Distinct Structural Features of the Peroxide Response Regulator from Group A Streptococcus Drive DNA Binding

Group A streptococcus (GAS, Streptococcus pyogenes) is a strict human pathogen that causes severe, invasive diseases. GAS does not produce catalase, but has an ability to resist killing by reactive oxygen species (ROS) through novel mechanisms. The peroxide response regulator (PerR), a member of ferric uptake regulator (Fur) family, plays a key role for GAS to cope with oxidative stress by regulating the expression of multiple genes. Our previous studies have found that expression of an iron-binding protein, Dpr, is under the direct control of PerR. To elucidate the molecular interactions of PerR with its cognate promoter, we have carried out structural studies on PerR and PerR-DNA complex. By combining crystallography and small-angle X-ray scattering (SAXS), we confirmed that the determined PerR crystal structure reflects its conformation in solution. Through mutagenesis and biochemical analysis, we have identified DNA-binding residues suggesting that PerR binds to the dpr promoter at the per box through a winged-helix motif. Furthermore, we have performed SAXS analysis and resolved the molecular architecture of PerR-DNA complex, in which two 30 bp DNA fragments wrap around two PerR homodimers by interacting with the adjacent positively-charged winged-helix motifs. Overall, we provide structural insights into molecular recognition of DNA by PerR and define the hollow structural arrangement of PerR-30bpDNA complex, which displays a unique topology distinct from currently proposed DNA-binding models for Fur family regulators.


Introduction
Group A streptococcus (GAS, Streptococcus pyogenes) is one of the most aggressive gram-positive human pathogens, responsible for a broad spectrum of diseases with diverse clinical manifestations [1,2]. Over 500,000 deaths caused by GAS are documented each year around the world [1]. GAS is renowned for its rapidly progressive, highly destructive ability to infect a wide variety of different tissues [3]. GAS infection can develop dramatically from a minor skin lesion to a lethal disease in just a few hours, if left untreated.
Unlike other gram-positive bacteria, GAS does not produce catalase, an oxidoreductase that, in other bacterial species, repairs damage to the bacteria when grown in an aerobic environment. Yet GAS does resist lethal reactive oxygen species (ROS) [4][5][6][7], through some other means. In addition, the survival and virulence of GAS correlates with the regulation of oxidative stress and metal homeostasis [5,8,9]. Therefore, there is considerable interest in how GAS adapts to ROS stress during infection.
PerR is the peroxide response transcriptional regulator that controls genes expression under oxidative stress and it is required for full virulence [5,9,10]. PerR belongs to the Fur (ferric uptake regulator) super family known to be dimeric, metal-binding regulators [11][12][13][14]. PerR has been identified as a repressor that blocks transcription initiation by binding to a specific site (the per box) in the promoter region of target genes [5,15]. One of the targeted genes, dpr, containing a putative per box sequence in its promoter region, has been found to be directly regulated by PerR in the A-20 GAS strain (M1 serotype) [5,[16][17][18]. Expression of dpr in response to a hydrogen peroxide challenge provides protection for bacteria by preventing the Fenton reaction [19,20]. To understand how dpr expression is regulated by PerR, and how PerR interacts with dpr promoter DNA, we have conducted a series of mutagenesis, biochemical and structural studies by combining protein crystallography and small-angle X-ray scatter-ing (SAXS). Our results have revealed the PerR-DNA interaction model and illustrated the DNA-binding mode of PerR that is distinct from all other regulators in Fur family.

