Structural and biochemical analyses of alanine racemase from the multidrug-resistant Clostridium difficile strain 630

Structures of C. difficile alanine racemase in complex with the PLP cofactor and cycloserine are presented.


Introduction
The Gram-positive spore-forming anaerobic bacterium Clostridium difficile was discovered in the feces of newborns by Hall and O'Toole in 1935, and was initially named Bacillus difficile owing to the difficulty of culturing it in vitro (Hall & O'Toole, 1935). The multiple strains of C. difficile are either dormant spore-forming or vegetative toxin-producing forms (Bartlett, 2009a). Both forms cause disease; however, the spore-forming forms are resistant to antibiotics while the vegetative forms are more susceptible to some antibiotics (Bartlett, 2009b). C. difficile was identified in 1978 as the agent responsible for pseudomembranous colitis, with hypervirulent strains, including strain 630, causing morbidity and mortality after exposure of patients to fluoroquinolinones or protonpump inhibitors (Carroll & Bartlett, 2011). C. difficile is a troubling nosocomial bacterium that is known to contaminate hospital air, surfaces and personnel, thus posing a great risk to immunocompromised patients (Carroll & Bartlett, 2011;Best et al., 2010). Drug-resistant C. difficile infections usually result in explosive diarrhea, exacerbating the spread of the bacteria in hospitals and the community (Burnham & Carroll, 2013;Best et al., 2010;Bartlett, 2009a). The virulence and mortality of C. difficile infection is on the rise; thus, there is an urgent need for new therapeutics to combat this pathogen (Burnham & Carroll, 2013).
Alanine racemase (Alr; EC 5.1.1.1) is the prototypical member of the fold type III superfamily of enzymes that are dependent on pyridoxal-5 0 -phosphate (PLP) for catalysis (Jansonius, 1998). Alr catalyzes the reversible racemization of l-alanine and d-alanine; the latter is a key building block in the biosynthesis of the peptidoglycan layer in bacterial cell walls (Walsh, 1989). Alr has been extensively studied as a target for antimicrobial drug development because of its role in an essential bacterial process and the lack of a homologous function in humans (Silverman, 1988). Genetic studies have established that bacteria carry either one or two alanine racemase genes (alr and/or dadX). The alr gene encodes the constitutively expressed biosynthetic enzyme that provides d-alanine necessary for peptidoglycan synthesis (Wasserman et al., 1983). The catabolic dadX gene, which is present mainly in Gram-negative bacteria, encodes a second alanine racemase isozyme with expression induced by high levels of l-alanine in the medium (Strych et al., 2000). The dadX gene product converts the abundant l-alanine to d-alanine, which is subsequently metabolized to pyruvate and used as an alternate carbon and energy source (Wasserman et al., 1983;Wild et al., 1985). Only one alr gene has been identified in the C. difficile genome, and as in other bacteria with only one alr gene this gene is likely to be essential for C. difficile growth (Monot et al., 2011).
Structural studies of Alrs from bacteria such as Geobacillus stearothermophilus (Shaw et al., 1997), Streptomyces levendulae (Noda et al., 2004), Pseudomonas aeruginosa (LeMagueres et al., 2003), Mycobacterium tuberculosis (LeMagueres et al., 2005), Escherichia coli (Wu et al., 2008), Bacillus anthracis (Au et al., 2008;Couñ ago et al., 2009), Streptococcus pneumoniae (Im et al., 2011) and Staphylococcus aureus (Scaletti et al., 2012) indicate that Alr is a homodimeric enzyme formed by head-to-tail association of two monomers. Each monomer consists of two domains: an eight-stranded / barrel at the N-terminus and -strands at the C-terminus. The dimer contains two active sites, which are formed by the association of the N-terminal domain of one monomer and the C-terminal domain of the other. In each active site, the PLP cofactor resides in the mouth of the / barrel, forming a Shiff base with a catalytic lysine. Structures of Alr in complex with substrate analogs (Fenn et al., 2003;Morollo et al., 1999;Stamper et al., 1998;Watanabe et al., 2002) and site-directed mutagenesis studies of the enzyme (Watanabe et al., 2002;Watanabe, Kurokawa et al., 1999;Sun & Toney, 1999) have elucidated the reaction mechanism and verified the key roles of active-site residues. In general, Alr utilizes a stepwise two-base mechanism whereby two highly conserved lysine and tyrosine residues at the active site act as acid/base catalysts, assuming the roles of either donating or abstracting an -hydrogen from the substrate.
Structures of Alr with inhibitors, such as alanine phosphonate and d-cycloserine, show that these inhibitors covalently bind to the PLP cofactor (Fenn et al., 2003).
The availability of structural information on Alrs from a range of bacterial species is crucial for the design of inhibitors that can serve as antibiotics. Although numerous structures of Alrs from aerobic and facultative anaerobes are available, very little is known to date about the homologous enzyme from the obligate anaerobe C. difficile. Thus, the successful purification, crystallization, structure determination and kinetic characterization of C. difficile alanine racemase (CdAlr) from the hypervirulent and multidrug-resistant strain 630 are presented here.

