The SrtB_ΔN26 from C. difficile 630 was cloned into a pMCSG7 vector by using a ligation-independent cloning method (Aslanidis and de Jong, 1990) and transformed into E. coli BL21 (DE3). A recombinant 6xHis-tagged SrtB_ΔN26 protein was induced by adding 0.5 mM isopropyl-β-D-thiogalactopyranoside when the cells reached an O.D.₆₀₀ of 0.5, and further incubated at 37°C for 4 h. Cells were centrifuged at 8000 rpm for 30 min at 4°C, resuspended in buffer A (20 mM HEPES pH 7.4, 200 mM NaCl and 20 mM imidazole), and disrupted by sonication on ice. Moreover, the supernatant was loaded into an Ni-NTA column (GE Healthcare Life Sciences) and contaminant proteins were eliminated through a washing procedure by using 60 mM imidazole in buffer A. SrtB_ΔN26 proteins were eluted with 300 mM imidazole in buffer A. Fractions containing SrtB_ΔN26 proteins were pooled and further purified through HiLoad 26/600 Superdex™ 75 size-exclusion chromatography (GE Healthcare Life Sciences). Subsequently, the proteins were dialyzed in buffer B (10 mM HEPES pH 7.4 and 150 mM NaCl) and stored at 4°C for further use.

The peptide substrate of Cd-SrtB_ΔN26, PPKTG was conjugated using a fluorophore, 5-[(2-aminoethyl) amino] naphthalene-1-sulfonic acid, and a quencher, 4-([4-(dimethylamino) phenyl] azo) benzoic acid. To determine the suitable concentrations of Cd-SrtB_ΔN26 and fluorescently labeled PPKTG peptide in the assay, a matrix of various enzyme and substrate concentrations in the total volume of 100 μL in FRET buffer (10 mM HEPES pH 7.4 and 150 mM NaCl) was reacted in a 96-well black polystyrene plate and was incubated at 37°C for 48 h. The fluorescence signal was monitored at an excitation/emission wavelength of 340/490 nm and recorded every hour during the first 8 h and then at 24, 36, and 48 h by using a Spectra-Max M3 plate reader (Molecular Devices). The optimal concentrations of Cd-SrtB_ΔN26 and fluorogenic peptide used in our reactions are 240 and 20 μM, respectively. Stock solutions of MTSET and AAEK1 were dissolved in the FRET buffer, and curcumin was dissolved in DMSO. Serial dilutions of inhibitors at the millimolar range were added into the FRET buffer. All experiments were conducted in triplicate. The data are presented as means and standard errors. The statistical significance of the inhibitory effect on enzymatic activity was calculated using GraphPad Prism software (GraphPad Software). Two-tailed unpaired Student t-tests revealed significant differences between Cd-SrtB_ΔN26 + PPKTG and Cd-SrtB_ΔN26 + PPKTG + inhibitors at different concentrations (^*p ≤ 0.05, ^**p ≤ 0.01, and ^***p ≤ 0.001).

Purified Cd-SrtB_ΔN26 proteins were concentrated to 8–10 mg/mL for crystallization trials. For sparse matrix screening, numerous commercial kits (Hampton Research and Emeralds BioSystems) were used for performing the crystallization setup of the vapor diffusion method by using a high-throughput platform (Digilab Genomic Solutions). Cd-SrtB_ΔN26 crystals were observed in sitting drops containing 0.5 μL of protein and 0.5 μL of various crystallization solutions at 25°C within 1 week. Diffraction quality crystals were obtained using the hanging drop method by mixing 1 μL of protein (10 mg/ml in 10 mM HEPES pH 7.4 and 150 mM NaCl) and 1 μL of solution (0.1 M citric acid pH 3.5, 24% PEG 3350 and 0.1 M glycine). Prior to data collection, the crystals were directly mounted on loops from mother liquor and flash-frozen in liquid nitrogen without an additional cryoprotectant treatment.

