Structural Analysis of an l-Cysteine Desulfurase from an Ssp DNA Phosphorothioation System

Apart from its roles in Fe-S cluster assembly, tRNA thiolation, and sulfur-containing cofactor biosynthesis, cysteine desulfurase serves as a sulfur donor in the DNA PT modification, in which a sulfur atom substitutes a nonbridging oxygen in the DNA phosphodiester backbone. The initial sulfur mobilization from l-cysteine is catalyzed by the SspA cysteine desulfurase in the SspABCD-mediated DNA PT modification system. By determining the crystal structure of SspA, the study presents the molecular mechanism that SspA employs to recognize its cysteine substrate and PLP cofactor. To overcome the long distance (8.9 Å) between the catalytic Cys314 and the cysteine substrate, a conformational change occurs to bring Cys314 to the vicinity of the substrate, allowing for nucleophilic attack.

C ysteine desulfurase, a pyridoxal phosphate (PLP)-dependent homodimer, strips sulfur from the L-cysteine substrate to generate L-alanine and a protein-bound persulfide intermediate on the active site cysteine residue. The persulfide sulfur is subsequently incorporated into a variety of sulfur-containing biofactors, such as tRNA thionucleotides, biotin, molybdopterin, lipoic acid, and iron-sulfur (Fe-S) clusters, which are essential for biosynthetic processes (1,2). Cysteine desulfurase activity was first characterized for the NifS protein from the nitrogen fixation (NIF) system of Azotobacter vinelandii and later found in sulfur formation (SUF) and iron-sulfur cluster assembly (ISC) machineries as the initial stage of sulfur trafficking catalyzed by the paralogous cysteine desulfurases, SufS (also known as CsdB) and IscS, respectively (3)(4)(5).
As a versatile sulfur donor, cysteine desulfurase activity is also essential to DNA phosphorothioate (PT) modification, in which the nonbridging oxygen atom in the DNA sugar-phosphate backbone is replaced by sulfur (6,7). The "writing" of a PT modification into DNA occurs in a sequence-selective and R P configuration-specific manner governed by Dnd or Ssp machineries in bacteria and archaea (6,8,9). Dnd modification systems consist of five proteins (DndA, B, C, D, and E) and confer PTs in 4-bp complementary motifs, e.g., 5=-G PS AAC-3=/5=-G PS TTC-3= (PS, phosphate-sulfur linkage) in Escherichia coli B7A and 5=-G PS ATC-3=/5=-G PS ATC-3= in Hahella chejuensis KCTC2396 (10,11). DndA is a cysteine desulfurase and can be functionally substituted by IscS in vivo, agreeing well with the observation that some bacterial genomes possess clustered dndBCDE but not dndA (12,13). The catalytic cysteine in DndA undergoes nucleophilic reaction with the sulfur atom of its cysteine substrate to form an activated persulfide and then transfers the sulfur to the [4Fe-4S] cluster of the DndC protein (13,14). DndC exhibits ATP pyrophosphatase activity and shows significant sequence homology to phosphoadenosine phosphosulfate reductase (13). DndB functions as a transcriptional regulator capable of sensing cellular ATP levels to control the transcription of the dndBCDE cluster (15,16). DndD has ATPase activity and has been proposed to provide energy during sulfur incorporation (17). DndE adopts a tetramer conformation and displays preferred binding affinity for nicked double-stranded (ds) DNA in vitro (18). Owing to the nuclease tolerance of PT linkage, DndABCDE coupled to DndFGH constitutes a defense barrier analogous to a methylation-based restriction-modification system, i.e., DndFGH restricts non-PT-modified invading DNA (19).
