Crystal structures of the Apo and penicillin-acylated forms of the BlaR1 beta-lactam sensor of Staphylococcus aureus.

Staphylococcus aureus is among the most prevalent and antibiotic-resistant of pathogenic bacteria. The resistance of S. aureus to prototypal beta-lactam antibiotics is conferred by two mechanisms: (i) secretion of hydrolytic beta-lactamase enzymes and (ii) production of beta-lactam-insensitive penicillin-binding proteins (PBP2a). Despite their distinct modes of resistance, expression of these proteins is controlled by similar regulation systems, including a repressor (BlaI/MecI) and a multidomain transmembrane receptor (BlaR1/MecR1). Resistance is triggered in response to a covalent binding event between a beta-lactam antibiotic and the extracellular sensor domain of BlaR1/MecR1 by transduction of the binding signal to an intracellular protease domain capable of repressor inactivation. This study describes the first crystal structures of the sensor domain of BlaR1 (BlaRS) from S. aureus in both the apo and penicillin-acylated forms. The structures show that the sensor domain resembles the beta-lactam-hydrolyzing class D beta-lactamases, but is rendered a penicillin-binding protein due to the formation of a very stable acyl-enzyme. Surprisingly, conformational changes upon penicillin binding were not observed in our structures, supporting the hypothesis that transduction of the antibiotic-binding signal into the cytosol is mediated by additional intramolecular interactions of the sensor domain with an adjacent extracellular loop in BlaR1.

in autolytic activation of the intracellular protease domain (10), allowing it in turn to catalyze cleavage of BlaI/MecI (directly or indirectly) at a site critical for dimerization. This cleavage event effectively destroys the ability of the repressor to bind DNA, permitting the transcription of not only blaZ/mecA, but blaI/mecI and blaR1/mecR1 as well. In this system, expression of these proteins is efficiently terminated when the signaling antibiotic levels are reduced (primarily through hydrolysis by surrounding ␤-lactamase enzymes (10)). Whereas the bla divergeon (i.e. blaZ, blaI, and blaR1) is plasmid-borne, constitutive ␤-lactamase expression has been observed in S. aureus strains possessing normal penicillinase plasmids (11). Several explanations for this observation have been proposed, including the involvement of an as of yet unidentified chromosomally encoded regulatory component known as BlaR2 (12,13).
BlaR1 and MecR1 from S. aureus share significant sequence identity (sensor domains, 43%; protease domains, 33%; and full-length proteins, 34%), and the regulatory genes of bla and mec have been shown to be interchangeable in vivo (14,15). Indeed, due to the potency of the MecI repressor, many methicillin-resistant S. aureus strains have MecI deletions that inactivate repression (16), incurring either constitutive expression of PBP2a or inducible expression regulated by BlaR1/BlaI. The recent observation that the absence of the bla or mec regulatory genes selects against PBP2a expression suggests a role for these genes in stabilizing dissemination of mecA to new host strains (17).
Crystal structures have been determined previously for the S. aureus repressor MecI (13,18) and the apo form of the BlaR1 sensor domain of Bacillus licheniformis (19). In addition, the NMR solution structure is available for the B. licheniformis BlaI DNA-binding domain (20). This study describes the crystal structures of the S. aureus BlaR1 ␤-lactam sensor domain (hereafter referred to as BlaR S ) in both its apo and penicillinacylated forms. These structures illuminate the active-site features that are responsible for the PBP activity of BlaR S and provide mechanistic insights into the role of the BlaR1 sensor domain in detecting ␤-lactam antibiotics and transducing the binding signal across the bacterial cell membrane.

