Three-dimensional Structure of MecI

Methicillin-resistant Staphylococcus aureus is the main cause of nosocomial and community-onset infections that affect millions of people worldwide. Some methicillin-resistant Staphylococcus aureus infections have become essentially untreatable by β-lactams because of acquired molecular machineries enabling antibiotic resistance. Evasion from methicillin challenge is mainly achieved by the synthesis of a penicillin-binding protein of low affinity for antibiotics, MecA, that replaces regular penicillin-binding proteins in cell wall turnover when these have been inactivated by antibiotics. MecA synthesis is regulated by a signal transduction system consisting of the sensor/transducer MecR1 and the 14-kDa transcriptional repressor MecI (also known as methicillin repressor) that constitutively blocks mecA transcription. The three-dimensional structure of MecI reveals a dimer of two independent winged helix domains, each of which binds a palindromic DNA-operator half site, and two intimately intertwining dimerization domains of novel spiral staircase architecture, held together by a hydrophobic core. Limited proteolytic cleavage by cognate MecR1 within the dimerization domains results in loss of dimer interaction surface, dissociation, and repressor release, which triggers MecA synthesis. Structural information on components of the MecA regulatory pathway, in particular on methicillin repressor, the ultimate transcriptional trigger of mecA-encoded methicillin resistance, is expected to lead to the development of new antimicrobial drugs.

in the cell wall of Gram-positive bacteria (8,9). BLAs target PBPs on the outer surface of the cytoplasmic membrane because of structural analogy with their natural substrate, Dalanyl-D-alanine-terminated peptides. PBPs are acylated by BLAs at their active-site serine residues, subsequently deacylating very slowly. This prevents PBPs from performing their regular functions efficiently, leading to cell wall loosening. This is followed by a non-lytic killing event and, finally, bacteriolysis (10,11). Bacterial resistance occurs mainly through excreted ␤-lactamases (penicillinases), which inactivate penicillins by hydrolysis of their ␤-lactam ring (12). MRSA, however, bases its resistance on the production of MecA (also known as PBP2a and PBP2Ј). This enzyme is a facultative PBP with reduced ␤-lactam affinity at the clinically achievable drug concentrations and is capable of essential cell wall construction when the housekeeping ␤-lactam-sensitive PBPs are switched off (13). In this way, MecA confers imperviousness to all BLAs. It is encoded by mecA, localized within 32-60-kb DNA regions called mec complexes or staphylococcal chromosome cassettes mec (SCCmec elements; Refs.14, 15), mobile elements that can also encode resistance to non-␤-lactam antibiotics. SCCmec elements have been acquired through horizontal transfer from an unknown heterologous source (16). MecA synthesis is regulated by a signal transduction system consisting of an integral-membrane zinc-dependent metalloprotease sensor/signal transducer, MecR1, and a constitutive transcriptional repressor, methicillin repressor MecI, both located immediately upstream from the mecA promoter on mec elements and counter-transcribed (4,17,18). The MecR1-MecI-MecA system has a close functional and structural relative encoded by the ␤-lactamase divergon (bla), the BlaR1-BlaI-BlaZ (also known as BlaR-BlaI-BlaP or PenJ-PenI-PenP) protein triad (17,19). In this case, the encoded effector is a ␤-lactamase, BlaP, also known as BlaZ or PenP (20).
MecI/BlaI repressors are encoded by the same operon as MecR1/BlaR1. They are ϳ125-amino acid proteins found in a number of bacterial genomes (Fig. 1) and are highly homologous to each other, with S. aureus MecI and BlaI sharing 61% and S. aureus MecI and B. licheniformis BlaI/PenI 43% sequence homology (23). These proteins are dimeric and may interact with cognate operators as dimers. These repressors have been proposed to consist of an N-terminal DNA-binding domain (DBD) featuring a helix-turn-helix (HTH) motif and a C-terminal dimerization domain (DD) (24,25). BlaI binds to two separate palindromic operators in S. aureus (R1-dyad and Z-dyad) encompassing the Ϫ10 region of blaZ and the Ϫ35 of blaR1-blaI, though independently and not cooperatively and not inducing any significant DNA bending (19,26). For B. licheniformis, binding to three bla/pen regions has been reported, two including the blaZ/P and one the blaR1-blaI promoters (25,27). Finally, MecI binds to an extended region containing two consecutive palindromes and covering the mecA Ϫ10 and the mecR1-mecI Ϫ35 promoter sequences (18). The regulatory regions of the bla and mec elements are 57% identical (17). Indeed, both repressor systems seem to be interchangeable to block transcription of both structural and regulatory genes in S. aureus (28,29). The sensor/transducers BlaR1 and MecR1, however, are not interchangeable, and they have distinct kinetics; whereas S. aureus BlaR1 induces BlaZ synthesis within minutes, MecR1 takes hours to induce synthesis of MecA (9).