Interestingly, we recently characterized a new type of Ssp system that exhibits different genetic organization, biochemical functions, and phenotypic behavior from Dnd systems. In contrast to the typical dsDNA PT modification mediated by Dnd systems, Ssp systems confer single-stranded (ss), high-frequency DNA PT modification (9). For instance, SspABCD confers ssDNA PT modification to 3-bp consensus sequences, e.g., 5=-C PS CA-3= in Vibrio cyclitrophicus FF75, and the PT levels are 3-to 10-fold higher than those in DndBCDE-expressing E. coli B7A and H. chejuensis KCTC2396 (10,20). In contrast to the simple self-nonself discrimination mechanism in the Dnd system, SspABCD functions together with SspE to provide protection against phage invasion in an unusual PT-dependent manner (9). In parallel with the companion ssDNA PT modification, SspE exerts its toxicity by introducing nicking damage on phage DNA and consequently impairs phage replication and disturbs phage propagation (9). Additionally, the redox and nucleophilic properties of PT sulfur render the PT modification a versatile player in maintenance of cellular redox homeostasis, epigenetic regulation, and environmental fitness (21,22).
We have recently determined the crystal structures of SspB and SspE in the Ssp systems, elucidating the essential role of SspB as a nickase in the ssDNA PT formation and the dual functions of SspE as a PT-stimulated nucleoside triphosphatase (NTPase) and a nicking endonuclease in phage resistance (9). To provide further insights into Ssp-mediated PT modification, we determined the crystal structure of SspA from V. cyclitrophicus FF75 at a 1.80-Å resolution. Our biochemical and structural studies provide the molecular details of how SspA recognizes the cysteine substrate and the PLP cofactor. Moreover, modeling and experimental data show that SspA undergoes a ( Fig. 1C), prompting us to investigate the mechanism of sulfur transfer in the Ssp PT system. The catalytic cysteine, Cys314, of SspA, corresponding to Cys327 in Streptomyces lividans DndA and Cys328 in E. coli IscS, is responsible for nucleophilic attack on the cysteine substrate (14,23). Indeed, the C314S point mutation in SspA abolishes d(C PS C) PT modification of V. cyclitrophicus FF75 (9). Here, we determined the crystal structure of the C314S mutant of SspA from V. cyclitrophicus FF75 in complex with its natural substrate, cysteine, at the resolution of 1.80 Å (see Table S1 in the supplemental material).
In this structure, SspA forms a symmetric dimer ( Fig. 2A), and its dimeric organization resembles that of other cysteine desulfurases, such as DndA and IscS. The fold of SspA is also similar to that of DndA (PDB code 3VAX) and IscS (PDB code 1P3W) with root mean square deviation (RMSD) values of 1.176 Å and 1.052 Å for 236 and 258 aligned C␣ atoms, respectively (Fig. 1B). The SspA protein can be divided into two regions as follows: a larger N-terminal region (residues 1 to 254) possessing the PLP cofactor-binding site and a smaller C-terminal region (residues 255 to 348) bearing the Cys314 active site (Fig. 2B). The large region mostly consists of a seven-stranded parallel ␤-sheet flanked by seven ␣-helices and several tightly packed ␣-helices. The small region harbors a three-stranded antiparallel ␤-sheet flanked by four ␣-helices. The distance between Cys314 and the PLP cofactor is approximately 14.4 Å, and the distance between Cys314 and the cysteine substrate is 8.9 Å (Fig. 2B).
Active site structure and the substrate-binding mechanism of SspA. In the structures of E. coli CsdB (PDB code 1C0N) and Synechocystis sp. strain PCC 6803 SufS (PDB code 1T3I), the catalytic cysteines are located on a short loop, while in S. lividans DndA (PDB code 3VAX), the catalytic cysteine resides on a ␤ strand (Fig. S1). In contrast, the Cys314 active site cysteine in SspA occurs in a short ␣-helix (residues 313 to 318) with only six amino acid residues in length. This Cys314-resident helix is flanked by 9-(residues 304 to 312) and 4-amino-acid (residues 319 to 322) loops, which may allow versatile movement of Cys314 (Fig. S1). In terms of the cysteine substrate, it is located in a positively charged surface pocket in SspA (Fig. 3A). The side chain guanidinium group of Arg340 and the side chain amide group of Asn150 make three hydrogen bonds with the carboxyl group of the cysteine substrate (Fig. 3B). Similarly to C314S, R340E and N150D mutations in pWHU732, expressing sspABCD from V. cyclitrophicus FF75, remarkably impaired the PT modification in E. coli Trans1-T1 (Fig. 3C), which confirmed the essential roles of these residues in catalyzing the desulfurization reaction of the cysteine substrate. Moreover, both R340 and N150 are conserved in DndA and IscS (Fig. 1B), suggesting similar binding modes for cysteine substrates.