GAS PerR Binds to dpr Promoter DNA through per Box
Recombinant 6xHis-tagged PerR protein was purified by Ni-NTA affinity column and size exclusion chromatography to ,95% purity judged by SDS-PAGE (data not shown). The gel filtration profile of PerR showed that the protein eluted at the volume corresponding to an apparent molecular weight of about 42 kDa ( Figure S1). This result is consistent with previous findings that PerR functions as a dimer [21]. PerR binding to the promoter region of dpr regulates the gene expression [18]. To confirm that the recombinant PerR protein possesses DNA-binding ability, dpr promoter DNA containing 419 bp (2403 to +16) was used to perform an electrophoretic mobility shift assay (EMSA). The PerR protein was able to bind the dpr promoter in a concentration dependent manner ( Figure 1A), suggesting that the recombinant PerR protein is functional in vitro. In order to identify the specific binding site of PerR on the dpr promoter, seven different PCRgenerated ,100 bp fragments of DNA within the dpr promoter at different locations were tested ( Figure 1B and 1C). The results showed that PerR binds to the dpr promoter from 2185 to 2135, which includes the previously identified per box binding site (from 2157 to 2143). The DNA-binding sequence includes a region of two 12 bp inverted repeats separated by a 21 bp spacer ( Figure 1D). These results demonstrated that the recombinant PerR binds to dpr promoter DNA through the per box.

Crystal Structure of PerR
We solved the crystal structure of PerR to 1.6 Å resolution by three-wavelength MAD phasing, using zinc as the anomalous scatter [22]. The structure represents the assembly of a homodimer in the asymmetric unit. Each subunit consists of six a-helices (a1 to a6) and six b-strands (b1 to b6) ( Figure 2A). Similar to Fur and Fur-like proteins, the GAS PerR monomer can be divided into an N-terminal DNA-binding domain (residues 1-94) and a Cterminal dimerization domain (residues 98-149 in chain A and residues 98-157 in chain B). The two domains are connected by a short linker (residues 95-97) [21,23,24]. The X-ray data and refinement statistics are summarized in Table 1. Validation of the structure by program MolProbity [25] showed that no phi-psi angles are in the disallowed regions of the Ramachandran map. The overall structure of PerR resembles the recently published structure (PDB code 4I7H) [21]. Superimposition of two PerR structures for the Ca atoms yields the root mean square deviation of 1.4 Å ( Figure 2B).

C4-type Zn-finger Motif and Zn-bound Regulatory Site
Initially two zinc sites were identified by program SOLVE from the X-ray MAD data and the atoms were located at the Cterminal C-X-X-C zinc finger motif composed by residues Cys104, Cys107, Cys144, and Cys147, in a tetrahedral coordination and consistent with the structure of PDB code 4I7H ( Figure 2C, left panel and Figure S2). Interestingly, two peaks of high electron density were apparent during the model building. We performed an X-ray fluorescence emission spectrum and identified the zinc content in the crystals. No other metal emission signals were detected. No exogenous zinc was added during the process of protein purification or crystallization, suggesting that an endogenous zinc is normally bound at the regulatory site, coordinated with residues His4, His6, Asn15, His19, His97, His99 in a pseudo octahedral geometry. The regulatory site of the recently reported PerR structure (PDB code 4I7H) was coordinated by a nickel ion, most likely acquired during Ni-NTA purification [21]. His4 and His6 are located on the loop of the Nterminus, Asn15 and His19 are on the a1 helix, His97 is located on the linker between b2 and b3 b-strands, and His99 is situated on the b3 b-strand ( Figure 2C, right panel and Figure S2).
To assess the importance of two metal-binding sites for the DNA-binding activity of PerR, we generated 10 mutants by sitedirected mutagenesis. Each residue coordinated by zinc was mutated to alanine or serine: H4A, H6A, N15A, H19A, H97A, H99A, C104S, C107S, C144S, and C147S. Two mutants, H44A and N101A, which are not involving in the metal binding from structural analysis, were used as controls in EMSA. The results showed that mutations in the residues of the zinc-finger motif or of the regulatory site, abolished the DNA binding ability of PerR, but not the control residues H44A and N101A ( Figure 3). Therefore, residues involved in metal-binding sites are critical for the DNA binding ability of GAS PerR.