Expression and purification
The amino-acid sequence of C. difficile strain 630 Alr (NCBI Reference Sequence YP_001089983.1) was used as a template for the chemical synthesis of DNA expression constructs encoding the 385-amino-acid protein as well as a point mutant carrying a lysine-to-threonine substitution at position 271, K271T (GenScript USA Inc., Piscataway, New Jersey, USA). These synthetic constructs were codon-optimized for expression in E. coli and cloned into the XbaI/BglII sites of pET-32a vector (Novagen, Madison, Wisconsin, USA). The expression plasmids were transformed into E. coli strain Rosetta 2 (pLysS) DE3 by heat shock. Positive colonies were identified by colony PCR with T7 promoter and T7 terminator primers (EMD Millipore, Rockland, Massachusetts, USA). Initial small-scale expression screens were performed on ten positive clones by culturing a single colony in 10 ml LB medium containing 50 mg ml À1 ampicillin to an OD 600 of 0.7. Protein expression was induced by the addition of IPTG to a final concentration of 0.5 mM for 16 h at 288 K. The highest expressing clone was selected for large-scale expression. Cell pellets were harvested by centrifugation, suspended in 20 mM Tris pH 7.5 with protease inhibitors (Roche) and lysed using a French press. The lysate was clarified by centrifugation prior to Alr purification by anion-exchange chromatography (Q Sepharose Fast Flow, GE Healthcare) and elution with an NaCl gradient in lysis buffer. This was followed by further purification by size-exclusion chromatography (HiPrep 16/60 Sephacryl S-300 HR) with 100 mM Tris pH 7.5, 1 mM PLP, 1 mM TCEP. The typical yield of purified recombinant CdAlr was 15 mg from 1.2 l of culture. Coomassie-stained SDS gels were used to assess the purity of the recombinant proteins, which were dialyzed into storage buffer (50 mM Tris pH 7.5, 10 mM PLP, 1 mM DTT) and stored at À80 C until use.

Enzyme kinetics
The rate of conversion of l-alanine to d-alanine was determined by measuring the amounts of both enantiomers by HPLC as described previously (Lee et al., 2013). The kinetics of alanine racemization in both the forward (l-alanine to d-alanine) and the reverse (d-alanine to l-alanine) directions research papers were determined for recombinant wild-type (CdAlr WT ) and K271T mutant (CdAlr K271T ) enzymes. The reaction mixtures consisted of l-alanine or d-alanine at a range of concentrations (0.2, 0.39, 0.78, 1.56, 3.12, 6.25, 12.5, 25 and 50 mM) in 50 mM sodium borate pH 8.5 buffer. The reaction was initiated by the addition of either CdAlr WT (480 ng) or CdAlr K271T (6 ng). The amount of each component and the reaction times were determined via optimization studies. As the reaction proceeded, aliquots were removed at 0.5 min intervals over a total of 4 min and the reaction was quenched with four volumes of methanol. The l-alanine and d-alanine enantiomers in the methanol extracts were separated based on their differential diastereomer stability via isocratic HPLC (Agilent 1100 Chemstation) using a Nucleosil Chiral-1 column (Machery-Nagel, Germany) with a mobile phase containing 0.5 mM copper sulfate solution at a flow rate of 0.3 ml min À1 at 333 K. The percentage of d-alanine product in the total alanine mixture was calculated from the peak areas in the HPLC profile at A 240 nm . The rates of production of l-alanine or d-alanine (mM min À1 ) were plotted against the substrate concentration and the slopes were used in the Michaelis-Menten analysis to obtain V max and K m (GraphPad Prism 6, GraphPad Software, La Jolla, California, USA). The k cat was calculated from the V max [k cat = V max (mM min À1 )/Alr concentration (mM)]. The results shown are the mean AE SD of three independent experiments.