Interaction interface between SspA and PLP. The PLP cofactor is located between the larger and smaller regions and is spatially closer to the larger one (Fig. 2B). The PLP cofactor is covalently attached to the side chain amino group of Lys201 in a deep surface pocket via the formation of an internal aldimine Schiff base. Several additional polar and nonpolar interactions are involved in recognition of the PLP cofactor. The imidazole ring of His99 formsstacking with the pyridine ring of PLP to make multiple van der Waals interactions. Additionally, at the bottom of the surface pocket, Asp175 forms two hydrogen bonds with the pyridine N1 atom of PLP. Gln178 from the bottom of the surface pocket forms a hydrogen bond with the hydroxyl group of the pyridine of the PLP. In addition, the phosphate group of PLP forms hydrogen bonds with the side chains of Ser71, Ser198, and His200 (Fig. 3D). These interactions are similar to those formed by PLP in the DndA and IscS structures, indicating a conserved molecular mechanism of PLP cofactor binding. Involvement of the multiple interactions ensures the fixation of PLP in the active site even when its internal Schiff base covalently bonded with Lys201 is broken in exchange of forming an external aldimine with the amino group of the cysteine substrate (23).
In the structure of SspA, the distance between the Cys314 catalytic cysteine and the cysteine substrate is 8.9 Å, exceeding the upper limit for effective nucleophilic attack (Fig. 2B). This long distance has been observed in almost all cysteine desulfurases examined (14, 23-25, 30, 31, 33). Movement of the catalytic cysteine toward the substrate cysteine to enable nucleophilic attack might be inevitable. Therefore, we first performed a molecular dynamics (MD) simulation on the structure of SspA in complex with its substrate cysteine. After the MD simulation, the larger region of SspA exhibited relatively little structural difference with an RMSD value between the crystal structure and the MD-simulated structure being 1.617 Å for 207 aligned C␣ atoms. In contrast, the smaller region of SspA underwent a relatively larger conformational change with an RMSD value of 1.820 Å for 79 aligned C␣ atoms. In addition, there was a relative motion between the larger and smaller regions of SspA. The Cys314 catalytic cysteine moved 5.5 Å toward the cysteine substrate (Fig. 4A), shortening the distance between them to 3.4 Å (Fig. 4B), which is within the range of nucleophilic attack. The conformational change of the smaller region and its motion toward the larger region accounted mostly for the overall structural difference of SspA with an RMSD value of 2.051 Å for 285 aligned C␣ atoms. The distance between Cys314 and PLP was also shortened accordingly (Fig. 4C), while the PLP-substrate cysteine distance did not substantially alter (Fig. 4D).
To corroborate the MD simulation results, we conducted a normal mode (NM) analysis, which is a convenient means to analyze intrinsic breathing motions and conformational changes of proteins (35,36). NM analysis also revealed that the catalytic Cys314 had an intrinsic vibrational motion toward the cysteine substrate and the PLP cofactor (Fig. 4E). In various cysteine desulfurases, the catalytic cysteines reside in different locations and different secondary structure elements, such as ␣-helices (e.g., SspA), ␤-sheets (e.g., DndA), and loops (e.g., IscS). Nevertheless, no matter which secondary structure element the catalytic cysteine resides in, NM analysis indicated that it persistently exhibited intrinsic motion toward PLP and the putative substrate-binding site ( Fig. 4F and G). Therefore, movement of the catalytic cysteine toward the substrate is an intrinsic property of cysteine desulfurases and independent of the relative locations or secondary elements where the catalytic cysteine is located.