Solution Structure of PerR by SAXS Studies
The overall conformation of our GAS PerR-Zn-Zn structure is substantially different from that of the B. subtilis PerR-Zn-Mn structure, which displays a DNA-binding-competent conformation [26] ( Figure S3). Our recombinant GAS PerR protein with a DNA-binding activity, however, exhibits a DNA-binding-incompetent conformation similar to the B. subtilis apo-PerR-Zn structure [24]. Therefore we performed SAXS studies to test whether our crystallographic model agrees with the conformation of PerR in solution. SAXS data were collected at beamline 23A of the National Synchrotron Radiation Research Center, Hsinchu, Taiwan. Scattering curves were collected for purified PerR proteins at various concentrations ranged from 20.0 to 177.6 mM and merged by program PRIMUS [27] ( Figure 4A, left panel). From experimental data, values of the radius of gyration (R g ), were computed from Guinier analysis of low-angle data, and also from the pair-distance distribution function P(r) for individual datasets, to check for consistency. An R g value of 24.860.4 Å was calculated from the Guinier region (qNR g ,1.3) of the scattering curve ( Figure 4A, inset in left panel). The best value for D max was ,77 Å , determined with GNOM [28] by trying a range of values and finding what value gave the best fit to the scattering curve when P(r) was back-transformed ( Figure 4A, right panel).
The degree of agreement between crystal structure and solution conformation can be assessed by the x-value, which measures the goodness-of-fit between the experimental scattering curve and the theoretical scattering profile calculated from structure model by FoXS [29,30]. Full-atomic modeling is essential for accurate fitting of the experimental data [31,32], so the missing residues in our PerR structural model were built with the program MODELLER [33]. The theoretically calculated SAXS profile from GAS PerR-Zn-Zn atomistic model fits well with experimental data, with x = 1.27 ( Figure 4B). The B. subtilis PerR-Zn-Mn atomistic model showed less well fit to GAS PerR experimental data at scattering angle q ranged from 0.2 to 0.3 Å 21 (corresponding to ,30-20 Å resolution), consistent with the conformational disparity between two crystal structures at N-terminus ( Figure S3). To model the low-resolution solution structure, molecular shapes were computed and averaged with the program DAMMIN [34] and DAMAVER [35]. The PerR crystal structure was superimposed onto the molecular SAXS envelope by program SUPCOMB [36] ( Figure 4C). SAXS studies show that the crystallographic structure of GAS PerR is indeed consistent with its solution formation. All experimental parameters from SAXS data are presented in Table 2 according to the publication guideline of small-angle scattering data [37].

DNA-binding Residues of PerR
Our PerR structure contains an N-terminal winged-helix motif consisting of a2-a4 and b1-b2. To identify the potential residues involved in the contact interface between PerR and dpr promoter, The dpr promoter from sequence 2403 to +16 was incubated with wide-type PerR at the concentrations indicated for 15 min at room temperature. The mixture were separated by 6% native PAGE, and the mobility shift of the promoter was resolved by ethidium bromide staining. (B) Seven DNA fragments were derived from dpr promoter. Segments A to G represents sequences as follows: A (284 to +16), B (2134 to 234), C (2184 to 284), D (2234 to 2134), E (2284 to 2184), F (2334 to 2234) and G (2403 to 2284). Putative per box was located at sequence between 2157 to 2143. (C) PerR binding to dpr promoter fragments was analyzed by EMSA as described in Figure 1A. (D) PerR binds to dpr promoter through sequence 2185 to 2135. A per box similar sequence with opposite direction was found at 2185 to 2174. doi:10.1371/journal.pone.0089027.g001 we superimposed the winged-helix motif of PerR with that of the Staphylococcus aureus BlaI-DNA complex structure (PDB code 1XSD) [38]. This study suggests that four residues in PerR Tyr67, Asn68, Lys71 and Lys83 might be involved in the interaction with DNA ( Figure 5A). The surface electrostatic potential of PerR contains a strong positively-charged patch including residues Arg21, Arg26, Arg31 and Asn69 ( Figure 5B). To determine whether these residues were involved in DNA interaction, each residue was substituted with an alanine. Mutants were constructed and purified up to 95% purity. All the gelfiltration profiles were consistent with wild-type PerR, suggesting that no mutation affected the protein dimerization ( Figure 5C). The mobility shift assay showed that alanine substitution of any of R21A, R26A, R31A, N68A or K71A abolished DNA-binding by PerR. Alanine substitution of any of Y67A, N69A and K83A also reduced but did not completely eliminate DNA binding  Figure 5D). This study strongly suggests that residues Arg21, Arg26, Arg31, Asn69, Tyr67, Asn68, Lys71 and Lys83, located on the positively charged patch, are involved in PerR interaction with DNA.