Based on biochemical data mainly from the bla divergon proteins, the MecR1/MecI system is thought to work as follows. Dimeric MecI constitutively blocks mecA transcription via inhibition of mRNA synthesis initiation or elongation. Through recognition of regulatory regions of the counter-transcribed mecR1-mecI operon, the repressor also regulates its own transcription and that of the signal sensor/transducer. MecR1 detects BLAs in the extracellular space via its PBP-like penicillin sensor. Upon protein acylation, a conformational change within MecR1 leads to autocatalytic activation of the integral-mem- brane metalloprotease domain. The active protease facing the cytosol specifically cleaves MecI, directly or indirectly. In the case of the BlaR1-BlaI-BlaZ system, some authors have postulated that a further, chromosome-encoded regulatory element, BlaR2, is required for BlaI inactivation (25,30). This system highlights the key role of methicillin repressor as the eventual transcriptional regulator of MRSA response (14,18,(31)(32)(33)(34).

EXPERIMENTAL PROCEDURES
Cloning and Overexpression of MecI-S. aureus strain N315 (MRSA) bacterial cells were grown in Luria Bertani broth and lysed with lysostaphin (25 g/ml). The DNA was extracted as described (35) and used as a template for PCR amplification of the MecI coding sequence. Taq polymerase (Fermentas) was used with oligonucleotides 5Ј-GTTAAT-ACCATGGATAATAAAACGTATG-3Ј as forward primer and 5Ј-AAA-CACAACTAGTTTATTTTTTATTCAAT-3Ј as reverse primer (purchased from Roche Applied Science). The reaction yielded a 393-bp fragment containing the correct coding sequence within restriction sites for NcoI (5Ј) and BcuI (3Ј). After digestion with these enzymes, the fragment was inserted into expression vector pP RO EX TM Hta (Invitrogen). DNA sequencing confirmed that the MecI coding sequence was properly placed after the sequence encoding the His 6 tag and the tobacco etch virus protease recognition site. Reactions with restriction endonucleases (Fermentas) and ligation with T4 DNA ligase (Roche Applied Science) were performed as recommended by the suppliers. The recombinant plasmid obtained was introduced into Escherichia coli DH5␣ cells by heat shock treatment. Freshly transformed cells were cultured overnight in Luria Bertani medium containing 100 g/ml ampicillin. For overexpression, 4 ml of these cultures was added to 500 ml of Luria Bertani medium, also 100 g/ml in ampicillin. Cells were grown at 37°C and induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside at an A 600 of 0.5. After 3 h of incubation at the same temperature, cells were harvested by centrifugation and frozen. The selenomethionine (SeMet) variant was obtained in the same way, except that the cells were added to 500 ml of minimal medium lacking methionine and implemented with 25 mg of SeMet (Sigma) 30 min before induction.