To experimentally corroborate our results of MD simulation and NM analysis, we exploited small-angle X-ray scattering (SAXS) in combination with an MD-based conformational search using the crystal structure of the SspA dimer. First, we collected synchrotron SAXS data of purified SspA protein in solution. The SAXS data were then averaged to generate an ab initio model of SspA (Fig. 5A), which agreed well with our crystal structure. Next, we performed MD simulation to generate 20,000 simulated trajectories of SspA in 1,000 ns. Three clusters of calculated structures, named A (8%), B (75%), and C (17%), were obtained from the 20,000 simulations using the Monte Carlo method (Fig. 5B) (37). The superposition comparison between the three clusters of structures and the crystal structure of SspA revealed that the conformation of the large Cysteine Desulfurase in DNA PT Modification ® region in each protomer of SspA barely changed (Fig. 5C). In contrast, a conformational change extending outward at C314 occurred in the small region of one SspA protomer. The short ␣-helix harboring the active site changed into a loop and moved toward the substrate (Fig. 5C). However, a similar conformational change was not observed in the other SspA protomer, SspA=, suggesting that the catalytic processes in the two protomers do not occur simultaneously.
We calculated theoretical scattering profiles from the crystal structure of SspA (green line in Fig. 5D) and from the weighted square of the simulated structure clusters (red line in Fig. 5D). In the low-q region (i.e., overall structure at the macroscopic level), the theoretical scattering profile calculated from the crystal structure of SspA and the experimental SAXS scattering data matched perfectly. However, in the high-q range (i.e., detailed fine structure at the microscopic level, magenta circle in Fig. 5D), the theoretical profile calculated from the crystal structure of SspA exhibited considerable deviation from the experimental data. In contrast, the theoretical scattering profile calculated from the MD-simulated structures of SspA (red line in Fig. 5D) displayed remarkable correction, especially at the high-q range, and agreed much more with the experimental SAXS data ( 2 ϭ 1.04 for the theoretical curve generated from the MD result versus 2 ϭ 3.70 for that from the crystal structure). Because the MD-simulated structure takes the mobility of the catalytic cysteine-resident loop into account while the crystal structure does not, our SAXS studies provide supportive experimental evidence that the catalytic cysteine of SspA moves toward the cysteine substrate.
SspA and SspD form a complex. IscS interacts with downstream sulfur-acceptor proteins, such as IscU, and their associations expedite the transfer of the activated sulfur atom (38,39). The conformational plasticity of the cysteine-containing loop is essential for IscS to transfer sulfur to multiple acceptor proteins, such as IscU and TusA (26,27). In the Dnd PT modification system, DndA promotes the assembly of the 4Fe-4S cluster in DndC, presumably by providing an activated sulfur atom (40,41). Therefore, we investigated whether SspA in the Ssp system associates with SspD, which harbors pyrophosphatase activity in vitro as DndC (9). Due to the insoluble expression of SspD from V. cyclitrophicus FF75, we expressed both SspA and SspD from Shewanella frigidimarina NCIMB400. Pulldown experiments clearly showed that His-tagged SspD formed a complex with untagged SspA (see Fig. S2 in the supplemental material). Therefore, the flexibility and mobility of the SspA active site cysteine-containing loop are expected not only to be essential for nucleophilic attack on the cysteine substrate but also to play an important role in transferring the activated sulfur atom to SspD and eventually incorporating sulfur into the DNA backbone with the assistance of other Ssp proteins.
Conclusions. In the most recently characterized Ssp system, SspABCD confers single-stranded PT via the DNA-nicking activity of SspB (9). The sulfur atoms required for PT modification are extracted from cysteines by cysteine desulfurase. SspA exhibits PLP-dependent cysteine desulfurase activity and is predicted to initiate the delivery of sulfur for ssDNA PT formation (9). In this study, the crystal structure of SspA in complex with its cysteine substrate was determined. Considering the similar overall fold and the conservation of substrate-binding residues, it is conceivable that the cysteine-binding molecular mechanisms of IscS, DndA, and SufS resemble that of SspA. Given that SspA and SspD form a complex, it is likely that SspD functions as the intermediate sulfur acceptor, analogous to the sulfur transfer from DndA to DndC, before the sulfur can be further supplied to PT bonds. In combination with the observations that both Ssp and Dnd systems employ cysteine desulfurases, i.e., SspA and DndA, as sulfur donors and their interactions with SspD and DndC pyrophosphatase, respectively, we find it plausible that Ssp-and Dnd-governed DNA PT modification systems may be derived from a common ancestor but that divergence occurred leading to distinct modes of DNA target selection. The structure showed that the distance between the Cys314 active site and the cysteine substrate is too great to allow Cys314 to form a persulfide bond with the PLP-bound cysteine substrate. Our study shows that cysteine desul-furases employ conformational changes to move the thiol group of their active site cysteines into the closer vicinity of their cysteine substrates to facilitate nucleophilic attack, thus providing a basis for further understanding of the molecular mechanism of sulfur transfer in PT systems.