Structural Characterization of the PerR-DNA Complex by SAXS Analysis
Currently, no structural information is available for the PerR-DNA complex. In order to resolve its molecular architecture, we have performed SAXS studies. To explore the stoichiometry of PerR-DNA complex formed in solution and to investigate if the stoichiometry varies when mixing different ratios of PerR and DNA, we collected SAXS data on samples containing mixtures of 1:1, 2:1, and 4:1 PerR:DNA ratios (Table 2 and Figure S4). The DNA is a 30 bp fragment of the dpr promoter ranging from 2135 to 2164 covering the per box. P(r) function for 1:1 PerR:DNA ratio with the maxima at the ,60 Å and maximal dimension ,100 Å strongly indicate hollow structural arrangement of PerR-DNA assembly ( Figure 6A). Additionally, P(r) with a shoulder at the ,80 Å indicates additional density layer surrounding overall ring-like conformations ( Figure 6A). Based on the information from P(r), mutagenesis studies, and surface electrostatic potential of PerR, we manually built an initial model of the PerR-DNA complex. The model consists of two PerR homodimers which are held together by two 30 bp DNA fragments interacting with the positively-charged patches from two adjacent dimers ( Figure 6B). Formation of the PerR tetramer-DNA complex was also confirmed by Porod Volume calculation indicating ,80 kDa large assembly for 1:1, ,50 kDa for 2:1, and ,40 kDa for 4:1 PerR:DNA ratios (Table 2) [39]. The proposed head-to-head tetrameric assembly of PerR in the model satisfies hollow assembly as indicate by P(r) shape and P(r) shoulders ( Figure 6A and 6B). The theoretical SAXS curve calculated from this PerR-30bpDNA model by program FoXS [29,30] showed good fit to the experimental SAXS data (x = 3.2) ( Figure 6C and Figure S5A). An alternative PerR-30bpDNA model with a compact assembly of PerR tetramer in head-to-tail arrangement showed poor fit to the experimental data (x = 7.0) ( Figure 6C). SAXS data of samples with 2:1 and 4:1 PerR:DNA ratios show the same Dmax , 125 Å ( Figure S4D), suggesting the complex formed from different mixtures is the same as the proposed PerR-30bpDNA model from 1:1 PerR:DNA ratio. To understand the dynamic assembly of PerR-DNA complex, Minimal Ensemble Search (MES) [40] was applied to define the population of the dissociated species from the proposed model. Four different models (unbound DNA, protein, partial complex and complex) were input into MES for selecting the ensemble to optimize the fit to the experimental SAXS profile ( Figure S5). The best ensemble was calculated to best fit the experimental curve by minimizing the discrepancy x between the experimental and calculated profile. The scattering from such a minimal ensemble is computed by weighting the individual scattering profile from selected state. For 1:1 PerR:DNA ratio, the fit was improved (x = 3.2 to 2.5) when the ensemble of 96% PerR-DNA complex and 4% PerR dimers was selected ( Figure  S5A). For 2:1 and 4:1 PerR:DNA ratios, the selected ensembles significantly improved the fits than the single PerR-DNA complex model ( Figures S5B and S5C). The excess of protein in the mixtures results in dissociate products of free protein and partial complex assembly, but does not form the higher-order PerR-DNA complex. Taken together, SAXS results allowed us to define the molecular assembly of PerR with a sequence-specific 30 bp DNA.