Purification and Crystallization of MecI-Pelleted cells were gently The total protein concentration (9 mg/ml) was measured according to Bradford (Bio-Rad) using lysozyme as a standard. EDTA and 1,4-dithio-D,L-threitol were added to the fractions containing the fusion protein to a final concentration of 0.5 and 1 mM, respectively. Tobacco etch virus protease was added at a 1:500 (protease:fusion protein) molar ratio to remove the His 6 tag. After overnight digestion at 22°C, the sample was subjected to gel filtration on a Superdex 75 HR 10/65 column (Amersham Biosciences) equilibrated with buffer C (20 mM Tris-HCl, pH 7.4, 0.2 M NaCl). 3 ml of MecI (14.9 kDa) in buffer C was obtained at a final concentration of 1.25 mg/ml (as calculated with the experimentally determined molar extinction coefficient, ⑀ ϭ 35413 M Ϫ1 cm Ϫ1 ) and stored at 4°C. The purity of the protein was assessed by SDS-PAGE and mass spectrometry, and its N terminus was found by chemical sequencing to be GAMDNK (amino acid one-letter code), including a glycine-alanine dipeptide from the tobacco etch virus protease recognition sequence prior to the first methionine of native MecI. The protein crystallized spontaneously in the chromatography collection tubes overnight at 4°C in the form of well-shaped, ordered crystals. The same crystals were obtained with SeMet-MecI. DNA Binding Assay-To test the DNA-binding capability of MecI, two complementary oligonucleotides (5Ј-CAAAATTACAACTGTA-ATATCGGAG-3Ј and 5Ј-GCTCCGATATTACAGTTGTAATTTT-3Ј) purchased from MWG Biotech AG) were annealed in buffer D (20 mM Tris-HCl, pH 7.4, 0.1 M NaCl) to render 180 nmol of a 25-bp doublestranded (ds) DNA comprising the sequence protected from DNase I attack of the Z-dyad of the bla divergon (19), with an additional 1-bp overhang on either side (C/G). Purified MecI (85 M) in buffer C was mixed with DNA in buffer D at a 2:1.1 protein:dsDNA molar ratio. The mixture was diluted 1:4 in 20 mM Tris-HCl, pH 7.4, and incubated 15 min at 4°C before being loaded onto a gel filtration Superdex 75 HR c Calculations were performed in P2 1 , assuming two dimers per asymmetric unit. d R deri ϭ ¥ hkl ʈF PH ͉ Ϫ ͉F P ʈ/¥ hkl ͉F P ͉ with respect to the data set chosen as a reference (f Љ max ). e The data set at the absorption peak (fЉ max ) was taken as reference.
with F obs and F calc as the observed and calculated structure factor amplitudes; free R factor , same for a test set of reflections (Ͼ 500) not used during refinement.
h Values in parenthesis refer to the outermost resolution shell if not otherwise indicated.
10/65 column equilibrated with buffer E (20 mM Tris-HCl, pH 7.4, 50 mM NaCl). A single peak was obtained and analyzed with a band-shift assay in a 5% acrylamide gel with 5% glycerol and 50 mM Hepes, pH 8.3, using annealed oligonucleotides as a negative control. Analytical Gel Filtration and Cross-linking Experiments-To assess the concentration-dependent aggregation states of MecI, a peak fraction of the last purification step, which rendered crystals, was diluted with buffer C and subjected to analytical gel filtration on a Superdex 75 HR 10/65 column previously calibrated with protein markers of known molecular mass and equilibrated with buffer F (50 mM Hepes. pH 7.4, 0.2 M NaCl). Resulting peaks revealed by SDS-PAGE a single band corresponding to a protein monomer.
140 pmol purified MecI was cross-linked at room temperature with glutaraldehyde at different concentrations in a total volume of 50 l of buffer for 2 min, followed by quenching with 12 l of 50 mM Tris-HCl, pH 6.8. The products were resolved on 15% SDS-PAGE and visualized by silver staining.
Structure Solution and Refinement-Native MecI crystals belong to the orthorhombic space group P2 1 2 1 2 1 , harbor one MecI dimer per asymmetric unit (Matthews parameter (V M ) ϭ 3.1 Å 3 /Da; 59.5% solvent content; Ref. 36), and diffract beyond 2.4 Å resolution. SeMetderivatized protein crystals are fairly isomorphous to the native ones and diffract to 2.8 Å resolution (Table I). A cryoprotecting protocol was established consisting of soaking crystals in a mixture containing crystal supernatant protein solution and increasing (5% steps) glycerol concentrations (up to 25% v/v) at 4°C, allowing the crystals to equilibrate for 15 min after each step. Complete diffraction data sets were collected from a single N 2 flash-cryo-cooled (Oxford Cryosystems) crystal, each on an ADSC Quantum4-charged coupled device detector at beamline ID29 of the European Synchrotron Radiation Facility, Grenoble, France. For the SeMet derivative, a multiple-wavelength anomalous diffraction experiment was carried out at three wavelengths, corresponding to the absorption maximum (12662.53 eV), the inflection point (12660.95 eV), and a hard remote wavelength (12900 eV). These values were determined from the crystal to be measured by means of a fluorescence spectrum around the theoretical selenium K-edge value. All diffraction data were processed with program MOS-FLM v. 6.2.2 and scaled, merged, and reduced with SCALA, within the CCP4 suite of programs (Ref. 37) ( Table I).