Complementation analysis
For complementation analysis, two plasmids in which the alr genes were constitutively expressed from the lac promoter were constructed by inserting synthetic genes encoding C. difficile wild-type Alr and K271T Alr into the XbaI-SacI sites of the pUC57 vector. The alr/dadX double-knockout E. coli strain MB2159 (Strych et al., 2001) was transformed with the plasmids by electroporation (Bio-Rad; 25 mF, 200 and 1.65 kV). Transformants were selected on LB medium in the absence of exogenous d-alanine.

Size-exclusion chromatography and multi-angle light scattering (SEC-MALS)
SEC-MALS experiments involved loading 10 mg of each protein sample onto a stationary phase comprised of a TSKgel Super SW mAb HTP column (Tosoh Biosciences, King of Prussia, Pennsylvania, USA) at a flow rate of 0.25 ml min À1 with an Agilent 1260 Infinity series HPLC. The mobile phase buffer was phosphate-buffered saline pH 7.4. Each experiment was performed in triplicate. The detector system consisted of a UV detector (Agilent) and a miniDAWN triple-angle lightscattering detector (Wyatt Technology) with an Optilab rEX differential refractometer (Wyatt Technology) connected in series. The refractometer provides a continuous index of protein concentration using a dn/dc (refractive-index increment) value of 0.185 ml mg À1 . Bovine serum albumin was used as an isotropic scatterer for detector normalization. Since the light scattered by a protein is directly proportional to its weight-average molecular mass and concentration, molecular masses were calculated from the light-scattering and interferometric refractometer data using the ASTRA 6.0 software.

Crystallization
Both the purified recombinant wild-type (CdAlr WT ) and K271T mutant (CdAlr K271T ) protein samples were concentrated to 32 mg ml À1 in buffer composed of 10 mM PLP and 1 mM TCEP in 50 mM Tris-HCl pH 8 prior to setting up crystal screens. Initial crystallization screens were carried out on samples of both proteins using commercial sparse-matrix screens from Hampton Research and Qiagen, including The Classics I, Classics II, Cryos, PEGs and AmSO 4 Suites and Index at both 293 and 298 K by vapor diffusion in sitting drops. Drops were prepared by mixing 2 ml protein solution with 1.5 ml reservoir solution. Large yellow plate-like crystals of both enzymes were obtained using high concentrations of polyethylene glycol (PEG) at both temperatures. These crystals were typically 0.4 Â 0.5 Â 0.1 mm in size. The CdAlr WT crystals were obtained in 0.2 M ammonium sulfate, 0.1 M bistris pH 6.5, 25%(w/v) PEG 3350. Virtually identical crystals were obtained at different pH and buffer conditions (Tris-HCl pH 8.5, dl-malic acid pH 7.0, HEPES pH 7.5 or bis-tris pH 5.5) or using different salts (magnesium chloride, sodium formate, lithium sulfate, sodium chloride or potassium bromide). The largest crystals of CdAlr K271T were obtained from 0.17 M lithium sulfate, 0.085 M Tris-HCl pH 8.5, 25.5%(w/v) PEG 4000, 20%(v/v) glycerol. The CdAlr co-crystals with cycloserine (CdAlr cycloserine ) were obtained in 100 mM cycloserine, 200 mM sodium formate, 20%(w/v) PEG 3350. Prior to data collection, all crystals that were grown in crystallization solutions with less than 15%(v/v) glycerol were incubated for approximately 30 s in cryoprotecting buffers comprised of the crystallization solution with 15%(v/v) glycerol.