Diffraction data were collected on beamline BL13B1 of the National Synchrotron Radiation Research Center (NSRRC; Hsinchu, Taiwan) and beamline 4.2.2 of the Advanced Light Source (Berkeley, CA, USA). Most of our sortase crystals did not diffract beyond 3 Å resolution. The best crystal diffracted to 2.67 Å resolution and native data were collected at BL13B1 of the NSRRC. Ninety frames were collected, each with 1° oscillation and were exposed for 30 s at the wavelength of 1.0 Å with the crystal-to-detector distance of 400 mm at a temperature of 100 K. The data were indexed, integrated, and scaled using HKL2000 (Otwinowski and Minor, 1997). The initial data were scaled to 2.67 Å resolution, but the I/σI decreased to 1.67 at the highest resolution shell (2.77−2.67 Å) suggesting that the data were effective at a resolution of approximately 2.8 Å. The crystallographic parameters and data collection statistics were summarized in Table 1.

The crystal structure of Cd-SrtB_ΔN26 was solved by molecular replacement method by program Phaser-MR (Mccoy et al., 2007) with the structure of Sa-SrtB (PDB 1QWZ) (Zong et al., 2004a) as a search model. Initially, the structure was determined at 3.5 Å resolution, and a polyalanine model was constructed. With the availability of better native datasets at higher resolutions, the model was manually rebuilt using COOT (Emsley et al., 2010) guided by 2Fo-Fc and Fo-Fc density maps. Computational refinement was conducted using REFMAC (Murshudov et al., 2011) and PHENIX (Adams et al., 2010), with 5% of the data flagged for cross-validation. We first carried out the refinement at 2.67 Å resolution, but the statistics were poor. The structural quality was improved when we systematically excluded the weak inflections by truncating data at different resolutions. Iterative model rebuilding and refinement were conducted. The final refinement statistics for the structural model at 2.8 Å resolution were summarized in Table 1. Coordinates and structure factors with the identifier 5GYJ have been deposited in the Protein Data Bank.

The peptide of sequence NPQC co-crystalized with S. aureus SrtB structure (PDB 4LFD) was positioned in the catalytic pocket of C. difficile SrtB by superimposing of S. aureus SrtB onto C. difficile SrtB (Cd-SrtB). It was then mutated into a set of peptides of our interest, including PPKT and NPQT. PPKTG and NPQTN were modeled by adding one more G and N, respectively, in the C-terminus using the package VMD 1.9.2 (Humphrey et al., 1996). We further replaced the P4 residue in PPKTG to give SPKTG and NPKTG. Missing loops of 27–28 (ML), 162–167 (ESDYDY), 210–216 (TYEFDDA), and 225 (I) in the Cd-SrtB were modeled by UCSF CHIMERA (Yang et al., 2012). Cd-SrtB−peptide complexes were solvated in TIP3P water molecules of 8Å thickness in all directions of a rectangular box (Jorgensen et al., 1983). One hundred and fifty millimolar sodium chloride were added as counter-ions to neutralize the system. Energy minimization and explicit-solvent MD simulations were performed by NAMD 2.10 package (Phillips et al., 2005; Huang and MacKerell, 2013) with CHARMM36 force field (Huang and MacKerell, 2013). Non-bonded interactions were carried out using a cut-off distance of 12 Å, with a switching distance of 10 Å. With periodic boundary conditions, the Particle Mesh Ewald method was employed for calculations of electrostatic energy (Darden et al., 1993). The Cd-SrtB, crystal waters and peptides (PPKTG, SPKTG, NPKTG, and NPQTN) were first restrained to the positions reported in the X-ray crystallography and then gradually released, first on the side chains and then the entire peptide and protein. No hydrogen atom is restrained at all time. After a 1.25 ns canonical ensemble (NVT) heating process and a short isothermal–isobaric (NPT) equilibration, the whole system was allowed for a productive run for 9 ns in a NPT ensemble at 310K and 1 atm, respectively controlled by solvent friction and Nosé-Hoover Langevin piston (Feller et al., 1995). MD simulations trajectories are further analyzed by VMD, MDAnalysis toolkit (Michaud-Agrawal et al., 2011) and in-house programs coded in python.

