Refinement of the Pa-PepP structure resulted in a final model with a free R-factor (Rfree) of 0.2206 at 1.8 Å. Sufficient electron density allowed us to model all residues from 1 to 444. The crystallographic statistics are summarized in Table 1. The monomer structure displays a two-domain organization where the N-domain (1–175) is composed of a mainly parallel beta-sheet core (B1–B6) flanked by seven alpha helices (A–G). In contrast, the catalytic C- domain (176–444) adopts a conserved “pita-bread” fold with six beta sheets (B7-B12) in antiparallel configurations (Figure 1A). This pita-bread fold is commonly found in N-terminal amido-, imido-, and amidino-scissile bond-cleaving enzymes, and serves as a structural basis for the metal-dependent catalysis (Lowther and Matthews, 2000; Besio et al., 2010).

Although the recombinant Pa-PepP purified from E. coli strain BL21 (DE3) exists as a monomer in solution (Figure 1B), tetrameric oligomerization was observed in crystal packing (Figure 1C). The monomers were arranged as a dimer with an extended loop contributing to the active site of the adjacent subunit and an average interface area of approximately 2050.9 A² per subunit. The dimer-of-dimers was generated by crystallographic symmetry operation (x–1/2, –y–1/2, –z), and the major interactions are contributed by the C-domain of each monomer, resulting in an ∼815.6 A² buried area per subunit. For most metalloaminopeptidases, the oligomeric state is essential for their biological functions. The loop from the adjacent monomer extends the substrate-binding site and thus enables the enzyme to cleave larger substrates. Thus, the antiparallel dimerization would be the basic active polymerization forms to its biological activity, and the different quaternary arrangement may reflect a possible regulatory mechanism for its activity.

Metalloaminopeptidases are well known for their metal-dependent catalysis mechanisms. Proline-specific aminopeptidases such as APPro and prolidase prefer dinuclear Mn(II) cluster as cofactor (Wilce et al., 1998; Graham et al., 2005). The residues coordinating the two Mn atoms (Mn_A and Mn_B) are highly conserved among APPro family members (Figure 2), and the non-equivalent roles of the two metal ions in catalysis have been extensively studied (Graham et al., 2005; Hu et al., 2007). To test the role of divalent metal cluster on Pa-PepP, we measured the relative activities of Pa-PepP in the presence of 1 mM Mn²⁺,Ca²⁺,Mg²⁺, Ni²⁺, Zn²⁺, and EDTA (Figure 3A). The addition of Mn²⁺ significantly restored the Pa-PepP activity, while the limited enhancement of activity was observed upon addition of Ca²⁺ or Mg²⁺. The exhibited basal activity may be a result of the partially pre-bound metal ions from the cell, and the limited effects of Ca²⁺ and Mg²⁺ may be due to their weak binding affinity to the active site. Previous studies on metal selection in Ec-PepP revealed that the Zn²⁺ ion has high affinity for APPro and inhibits the hydrolysis reaction by occupying a third metal binding site (Graham et al., 2005; Hu et al., 2007). Although we could not measure the effect of Zn²⁺ ion on Pa-PepP because of the formation of zinc hydroxide precipitates upon adding ZnCl₂ to the alkaline enzyme buffer, we observed the complete inhibition of enzyme activity in the presence of both the Ni²⁺ ion and the chelating agent EDTA. These results suggested that the Ni²⁺ may also possess high affinity to APPro and tends to bind to those positions equivalent to Zn²⁺ ions in Ec-PepP.