Data collection and structure determination
X-ray diffraction data sets were collected at the Baylor College of Medicine core facility using a Rigaku HTC detector and a Rigaku FR-E+ SuperBright microfocus rotating-anode generator with VariMax HF optics. Each data set was collected at 100 K from a single crystal with exposure times of 45 s for 0.5 oscillations using the CrystalClear (d*TREK) package (Pflugrath, 1999). The data were processed using iMosflm (Leslie, 2006;Battye et al., 2011). For structure determination, data were collected from three crystals. The first data set was for the complex of the wild-type enzyme with the PLP cofactor (CdAlr PLP ). The second data set was for a co-crystal with the inhibitor cycloserine (CdAlr cycloserine ), while the third was for the K271T mutant enzyme in complex with PLP (CdAlr K271T ). Both the CdAlr PLP and the CdAlr K271T crystals belonged to the monoclinic space group P2 1 . The former has unit-cell parameters a = 85.2, b = 93.3, c = 107 Å , = = 90, = 91 , while the latter has unit-cell parameters a = 49.97, b = 154.3, c = 55.8 Å , = = 90, = 113 . The CdAlr cycloserine crystal belonged to the orthorhombic space group P2 1 2 1 2 1 , with unitcell parameters a = 57.9, b = 139.1, c = 144.9 Å , = = = 90 . Despite having visible reflections to 2.26 Å resolution, the CdAlr cycloserine data were less than 60% complete at high resolution. Attempts to collect more complete data just increased the redundancy of the data without increasing the completeness owing to the marked diffraction anisotropy of the crystal. According to Wlodawer and coworkers, marked anisotropy of diffraction occurs when the reflection intensity in one direction extends to higher resolution than in the other directions; since contemporary refinement programs can effectively handle the refinement of this type of data, the best remedy in this situation is to use all the weak high-resolution data, i.e. the maximum possible number of reflections (Wlodawer et al., 2013). Setting an arbitrary cutoff to maintain an overall completeness of greater than 80% results in a significantly less informative 3.4 Å resolution structure. Additionally, the overall R merge of the three data sets are artificially inflated by the high multiplicity of the data sets, since multiplicity and R merge are interdependent variables (Wlodawer et al., 2013).
All structures were solved by molecular replacement with Phaser (McCoy et al., 2005;Storoni et al., 2004). The molecularreplacement search model was a monomer of alanine racemase from S. aureus (PDB entry 4a3q) stripped of all waters and ligands (Scaletti et al., 2012). Molecular replacement was followed by iterative cycles of model building with Coot (Emsley et al., 2010) and structure refinement with REFMAC5 (Murshudov et al., 2011) within the CCP4 package . In the initial refinement stages, noncrystallographic symmetry (NCS) restraints were utilized. Additionally, the refined coordinates were subjected to PDB-REDO (Joosten, Salzemann et al., 2009;Joosten, Womack et al., 2009;Joosten et al., 2012) to optimize the model, followed by several additional cycles of manual building to fix out-of-register errors and any other badly fitting regions that were introduced after PDB-REDO. Data-collection and refinement statistics are shown in Table 1. The atomic coordinates and structure factors have been deposited in the Protein Data Bank as entries 4lus (CdAlr PLP ), 4luy (CdAlr K271T ) and 4lut (CdAlr cycloserine ).