EXPERIMENTAL PROCEDURES
Cloning, Overexpression, and Purification-BlaR1 was cloned from a plasmid containing the bla divergeon (p184R6H) as described previ-ously (10). This plasmid was originally derived from a ␤-lactamaseexpressing strain of S. aureus (67-0) isolated from a patient at San Francisco General Hospital Medical Center (21,22). The ␤-lactam sensor domain (amino acids 330 -585, i.e. BlaR S ) was subcloned into the pET41b(ϩ) vector (Novagen) and transformed into Escherichia coli Rosetta (DE3) (Novagen). Cells were grown from an overnight culture at 37°C until A 600 ϳ 0.6, heat-shocked at 42°C for 1 h, cooled to 15°C, and then induced to overexpress BlaR S overnight using 1 mM isopropyl ␤-D-thiogalactopyranoside. This same expression procedure was adapted for the production of selenomethione (SeMet)-substituted BlaR S using previously described protocols (23). The cells were lysed by high pressure homogenization using an Avestin EmulsiFlex-C5. The BlaR S protein was purified from the soluble cell fraction using three chromatographic steps performed at 4°C. The soluble cell lysate was first bound to Fractogel EMD SO 3 Ϫ resin (Novagen), pre-equilibrated in 20 mM HEPES (pH 7.5) and 50 mM NaCl, and eluted with 0.8 M NaCl. The eluate was dialyzed overnight at 4°C in buffer containing 0.2 M NaCl and then filtered using 0.22-m filters. The filtrate was further purified using a Mono S HR 5/5 cation exchange column (Amersham Biosciences) and a 0.2-0.8 M NaCl elution gradient. Elution fractions corresponding to the principal A 280 peak were combined, concentrated, and passed through a Superdex 200 HR 10/30 size exclusion column (Amersham Biosciences) equilibrated in 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl as the storage buffer. The single A 280 peak was collected and concentrated to 20 -40 mg/ml as estimated using a predicted molar absorption coefficient of 60280 M Ϫ1 cm Ϫ1 (24). Concentrated BlaR S was either used fresh (stored at 4°C) or flash-frozen in liquid N 2 and stored at Ϫ80°C.
Crystallization and Data Collection-Crystals of apo-and acyl-BlaR S were obtained using the hanging drop vapor diffusion method. For apo-BlaR S crystals, 1 l of a 20 mg/ml protein solution was added to equal volumes of reservoir solution consisting of 18 -23% polyethylene glycol 3350 and 0.2 M NaH 2 PO 4 . Equilibration over a 0.5-ml reservoir at 18°C for 1-2 weeks produced single crystals reaching dimensions of up to 0.7 ϫ 0.5 ϫ 0.4 mm. For acyl-BlaR S , benzylpenicillin was incubated with BlaR S at final concentrations of 20 mg/ml protein and 10 mM benzylpenicillin for 15 min at room temperature before initiating crystallization. Native acyl-BlaR S crystals were then grown using 1 l of this protein solution mixed with an equal volume of reservoir solution consisting of 26 -29% polyethylene glycol 3350, 0.2 M NaCl, and 0.1 M BisTris (pH 6.6). Crystals were typically observed after 1 week at 18°C. SeMet crystals of acyl-BlaR S were obtained by streak-seeding native crystals into fresh drops containing SeMet-substituted BlaR S , benzylpenicillin, and crystallization reagents at the same concentrations used to grow the native crystals. Crystal clusters formed in 3-5 days with maximum dimensions of 0.4 ϫ 0.2 ϫ 0.15 mm for the individual crystals of the cluster. Single SeMet acyl-BlaR S crystals were obtained by gently breaking the clusters.
Diffraction data were collected at 100 K using cryoprotectant solu- tions of 35% polyethylene glycol 3350 and 0.3 M NaH 2 PO 4 for the apo-BlaR S crystal and 35% polyethylene glycol 3350, 0.5 M NaCl, and 0.1 M BisTris (pH 6.6) for the SeMet acyl-BlaR S crystal. All data sets were collected at the National Synchrotron Light Source on beamline X8-C using an ADSC Quantum Q4R CCD detector. Data were processed using the HKL package (25) and programs from the CCP4 software suite (26). The apo crystal was of space group P2 1 , with four molecules/ asymmetric unit and unit cell dimensions of a ϭ 59.9, b ϭ 104.9, and c ϭ 90.3 Å and ␤ ϭ 107.7°. SeMet acyl-BlaR S crystallized in space group P4 3 2 1 2, with two molecules/asymmetric unit and unit cell dimensions of a ϭ b ϭ 88.4 and c ϭ 125.1 Å. Statistics for data collection and processing are summarized in Table I.