The heavy atom model of the SeMet derivative could not be determined in P2 1 2 1 2 1 by any procedure. All tests for hemihedral twinning were negative. Consequently, the diffraction data were reprocessed as P2 1 , choosing the short axis as unique axis b. With this setting, 10 of 16 theoretical selenium sites were found with SOLVE v. 1.18 (38), giving a figure of merit of 0.33 for the resolution range of 50 -2.9 Å (Table I). Subsequent density modification by means of DM within CCP4 increased this value to 0.66, rendering a map where some helical segments could be identified. The four molecules in the P2 1 asymmetric unit were delimited, and non-crystallographic symmetry operators were calculated. A second step, including averaging, delivered a Aweighted electron-density map (Fig. 3a) that enabled straightforward tracing of the whole polypeptide chain employing a Silicon-Graphics graphic work station and program TURBO-FRODO (Biographics, Marseille, France). A model comprising a dimer was submitted to a molecular replacement calculation, employing program AMORE (39) against the high-resolution native data set (15-4 Å resolution range). A unique solution was obtained at 81.3, 67.9, 105.2, 0.2248, 0.4739, 0.4103 (␣, ␤, ␥, in Eulerian angles; x, y, z, as fractional unit-cell coordinates) with a correlation coefficient in structure-factor amplitudes (CC F ) of 72.4% and a crystallographic R factor of 34.5% (for definitions, see Table  I and Ref. 39); second highest peak, CC F 49.6, R factor 47.0%). This calculation confirmed P2 1 2 1 2 1 and ruled out those other primitive orthorhombic space groups. Manual model building alternated with crystallographic refinement utilizing CNS v. 1.1 (40) and REFMAC5 (including TLS refinement) within CCP4, until the final model was obtained. It features protein residues Lys 4A -Lys 123A and Met 1B -Lys 122B corresponding to monomers A and B, two glycerol molecules (Gol201W and Gol202W), one chloride anion (Cl1203W), and 69 solvent molecules (Hoh204W-Hoh272W). All residues are placed in allowed regions of the Ramachandran plot, except for Lys 23 of each polypeptide chain (⌽ 23A ϭ 64°, ⌿ 23A ϭ Ϫ61°; ⌽ 23B ϭ 72°, ⌿ 23B ϭ Ϫ57°). These residues and their side chains are, however, unambiguously defined by electron density at the end of helices ␣1.
Miscellaneous-Figures were prepared with TURBO-FRODO, SE-TOR (41), and GRASP (42). Superimpositions were performed with TURBO-FRODO, cavities were ascertained with GRASP, close contacts and interaction surfaces were calculated with crystallography NMR software. Three-dimensional protein structure comparisons were done with the DALI server at www.ebi.ac.uk/dali. The final coordinates have been deposited with the Protein Data Bank (PDB).

RESULTS AND DISCUSSION
Oligomerization Behavior and DNA Binding of Methicillin Repressor in Vitro-It has been shown that BlaI from both S. aureus and B. licheniformis forms dimers in solution (24 -26). For MecI, oligomerizing behavior was observed and mixed BlaI/ MecI heterodimers have been described (9,18). Analytical sizeexclusion chromatography on a calibrated column showed a concentration-dependent monomer/dimer equilibrium in the conditions assayed (pH 7.4, 0.2 M NaCl). At concentrations above 1 mg/ml (67 M), which eventually lead to crystallization, the protein is almost completely dimeric (Fig. 2a). Under progressive dilution, a peak corresponding to a monomer appears that becomes single at 4 M. These values are in accordance with B. licheniformis BlaI, for which a dissociation constant of 25 M was estimated by ultracentrifugation (25) and are corroborated by cross-linking experiments. Fig. 2b shows that increasing concentrations of glutaraldehyde resulted in the formation of dimers with an apparent molecular mass of ϳ30 kDa.