To understand the role of the P4 residue in the substrate peptides, we analyzed the contact frequency of the P4 residue of a peptide with Cd-SrtB_ΔN26. At every frame, a residue in Cd-SrtB_ΔN26 situating within 4.0 Å from the P4 residue of the substrate peptides is marked as a contact. For each contacting residue, the contacting percentage is defined as the number of frames that the residue is in contact divided by total number of frames in MD simulations.

To calculate the root-mean square fluctuations (RMSF) for a residue, we first iteratively superimposed the peptides in all MD snapshots to their mean positions by Kabsch's approach (Kabsch, 1976) until the process converges. RMSF of a residue is calculated as ∑i=1N∑k=1M(Xi,k - Xi¯)2NM, where N is the number of heavy atoms in this residue; M is the total number of frames; X_{i, k} is the i-th heavy atom in the k-th frame and X̄i  is the mean position for atom i over all the frames.

The recombinant 6xHis-tagged C. difficile sortase enzyme with a deletion of 26 residues at the N-terminal transmembrane domain, designated as Cd-SrtB_ΔN26, was overexpressed in Escherichia coli BL21 (DE3) and purified using an Ni-NTA affinity column (Supplementary Figure 1A). Size exclusion chromatography revealed that Cd-SrtB_ΔN26 was eluted at a volume corresponding to an apparent molecular weight of approximately 24 kDa (Supplementary Figure 1B), suggesting that Cd-SrtB_ΔN26 exists as a monomer in solution.

To confirm whether the recombinant purified Cd-SrtB_ΔN26 retains protease activity, we constructed a fluorescently labeled peptide to observe the fluorescence signal after Cd-SrtB_ΔN26 cleaves the substrate peptide in vitro (Figure 1A). In this construct, the known peptide substrate PPKTG (van Leeuwen et al., 2014) is sandwiched between a fluorophore and quencher. When the peptide remains intact, the intrinsic fluorescence is considerably reduced because of the proximity between the fluorescence donor and quenching acceptor. When the peptide is cleaved, the un-quenched fluorophore gives an enhanced fluorescent signal. To assess whether the previously described sortase inhibitors can inhibit the catalytic activity of Cd-SrtB_ΔN26, MTSET, AAEK1, and curcumin (Maresso et al., 2007; Hu et al., 2013; Donahue et al., 2014) (Supplementary Figure 2) were examined using the fluorescence resonance energy transfer (FRET)-based assay. The concentration-dependent inhibitory effects of MTSET (Figure 1B), AAEK1 (Figure 1C), and curcumin (Figure 1D) on the cleavage activity of the recombinant Cd-SrtB_ΔN26 were observed as the fluorescence signals were reduced when the inhibitors were added to the reactions comprised of Cd-SrtB_ΔN26 and fluorogenic peptides. The results further verified the recombinant Cd-SrtB_ΔN26 is catalytically active.

Cd-SrtB_ΔN26 comprises 198 residues with a 6xHis tag at the C-terminus. Cd-SrtB_ΔN26 crystallized in space group I23, with the unit cell parameters a = b = c = 121.25 Å and α = β = γ = 90°. The crystal structure of Cd-SrtB_ΔN26 was solved at 2.8 Å resolution by using the molecular replacement method, revealing one molecule in the crystallographic asymmetric unit. Most of the electron density was visible and interpretable for reliable model building. However, the density map for residues 27 and 28, 162–167, 210–216, and 225 as well as the C-terminal 6xHis tag was disordered. The crystallographic data and refinement statistics are summarized in Table 1. Validation of the Cd-SrtB_ΔN26 structure by using the MolProbity program (Chen et al., 2010) revealed no phi–psi angles in the disallowed region of the Ramachandran map.