Recombinant CdAlr is functionally active
Analysis of purified recombinant CdAlr WT by Coomassie-stained SDS gels and subsequent mass spectrometry revealed a major species corresponding to the full-length protein (molecular weight 43 306) and two minor species of molecular weights 30 407 and 12 917 (Fig. 1a). Both minor species interact tightly with the full-length protein and cannot be separated by additional size-exclusion chromatography. Subsequent peptide sequencing revealed that the minor species corresponded to amino acids 1-271 and 272-382 and that the two peptides resulted from a cleavage of unknown nature at position 271. To prevent this cleavage, the lysine at position 271 was mutated to a threonine: K271T. Thr was chosen because homologous Alrs from S. aureus, M. tuberculosis, B. anthracis and G. stearothermophilus that are readily purified and crystallized have a Thr at position 271. The K271T single substitution enabled the production and purification of homogeneous full-length CdAlr protein as assessed by SDS-PAGE (Fig. 1a) and mass spectrometry (data not shown).
The biological activity and functionality of the synthetic constructs were determined by evaluating their ability to complement an E. coli d-alanine auxotroph. When transformed into an E. coli alr/dadX mutant, the plasmids encoding the wild-type and K271T mutant CdAlr proteins were able to complement the d-alanine auxotrophy and restore its growth in d-alanine-deficient medium (Fig. 1b), demonstrating that both recombinant proteins are biologically active.
After confirmation of their biological activity, the racemase activity of recombinant wild-type (CdAlr WT ) and K271T   (hkl)i are the intensity of the ith measurement and the mean intensity of the reflection with indices hkl, respectively. ‡ R cryst = P hkl jF obs j À jF calc j = P hkl jF obs j, where F obs are observed and F calc are calculated structure-factor amplitudes. § The R free set uses 5% of randomly chosen reflections. mutant (CdAlr K271T ) enzymes were assessed by testing their ability to catalyze the interconversion of l-alanine and d-alanine. Both recombinant enzymes were catalytically active (Fig. 1c). Purified CdAlr WT catalyzed the interconversion of l-alanine to d-alanine in vitro, with a K m of 17.6 mM for l-alanine and a V max of 82.1 units mg À1 in the forward reaction and a K m of 7 mM for d-alanine and a V max of 26.7 units mg À1 in the reverse reaction. CdAlr K271T had a K m of 13.8 mM for l-alanine and a V max of 4533 units mg À1 in the forward reaction and a K m of 5.4 mM for d-alanine and a V max of 1797 units mg À1 in the reverse reaction. CdAlr K271T had a significantly higher catalytic activity than CdAlr WT ( Table 2). The lysine-to-threonine substitution is relatively conservative in that both are polar and hydrophobic residues, thus the observation that the smaller (molecular mass 114 versus 146) uncharged amino acid threonine has such a large impact on the catalytic activity of CdAlr is somewhat puzzling. As mentioned previously, the main difference between the recombinant CdAlr WT and CdAlr K271T protein samples is the presence of two fragments in the former (Fig. 1a). The fragments are likely to be enzymatically inactive since the head-totail association of two monomers is necessary to form the dimeric active sites of Alr, in which both the N-terminal and C-terminal residues are essential for catalytic activity. Additionally, these inactive species are likely to associate with fulllength monomers to form inactive or partially active dimers, which could explain the inability to separate the cleaved species by size-exclusion chromatography as well as the apparent monodisperity of the sample via SEC-MALS. Therefore, the CdAlr WT preparation is likely to contain a heterogeneous population of dimers made up of both fulllength and truncated monomers, which could account for the need to use substantially more CdAlr WT than CdAlr K271T in the enzymatic assays as well as the lower catalytic activity.  The catalytic activities of CdAlr WT and CdAlr K271T were compared with those of several Alr homologs that have been reported by others. As summarized in Table 2, CdAlr K271T appears to be significantly more active than CdAlr WT and the Alrs from G. stearothermophilus, S. lavendulae, M. tuberculosis and M. smegmatis, as reflected by a 50-fold to 1000-fold higher k cat (Table 2). Significant differences in catalytic activities among Alrs with high sequence identities have been reported (Strych et al., 2000). We previously reported that the Alr of M. smegmatis, which shares 65% sequence identity with its homolog from M. tuberculosis, is 20-fold more active (

Table 2
Comparison of kinetic parameters of selected Alrs. NR, not reported. et al., 2013). The molecular basis for the variations, although still remaining elusive, has been attributed to differences in the monomer-dimer equilibrium (Ju et al., 2011). The more active isoforms are thought to exist predominantly as catalytically active dimers owing to the increased propensity of the monomers to associate, while the less active isoforms, with their lower monomer association constants, are largely monomeric. Accordingly, the high catalytic activity of CdAlr K271T could arise from the formation of a tight dimer that hardly dissociates. The biological significance of Lys271 in CdAlr instead of the Thr commonly found at this position in Alr homologs, including those from S. aureus, M. tuberculosis, B. anthracis and G. stearothermophilus, is not known. The confusion is further compounded by the observation that the S. lavendulae homolog (PDB entry 1vft; Noda et al., 2004), which has a similar k cat as CdAlr, has a His in the corresponding position. A more detailed analysis of the monomer association constants among the various Alrs is needed to shed light on the differences in the catalytic activities. It is also unclear whether the native Alr from the bacterium C. difficile undergoes cleavage similar to that observed for the recombinant enzyme produced in E. coli; this is worth exploring in order to determine whether C. difficile uses this mechanism to modulate Alr activity during anaerobic growth.