A 43-bp region within the mec regulatory region fragment encompassing two consecutive 15-bp inverted repeats and the promoter-operator sequences for mecR1-mecI and mecA is protected by methicillin repressor and by BlaI from DNase I attack (9,18). Though MecI represses mecA transcription more strongly than BlaI, no differences are found with shorter synthetic mec and bla operator sequences (43). The protected region corresponds to two successive palindromes and suggests binding of several MecI molecules to this elongated stretch of dsDNA. Other studies have revealed that both S. aureus BlaI and MecI bind identically and protect two regions of 25 bp (R1-dyad; affects blaI-blaR1 promoter) and 24 bp (Z-dyad; around blaZ promoter) of the bla divergon (19). In this latter case, a 5-bp spacer is found between the two regions. With the aim of obtaining crystals of MecI in complex with dsDNA, the better delimitated and shorter sequence around the Z-dyad of the bla divergon was chosen to perform a band-shift assay. To this end, protein and annealed 25-bp oligonucleotide were mixed in an ϳ2:1 (protein:DNA) molar ratio and subjected to size-exclusion chromatography. A single peak was obtained, with no indication of excess of either protein or DNA. The purified complex analyzed by native PAGE showed different migratory properties from control DNA (Fig. 2c). Taken together, these results indicate that MecI binds to the 25-bp dsDNA as a dimer, confirming previous studies.
MecI-DBD starts with a segment in extended conformation and is stabilized by intramolecular interactions from Tyr 6 onwards. At Ser 10 , the chain enters helix ␣1. This regular secondary structural element is followed by a short segment in extended conformation, loop ␣1␣2 (Tyr 24 -Ser 26 ). At Ala 25 , a double main-chain interaction anchors this stretch with Tyr 69 of a ␤-ribbon (see below). Helix ␣2 starts at Ser 26 , with its axis rotated by 60°clockwise with respect to ␣1. At Lys 38 , helix ␣2 ends, and the main chain starts at Ser 41 with helix ␣3 after a short three-residue spacer. This helix, running until Lys 55 , runs with its axis rotated 135°counterclockwise away from ␣2. Both helices ␣1 and ␣2 contribute to the presentation of the following helix to the molecular surface. A glycine-mediated change in the direction of the main chain after ␣3 leads to the first residue of a ␤-ribbon structure. This is made up by strand ␤1 (Phe 57 to Asp 63 ), followed by a tight 1,4 turn of type I and a second strand, ␤2 (Ile 66 -Ser 71 ). A total of seven inter-main-FIG. 3. Structure of methicillin repressor MecI. a, initial experimental F obs -type a -weighted electron density map after density modification and averaging, superimposed with the final refined model around helix ␣5 of each protomer (contour level 1 above average). b, C␣ plot of MecI. Both molecules within a dimer are shown, in green and white, after optimal superimposition of their DNA-binding domains (DBD). The dimerization domains (DD) display a certain degree of flexibility. Selected residues are labeled. c, ribbon plot of a MecI dimer with DBDs on top and DDs at bottom in frontal view. Regular secondary structure elements are labeled in orange, and the scissile bond (Asn 101 -Phe 102 ) is indicated by red arrows (see Fig. 1). Close-up view of residues participating in close contacts (below 4 Å) between the protomers. Some residues are labeled. d, same as panel c but after a vertical 90°rotation (clockwise). chain hydrogen bonds stabilize this ribbon, conferring it a rigid structure. MecI-DBD finishes at Val 73 after ␤2.
The overall fold topology of MecI-DBD is ␣1-␣2-␣3-␤1-wing-␤2, in accordance with a winged-helix architecture, first identified in the complex of hepatocyte nuclear factor-3-DBD with DNA (44). It is common to many prokaryotic and eukaryotic transcription factors and had been partially anticipated for BlaI (25). This fold is characterized by an ␣1-␤1-␣2-␣3-␤2-wing1-␤3-wing2 topology (45). In our structure, ␤1 is reduced to just one residue (Ala 25 , see above), while the position of the second wing is occupied by the first helix of the MecI-DD (see below). The winged helix topology encompasses a central HTH motif, identified for the first time in bacteriophage cro repressor (46), modified by variations in the length of the connecting turn and the angle between the helices. It is engaged in DNA major-groove recognition and made up in MecI by helix ␣2, loop ␣2␣3, and helix ␣3 ("recognition helix"). The fold further harbors a "wing," in most cases engaged in minor-and/or majorgroove interactions (45,47) and constituted in MecI by the tip of the ␤-ribbon. The MecI-DBD globular structure is compact and maintained by an extended hydrophobic core traversing the domain (Fig. 3b) MecI-DD starts at Glu 74 and is a right-handed superhelical tail consisting of three consecutive helices, ␣4 -␣6 (Fig. 3, b-d).