The Cd-SrtB_ΔN26 structure possesses the sortase-unique protein fold, comprising eight β-strands (β1–β8), three α-helices (H1, H4, and H5), two 3₁₀-helices (H2 and H3), and several loops (Figure 2A). Resembling other sortase structures (Frankel et al., 2007; Kang et al., 2011; Jacobitz et al., 2014), the central β-barrel of Cd-SrtB_ΔN26 is formed by strands β1, β2, β5, and β6 on one side and by strands β3, β4, β7, and β8 on the other side. The characteristic N-terminal helix bundle, absent in SrtA structures and unique to SrtB, is composed of a 13-residue α-helix (H1), a 3₁₀-helix (H2), and a loop. The other 3₁₀-helix (H3) and a short α-helix (H4) are positioned between β4 and β5; H5 is inserted between the longest β-strand β6 and β7. The loop connecting β7 and β8 that has been postulated for accommodating peptidoglycan substrate binding was not visible in our structure, implying the flexibility of the large loop. Consistent with previous studies on sortase structures, β4, β7, and β8 of the β-barrel forming the Cys–His–Arg triad of Cd-SrtB_ΔN26 appears in a concave surface (Figure 2B). The catalytic residues Cys209 and His116 are located slightly beyond the C-terminal ends of β7 and β4, whereas Arg217 is anchored at the beginning of β8 (Figure 2C). A crystal structure of a catalytically inactive C. difficile mutant SrtB_{ΔN32, C226A} (PDB 4UX7) was published (Chambers et al., 2015). Superimposition of the catalytic residues of SrtB structures from C. difficile (PDB 5GYJ and 4UX7) (Chambers et al., 2015), S. aureus (PDB 1NG5) (Zhang et al., 2004), B. anthracis (PDB 1RZ2) (Zhang et al., 2004), and S. pyogenes (PDB 3PSQ) (Kang et al., 2011) shows the conservation of the active site (Supplementary Figure 3).

Cd-SrtB_ΔN26 is structurally equivalent to Cd-SrtB_{ΔN32, C226A} (Chambers et al., 2015). However, visualizing the sulfhydryl group of the catalytic cysteine residue in Cd-SrtB_ΔN26 is essential for facilitating our understanding of the substrate-specific catalysis of Cd-SrtB.

To gain structural insights into how Cd-SrtB_ΔN26 recognizes PPKTG, we performed computational modeling based on the crystal structure of the Sa-SrtB–NPQT^* complex (PDB 4LFD) (Jacobitz et al., 2014) for predicting the Cd-SrtB_ΔN26–PPKTG structure. In the Sa-SrtB–NPQT^* structure, the substrate-binding pocket is delineated by a groove near the active site residues within the strands β4 and β7 and within loops β2/β3, β6/β7, and β7/β8. The NPQT^* peptide was bound to Sa-SrtB in an “L-shaped” structure via hydrophobic interactions with Leu96, Tyr128, Tyr181, and Ile182 and via hydrogen bonds with Asn92, Thr177, Glu224, and Arg233. Moreover, the almost superimposable hydrophobic residues from S. aureus with the corresponding residues from C. difficile underlies the importance of their function (Supplementary Figure 4) and imply that the Cd-SrtB substrate may be positioned in a similar pattern within the hydrophobic groove. Therefore, we superimposed the structure of Sa-SrtB–NPQT^* onto Cd-SrtB_ΔN26, mutated NPQT^* to PPKT and added a glycine in the C-terminus in silico using the software VMD 1.9.2 (Humphrey et al., 1996) as an initial model of the Cd-SrtB_ΔN26–PPKTG complex. In addition, the missing residues and loops in the Cd-SrtB_ΔN26 structure, including the N- and C-terminal residues (27, 28, and 225), and residues located on the β6/β7 (162–167) and β7/β8 (210–216) loops were modeled using UCSF CHIMERA (Yang et al., 2012). To refine the docking pose of the PPKTG in the catalytic pocket of Cd-SrtB_ΔN26, energy minimization and MD simulations were conducted by gradually releasing the restraints on PPKTG, first on the side chains and then on the entire peptide, whereas the Cd-SrtB_ΔN26 residues and crystal waters were restrained to their atomic positions in the Cd-SrtB_ΔN26 structure at all times during the simulations.