Structure Solution and Refinement-The structure of acyl-BlaR S was determined using single anomalous dispersion with the peak data to locate the initial selenium sites in SOLVE (27), followed by threewavelength multiple anomalous dispersion to generate more accurate phases. Although 100% incorporation of SeMet into BlaR S was indicated by mass spectrometry, only 15 of 18 possible selenium sites were located. Phases were improved with density modification using RE-SOLVE (28). The initial model was automatically built with RESOLVE (50% complete) (29) and manually rebuilt using XTALVIEW (30). Iterations of refinement with CNS (25) and REFMAC (26) resulted in the final model. Of the 255 residues in the BlaR S construct, 242 residues were modeled in chain A (residues 4 -13, 18 -201, and 204 -251), and 241 residues were modeled in chain B (residues 8 -26 and 31-252). Model quality was analyzed using PROCHECK (84% in the most favorable region of the Ramachandran plot) (32). Asn 388 is a well ordered active-site residue, but adopts a disallowed main chain conformation due to its juxtaposition to benzylpenicillin-acylated Ser 389 . The apo-BlaR S structure was solved by molecular replacement using the acyl-BlaR S structure as a starting model and the program Molrep (33). Model rebuilding was performed with XTALVIEW using a prime-andswitch map generated by RESOLVE to reduce model bias. The model was refined and analyzed for quality as described above. The resulting model consisted of 244 residues for chain A (residues 8 -166 and 169 -253), 247 residues for chain B (residues 8 -254), 243 residues for chain C (residues 5-13, 15-76, 78 -84, 86 -202, and 206 -253), and 240 residues for chain D (residues 7-76, 80 -82, 84 -166, and 169 -252). In each of the four molecules of the apo form, a pyrophosphate was also modeled into the active site. This ligand is only partially occupied in three of the four molecules of the asymmetric unit, as revealed by the close proximity of a partially occupied water molecule to one of the phosphorus atoms. Considering that the crystallization conditions for apo-BlaR S contained 0.2 M sodium phosphate, the unknown electron density could also be modeled as two partially occupied phosphate ions, but the staggered geometry of the tetrahedral phosphorus atoms and the bent P-O-P bond indicate a molecule of pyrophosphate, introduced as a sodium phosphate impurity. The final apo-BlaR S model had 88% of the residues in the most favorable region of the Ramachandran plot and none in the disallowed region. The multiple anomalous dispersion phasing and model refinement statistics for both structures are provided in Table I. Fig. 3 was prepared with MOLSCRIPT (34) and rendered with Raster3D (35). All other protein graphics (Figs. 5-7) were prepared and rendered with PyMOL (36).
Static Light Scattering-Static light scattering experiments were performed at 25°C on a Superdex 75 HR 10/30 size exclusion column (Amersham Biosciences) using 50 mM HEPES (pH 7.5) and 100 mM NaCl. Protein concentration was 1 mg/ml. Refractive index and miniDAWN light scattering detectors (Wyatt Technology Corp.) were calibrated using bovine serum albumin (Sigma).
Mass Spectrometry-Apo-and acyl-BlaR S samples were prepared by briefly incubating the pure BlaR S protein at room temperature in the presence and absence of 10 mM benzylpenicillin. These samples were then injected onto an ultrafast microprotein analyzer (Michrom BioResources, Pleasanton, CA) equipped with a 1 ϫ 50-mm PLRP-S reverse phase column. The protein was subsequently eluted with a 2-90% gradient of acetonitrile in water and 0.05% trifluoroacetic acid at a flow rate of 50 l/min over 5 min. Mass analysis of the eluted protein was performed using a PE-Sciex API 300 triple-quadrupole mass spectrometer scanned over a mass-to-charge ratio range of 300 -2200 Da with a step size of 0.5 Da and a dwell time of 1.5 ms/step. The ion source voltage was set at 5.5 kV, and the orifice energy was 45 V. Protein molecular masses were determined from these data using the deconvolution software supplied by PE-Sciex.
␤-Lactamase Assay-A modification of the whole cell assay method of Kernodle et al. (37) was used. S. aureus cells were grown overnight at 37°C in 2.5 ml of brain/heart infusion broth containing 10 g/ml chloramphenicol to maintain the plasmid vector with and without 10 g/ml CBAP as inducer on a reciprocal shaker platform at 250 rpm. 1-ml volumes were removed; cells were pelleted by centrifugation at 10,000 ϫ g for 3 min; the broth was discarded; and cells were resuspended in 0.1 M sodium phosphate buffer (pH 6.0) to achieve a cell suspension of A 570 ϭ 0.95-1.05. A 0.9-ml volume of bacterial suspension was added to 0.1 ml of 1 mM cephaloridine (final cephaloridine concentration of 100 M) in sodium phosphate buffer and incubated at 37°C. At 30 and 60 min, 1-ml samples were centrifuged to pellet cells. The cephaloridine concentration in the supernatant was determined spectrophotometrically at 254 nm using buffer and the ␤-lactamase-negative control strain RN4220 as the blank. ␤-Lactamase activity was detected as a decrease in cephaloridine concentration.