Discussion
PerR is a peroxide sensor and plays a crucial role in GAS oxidative resistant response. The biological functions of PerR in peroxide sensing, metal binding, and gene regulation have been well studied. However, the molecular mechanism by which PerR recognizes DNA has not yet been clearly elucidated due to the lack of structural information of PerR-DNA complex. In this work, we provide the first structural insight for PerR-DNA complex and present a model which illustrates the physical association of PerR with its cognate promoter DNA.
Our high-resolution PerR crystal structure revealed that an endogenous zinc is bound at the regulatory site (Figure 2), unlike what was found in two other PerR crystal structures [21,26]. Zinc ion is the second most common transition metal in the human body, playing important roles in the host immune system [41,42]. Zn(II) is also an essential nutrient for bacteria, but exhibits toxicity at high concentrations. PerR regulates several zinc homeostasisrelated genes, such as pmtA [8], adcR and adcB [43]. A PerRdeficient mutant shows hyper-resistant to zinc toxicity [8]. AdcRCB (SPy0092-SPy0094) was identified as a transport complex for streptococcal Mn 2+ /Zn 2+ import [44,45]. PerR-dependent expression of AdcR regulon is important in metal homeostasis [43]. Collective information suggests that GAS uses zinc as a signal to activate the PerR regulon to cope with toxicity from high zinc concentrations. GAS PerR exhibits different dimeric conformation compared to the DNA-binding-competent conformation of the B. subtilis PerR-Zn-Mn ( Figure S3), suggesting an alternative DNA-binding mode. To exclude the conformational difference is not due to the crystallographic artifact, we performed SAXS studies which confirmed the crystal structure indeed reflects the PerR solution conformation (Figure 4).
A recently published paper indicated that H44 is important in pmtA promoter DNA binding [21]. However, our EMSA analysis showed that mutation of H44A did not affect dpr binding ability of PerR. To define the DNA-binding interface of PerR, we constructed eight alanine-substituted mutants and confirmed that residues on the winged-helix motif and the positive surface patch are involved in DNA recognition. H44 is distant from our identified DNA-binding residues, consistent with the EMSA results that H44 is not involved in the dpr promoter DNA recognition ( Figure 3B).
Besides PerR, the Fur family includes iron sensor Fur [46], zinc sensor Zur [47,48], nickel sensor Nur [49] and manganese sensor Mur [50], which are all structurally similar [23,51,52]. How these regulators with structural resemblance recognize divergent DNA sequences is a long-standing fundamental question. Several protein-DNA complex models were proposed for the Fur family, suggesting that regulators interact with DNA by N-terminal helixturn-helix/winged-helix in a DNA-binding-competent homodimeric conformation [53,52,21]. Here, we describe a PerR-30bpDNA complex model and delineate the hollow structural architecture with PerR tetramer wrapped by two DNA fragments ( Figure 6). In the model, the DNA-binding interface is contributed from two adjacent PerR homodimers in tetrameric assembly. This distinct DNA-binding feature explains that PerR dimer alone is in the DNA-binding-incompetent conformation. In conclusion, this study confers insights into the molecular basis of DNA recognition by PerR and offers a structural framework of PerR-DNA complex for Fur family.

Protein over Expression and Purification
The full-length PerR of GAS serotype M1 A20 was cloned into a pET-21b vector between NdeI and XhoI restriction sites and transformed into E. coli BL21 (DE3). Expression of the recombinant 6xHIS-tagged PerR protein was induced by adding 0.5 mM IPTG (isopropyl-b-D-thiogalactopyranoside) to the culture when cells reached an O.D. 600 of 0.6, and further incubated at 25uC for 4 hr. Cells were centrifuged at 8,000 rpm for 20 min, resuspended in buffer A (20 mM Tris-HCl, 200 mM NaCl, pH 7.5) and disrupted by sonication on ice. Supernatant was loaded into a Ni-NTA column (GE Healthcare) and unbound proteins were washed away with 60 mM imidazole in buffer A. PerR proteins were eluted with 300 mM imidazole in buffer A. Fractions containing PerR proteins were pooled and further purified by Superdex TM 75 size exclusion chromatography (GE Healthcare). Subsequently, proteins were dialyzed in buffer B (20 mM Tris-HCl, 150 mM NaCl, 0.5 mM EDTA, 5% glycerol, and 3 mM DTT, pH 7.5). Purified proteins were stored at 4uC for further use. The PerR mutant plasmids were constructed by the method of Dpn Idependent site-directly mutagenesis (SDM) or overlap extension PCR (OE-PCR) using wild-type PerR as a template. All primers used for construction of PerR and PerR mutant plasmids are listed in Table S1. Overexpression and purification of PerR mutant proteins were performed as described above.

