Remodeling of Conformational Dynamics Enhances Catalytic Activities of M1 Zinc-metallopeptidases from Lanthipeptide Biosynthesis

Lanthipeptides are an important group of natural products with diverse biological functions, and their biosynthesis requires 13 the removal of N-terminal leader peptides (LPs) by designated proteases. LanP M1 enzymes, a subgroup of M1 zinc- 14 metallopeptidases, are recently identified as bifunctional proteases with both endo- and aminopeptidase activities to 15 remove LPs of class III and class IV lanthipeptides. Herein, we report the biochemical and structural characterization of 16 EryP as the LanP M1 enzyme from the biosynthesis of class III lanthipeptide erythreapeptin. We determined X-ray crystal 17 structures of EryP in three conformational states, the open , intermediate and closed states and identified a unique inter- 18 domain Ca binding site as a regulatory element to modulate its domain dynamics and proteolytic activity. Inspired by the 19 regulatory Ca binding, we develop a strategy to engineer LanP M1 enzymes for enhanced catalytic activities by strengthening 20 inter-domain associations and driving the conformational equilibrium toward their closed forms. for Several and to inoculate separate 5 mL cultures of LB-kanamycin were grown at 37 °C for 12 h, and plasmids isolated using an Omega Biotech Plasmid Mini products confirmed by sequencing. Mutagenesis AplP EryP. Mutagenesis of AplP and EryP carried out by a two-stage protocol, which starts with the generation of primers using two parallel asymmetric PCRs, followed by an exponential whole-plasmid amplification. amplification by 30 of denaturing (95 °C for 30 s), annealing (65 °C 30 s), and extending (72 °C, 1 min/kb) using high fidelity Phanta ® DNA Polymerase and confirmed by 1% agarose gel electrophoresis. PCR purified using an Omega Biotech Cycle Pure Kit. DNA fragment digested by DpnI in NEB 482 buffer (New England Biolabs) for 3 h at 37 °C. DH5α cells transformed with 2.5 μL of the digested product by heat shock, on LB-agar plates and grown for 15 h at 37 °C. Single colonies picked to inoculate separate 5 mL of LB-chloramphenicol which were grown at 37 °C for 12 Plasmids isolated an Omega Biotech Plasmid Kit. of plasmid products were confirmed by sequencing.


INTRODUCTION 22
M1 family zinc-metallopeptidases are widely distributed in many species and regulate a diverse range of biological 23 processes. 1 For instance, Escherichia coli (E. coli) aminopeptidase N (ePepN) is involved in ATP-dependent downstream 24 processing during cytosolic protein degradation. 2, 3 Thermoplasma acidophilum tricorn interacting factor F3 (F3) acts 25 downstream of the proteasome to produce di-and tripeptides. 4 Human endosomal insulin-regulated aminopeptidase (IRAP) 26 is a key regulator in endocytic trafficking and signaling, 5 while the endoplasmic reticulum aminopeptidases (ERAP1 and 27 ERAP2) are responsible for antigen-trimming. 6,7 Recently, a new biological function of M1 metallopeptidases is revealed 28 in the biosynthesis of bacteria-derived natural product lanthipeptides, a subfamily of ribosomally synthesized and 29 posttranslationally modified peptides (RiPPs). 8 As the most studied and rapidly growing family of RiPPs, lanthipeptides 30 are featured with diverse structures and biological activities, including antimicrobial, anti-fungal, anti-virus and 31 antiallodynic properties. 8 The precursor peptide of a lanthipeptide is usually composed of an N-terminal leader peptide and 32 a C-terminal core peptide, which undergoes enzymatic cyclization during biosynthesis. The N-terminal leader peptide 33 plays important roles during the biogenesis of lanthipeptides as recognition elements for modification enzymes, as 34 secretion signals during the exportation or for self-immunity. 9 The removal of leader peptides is often the last but essential 35 step to produce the mature lanthipeptides. If additional modification is required on the N-terminus of lanthipeptides, such 36 as in the biosynthesis of lipolanthines and goadvionins, 10, 11 precise and efficient removal of the N-terminal leader peptide 37 becomes even more essential.