To identify the binding state of Mn²⁺ ions in active site, 10 mM MnCl₂ was added during the crystallization experiment, along with 10 mM proline. In the final structure, there are three Mn atoms in each monomer (Figure 4). Mn_A is liganded by the side chains of His354, Asp271, Glu384, Glu408, and Mn_B interacts with Asp260, Asp271, Glu408. These interactions are highly conserved among those reported APPro structures that resulted in a distorted trigonal-pyramidal coordination network. The smaller anomalous signals and 0.5 occupancies of the Mn_B site confirmed one of the structural features in metal-dependent aminopeptidase: the metal binding site B has relatively weaker binding affinity comparing to the site A (Hu et al., 2007). Additionally, the third Mn atom (Mn_C) is located in a position previously identified as being occupied by water in Zn-load Ec-PepP structures (Graham et al., 2005). The strong anomalous peak at this position verified its identity as a tightly bound Mn²⁺ ion rather than a water molecule. It is interesting to note that this Mn_C position is different from the third metal binding site found in Ec-PepP, in which the Zn_C is coordinated by His243, His361, and the hydrolysis product Pro residue (Graham et al., 2005). In Pa-PepP, Mn_C interacts through one solvent molecular (W) with Mn_A and Mn_B, with a metal-separation distance of 3.1–3.7 Å. The Pro residue lies immediately above the solvent molecular W but does not participate in the interaction with Mn_C. Remarkably, Mn_C occupies the S1 subsite that ordinarily binds the substrate, providing an explanation for the inhibition effect of excessive metal binding. To verify the inhibitory effect of Mn²⁺ ion, we performed an activity assay in the presence of different amounts of Mn²⁺. The result is consistent with our structural analysis (Figure 3B): there is a Mn²⁺ ion concentration-dependent activity regulation pattern, in which 0.5–1 mM of Mn ion is the optimal concentration for 20 nM (1 μg/ml) Pa-PepP, and inhibitory effects are observed when Mn²⁺ ion exceeded 5 mM. Inhibition was not observed at ion concentrations up to 20 mM in the assays with Ca²⁺ and Mg²⁺ (Supplementary Figure S1).

Several APPro structures have been reported to date, including APPro from Escherichia coli (Ec-PepP, 1M35) (Wilce et al., 1998), Streptococcus thermophilus (3IL0), Bacillus anthracis (3IG4), Yersinia pestis (4PV4), Thermotoga maritima (2ZSG), Streptococcus pyogenes (3OVK), human (hAPP1, 3CTZ) (Li et al., 2008), Caenorhabditis elegans (CeAPP1, 4S2R) (Iyer et al., 2015) and Plasmodium falciparum (PfAPP, 5JQK) (Drinkwater et al., 2016). Ec-PepP shares the highest sequence identity with Pa-PepP and both of them are prokaryotic APPro (Gonzales and Robert-Baudouy, 1996). All of these APPros share a common catalytic domain structure that contains a metal center flanked by the S1’-S1 pockets (Figures 2, 5A). Previous structural studies on Ec-PepP in complex with the substrate or inhibitor have revealed those pockets (Graham et al., 2004; Graham and Guss, 2008). We modeled the Val-Pro-Leu bound Pa-PepP complex by superposing Pa-PepP structure with the substrate-bound Ec-PepP structure (PDB code: 2BN7) (Figure 5A). The shallow hydrophobic S1’ pocket could be defined by the pre-bound Pro residue, Arg406, Tyr388, and His350. S1’ sites are conserved in APPro and prolidase but not in other metalloaminopeptidases, which implies that they play critical roles in regulating the activity of proline-specific peptidases. Additionally, in prokaryotic APPros and prolidases, a tryptophan (Trp88’ in Pa-PepP) from an adjacent subunit extends the boundary of S1’ site and contributes to the P1’ Pro binding (Yoshimoto et al., 1994; Jao et al., 2006). At the S1’-S1 junction, His243 and His361 interact with the main-chain carbonyl of P1’ Pro and P1 Val, respectively, indicating their specific roles in recognizing and stabilizing the stereospecific scissile X-Pro bond. In the S1 pocket, the side chain of P1 Val faces Tyr229, Ile232, Arg245, while its main chain replaces the Mn_C ion and is involved in the Mn_A and Mn_B ion mediated interactions. In metallomainopeptidases, the dinuclear Mn center is thought to be essential for nucleophilic attack (Graham et al., 2005). To verify the impacts of the critical motifs revealed by structural analysis, we constructed Ala substitutions on those motifs (Trp88 for S1’ pocket, His243 for scissile bond stabilizing, Glu384 for metal binding) and verified that the expression level and stability are comparable between the wild type and all the mutants (Supplementary Figure S2). The Pa-PepP activity assay shows that all the mutations would eliminate enzymatic activity, which further confirmed that the evolutionary conservation in those regions is strongly associated with the essential functional roles of APPro family (Figure 5B).