MATERIALS AND METHODS
Bacterial strains and plasmids. The bacterial strains and plasmids used in this study are listed in Table S2 in the supplemental material.
PT detection by LC-MS/MS. PT detection by liquid chromatography-tandem quadrupole mass spectrometry (LC-MS/MS) was conducted as previously described (20). Briefly, 20 g of DNA was digested with nuclease P1 (US Biological, USA) followed by dephosphorylation by alkaline phosphatase (Sigma, USA). The enzymes were removed by ultrafiltration using a Nanosep 10K-column (Pall, USA). The filtrate, containing chemically synthesized d(G PS A) S P as reference, was loaded onto a Thermo Hypersil Gold aQ column (150 ϫ 2.1 mm, 3 m). The column was coupled to a Thermo TSQ Quantum Access Max mass spectrometer for PT modification detection.
Construction of pWHU730 and derivatives. A 5,844-bp fragment harboring sspBCD was amplified from the genomic DNA of V. cyclitrophicus FF75 using the sspBCD-F/sspBCD-R primer pair (Table S3). The fragment was digested with KpnI and BamHI and ligated with pBluescript II SK(ϩ), which had been digested using the same enzymes, generating pWHU730. The pWHU4393 to pWHU4395 plasmids, expressing SspA variants and SspBCD from V. cyclitrophicus FF75, were constructed using overlap extension PCR (OE-PCR). In terms of pWHU4393, two fragments, one of 877 bp and the other of 1,035 bp, both bearing the SspA N150D mutation, were amplified from the genomic DNA of V. cyclitrophicus FF75 using the sspA-F/sspA-N150D-R and sspA-N150D-F/sspA-R primer pairs, respectively (Table S3). The two PCR products, sharing 17-bp overlapping ends, were combined and fused in a subsequent OE-PCR using the sspA-F/sspA-R primer pair. The resulting fusion product was digested with XbaI and SacII and ligated with pWHU730, which had been treated with the same enzymes to yield pWHU4393. Similarly, pWHU4394 and pWHU4395 were constructed with the sspA-C314S-F/sspA-C314S-R and sspA-R340E-F/ sspA-R340E-R primer pairs, respectively (Table S3).
Construction of pWHU4396 and pWHU4397. A 1,633-bp (dndA) fragment was amplified from the genomic DNA of ZS-139 using the dndA-ZS139-F/dndA-ZS139-R primer pair (Table S3). The PCR fragment was ligated with pMMB67 that had been linearized by BamHI and SalI via an in vitro recombination method using the Hieff Clone Plus one-step cloning kit (Yeasen, China), generating pWHU4396. Similarly, a 1,557-bp fragment containing sspA from FF75 was cloned into pDSK519, generating pWHU4397.
Construction of plasmids for protein purification. A 1,089-bp (sspA) fragment was amplified from the genomic DNA of FF75 using the 28a-sspA-75-F/28a-sspA-75-R primer pair (Table S3), which was ligated with pET28a linearized by NdeI and XhoI using a recombination method in vitro, generating pWHU4388. pWHU4389, pWHU4391, and pWHU4392 were cloned in the same way using pWHU4394 and the genomic DNA of NCIMB400 as the PCR templates.