These are not totally equivalent within each monomer and are somewhat flexible, in particular around the C-terminal helix. Helix ␣4 protrudes from the DBD moiety and continues to Tyr 91 . At this residue, a tight 1,4 turn of type II features loop ␣4␣5, prior to ␣5 (Phe 95 -Glu 106 ) with its axis rotated clockwise 90°with respect to ␣4. At its end, loop ␣5␣6 consists of two residues and links it with the C-terminal helix ␣6. This helix runs roughly antiparallel to ␣5 and extends to Asn 121 . The side chains of the three helices facing the central superhelical axis are mainly hydrophobic (Tyr 80 , Phe 86 , Ile 87 , Val 90 , and Tyr 91 from ␣4; Leu 98 , Val 99 , Phe 102 , Val 103 from ␣5; and Leu 108 , Ile 113 , Leu 116 , Ile 119 , and Leu 120 from or before ␣6). This structure, together with the lack of stabilizing interactions between these helices, explains why dimers predominate at higher protein concentrations, whereas only monomers are found in diluted samples.
The functional and stable structural unit is a dimer, as present in the asymmetric unit, with the overall shape of a triangle, 75 Å in width, 45 Å in height, and 35 Å in depth (Fig.   3c). A total of 63 intermonomeric close contacts (Ͻ4 Å) are observed, among them 18 hydrogen bonds, 20 van der Waals contacts, and 3 salt bridges. Dimerization occludes 3,677 Å 2 or 40% of the total surface area of a monomer and is almost perfectly self-complementary (see Fig. 4a). Both protomers display close structural similarity, as indicated by a root mean square deviation of 0.47 Å for the 90 C␣ atoms deviating less than 1 Å (Fig. 3b). Among the DBDs, this equivalence is almost total, whereas the DDs display a certain degree of variability with greater deviations, in particular around the C-terminal helices ␣6. Within the dimer, the globular DBDs are adjacent to each other but not in contact, apparently well situated to function independently (Fig. 3, c and d; see below). The dimeric structure is mainly maintained by the DDs, which display a previously unobserved arrangement. They are closely juxtaposed and intertwine like two surface-and side-chain-complementary protein superhelices in such a manner that three layers are formed, each made up by the equivalent helices of each monomer running in an antiparallel manner. Within each layer, the interior side chains interdigitate in a zipper-like fashion (Figs. 3, a, c, d and 4a). The first layer, consisting of helices ␣4, is rotated 90°counterclockwise with respect to layers two and three, which are constituted by helices ␣5 and helices ␣6, respectively, which are parallel. This three-layered structure is held together by an elongated hydrophobic core running from the interface between helices ␣4 on one end to the interface between helices ␣6 at the other end. In particular, residues Phe 86 , Ile 87 , Tyr 91 , Phe 95 , Leu 98 , Val 99 , Phe 102 , Val 103 , Leu 108 , Ile 113 , Leu 116 , Ile 119 , and Leu 120 participate in van der Waals contacts. This inner core is, however, not completely solid but shows two cavities, of 117 and 22 Å 3 (Fig. 4e), which may account for the certain flexibility observed within the DDs. Besides residues from the DDs, two side chains at the beginning of helix ␣1 of each monomer, Trp 13 and Asn 17 , also participate in the dimerization interface, mainly with the last turn of helix ␣4. In particular, Asn 17 C-caps it through its side chain.
Similarities with Other Proteins and Working Hypothesis for the Recognition of Operator DNA-MecI shares structural similarity with DNA-binding proteins harboring the winged helix motif (45,47). The closest similarity is encountered with MarR and SmtB, despite negligible sequence similarity (12 and 6%, respectively, in the aligned stretches). MarR is a regulator of multiple antibiotic resistance from E. coli (PDB accession code 1jgs; Ref. 48) and SmtB, a metal-tuneable trans-acting dimeric transcriptional regulator from Synechococcus that represses its own synthesis and that of metallothionein (PDB accession code 1smt; Ref. 49). The topological equivalence is restricted, however, to the winged helix DBDs and ranges from helix ␣1 to the beginning of the dimerization helix ␣4 of MecI (Fig. 4c). The dimeric arrangement is different in all three proteins. In both MarR and SmtB, the recognition helices are not properly positioned for operator recognition in the structures described. On the other hand, the DD of MecI displays a novel fold that merely bears a topological similarity with a three-helix segment of a functionally unrelated archaeal endonuclease, I-DmoI. This three-helical segment does not, however, form a distinct folding unit in I-DmoI but belongs to two separate domains (PDB accession code 1b24).