Electrophoretic Mobility Shift Assay (EMSA)
PerR binding to promoter DNA containing a putative PerRbinding sequence was performed as previously described [18]. Various segments of dpr promoter DNA were amplified by PCR using wild-type A20 as a template (Table S2). Purified PerR and promoter DNA were incubated in 20 ml binding buffer (20 mM Tris pH 8.0, 50 mM KCl, 5% glycerol, and 50 mg/ml bovine

Crystallization
Purified PerR proteins were concentrated to 8-10 mg/ml in buffer B for crystallization trials. Various commercial kits of Hampton Research and Emeralds BioSystems were initially screened for crystallization conditions of the PerR protein, by the vapor diffusion method, performed with a Honeybee 961 robot (Digilab Genomic Solutions). PerR crystals were obtained in sitting drops containing 0.5 ml of protein and 0.5 ml of various precipitation solutions at 25uC within three days. Diffractionquality crystals were obtained by the hanging drop diffusion method mixing 2 ml of protein and 2 ml of solution (0.1 M Tris-HCl pH 8.6, 0.25 M MgCl 2 , and 27% PEG 4000). Crystals were flash-frozen using 30% PEG 4000 as a cryoprotectant in a stream of nitrogen gas at ,100 K prior to data collection.

X-ray Data Collection and Processing
Native diffraction data were collected to 1.5 Å resolution at beamline BL13B1 equipped with ADSC Q315 CCD detector at National Synchrotron Radiation Research Center (NSRRC) in  Hsinchu, Taiwan. 250 frames were collected; each with 1u oscillation and exposed for 2 sec at the wavelength of 1.0 Å with the crystal-to-detector distance of 200 mm at the temperature 100 K. Diffraction data were processed by HKL2000 program [54]. The Zn multi-wavelength anomalous (Zn-MAD) data sets were collected at beamline 4.2.2 equipped with a NOIR-1 detector at the Advanced Light Source (ALS) of the Lawrence Berkeley National Laboratory, Berkeley, USA. A three-wavelength MAD dataset was collected to 1.8 Å resolution at wavelengths of 1.28276 Å , 1.28296 Å , and 1.25700 Å , corresponding to Zn absorption peak, edge, and remote high energy. 360 frames were collected for each wavelength; each frame was exposed for 1 sec with the distance of 100 mm from crystal to detector. Diffraction data were processed with d * TREK [55]. Crystal parameters and data collection statistics are summarized in Table 1.

Crystal Structure Determination and Refinement
Zinc sites were determined from anomalous difference Patterson maps computed from the MAD data using the automated search algorithms of the program SOLVE [56]; two zinc sites were initially identified in the asymmetric unit. MAD phases gave an initial overall figure of merit of 0.45 for all data to 2.5 Å . The experimental electron density map was improved by density modification with the program RESOLVE [56]. Subsequent model building was carried out with the program Coot [57]. A peak of high electron density in each monomer was recognized as a metal ion in the structure and confirmed to be zinc by crystal Xray fluorescence (described below). Multiple cycles of simulated annealing, positional, and individual isotropic B factor refinements against native data to 1.6 Å were performed with REFMAC [58] and PHENIX [25], alternating with manual model rebuilding in Coot [57]. A total of 185 water molecules were included in the final model and in the final stage of refinement. The refinement statistics are summarized in Table 1. All structure representation figures are generated by the program PyMOL (http://www. pymol.org).