RESULTS AND DISCUSSION
Functional Characterization of Recombinant BlaR S -The Cterminal extracellular sensor of the BlaR1 signal transducer from S. aureus (BlaR S ) was expressed in the cytoplasm of E. coli as a soluble domain, spanning amino acids 331-585 and lacking an N-terminal Met. To verify the activity of our recombinant protein, the ability of BlaR S to be covalently modified by ␤-lactam antibiotics was demonstrated using mass spectrometric analysis before and after incubation with benzylpenicillin. This experiment yielded a homogeneous mass shift of 333 Da, corresponding closely to the expected mass shift of 334 Da for a benzylpenicillin adduct. Similar deletion mutants containing the sensor domain of BlaR1 have been expressed in E. coli previously for both S. aureus (38) and B. licheniformis (39) and have been shown to fully retain their activities as highly sensitive PBPs.
Overall Fold and Oligomerization State-BlaR S was crystallized in the absence and presence of benzylpenicillin to facilitate determination of the three-dimensional structures of the apo-and acyl-BlaR S proteins by x-ray crystallography. The crystal structure of acyl-BlaR S was solved first using multiple anomalous dispersion data collected from a SeMet-substituted crystal. The final model of acyl-BlaR S consisted of two molecules/asymmetric unit, which superposed with a root mean square deviation (r.m.s.d.) of 0.58 Å on the 232 commonly observed C-␣ atoms. The model was refined to 2.4-Å resolution with R and R free values of 22.0 and 27.5%, respectively ( Table  I). The atomic coordinates from the acyl-BlaR S structure were then used to solve the crystal structure of apo-BlaR S by molecular replacement. The resulting model had four molecules in the asymmetric unit and was refined to 1.8-Å resolution with R and R free values of 18.6 and 22.7%, respectively (Table I). The four molecules superposed closely with r.m.s.d. values of 0.33-0.43 Å on the 234 -244 commonly observed C-␣ atoms.
The main chain fold of BlaR S consists of two domains, one helical and one mixed ␣/␤ with the penicillin-binding site residing at the interface (Fig. 2). The mixed ␣/␤ domain includes a seven-stranded ␤-sheet, composed of six antiparallel strands and a short parallel strand, sandwiched between a pair of ␣-helices on either side. The helical domain is composed of six helices with the active site situated in an interdomain cleft centered on helix C, which consists of a single 3 10 -helical turn followed by three ␣-helical turns. An ⍀-loop connects helices G and H and forms one edge of the active site. There is no obvious aromatic or hydrophobic patch on the surface of BlaR S , and addition of detergent or lipid was not required for solubility or activity, suggesting no close association between the sensor domain of BlaR1 and the bacterial membrane.
As observed previously for the B. licheniformis BlaR1 sensor domain (19), the S. aureus BlaR S structure resembles that of the class D ␤-lactamases, with overall r.m.s.d. values in the range of 1.17-1.41 Å on 192-200 common C-␣ residues for the available OXA structures. Overlapping the apo form of the sensor domains of B. licheniformis and S. aureus gives similar r.m.s.d. values of 1.23-1.27 Å on 214 common C-␣ atoms (36% amino acid sequence identity). The closest match of BlaR S and the class D enzymes occurs with OXA-1 from E. coli, which shares 28% amino acid sequence identity with BlaR S . This is consistent with the fact that OXA-1 is the only class D ␤-lactamase that prefers a monomeric state (40). Other class D ␤-lactamases have been shown to exist predominantly as dimers in solution, an oligomeric form that promotes maximum catalytic activity and that can be mediated by ion binding (41). BlaR S was observed as a monomer in either crystal form as well as in solution as determined by static light scattering at similar protein concentrations (data not shown). Examination of the BlaR S structure indicates that it has lost the prominent di-meric interface observed in the class D crystal structures. A comparison of BlaR S with the OXA-10 dimer reveals that many of the residues responsible for stabilizing the OXA dimer are different in BlaR S , precluding the formation of two salt bridges, three hydrophobic interactions, and three metal ligands. Although a pseudo 2-fold symmetry axis that resembles the OXA dimer is created by crystal packing in both of our structures, the strands of the intermolecular ␤-sheet meet at a steep angle of ϳ30°, and only a total of four hydrogen bonds join the two molecules.