Biochemical studies indicate that dimeric MecI binds to its operator dyads in such a way that each monomer recognizes one half of the operator, while the central sequence is unprotected. This goes along with a lack of DNA bending upon protein binding (19). Within the dimer, the putative recognition helices ␣3 of the HTH motif are solvent-presented in parallel with a spacing of about 30 Å, geometrically appropriate to interact with the B-DNA major groove, and the wings are positioned so as to interact with the adjacent minor groove (Fig.  3c). The surface potential indicates that the exposed surfaces of helices ␣3 and those of the lateral wings are positively charged, thus electrostatically suited to recognize the B-DNA backbone (Fig. 4, a and b) as observed in other winged helix DNA-binding proteins (45). Despite limited overall topological similarity, the positioning, orientation, and distance of the recognition helices of the winged HTH motifs are equivalent to those observed in the structure of Corynebacterium diphtheriae toxin in its complex with its cognate tox operator. This protein shares with MecI the general winged helix DBD architecture (Fig. 4d) but shows a completely unrelated nickel-dependent DD (PDB accession number 1ddn; Ref. 50). In the latter complex, one diphtheria toxin dimer binds on one side to the major groove of a 19-bp dsDNA at two successive turns of the double helix, and a second dimer binds independently on the opposite side of the DNA helix. The wing, shorter than in MecI, only establishes one interaction with the phosphodiester backbone in the minor groove, mediated by an arginine. Diphtheria toxin binds its cognate operator without causing significant global bending, with equivalent positions of the recognition helices 28 Å away. A similar value (30 Å) has been reported for a HTH replication terminator protein in a DNA complex (PDB accession number 1f4k).
Based on these similarities, a working model for the interaction of MecI with its cognate dsDNA can be constructed using the dimer in our unliganded structure and canonic B-DNA encompassing the target sequence of the bla Z-dyad ( Fig. 4f; Ref. 19). According to this model and on looking on just one half of the operator, Lys 65 and Ile 66 from the tip of the wing could be engaged in interactions with the phosphate backbone or with the O-2 atom of one of the three consecutive thymine bases found in the 3Ј-end of the complementary strand of the operator (Fig. 4f, 1). This would imply only a minor rearrangement of the wing to fit snugly into the dsDNA minor groove, as observed in the DNA-bound structure of the DBD of signal transduction transcriptional activator PhoB (51). At the N terminus of helix ␣1, the phosphate backbone around positions 11-12 of the coding strand could complement the side chain of Ser 9 in Ncapping the regular secondary structure element; the same could occur with the N terminus of helix ␣2 with the complementary DNA backbone around position 6 (Fig. 4f, 2 and 4). Alternatively, these serine residues could undergo a 180°rotation about their 1 angles and be engaged in phosphate recognition themselves. Also, the N-terminal Lys 4 could play a role in backbone recognition, as anticipated by BlaI mutant studies (26).
The main interaction with DNA would be established by the recognition helix ␣3, at the center of one dyad half. In particular, solvent-exposed side chains in the unbound structure would be engaged in nucleoprotein complex stabilization as follows (Fig. 4f, 3). Lys 43 could target the phosphodiester backbone or a guanine or thymine base at complementary strand positions 7-9; Arg 46 could contact the backbone or codingstrand thymine at position 6; Thr 47 and Thr 50 , placed at the center of the recognition helix, could recognize thymine O-4 or guanine O-6 atoms of complementary strand positions 8 -10; Arg 51 could interact with the backbone or with complementary strand bases; finally, Lys 54 and Lys 55 could bind backbone phosphate oxygen atoms.
Interactions of the left-hand side of the operator would be symmetric. This model would not entail any significant bending, and it would leave the central guanine at position 13 freely accessible, in accordance with biochemical experiments (19). The DBDs would bind independently, with no interactions between them. Oligomerization on longer dsDNA is not ruled out, as in the case of the segment encompassing the two consecutive operators within the mec region (see above). However, cooperativity seems unlikely, because it would have to account for intensive interactions between adjacent dimers via the wings, which are more probably engaged in DNA recognition. Some interactions between adjacent wings can, however, not be ruled out.