38
A subgroup of M1 zinc-metallopeptidases, named LanPM1, function as key proteases to remove the N-terminal leader 39 peptides (LPs) of class III and class IV lanthipeptides during their maturation. 12 AplP from the biosynthesis of class III 40 lanthipeptide NAI-112 represents the first example of LanPM1 proteases and removes the leader peptide of precursor 41 peptide AplA after cyclization and glycosylation at its core peptide. 13 As M1 metallopeptidases usually prefer linear

55
LanPM1 enzymes are usually highly active as endopeptidases, but significantly less efficient as aminopeptidases during 56 the processing of LP overhang residues. 13,16 The inefficient aminopeptidase activities of LanPM1 enzymes often result in 57 incomplete LP removal both in vitro and in vivo (Fig. 1a), which impedes the heterologous production and bioengineering 58 of corresponding lanthipeptides. 9, 12 Thus, to understand the enzymology underlining the dual functionality of LanPM1 59 proteases and to engineer LanPM1 enzymes with improved aminopeptidase activities for lanthipeptide production are highly 60 appealing. Herein, we report the biochemical and structural characterization of EryP, which is the LanPM1 enzyme from

87
Next, we examined the proteolytic activity of EryP toward cyclized lanthipeptide substrates. Since cyclized EryA 88 peptide (EryAcyc, Fig. 1b), the native substrate of EryP, was not accessible, we employed cyclized AplA, the precursor 89 peptide of NAI-112, as an alternative substrate. Cyclized AplA (AplAcyc) shares structural similarity with EryAcyc in the 90 leader peptide and the core peptide, therefore serving as a good substrate mimic (Fig. 2c, Supplementary Fig. 4). As 91 expected, EryP cleaved AplAcyc at multiple sites in the segment of AplALP peptide after 90 min incubation by generating 92 AplAcyc-CP(-6) as the detectable core peptide product ( Supplementary Fig. 5). Prolonged incubation led to stepwise removal 93 of overhang residues in AplAcyc-CP(-6) and eventually yielded AplPcyc-CP as the final product (Fig. 2d) To understand the substrate recognition mechanism of EryP, we performed microscale thermophoresis (MST) to 102 determine the binding affinity between EryP and peptide substrates. Results showed that EryP binds to EryALP with a KD 103 of (40.8±5.8) μM, whereas its binding with EryALP(-15) is too weak to measure (Table 1). EryP binds to full length AplAcyc 104 and AplALP with KD values of (0.503±0.109) μM and (7.98±2.72) μM, respectively, indicating that the cyclized AplA core to tricorn-interacting factor F3 (TIFF3) and ePepN of the M1 metalloprotease family, EryP folds into four distinct domains: 114 domain-I (residues 1-201), domain-II (residues 202-456), domain-III (residues 457-552) and domain-IV (residues 553-115 860) (Fig. 3a). The barrel-like N-terminal domain-I is packed by a central 12-stranded β-sheet and two smaller antiparallel 116 β-sheets. The following catalytic domain-II adopts a thermolysin-like fold characterized by an α/β lobe on the top of an α-117 helices lobe. A Zn ion binding motif, HEXXH(X)18E, which is highly conserved among M1 metallopeptidases, is located 118 at the deep groove between the two lobes ( Supplementary Fig. 6). Domain-III presents a barrel-like structure formed by 119 eight β-strands. The C-terminal domain-IV consists of 16 α-helices that are organized in an antiparallel manner by forming 120 a bowl-shaped structure, which might regulate the accessibility of the catalytic pocket in domain-II. 6

121
In the closed state of EryP, a zinc atom resides at the bottom of active-site cleft in catalytic domain and is coordinated 122 by NE2 atoms of two histidine residues (H306 and H310) and OD atoms of E329 (Fig. 3b, Supplementary Fig. 6). 20 A 123 catalytic water molecule lies immediately (2.0 Å) next to the Zn ion, resulting in a distorted trigonal-pyramidal 124 coordination network (Fig. 3b). The presence of a catalytic water molecule in the active site prior to peptide substrate 125 binding is consistent with previous studies of M1 metallopeptidases, such as thermolysin, and is considered to be essential

131
The interactions between short peptide substrates and the reaction center were visualized by docking a dipeptide Leu-