Structures of S. aureus Apo-and Acyl-BlaR S
of partially occupied pyrophosphate that hydrogen bonds with the O-␥ of Ser 389 (2.4 Å), the N-of Lys 526 (2.9 Å), the side chains of Thr 527 and Thr 529 , the backbone carbonyl of Thr 529 , and the amide nitrogens of Ser 389 and Thr 529 that comprise the oxyanion hole.
In the acyl-BlaR S crystal, benzylpenicillin is unambiguously observed as a covalent adduct of Ser 389 in both molecules of the asymmetric unit, with C-7 bound to the O-␥ of the proposed nucleophile Ser 389 via an ester linkage (Fig. 5A). The adduct has been refined at full occupancy in each molecule, with average B-factors of 34.8 and 37.8 Å 2 (similar to the average B-factor of 31.1 Å 2 observed for the surrounding active-site residues). The backbone nitrogens of Thr 529 and Ser 389 form hydrogen bonds with the carbonyl oxygen of the adduct ester, creating the oxyanion hole typical of ␤-lactamases and serine proteases. The thiazolidine methyls are stabilized by hydrophobic interactions with the side chains of Phe 421 and Thr 529 , whereas hydrogen bonds with the side chain of Asn 439 and the backbone carbonyl of Thr 529 position the adduct amide. The thiazolidine carboxylate is fixed by hydrogen bonds with the side chains of Thr 527 and Thr 529 , a feature typical of PBPs. Interestingly, the extra arginine residue utilized by the various classes of serine ␤-lactamases to form electrostatic interactions with the thiazolidine carboxylate of the substrate is absent in the BlaR S structure, a scenario also typical of PBPs. The binding mode of benzylpenicillin is highly similar in both molecules of the asymmetric unit, except for the less ordered side chain phenyl substituent (Fig. 5B), which adopts alternative conformations in each case (making hydrophobic interactions with the side chains of Ile 531 , Thr 529 , Phe 421 , and Tyr 536 or with the side chains of Ile 531 and Met 476 and the C-␤ and C-␥ of Glu 477 ).
Role of Lys 392 as the General Base in Acylation-The active sites of the PBPs and class A, C, and D ␤-lactamases center on a common serine nucleophile and are well conserved, but the general base involved in catalysis is apparently different in each (e.g. Lys 73 (42)(43)(44) and/or Glu 166 (45)(46)(47) in class A ␤-lactamases, Tyr 150 in class C ␤-lactamases (48), carboxylated Lys 70 in class D ␤-lactamases (49), and unprotonated Lys 392 in PBPs (50, 51)) (Fig. 6). Whichever the class, the role of the general base in acylation is to activate the nucleophile that attacks the ␤-lactam ring by deprotonation of the catalytic SXXK serine. Deacylation requires either an additional general base to activate water for hydrolysis of the acyl-enzyme intermediate (e.g. class A ␤-lactamases) or a mechanism for deprotonating/regenerating the first general base (e.g. class A, C, and D ␤-lactamases). In BlaR S , the close proximity of Lys 392 (2.4 -2.8 Å) and Ser 437 (3.2-3.5 Å) to the Ser 389 nucleophile suggests two possible candidates for a general base in acylation, the former arguably more suitable in terms of distance and potential pK a to act either in an unprotonated state similar to the PBPs (19) or in a carboxylated state similar to the class D ␤-lactamases (38).
Carboxylation is favored at basic pH, but has been observed in crystal structures of the class D ␤-lactamases at as low as pH 6.0 (49) and perhaps even at pH 5.5 with low occupancy (40). The structures of apo-and acyl-BlaR S presented here were determined at pH 4.7 and 6.6, respectively. In either case, carboxylation of Lys 392 was not observed at any contour level in our electron density maps. In the class D ␤-lactamase structures, the non-carboxylated Lys 70 (corresponding to Lys 392 in BlaR1) seems to encourage an "inactive" conformation of Ser 115 (Ser 437 in BlaR1), in which the serine hydroxyl (presumed to shuttle a proton from the carboxylated lysine to the leaving group nitrogen of the ␤-lactam substrate) is somewhat displaced from its typical position in the active site (49,52). In contrast with many of the non-carboxylated structures of the class D ␤-lactamases, Kerff et al. (19) noted that the active site of the apo form of the B. licheniformis BlaR1 sensor domain (which also lacks a carboxylated active-site lysine despite the fact that the crystals were grown at pH 7.0) closely resembles the "active" conformation. Similarly, we observed the active sites of both our apo-and acyl-BlaR S forms to adopt the active conformation. These results are surprising since Lys 392 of BlaR S has been shown to specifically bind CO 2 with a dissociation constant of 0.6 M (38), only a 2-3-fold increase versus the class D ␤-lactamase OXA-10 (49). It should be noted, however, that the carboxylation of Lys 392 has not yet been directly shown to be a requirement for the ␤-lactam binding activity of BlaR1 in S. aureus, and preliminary evidence in B. licheniformis suggests no significant increase in acylating activity under conditions that would promote carboxylation of Lys 392 in that system (19).