In contrast to the prolidases that cleave dipeptides only, APPro is capable of modifying protein substrates. The relatively larger substrate binding ability of APPro is partly due to the extension of the substrate binding site beyond the S1’–S1 region (Drinkwater et al., 2016). The modeled tripeptide-bound Pa-PepP structure indicated that Arg153 and Arg351 accommodate the carboxylate group of the P2’ Leu, and guide the C-terminus of the substrate toward the tetramer surface (Figure 5A). Furthermore, the R351G mutation results in an increase in enzyme activity. The similar Arg153/Arg370 basic patch is also found in Ec-PepP, in which the non-basic residue substitution on either of them leads to an obvious decrease in K_m and an increase in k_cat/K_m (Jao et al., 2006). Thus, this basic patch, which is located at the entrance of catalytic cave, can play an important role in specificities of protein substrate recognition and orientation.

The conserved structural properties in active site suggest a common catalytic mechanism for APPro family members, while the differences in the substrate-binding regions indicate the specific functions for each APPro member. Next to the P2’ binding site, a surface loop (residue 365–377 in Pa-PepP) between helix α₁₅ and sheet β₁₀ was identified as a candidate for specific functions due to the diversity in its sequences between APPro members (Figure 2). To investigate its impact on enzyme catalytic activity, we mutated residues RVGGEW (Pa-PepP 367-372) to GQDRS (Ec-PepP 368-372). Equivalent activities were observed between the corresponding mutant and the wild-type (Figure 5B). Since, in the activity assay, we used a non-specific tripeptide analog instead of the native substrates, these results may not accurately reflect the actual substrate-binding situation.

To further elucidate the impacts of Pa-PepP mutations on the bacterium virulence, three mutants with different effects on enzymatic activity were made in P. aeruginos PA14 for use in a Hela-cell invasion assay (Figure 5C). The PA14 bacterial growth in vitro was not affected by pepP deletion or any mutations in pepP (data not shown). The results showed that Pa-PepP knockout PA14 (PA14-ΔpepP) strain lost half of its ability to enter mammalian cells compared to wild-type PA14, and the cell invasion ability for PA14-ΔpepP strain was restored when introducing a complementary plasmid pRK415 for Pa-PepP expression, which confirmed the contribution of Pa-PepP to P. aeruginos virulence (Feinbaum et al., 2012). The enzymatic inactive mutant (H243A) exhibited an attenuated invasion ability, suggesting a role for the cleavage activity of Pa-PepP in its biological function. Strikingly, the PA14-R351G mutant also displayed weak cell internalization efficiency comparable to PA14-ΔpepP, revealing that the specificity of protein-substrate recognition is indispensable to the virulence-related function. The mutant PA14-Pa-PepP-367-372 also presented an attenuated invasion ability, verifying the critical role of 367-372 loop in the biological function of Pa-PepP, even though it does not have an obvious impact on enzymatic activity.

Full-length pepP was amplified by PCR using gene-specific primers (Supplementary Table S1) from the P. aeruginosa PA14 genome DNA on Gene amplification Machine (Gene Touch, BIOER, HangZhou, China). Full-length pepP containing six C-terminal histidine residues (LEHHHHHH) was homologous recombined with the linearized pET-22b (+) using a ClonExpressTM II One Step Cloning Kit (Vazyme). All point mutants were generated using the QuickChange (I-5TM2^∗High Fidelity Master Mix, MCLAB) PCR-based method, on the pET-22b (+) construct. The recombinant plasmid was transformed into E. coli strain BL21 (DE3) for protein expression. The bacterial culture was grown in LB medium in the presence of 100 μg mL^-1 ampicillin and incubated with shaking at 310 K until the OD600 reached 0.9 (ZhiChu, ShangHai). The culture was cooled to 289 K before protein expression was induced with 0.2 mM IPTG for 20 h (Bao et al., 2013). Following induction, the bacteria were collected and resuspended in a lysis buffer consisting of 25 mM Tris–HCl pH 8.5, 10 mM NaCl, 5 % glycerol and 1 mM phenylmethanesulfonyl fluoride (20 g of cells/100 mL buffer, Sigma–Aldrich) and lysed by sonication. The lysate was cleared by centrifugation at 11000 g for 45 min and then the supernatant was loaded onto a 2 mL Ni–NTA affinity resin (Qiagen) for 2 L culture. The Ni–NTA column was washed with ten column volumes of the lysis buffer supplemented with 20 mM imidazole. The target protein was eluted with the same buffer in the presence of 200 mM imidazole. Fractions were pooled and determined by SDS–PAGE, followed by further purification on size-exclusion chromatography Superdex 200 column (GE Healthcare), which was pre-equilibrated with the buffer consisting of 25 mM Tris–HCl pH 8.5, 10 mM NaCl. Fractions containing Pa-PepP were pooled and concentrated to a concentration of approximately 16 mg mL^-1 using a Centricon filter (10 kDa cutoff; Millipore, Billerica).