Recombinant CdAlr is a dimer in solution
In order to determine the oligomeric form of the protein in solution, the absolute molecular masses of both CdAlr WT and CdAlr K271T were measured by size-exclusion chromatography  Structure of C. difficile alanine racemase. (a) Monomer of C. difficile alanine racemase. Ribbon diagram with -helices in orange and -strands in green; the PLP cofactor bound to Lys39 is shown as magenta sticks. (b) Dimer of CdAlr K271T . One monomer is shown in blue ribbon and the second is depicted as a green ribbon and surface. The PLP adduct, Llp39, is shown as magenta sticks. (c) Superposed ribbon structures of monomers of CdAlr PLP (blue), CdAlr K271T (grey) and CdAlr cycloserine (orange). Llp39 is shown as magenta sticks and DCS is shown as blue sticks. and multi-angle light scattering (SEC-MALS). Both proteins gave single monodisperse peaks on the sizing column (Fig. 2). The molecular masses were determined to be 81.09 AE 2.04 and 80.53 AE 1.98 kDa for CdAlr WT and CdAlr K271T , respectively, indicating that both proteins form dimers in solution. These values are slightly smaller than the theoretical molecular weights of 86.6 and 86.4 kDa for the dimers of CdAlr WT and CdAlr K271T , respectively. The differences in mass are comparable to the dynamic light-scattering studies of the homolog from S. aureus, which revealed a dimer with 98.6% of the theoretical molecular mass (Scaletti et al., 2012).

Overall structure of CdAlr
Three crystal structures of CdAlr are reported in this study. The first is the structure of the enzyme in complex with the PLP cofactor (CdAlr PLP ). The second structure is that of the co-crystal with the inhibitor cycloserine (CdAlr cycloserine ). The third structure is that of the K271T mutant enzyme in complex with PLP (CdAlr K271T ). Uniquely, the CdAlr PLP crystal structure is a tetramer, while the other two structures are dimers. The CdAlr PLP structure gives the highest resolution data, but it also has the highest level of disorder and lacks several residues in proximity to the binding cavity and the dimer interface. Key disordered regions in monomer A include all residues from Lys258 to Thr279; in monomer B the disordered residues are Ser133-Gly139 and Glu173-Glu177, while in monomer D residues Gly197-Ile200 are disordered. Some of these residues were disordered in other Alr structures (Au et al., 2008;Couñ ago et al., 2009;Im et al., 2011;Scaletti et al., 2012). The CdAlr PLP tetramer is a dimer of two almost identical dimers, with an r.m.s.d. of 0.274 Å for the superpositioning of main-chain atoms of the dimers. Each dimer is structurally similar to the homodimers of CdAlr K271T (r.m.s.d. of 0.288 Å to one dimer and 0.361 Å to the other) and CdAlr cycloserine (r.m.s.d. of 0.397 Å to one dimer and 0.511 Å to the other). The structure of CdAlr K271T is the most complete structure and each monomer has ordered main-chain and sidechain electron-density maps for 383 of 385 amino acids with the two disordered residues are at the N-terminus.
The topology of CdAlr K271T was chosen to describe the overall enzyme topology since it is the most complete structure. The secondary structure of each monomer is approximately 25% strands, 26% -helices, 5% 3 10 -helices and 44% loop and turns. There are five /-motifs, four -hairpins, four -bulges, four strands, 11 -helices, six 3 10 -helices, 13 helixhelix interactions, 25 -turns and four -turns (Fig. 3). Each monomer is made up of two domains, an N-terminal eightstranded /-barrel PLP-binding domain and a C-terminal domain, linked by a short hinge region (Fig. 4a). The homodimer is the minimal functional unit of CdAlr since each active site is formed at the interface of the N-terminal domain of one monomer and the C-terminal domain of the second monomer. The two monomers form a tight interface across the twofold axis of symmetry (Fig. 4b). The interacting interface between the monomers is 3387 Å 2 (66 amino-acid residues) from one monomer. The interface is mediated by 31 hydrogen bonds and corresponds to 19% of the surface area, which is comparable to other Alrs (Table 3b).