Crystal X-ray Fluorescence
At beamline 4.2.2 of ALS, X-ray fluorescence from a single crystal was excited with radiation of energy 11,000 eV, and the fluorescence emission spectrum from the crystal was recorded with a silicon drift diode coupled to a multichannel analyzer. Peaks at the energies expected for the K a (8630 eV) and K b (9572 eV) emissions of zinc were observed, at the appropriate amplitude ratios. No other significant emission signals were seen.

SAXS Data Collection
Data were collected at the NSRRC SWAXS beamline 23A, Hsinchu, Taiwan. PerR protein samples and PerR-DNA com-PerR proteins to dpr promoter (2403 to +16, top panel) and dpr segment (2185 to 2135, middle panel). A coomassie blue stained SDS-PAGE with equal amounts of wide-type or mutant PerR is shown below EMSA gels. doi:10.1371/journal.pone.0089027.g005 plexes were dialyzed against corresponding buffers prior to data collection. At beamline 23A, 60 ml of sample were loaded into thermostatic cells at 15uC. With 14 keV X-rays (wavelength # = 0.8857 Å ) and a sample-to-detector distance of 2.5 m, the scattering wave vector q, defined by 4pl 21 sinh with scattering angle 2h, is covered from 0.0085 Å 21 to 0.37 Å 21 . Data were collected with 1 frame per 60 s for 5 successive frames using a Pilatus 1M-F detector; frames were selectively combined for data reduction. Sample solution scattering was subtracted by the corresponding buffer solution scattering collected under identical experimental conditions, and was scaled to the absolute intensity scales [59].

SAXS Data Analysis and Modeling
Following data reduction, data were analyzed with the programs ATSAS 2.4 [60] and SCATTER. The values of radius of gyration (R g ) were first computed from Guinier plots [61] to test for aggregation. Scattering curves of PerR measured at different protein concentrations were merged using the program PRIMUS [27] and SCATTER. For PerR and PerR-DNA complex, distance distribution functions P(r) were computed as the Fourier transform of the scattering profiles using the program GNOM [28] and SCATTER. Theoretical SAXS profiles were calculated using program FoXS [29,30]. SAXS envelopes of PerR were computed with the program DAMMIN [34]. Ten runs were performed to verify the stability of the solution and the envelopes were averaged by DAMAVER [35]. Atomistic model of PerR-30bpDNA complex was built by combining two PerR homodimer to satisfy hollow assembly as indicated by P(r) shape ( Figure 6) and P(r) shoulders. Minimal Ensemble Search (MES) [40] was applied to select the assembly of the mixed states for the best fit to experimental data.

PDB Accession Code
Coordinates and structure factors have been deposited in the Protein Data Bank with the identifier 4LMY. Figure S1 Size-exclusion chromatographic profile of the recombinant 6xHis-tagged PerR protein. Elution volumes of standard molecular weight markers (black) and purified 6xHistagged PerR (red). (TIF) Figure S2 Zinc finger motif and regulatory site of PerR structure. Superposition of GAS PerR-Zn-Zn (magenta; PDB code 4LMY) and PerR-Zn-Ni (green; PDB code 4I7H) at zincfinger motif (left) and metal-bound regulatory site (right). Zn/Ni ion is colored in yellow. (TIF) Figure S3 Crystal structures of GAS PerR-Zn-Zn and B. subtilis PerR-Zn-Mn. Superposition of GAS PerR-Zn-Zn (magenta; PDB code 4LMY) and B. subtilis PerR-Zn-Mn (cyan; PDB code 3F8N). (TIF) Figure S4 Small-angle X-ray scattering data of the PerR-DNA complex. Experimental scattering profiles of (A) 1:1, (B) 2:1, and (C) 4:1, PerR:DNA ratio. Insets show the Guinier plot with linear fit and the corresponding radius of gyration (R g ) values. (D) P(r) functions of 1:1 (red), 2:1 (blue), and 4:1 (green), PerR:DNA ratios. P(r) functions are normalized to unity of their maxima. The maximal distance estimated from P(r) functions is ,125 Å for all ratios. Distinct disappearing of r ,60 Å P(r) maxima for the 2:1 and 4:1 ratios indicates dynamic protein-DNA assembly as shown in the Figure S5.