Protein expression and purification. The genes encoding the full-length SspA from V. cyclitrophicus FF75 as well as SspA and SspD from S. frigidimarina NCIMB400 were subcloned into the pET28a vector with N-terminal 6ϫHis tags. The resulting plasmids were transformed into E. coli BL21(DE3) competent cells. Transformed cells were cultured at 37°C to an optical density at 600 nm (OD 600 ) of 0.8, and protein expression was induced with 0.2 mM isopropyl-␤-D-1-thiogalactopyranoside followed by incubation for 14 to 18 h at 16°C. After collection and resuspension, cells were disrupted using a cell homogenizer (JNBIO, Guangzhou, China). Cell debris was removed by centrifugation at 14,000 ϫ g for 45 min at 4°C, and the supernatant was loaded onto a Ni 2ϩ -nitrilotriacetic acid (NTA) affinity chromatography column (GE Healthcare, Uppsala, Sweden). The eluted protein was further purified by size exclusion chromatography on a Superdex 200 gel filtration column (GE Healthcare, Uppsala, Sweden) in a buffer containing 10 mM Tris (pH 8.0), 100 mM NaCl, and 2 mM dithiothreitol. Peak fractions were pooled and concentrated to 10 mg/ml for crystallization. The C314S point mutant of the SspA protein was purified using the same procedure as that used for the wild-type SspA protein.
Crystallization, data collection, and structure determination. Crystals of SspA C314S in complex with the cysteine substrate were grown at 14°C by the hanging-drop vapor-diffusion method with 1 l of protein mixed with 1 l of reservoir solution containing 1.8 M ammonium citrate (pH 7.0) (Hampton Research, USA). Crystals were cryoprotected in this crystallization buffer supplemented with 25% glycerol. The diffraction data were collected at the BL19U1 beamline at the National Center for Protein Sciences Shanghai (Shanghai, China) at 100 K. Diffraction data were processed using HKL3000 (42). The structure of SspA with a cysteine substrate was determined at 1.80 Å by the method of molecular replacement using the PHASER CCP4 program. The structure of Streptomyces lividans DndA (PDB code 3VAX) was utilized as the searching model after the mode-building by Coot and refinement by the REFMAC program of CCP4 (43,44). The crystals belong to the P6 3 22 space group, and two molecules of the SspA-cysteine complexes were contained in each asymmetric unit. The final refined model had an R work /R free of 16.58%/19.06%. The PROCHECK CCP4 program was used to evaluate the quality of the structure model, which indicated that the model exhibited good stereochemistry based on a Ramachandran plot.
MD simulation. The MD simulations were performed by the ff99Sbildn force field and AMBER 12 package (45,46). From the X-ray crystal structure, the atomic coordinates of the SspA/Cys complex were obtained. The force field and AM1-bcc charges of the ligands were handled by the Antechamber module (47). System neutrality was maintained by adding counter-ions. The bonds involving hydrogen atoms Cysteine Desulfurase in DNA PT Modification ® were constrained by the SHAKE algorithm (48). A truncated octahedron box of TIP3P water models was applied in all systems with solvent layers 10 Å between the box edges and the solute surface (49). The long-range electrostatic interaction was evaluated by the partial mesh Ewald (PME) method (50). To relieve any structural clash, 1,000-step steepest descent minimization was used in the solvated system. Heating to 298 K and brief equilibrating for 20 ps in the NVT ensemble were performed with PMEMD of AMBER 12. Langevin dynamics with a time step of 2 fs were performed in the heating and equilibrating runs with friction constants of 1 ps. The simulation was performed at 298 K for nonspecific as well as specific systems with 100 ns for every system. CPPTRAJ was applied to process the trajectories (51), and the figure was plotted by OriginPro 9.1. NM analysis. NM analysis was conducted as previously described (36). Briefly, the structural coordinates of SspA, DndA (PDB code 3VAX), and IscS (PDB code 4EB7) were submitted for NM analysis using the Elastic Network Model server (http://www.sciences.univ-nantes.fr/elnemo/start.html), which can compute the low-frequency normal modes of a protein. The first vibrational mode (i.e., the seventh NM with the lowest vibrational frequency is the most important vibrational mode) generated by the server was selected for further analysis.