Active site of CdAlr
All three CdAlr structures contain a ligand in the active site, and there is clear unambiguous density for the ligands in all cases (Figs. 5a, 5b and 5c). Two structures (CdAlr K271T and CdAlr PLP ) contain the cofactor PLP, while one (CdAlr cycloserine ) contains the covalent adduct of the inhibitor cycloserine and the PLP cofactor. In each active site of CdAlr K271T  The Alr active site. Stereoview of cofactor PLP and active-site residues in a 2F o À F c electrondensity map contoured at 1.2 for (a) CdAlr PLP (PDB entry 4lus) and (b) CdAlr K271T (PDB entry 4luy).
Leading to the active site is the substrate entryway, which is comprised of conserved residues that have been considered to be potential targets for drug design (LeMagueres et al., 2005;Wu et al., 2008). As was the case in other Alrs, the residues lining the substrate entryway are from two monomers, and include residues from across the dimer interface (Fig. 6a). Two monomers contribute to the substrate entryway. One monomer contributes the key residues Asp172, Tyr356, Thr353, Ser234, Pro233 and Ile354, while another monomer across the dimer interface contributes Lys/Thr271, Arg312, Arg293 and Tyr287 to the entryway (Fig. 6a). Lys271 lies on a loop in the substrate entryway of the active site and it has previously been shown that mutations in this region affect the activity of EcAlr (Wu et al., 2008). Interestingly, the residues in the substrate entryway of EcAlr superpose quite well with those of CdAlr and are well conserved (Fig. 6b). The K271T mutation results in a more open and accessible entryway (Fig. 6a), which may contribute to the observed increased activity.

Comparison with other alanine racemase structures
The structures that were most similar to CdAlr were identified using the Structure Similarity option in PDBeFold (http://www.ebi.ac.uk/msd-srv/ssm/). The criteria for identifying similar structures included the r.m.s.d. for alignment of all main-chain atoms, alignment length and number of gaps. Although it only has 37% sequence identity to CdAlr, the crystal structure of the G. stearothermophilus (GsAlr) Y265F mutant alanine racemase in complex with cycloserine (PDB entry 1xqk; Fenn et al., 2005) has the greatest overall threedimensional structural alignment with the CdAlr structures. Other complexes and mutants of GsAlr also ranked high in terms of structural similarity. The corresponding PDB entries are 1l6g (Watanabe et al., 2002), 1epv (Fenn et al., 2003) and 2sfp (Fenn et al., 2003). Other proteins that share significant structural similarity include the B. anthracis homolog (PDB entry 3ha1; Couñ ago et al., 2009) as well as the S. aureus homologue (PDB entry 4a3q; Scaletti et al., 2012). These Alrs share $37% sequence identity with CdAlr. The monomers of all these structures superpose well and have r.m.s.d.s for alignment of all main-chain atoms varying from 1.45 to 1.62 Å against the three CdAlr structures, despite the observation that CdAlr has longer helices, strands and loops than many of the other Alrs.
The N-terminal and C-terminal domains of all reported Alr structures are connected by a short hinge region, and it was previously revealed that the hinge angles between the N-and C-terminal of Alrs vary (LeMagueres et al., 2003;Im et al., 2011). To compare the hinge angles of CdAlr with those of representative Alrs, the angles were calculated using Peter Sun's applet Hinge ( (Table 3a). There does not seem to be any correlation between the size of the hinge angle and the number of atoms in each domain. Additionally, there does not appear to be any correlation between hinge-angle size and catalytic activity (Tables 2 and 3a). CdAlr has a larger monomer surface area than many of the other Alrs, despite having a similar number of amino-acid residues (Table 3b). The regions that are most different among all of the Alrs are in the N-and C-terminal loops as well as in the loop regions proximal to the dimer interface and the active site (Au et al., 2008;Couñ ago et al., 2009;Kanodia et al., 2009;Im et al., 2011;Scaletti et al., 2012). These variations in the dimer interface contribute to differences in the buried surface area of the different Alrs (Table 3). The size of the dimeric buried surface area does not correlate to Alr enzymatic activity, as measured in terms of the ability to interconvert l-alanine to d-alanine (Tables 2 and 3b).