With that in mind, the active site of BlaR S differs from those of class D ␤-lactamases such as OXA-10 in several interesting ways (Fig. 6B). The carboxylate of the carboxy-Lys 70 of OXA-10 is held in position by hydrogen bonding interactions with Ser 67 , Trp 154 , and Asn 73 (through one molecule of water) and hydrophobic interactions with Phe 120 , Ile 155 , and Val 117 . Two of these interactions have been disrupted in the BlaR S active site. Strictly conserved in the OXA structures, Val 117 has been replaced by Asn 439 in the BlaR S structure, which now forms a hydrogen bond with the N-of Lys 392 (an interaction found in all class A and C ␤-lactamases as well as in the PBPs, none of which utilize a carboxylated lysine). Likewise, Asn 73 from the structure of OXA-10 has been substituted with the aliphatic Leu 395 in BlaR1, eliminating the possibility for hydrogen bonding with Lys 392 in a putative carboxylated state. Although the other available structures of class D ␤-lactamases show mutations at this position, they represent residues capable of hydro- gen bonding (i.e. His 73 in OXA-2 (Protein Data Bank code 1k38) and Ser 73 in OXA-13 (52)) or utilize a neighboring residue as a hydrogen bond donor (i.e. Ser 120 in OXA-1 (40)). Collectively, these differences in BlaR S create an environment that may discourage carboxylation of Lys 392 . It is fascinating that these two distinctions are conserved between BlaR S and class A ␤-lactamases such as SHV-2 (Fig. 6C) as well as PBPs such as PBP2x (Fig. 6D). In this sense, the active site of BlaR S best resembles a hybrid between those of the PBPs and the class D ␤-lactamases.
Formation of the Stable Penicilloyl Adduct-BlaR S shares the greatest resemblance with the class D ␤-lactamases in terms of its fold and sequence identity, but its sluggish deacylation activity is that of a PBP. Although carboxylation of Lys 392 was not observed in the crystal structures of BlaR S , the possibility for carboxylation of this residue under physiological conditions cannot be dismissed at this time. As such, a discussion of the mechanism of BlaR S must consider both scenarios.
In the observed case, in which Lys 392 is not carboxylated, the mechanism resembles that of the PBPs and requires an unpro-tonated Lys 392 (Fig. 7A) (51). Deprotonation of the lysine could be accomplished through the relatively hydrophobic environment of the Lys 392 side chain consisting of Leu 395 , Met 434 , Phe 442 , Trp 475 , and Met 476 . Considering the hydrogen bond network surrounding Lys 392 in the apo-BlaR S structure, including a tight hydrogen bond with the proposed nucleophile Ser 389 , a proton shuttling scheme similar to that proposed for the PBPs can be envisaged (Fig. 7A). The pathway is initiated by positioning of the ␤-lactam antibiotic in the active site and abstraction of a proton from Ser 389 by Lys 392 . The tetrahedral transition state (stabilized by the backbone nitrogens of Thr 529 and Ser 389 ) collapses to break the scissile ␤-lactam amide. The close proximity of Ser 437 to two protonated lysines (Lys 392 and Lys 526 ) facilitates the abstraction of its hydroxyl hydrogen by the lone pair electrons of the ␤-lactam amide nitrogen. Ser 437 finishes the cycle by abstracting a proton from Lys 392 . Following acylation, the N-of Lys 392 is observed rotated away from the adduct. With Ser 437 protonated and in the absence of a suitable general base for deacylation, BlaR S is stabilized in an acylated state. In an alternative mechanism and as in the class D ␤-lactamases, the carboxylation of Lys 392 in BlaR1 could provide a general base not only for acylation, but for deacylation as well. For this reason, a mechanism utilizing carboxylysine for acylation in BlaR1 must include a mode of preventing regeneration (i.e. deprotonation) of this residue as the general base for deacylation. Following a mechanistic scheme for the class D ␤-lactamases (40), deprotonation