserved residues in C-DES (Asp-197 and Gln-200), which are at hydrogen-bonding distance from the cofactor (8), may provide the structural elements for the stabilization of the dipolar species. A similar pattern of residues around the cofactor has already been observed in some NifS homologues (53,76,77). No 470 and 500 nm bands are present in the spectral components, indicating that both the ␣-aminoacrylate and quinonoid species do not accumulate.
Both the single wavelength traces and the amplitudes of the basis spectra retrieved by SVD analysis show a biphasic behavior, well described by lifetimes of about 7 and 41 ms. These rates are almost an order of magnitude higher than those measured in the reaction with cystine, although in this case the substrate concentration might not be saturating. Despite the apparent faster reaction rates, cysteine is a poor substrate, and the removal of the ␣ proton from the external aldimine and the release of sulfide to eventually give ␣-aminoacrylate are ratelimiting. The better properties of cysteine persulfide as a leaving group with respect to sulfide, based on acid-base characteristics (78,79), could account for the observed behavior. As stated above, even if this reaction appears to be simpler than that with cystine, the two distinct kinetic phases cannot be directly attributed to defined catalytic steps (69). Finally, it should be noted that the specific activity of C-DES with cysteine is only marginally lower than that described for group II desulfurases, usually regarded as very modest catalysts, whereas the activity with cystine is exceedingly higher than the activity reported for cysteine desulfurases of the NifS family (32,34,36,80).

Comparison between C-DES and Other PLP-dependent Enzymes Catalyzing ␤-Replacement/Elimination Reactions-
The reaction chemistry of the PLP-dependent enzymes catalyzing ␤-elimination/␤-replacement reactions tryptophan synthase (see Refs. 66 and 81 and references therein), O-acetylserine sulfhydrylase (see Ref. 65 and references therein), cystathionine ␤-synthase (see Refs. 82 and 83 and references therein), tryptophan indole-lyase and tyrosine phenol-lyase (see Ref. 84 and references therein) has been determined from pre-steady-state studies. For the first three enzymes, the catalytic reaction is composed by two distinct half-reactions as follows: the ␤-elimination is followed by a ␤-addition where nucleophilic agents, indole, sulfide, and homocysteine, respectively, react with the ␣-aminoacrylate Schiff base to form the final product, L-tryptophan, L-cysteine, and L-cystathionine, respectively. For the latter two enzymes, the ␤-elimination  7). B and D, the two significant basis spectra of the U matrix multiplied by their singular values. C and E, the corresponding columns of the V matrix, which represent the amplitude of the basis spectra as a function of time. Gray lines are the results of simultaneous fits to the data using a sum of exponentials (Equation 2). Fitting parameters are reported in Table 1. leads to the final products indole and ammonium pyruvate and phenol and ammonium pyruvate, respectively.
The accumulation of catalytic intermediates and the stabilization of distinct tautomers of the cofactor have been found to be critically influenced by the protein matrix surrounding the coenzyme that confers specificity toward the substrate and modulates the reactivity of PLP. With the exception of yeast cystathionine ␤-synthase, the external aldimine accumulates at detectable levels in the reactions of the above-mentioned PLPdependent enzymes. Furthermore, in the reverse reaction of the truncated form of cystathionine ␤-synthase, the tautomers of the ␣-aminoacrylate, absorbing at 320 and 460 nm, concomitantly form, as observed in C-DES in the presence of cystine. In C-DES the role played by the protein matrix in modulating specificity and reactivity toward substrates is also manifested in the discrimination between cysteine and cystine leading to the productive accumulation of the ␣-aminoacrylate Schiff base only in the presence of the latter ligand. These observations are reinforced by the very low affinity of cysteine for C-DES (K d ϭ 100 M, Fig. 2B), if compared with the affinity of the amino acid for cysteine desulfurases as, for example, the SufS homologue slr0077, with a dissociation constant of the order of 10 M (14). Furthermore, tight substrate selection prevents the desulfuration of cysteine via a mechanism that would ultimately lead to the accumulation of the toxic molecule hydrogen sulfide (4). Evolution in cyanobacteria of two distinct mechanisms for sulfur mobilization from cysteine and cystine in the versatile sulfane form might allow adaptation to fluctuating conditions associated with light and dark cycles to changes in the ecosystem and to cell differentiation (85).