Limited Proteolysis Inactivates Methicillin Repressor-BlaI and MecI are specifically cleaved at a single bond between Asn 101 and Phe 102 by activated BlaR1 or MecR1 (34). This limited proteolysis explains why both MecR1/BlaR1 (which are also autoproteolytically activated) and MecI/BlaI require synthesis induction, as they are turned over during the signal transduction process (26). Cleavage disables dimerization and the repressing capability of MecI, thus allowing transcription of mecA. Once the extracellular BLA concentration diminishes, proteolytic cleavage of MecI ceases and the intracellular concentration of the full-length inhibitor molecule increases, suppressing unnecessary MecA synthesis (9,25,26,30,31,33,34). The scissile bond is strongly conserved among members of the MecI/BlaI family (Fig. 1) and fully affects MecI-DD because it is located in the middle of helix ␣5 (Fig. 3, c and d). Interestingly, the scissile bond is not directly accessible from the exterior. This may explain the requirements for another component, such as BlaR2, which may melt ␣5 locally and thus facilitate access by MecR1/BlaR1 metalloproteinase (18,30,52). After proteolytic cleavage, the interaction surface would be reduced to 43% of the intact dimerization surface, with only 28 close contacts. This drastic reduction in interactions together with the two cavities observed (Fig. 4e) in MecI-DD, contributing to a certain sponginess (see above), would ultimately result in the dimer coming apart, as the DBDs recognizing each half of the dyad do not interact. A further conceivable mechanism foresees that the repressor is not stoichiometrically cleaved, in accordance with previous suggestions (26,33). In this case, a heterodimer containing a wild-type monomer and a free DD could be present in the cytoplasm, which should be unable to bind DNA. In either case, failure to recognize its target sequences would result in transcription of mecA but also of mecR1 and mecI. CONCLUSION Because a protein HTH motif recognizes the major groove of dsDNA only through one face by means of its recognition helix, only five or six base pairs may be directly contacted. Accordingly, DNA-binding proteins harboring one copy of this motif need to oligomerize to recognize larger operators. A further restriction is introduced by the nature of the cognate sequence. Palindromes or pseudosymmetric operators mostly require binding protein molecules to homo-oligomerize employing a dyad, which requires self-complementary protein interaction surfaces. Transcription factors binding DNA as dimers may either dimerize first and then associate with DNA or two monomers can bind sequentially and assemble their dimerization interface while binding DNA. Based on equilibrium ultracentrifugation and chemical cross-linking experiments, it has been suggested that B. licheniformis BlaI/PenI binds as a preformed dimer and cooperatively when recognizing the different operators within the bla divergon (24,25). The dimeric arrangement of methicillin repressor described here confirms our hypothesis and the findings of others (19, 24 -26) that S. aureus MecI (and, therefore, probably also BlaI and other members of the MecI/ BlaI family of (putative) transcriptional regulators; see Fig. 1) may also bind as a preformed dimer. An additional argument reads that it is actually dimer disruption through limited proteolysis, which disables DNA binding. For BlaI it has been shown that it fails to retard a synthetic dsDNA encompassing just one half site of the palindrome (26).Therefore, it is unlikely that a monomer can bind to an operator half site and then recruit the second monomer. More likely, MecI interacts as a preformed dimer with target DNA with each DBD recognizing one half of the pseudo-palindromic sequence and without any interaction between them. In the case of mec, the two dyads recognized by MecI are consecutive, with a distance of 10 bp between their centers (18). This positions the operators on the same face of the DNA helix. If each operator is recognized by a repressor dimer, interaction should be possible between the wings in the minor groove. In the case of bla, there is a complete turn between the recognized R1-and Z1-dyads. Interaction between the dimers thus seems unlikely, in particular if the dsDNA is not bent (19). This is in agreement with the separation of the probable recognition helices ␣3.
The current structure of methicillin repressor, the ultimate transcriptional regulator of mecA-mediated methicillin resistance in MRSA, provides the first step in the understanding of the molecular basis of its function and underlying regulatory process, as well as those of related proteins of the BlaI/MecI family. This information may work the switches to develop novel therapies as drugs disrupting this regulatory pathway responsible for resistance rescue the effectiveness of BLAs against MRSA.