by the carboxylate of carboxy-Lys 392 could activate Ser 389 for nucleophilic attack of the ␤-lactam carbonyl (Fig. 7B). Were BlaR S an OXA-like hydrolase, a deprotonated Ser 437 could then subsequently deprotonate carboxy-Lys 392 to regenerate its nucleophilicity and to permit deacylation by activation of a bound water (53). Without a structure showing a carboxylated Lys 392 in BlaR1, it is difficult to rationalize how the position of the carboxylysine carboxylate would be perturbed in BlaR1 versus a class D ␤-lactamase. Indeed, the effect of the lack of a stabilizing hydrogen bond at Leu 395 and the substitution of asparagine for valine at position 439 in BlaR1 can only be surmised. Still, one possibility is that Asn 439 may prevent deprotonation of carboxy-Lys 392 by accepting a hydrogen bond from the now protonated carboxylic acid group (Fig. 7B). A molecule of water bound to Lys 426 ultimately provides the proton for the shuttle instead, generating a stabilized acyl-enzyme. Although Asn 439 is not absolutely conserved in BlaR1/MecR1, this position appears to always be occupied by a polar residue capable of accepting a hydrogen bond (e.g. Thr 452 in the B. licheniformis BlaR (19)).
Structural Differences between Apo-and Acyl-BlaR S -The structures determined in this study provide us with the first opportunity to compare the two forms of the sensor domain of the BlaR1 receptor. The gross apo-and acyl-BlaR S structures are highly similar, possessing a r.m.s.d. in the range of 0.59 -0.74 Å on the 231-239 C-␣ atoms. The four molecules that form the asymmetric unit of the apo-BlaR S structure are shown superposed along with the two molecules that compose the asymmetric unit of the acyl-BlaR S structure (Fig. 8A). It is clear from this superposition that the only regions that differ in position between the apo-and acyl-BlaR S structures are surface-exposed loops that likely vary due to thermal motion/ mobility as opposed to significant conformational differences between the apo-and acyl-BlaR S forms. In the case of acyl-BlaR S , these loops exhibit weak electron density and higher than average B-factors. These disordered regions include (i) the ␤-hairpin connecting strands 5 and 6 (residues 531-537), (ii) the large loop connecting ␣-helices C and E and including the short ␣-helix D (residues 406 -427), and (iii) the N-terminal region up to ␤-strand 2 (residues 334 -360). Although these loops are substantially better ordered in the apo structure, they make numerous stabilizing interactions with neighboring molecules in the P2 1 crystal lattice.
The differences between the active sites of apo-and acyl-BlaR S are similarly subtle (Fig. 8B). Indeed, besides the actual acylation of Ser 389 , the largest structural difference appears to be a slight repositioning of ␤-strand 5 due to hydrogen bonds with the benzylpenicillin adduct. The change in position of Ile 531 is not surprising since it composes the C-terminal end of ␤-strand 5 and the beginning of a hairpin turn that is poorly ordered in the acylated structure. A rotation about the 1 of Thr 527 is a consequence of a hydrogen bond with the carboxyl- FIG. 8. A, overlay of apo-and acyl-BlaR S represented as strings passing through main chain C-␣ atoms. The blue strings correspond to the four molecules in the asymmetric unit of the apo-BlaR S structure, whereas the green strings correspond to the two molecules in the asymmetric unit of the acyl-BlaR S structure. The benzylpenicillin adduct is shown as yellow and purple sticks to indicate the variable position of the phenyl group. term, terminus. B, overlay of molecules from the respective asymmetric units of apo-BlaR S (tan) and acyl-BlaR S (green). The superposition was generated with Swiss-Pdb Viewer Magic Fit (31). ate of the benzylpenicillin adduct. Acylation also reorients the N-of Lys 392 away from Ser 389 , breaking the hydrogen bond observed in the apo structure.