Comparison of CdAlr with E. coli Alr
In order to selectively target CdAlr for drug development, its differences from Alr from gut bacteria such as E. coli have to be identified. Structures of E. coli Alr (EcAlr) have been reported, including those of mutants as well as the complex with cycloserine (Wu et al., 2008). The dimers of EcAlr and CdAlr are quite similar, having a main- Substrate entryway. (a) A stereoview of the substrate entryway of CdAlr K271T superposed with CdAlr cycloserine . (b) The same stereoview of the substrate entryway of CdAlr K271T superposed with that of EcAlr. In both figures residues from one monomer of CdAlr K271T are shown in pink, while residues from the second monomer are shown in blue and Lys271 from CdAlr WT is shown in yellow. Corresponding residues in EcAlr are colored green and aquamarine, respectively. The transparent surface and ribbon diagram shown are of CdAlr K271T. chain r.m.s.d. of 1.9 Å for the cycloserine structures or the PLP structures. Interestingly, when the monomers are superposed based on the conserved N-terminal domain, the large difference in the hinge angle is evident in the displacement in the C-terminal domains (Fig. 7a). In addition to the difference in the hinge angle, the loops in proximity to the binding cavity also differ in EcAlr and CdAlr. Despite these differences, the active sites are well conserved and superpose quite well (Fig. 7a). A closer comparison of the active sites reveals a conserved network of residues involved in interactions with either PLP or DCS (Figs. 7b and 7c). Additionally, Lys130 is carbamylated in both structures. Residues in the substrate entryway of EcAlr that are believed to be important for catalytic activity are conserved and superpose well with their counterparts from CdAlr (Fig. 6b). The substrate entryway of EcAlr is less open than that of CdAlr because of the orientation of the equivalent residues to Arg312 and Asp172 that close off the entryway (Fig. 6b). The major difference between the structures is that the hinge angle for EcAlr is substantially larger than that of CdAlr ( Fig.  7a and Table 3a). Further studies are needed to determined whether this difference can be exploited for inhibitor design.

Conclusions
The biochemical and structural studies of CdAlr presented here reveal that a mutant form of the enzyme has an unusually high catalytic activity despite an overall conserved fold with other members of the Alr family. The high catalytic activity of CdAlr K271T is intriguing, and further studies of this mutant are needed to elucidate the structural basis for the observed difference. While the overall structure of CdAlr is similar to that of Alr from E. coli, further studies are required to determine whether the differences in the hinge region and the catalytic activity can be exploited for the design of CdAlr-specific inhibitors. The reported CdAlr crystal structures provide templates for future structure-based drugdesign studies of the enzyme, with the ultimate goal of developing new antibiotics for multidrug-resistant C. difficile.
This project was supported by a grant (5U01AI082081) from the National Institute of Allergy and Infectious Diseases (KA) and startup funds provided by the National School of Tropical Medicine at the Baylor College of Medicine (OAA).

Figure 7
Comparison of C. difficile and E. coli Alr. (a) Superposed ribbon diagram of CdAlr K271T (gray) with EcAlr (cyan). The larger hinge angle of EcAlr is obvious from the displacement of the corresponding loops in the C-terminal domain, indicated by arrows. The N-terminal residue (Lys3) and C-terminal (Lys385) residue of the CdAlr K271T structure are labeled. Also labeled are the N-terminal (Met1) and C-terminal (Asp359) residues of the EcAlr structure. (b) A stereoview of the superposed active sites of CdAlr K271T (gray) and EcAlr (green). (c) A stereoview of the superposed residues in the active site of CdAlr cycloserine (gray) and EcAlr in complex with cycloserine (green).