The results from the computational modeling and unrestrained MD simulations suggest that PPKTG stays in the active site, forming a L-shape with a bend toward the N-terminus, resembling the structure of Sa-SrtB–NPQT^* complex (Jacobitz et al., 2014) (Figure 3A). The sulfhydryl group of Cys209 is 5.0 Å from the carbonyl carbon of the threonine residue at the P1 position (Schechter and Berger, 1967), which is slightly further apart as compared with that of NPQT^* in Sa-SrtB (Figure 3B). The side chain of Arg217 is hydrogen bonded to the hydroxyl oxygen of P1 Thr. The P2 Lys forms a hydrogen bond with Ser163 and salt-bridge interactions with Asp164. Moreover, the prolyl ring of P4 Pro noncovalently interacts with the aromatic ring of Tyr167. To assess whether Ser163 and Tyr167 are involved in substrate interactions as predicted in the structural model, two mutants that replace Ser163 and Tyr167 with alanine were generated in Cd-SrtB_ΔN26 by site-directed mutagenesis. By performing the FRET-based assay, we observed that the cleavage activity of mutants Cd-SrtB_{ΔN26, S163A} and Cd-SrtB_{ΔN26, Y167A} was substantially reduced compared to wild-type Cd-SrtB_ΔN26 (Figure 3C). The results indicate that the alanine substitution of Ser163 and Tyr167 did affect the interactions between Cd-SrtB_ΔN26 and PPTKG, resulting in the reduced florescence signals.

To have a better understanding of specific recognition of the substrate peptide PPKTG by Cd-SrtB, we also constructed models of Cd-SrtB_ΔN26–SPKTG, Cd-SrtB_ΔN26–NPKTG, and Cd-SrtB_ΔN26–NVQTG complexes in the same way as constructing model of Cd-SrtB_ΔN26–PPKTG complex. Subsequently, we performed molecular dynamics (MD) simulations to analyze the residues in Cd-SrtB_ΔN26 and the contact frequency of those residues with different substrate peptides (Supplementary Figure 5). It is observed that the P4 residues in PPKTG and SPKTG are stabilized by residues of Cd-SrtB_ΔN26 located in binding pocket (Tyr101–Arg102, Ser163–Tyr167 for PPKTG, and Asp164–167, Phe213–Asp214 for SPKTG) (Figure 4). Our results show that the P4 residue of PPKTG in Cd-SrtB_ΔN26–PPKTG complex interacts with Asp166 for about 80% of the simulations time (10 ns) and forms hydrogen bonds with Asp166; while the P4 residues in NPKTG and NVQTG do not specifically interact with any residue in Cd-SrtB_ΔN26 (Figure 4 and Supplementary Figure 5). Moreover, we assumed that a peptide would be subjected to a conformation that allows Cd-SrtB to achieve a better catalytic efficiency if the distance (DIS_Cys-Thr) between the sulfhydryl group of cysteine residue in Cd-SrtB and the carboxyl carbon of threonine residue in peptide is relatively short (Donahue et al., 2014; Chambers et al., 2015). We therefore examined distance distributions for four peptides of interest throughout 10 ns simulations. The distances are found to be 5.91 ± 0.53 Å, 5.81 ± 0.75 Å, 6.94 ± 0.74 Å, and 7.11 ± 0.77 Å for Cd-SrtB_ΔN26–PPKTG, Cd-SrtB_ΔN26–SPKTG, Cd-SrtB_ΔN26–NPKTG, and Cd-SrtB_ΔN26–NVQTG, respectively (Figure 5). To further explore the role of the P4 residues of peptides in substrate specificity, we calculated their root-mean square fluctuations (RMSFs) of the peptides in Cd-SrtB_ΔN26–PPKTG, Cd-SrtB_ΔN26–SPKTG, Cd-SrtB_ΔN26–NPKTG, and Cd-SrtB_ΔN26–NVQTG. The RMSFs are 0.19, 0.38, 0.48, and 0.58 Å for PPKTG SPKTG, NPKTG and NVQTG, respectively. The higher stability of P4 residues in (P/S)PKTG seen in our dynamic simulations correlates to the shorter DIS_Cys-Thr and previously shown higher reaction activity (Donahue et al., 2014; Chambers et al., 2015).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.