Initial crystallization experiments were carried out using four commercial crystallization screens from Hampton Research and Rigaku (Index HT, Crystal Screen HT, WIZARD HT, XTAL QUEST HT). Crystallization screens were conducted as previously described with some modifications (Bao et al., 2009). Briefly, crystallization initially screens were carried out using a Mosquito liquid dispenser by hanging-drop vapor-diffusion method at 291 K in 96-well plates. The 200 nL mixing drop containing protein solution and reservoir buffer by 1:1, with a final protein concentration of 16 mg ml^-1 in 25 mM Tris–HCl pH 8.5, 10 mM NaCl. The final optimized crystals for the recombinant protein were obtained by mixing 2 μl of the protein sample containing Mn²⁺ and proline with an equal volume of the reservoir solution containing 30% PEG400, 100 mM sodium cacodylate pH 6.5, 200 mM lithium sulfate. Crystals grew in approximately 2–3 days and were transferred to a cryo-protectant solution (reservoir solution with 6% PEG400) prior to flash-cooling in liquid nitrogen. X-ray data were collected with a CCD camera on BL-17U stations of the SSRF, China. The diffraction data were indexed, integrated, and scaled using the HKL2000 program suite (Otwinowski and Minor, 1997).

Data processing and scaling were carried out using the HKL2000 software package. The data were processed to a resolution limit of 1.847–1.783 Å (Rmerge = 0.099) in space group P2, with unit-cell parameters a = 111.197, b = 123.432, c = 149.485 Å. The phase problem was solved by molecular replacement using PHENIX with aminopeptidase P from E. coli (PDB entry 1az9; space group P 6422; resolution 1.9 Å) as a template. The process of structure building and refinement was monitored using the COOT (Emsley et al., 2010). Water molecules were automatically added by PHENIX (Adams et al., 2010).

Differential scanning calorimetry (DSC) is a promising thermoanalytical technique used for evaluating the stability of proteins as well as other biomolecules. The enthalpy value determined by DSC can provide direct information about the energetics of thermally induced processes, and the measured melting temperature(Tm) reflects thermal stability of proteins. The assays were conducted on the MicroCal VP-Capillary DSC System (Malvern). Protein samples were prepared with a final concentration of 1 mg/ml in 400 μl buffer and each sample has another 400 μl blank buffer as reference which were loaded in pairs. Operating parameters (such as pre-scan equilibration time, scan rate, temperature programming) were set and at least three buffer-buffer scans were performed before the sample was scanned. And DSC automated data analysis was conducted after experiments ran out. The DSC curves can read by Origin (Origin Pro 7.5).

To construct pepP mutants of P. aeruginosa, a two-step allelic exchange bacterial genome engineering strategy was employed (Hmelo et al., 2015; Sana et al., 2015). Briefly, in the first step of allelic exchange, the suicide vector pEX18Gm was integrated site-specifically into the chromosome of P. aeruginosa by homologous recombination, resulting in antibiotic-resistant single-crossover mutants. Then, in the second step of allelic exchange, a double crossover event occurred through a second homologous recombination, and the mutant was isolated using sucrose-mediated counter-selection. The corresponding mutants were finally identified by PCR and DNA sequencing. Specifically, PCRs were performed to amplify the target fragment sequences with upstream (800 bp) and downstream (800 bp) from P. aeruginosa chromosomal DNA, while the suicide plasmid pEX18Gm was linearized with gene-specific primers. The two PCR products were recombined with ClonExpress^® II One Step Cloning Kit (Vazyme), The resulting plasmid, pEX18-Gm-pepP, was then performed Site-directed mutagenesis or deletion. All these primers are listed in Supplementary Table S1. These vectors were then transformed into E. coli S17-1 and then mobilized into P. aeruginosa strains PA14 by conjugation, in order to transfer suicide plasmids from the E. coli donor S17-1 to the P. aeruginosa recipient PA14. Colonies were first screened using gentamicin resistance plates to get single-crossover mutants. And then the double-crossover mutants were screened by No-salt LB (NSLB) agar with 15% (wt/vol) sucrose. The pepP gene replacement mutant strain was further confirmed by PCR and DNA sequencing.