Molecular dynamics simulation for the SspA dimer. The protein structure was obtained by X-ray crystallography. The protein molecule was centered in a cubic box sized 11.03 nm with the CHARMM 36 force field and TIP3P water model, applying GROMACS 2016.3 as the MD engine (52)(53)(54)(55)(56)(57)(58). The force field parameter of the PLP ligand was generated by the CHARMM General Force Field (CGenFF) (59). Systems were neutralized with 0.1 M NaCl. van der Waals interaction was truncated at 1.2 nm with the Lennard-Jones potential switched to zero gradually at 1.0 nm. Electrostatic interaction was evaluated by the PME method with a Coulomb cutoff of 1.2 nm (60). The LINCS algorithm was used to constrain bonds involving hydrogen atoms, allowing a time step of 2 fs (61). The system was first energy minimized using the steepest descent steps with a maximum force of 10.0 kJ · mol Ϫ1 · nm Ϫ1 and a maximum of 50,000 steps, and it was then equilibrated in the NVT ensemble at a temperature of 300 K for 300 ps and in the NPT ensemble at P ϭ 10 5 Pa for 20 ns. Temperature coupling and pressure coupling were performed leveraging the velocity-rescale and Parrinello-Rahman algorithm, respectively, with a coupling time of ϭ 1 ps (62,63). The MD simulation of NPT was conducted for 1 s. The trajectories were recorded every 50 ps with 20,000 frames. SAXS data collection. Synchrotron SAXS measurements in this study were performed at the BL19U2 beamline at the National Center for Protein Science Shanghai (Shanghai, China). The useful data, which were in the q range of 0.01 to 0.35 Å Ϫ1 (q ϭ 4sin/, 2 is the scattering angle), were collected by a Pilatus detector. Measurement was conducted in a vacuum with an exposure time of 20 s in 21-s frames. The samples were monitored for possible radiation damage, and no radiation effects were detected. The X-ray wavelength was 1.03 Å. To remove aggregates and sediments, all samples were centrifuged at 12,000 ϫ g for 20 min immediately prior to measurements. In the scattering curves, only the most informative part (between 0.01 Å Ϫ1 and 0.2 Å Ϫ1 ) could be used for structural analysis as a result of the considerable experimental noise at higher scattering angles. To determine the effects of concentration, samples containing six different concentrations (0.5, 1, 3, 5, 7, and 9 mg/ml) of SspA protein were prepared and measured. The cysteine substrate, which was present in a 10-fold (molar ratio)-greater concentration than SspA, was added directly before measurements to avoid air oxidation. During the measurements, there was no concentration dependence or aggregates observed.
SAXS data processing. The ATSAS program package was used to process all SAXS data (64). After subtracting the scattering of buffer from the signals of proteins, data were extrapolated to zero concentrations with standard procedures and the PRIMUS program (65). Both calculations and reconstructions used the resultant curves. Three-dimensional reconstructions of SspA were performed using DAMMIN and then compared with the SspA crystal structure by the SUPCOMB program (66,67). The data set was fitted against predicted scattering profiles, which was calculated from atomic coordinates by the CRYSOL program (68).
Clustering method. Twenty thousand structures were clustered as described previously (37) based on the C␣-C␣ distance similarity criterion as follows: where d m,n i is the Euclidean distance of the ␣-carbon atoms of residues m and n in structure I and N 2 is the number of residue pairs.
Three clusters were determined from 20,000 structures with a cutoff of 1.5 Å. The average scattering profile of each cluster was then calculated by CRYSOL (69). The profilers were further regarded as the basis to fit the experimental scattering profile by minimizing the 2 with 20,000 steps of Metropolis Monte Carlo as follows: ͑ q s͒ where iϭ1 N P i ϫ I i ͑q s ͒ is a sum over different structures for a defined wave vector, q S ; the summation Α S is performed over different q vectors; L is the number of q vectors; I i (q) is the averaged SAXS profile of each cluster; and I exp (q s ) is the experimental SAXS profile. The populations of different clusters were obtained by fitting against the experimental SAXS profiles using a Monte Carlo procedure described previously (70). Molecular graphics. All protein structure figures were generated by the PyMOL program (71).
Data availability. The atomic coordinates and structure factors of SspA-C314S in complex with the cysteine substrate have been deposited in the Protein Data Bank with the accession number 6M4J.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, TIF