Transmembrane Signal Transduction-There are no major conformational differences observed between the structures of the apo-and acyl-BlaR S forms to provide an explanation as to how ␤-lactam binding at the extracellular face of BlaR1 initiates a signal capable of being transduced to the BlaR1 cytosolic protease domain. Based on the results of previous circular dichroism experiments that indicated an enhancement of secondary structure upon acylation (38), it was expected that acylation of BlaR S by benzylpenicillin would result in a change in conformation that could alter the interaction of the BlaR1 sensor domain with the transmembrane domain (Fig. 1). Although we cannot rule out the possibility that the unexpected binding of pyrophosphate in the active site of our apo-BlaR S structure triggered the "acyl" conformation (indeed, several of the interactions of pyrophosphate with the BlaR S active site closely resemble the interactions made with the benzylpenicillin adduct), if this were the case, distinct conformational differences should then be observed between these structures and the structure of B. licheniformis apo-BlaR S (19). However, superpositions of the B. licheniformis apo structure with the structures presented here reveal no significant conformational differences (r.m.s.d. of 1.23-1.27 Å on 214 C-␣ atoms for apo-BlaR S using pairwise comparisons with the molecules of the asymmetric unit and r.m.s.d. of 1.07-1.22 Å for the same set of comparisons in acyl-BlaR S ). Moreover, recent experiments using B. licheniformis BlaR S revealed no significant alterations in secondary or tertiary structure upon acylation with ␤-lactam antibiotics using a battery of biophysical techniques, including circular dichroism, fluorescence spectroscopy, Fourier transform infrared spectroscopy, and deuterium/hydrogen exchange kinetics (54). Although an argument can be made that the carboxylation of Lys 392 is required not only for acylation, but for maintaining the acyl conformation, no significant conformational changes have been observed in the class D ␤-lactamases upon ␤-lactam acylation. Accordingly, it is more likely that some component of the BlaR1 "sensor" remains missing. Recent evidence points to the 56-residue extracellular loop (L2) connecting the second and third transmembrane helices in BlaR1 as the missing trigger (Fig. 1). Phage display experiments with components of B. licheniformis BlaR1 predict an interaction between the L2 loop and BlaR S that is interrupted by the acylation of BlaR S with ␤-lactam antibiotics (54).
To address this possibility in vivo in the S. aureus BlaR1 system, we have performed site-specific mutagenesis on highly conserved residues within the L2 loop. Mutation of either Pro 49 or Pro 52 to alanine in the L2 loop was observed to negate inducibility of ␤-lactamase expression (Table II). In fact, ␤-lactamase activity was not even detectable in the proline mutants because the cephaloridine concentrations were virtually identical to those of the ␤-lactamase-negative RN4220 control. This result suggests the disruption of a tonic interaction between BlaR S and the L2 loop that determines basal levels of ␤-lactamase production, as was observed for the wild-type ZRI transformant, which hydrolyzed some cephaloridine in the absence of inducer. These prolines constitute part of a conserved PXXP sequence motif located at the N terminus of the L2 loop. We propose that the conformational rigidity inherent to proline residues serves to anchor the L2 loop appropriately for interaction with and regulation of the sensor domain. The lack of apparent conformational differences between the apo and acylated forms of BlaR S supports the prediction that the L2 loop is responsible for modulation of BlaR1 activity by interaction with the BlaR S active site. Indeed, integrity of the L2 loop appears to be required for transmitting the ␤-lactam-binding signal and de-repressing ␤-lactamase expression. We have also observed that the fusion of either a Myc or hexahistidine tag to the C terminus of BlaR1 confers a phenotype of constitutive high level ␤-lactamase expression (Table II), perhaps because either tag interferes with the binding of the L2 loop to the BlaR1 sensor domain.
The first structures of the sensor domain of apo-and acyl-BlaR1 from S. aureus have provided mechanistic predictions that can be probed further using active-site mutants. Of special import is concluding whether or not carboxylation of Lys 392 is required for acylation because this detail will clarify the PBP activity of BlaR S . The absence of a conformational change upon acylation of BlaR S strongly suggests that the L2 loop plays an integral role as a molecular trigger in signal transduction. This idea is supported by in vivo site-directed mutagenesis of the L2 loop. From the perspective of structure-based drug design, a detailed description of this interaction could inspire new drug leads. In the context of the entire BlaR1 signal transduction process, many questions remain that would clearly benefit from future structural studies. An understanding of the regulation machinery of ␤-lactam resistance not only will provide the possibility of artificially down-regulating resistance in dangerous pathogens such as S. aureus, but will resolve a novel and fascinating mode of regulating protein expression.