To construct the complementation plasmid pRK415-pepP, PCR-amplified pepP was cloned into the EcoRI and HindIII sites of plasmid pRK415, giving rise to the plasmid pRK415- pepP (Lin et al., 2017). The recombinant plasmid and plasmid pRK415 were transformed into E. coli S17-1, respectively, and then mobilized into P. aeruginosa PA14-ΔpepP by conjugation to transfer pRK415-pepP from the S17-1 to PA14-ΔpepP. The PA14-ΔpepP strain carried pRK415-pepP or pRK415 plasmid were screened by Pseudomonas isolation agar (PIA) with 150 μg/ml tetracycline. For expression of PepP in the P. aeruginosa strains, the PA14-ΔpepP strain carried pRK415-pepP were induced by addition of 1 mM IPTG and then conducted cell invasion assays.

To enumerate bacteria internalized by HeLa cells to verify bacterium virulence, gentamicin survival assays were conducted with slight modifications (Chi et al., 1991). Briefly, mammalian HeLa cells (obtained from ATCC) were grown in DMEM medium, containing 10% (v/v) FBS (Gibco, Auckland, New Zealand) and 1 % antibiotics (penicillin and streptomycin) in 5% CO₂ at 37°C. Suspension cultures of HeLa cells were seeded at 2–5 × 10⁵ cells per well in 12-well tissue culture plates for the overnight at 37°C. Cells were washed three times with PBS (pH 7.2) and changed to antibiotic-free medium immediately before infection. Cells were infected with exponential phase P. aeruginosa strains PA14 and mutant strains at a MOI of 10 for 1 h in 5% CO₂ at 37°C. The cells were washed twice with PBS and incubated for an additional 1 h in DMEM medium containing 150 μg/ml of gentamicin in order to kill extracellular bacteria. The monolayers cells were then washed three times with PBS and lysed with 0.5% Triton X-100 for 10–20 min, and appropriate dilutions were plated on LB plates to determine the number of viable intracellular bacteria.

The enzyme activity of wild type Pa-PepP and the mutants were spectrophotometrically determined using the quenched fluorescent substrate Lys (Abz) – Pro–Pro–pNA (synthesized by GL Biochem Shanghai Ltd., China) as the substrate. Prior to the assay, wild type Pa-PepP and the mutant enzyme are dialyzed with EDTA overnight to fully remove the pre-bound metal ions. And all enzyme preparations were freshly diluted with ice-cold 50 mM Tris (pH 8.5) and 100 mM NaCl and treated with MnCl₂ for 10 min at 37°C. After addition of substrate, the final reaction conditions were 50 mM Tris (pH 8.5), 100 mM NaCl, 1 mM MnCl₂, 0–250 μM substrate, and 20 nM Pa-PepP (final assay volume 100 μL). The assay was allowed to proceed for 5 min at 37°C in the Thermo Scientific Varioskan Flash plate reader. The appearance of fluorescent product (λex = 301 nm, λem = 410 nm) was monitored at 10 s intervals. Since the pure fluorescent product was not available to quantitate the changes in fluorescence, the kinetic results are presented as activities in relative fluorescence units. The kinetic parameters Km (Michaelis constant) for all the assays were obtained by fitting experimental data to the Michaelis-Menten equation by non-linear regression using the program OriginPro 7.5 (OriginLab Software). The data were showed in Supplementary Tables S2 and S3.

Statistical analysis was analyzed by 2-tailed Student’s t-test. In all statistical analysis, P-values < 0.05 were considered to be statistically significant.