132
Glu into the active site of EryP in the closed state ( Supplementary Fig. 8). The highly conserved G272AME275N motif in 133 EryP binds to the dipeptide substrate through multiple hydrogen bonds in the docking model. In particular, residue E275 134 directly binds to the amino group at the N-terminus of the dipeptide with additional binding from residues E132 and E329 135 ( Supplementary Fig. 8). Furthermore, this docking model suggests that both S1 and S1' pockets are spacious to 136 accommodate amino acid side chains of various sizes in peptide substrates, allowing the enzyme to sequentially cleave 137 various residues from the leader of EryA ( Supplementary Fig. 8).    conformations with domain-IV in close distance with domain-I and II (Fig. 4a). When the Ca 2+ ion was removed from 183 EryP, a significantly wider distribution of conformations was observed, a portion of which deviated from the closed 184 conformation (Fig. 4a). When Leu-pNA, the favorite aminopeptidase substrate of EryP, was docked into the active site to 185 mimic the substrate-bound EryP, the binding of Ca ion showed an even stronger effect on locking EryP in the closed 186 conformations (Fig. 4b). Collectively, these data suggest that by bridging residues from domain-I, II and IV, the Ca ion 187 shifts the conformational equilibrium of EryP toward the closed conformation.     overhang residues is the rate-limiting and often incomplete step during the biosynthesis of class III lanthipeptides, and 212 therefore LanPM1 proteases with enhanced aminopeptidase activities are highly desired. 12,15,16 The correlation between 213 structural dynamics and enzymatic activities of EryP provides an opportunity to create an engineered EryP mutant with 214 improved aminopeptidase activity by stabilizing the tri-domain association. The Ca binding residues in EryP provide a 215 promising engineering site to incorporate mutations for enhanced inter-domain network. Based on the closed structure of 216 EryP, we replaced the Ca 2+ coordinating residue E802 from domain-IV with an arginine (R) residue, which potentially 217 forms charge-charge interactions with residues E112 and E384 to strengthen the association between domain-I, II and IV.

218
To confirm the engineering effect, we determined the crystal structure of EryPE802R in 1.77 Å resolution (Supplementary 219 Fig. 13). In line with our expectation, the overall structure of EryPE802R displays a more compact conformation than EryP 220 (inter-domain angle θ=48.18, and the torsion angle φ=27.77) mainly due to the newly formed interactions introduced by 221 the E802R mutation (Fig. 5a). Specifically, residue R802 inserts into the cleft of domain II and forms a salt-bridge with 222 residue E384, as well as cation-π interaction with residue F387 from domain II ( Fig. 5b; Supplementary Fig. 13). Additional 223 interaction between R802 and T133 from domain I was observed, which pulls domain I closer to domain IV and increases 224 their domain interface from 58.2 to 352.3 Å 2 (Fig. 5a). MD simulation based on the structure of EryPE802R showed that 225 domain-IV of EryPE802R is in significantly closer contact with domain-I and II than that of EryP, and conformations of 226 EryPE802R are largely maintained in the closed state (Fig. 4a, b). Together, these data support that the E802R mutation  in the reaction buffer, supporting the notion that the E802R mutation is a reasonable design which could replace the 239 enhancement effect induced by Ca binding (Supplementary Fig. 15). Importantly, EryPE802R showed significantly higher 240 aminopeptidase activity towards peptide AplAcyc-CP(-4) than EryP during the removal of the N-terminal Asp residue (Fig.   241 5e). Together, our data showed that the E802R mutation stabilizes the overall structure of EryPE802R in a more compact 242 closed conformation than EryP, which leads to improved aminopeptidase activity.

243
To further validate the origin of the enhanced aminopeptidase activity of EryPE802R, we performed MD analysis of the 244 substrate binding of EryP and EryPE802R to Leu-pNA ( Supplementary Fig. 16). Typically, three major factors govern the 245 course of amide bond hydrolysis: the activation of an amide bond by Zn ion, the nucleophilic attack by the hydrolytic 246 water and the stabilization of the resulting tetrahedron intermediate. 29 MD analysis indicated that the average distance 247 between the Zn ion and the amide oxygen of Leu-pNA bound in EryPE802R was ~1.5 Å shorter than that in EryP, suggesting 248 a stronger amide activation and stabilization of the tetrahedron intermediate ( Supplementary Fig. 16). In addition, the 249 hydrolytic water is ~0.5 Å closer to the amide α-carbon, which increases the likelihood of nucleophilic attack to occur.

250
Finally, the phenolic oxygen of Y392 displays an average distance of 2.0 Å from the amide oxygen, which is ~2.0 Å shorter 251 than that in EryP, leading to a significantly stronger stabilization effect of the tetrahedron intermediate. These results show 252 that the active site constellation of EryPE802R is more optimized for amide hydrolysis than that of EryP.  Fig. 17). Thus, we prepared AplPR98E-A368E-A779R and evaluated its 270 catalytic activities. Gratifyingly, the aminopeptidase activity of AplPR98E-A368E-A779R was enhanced by 2.6-fold compared 271 with AplP toward Ala-pNA ( Supplementary Fig. 18). Furthermore, AplPR98E-A368E-A779R displayed significantly higher 272 endopeptidase activity toward its native peptide substrate AplAcyc than wild-type AplP, but with altered preference to 273 cleavage sites (Fig. 6a). During prolonged incubation, AplP generated AplAcyc-CP(-4) as the major product but no further 274 trimming was detected, indicating that residue Glu(-4) is a challenging residue to remove (Fig. 6b). In contrast, AplPR98E-   Table   286 3&4.

DISCUSSION 288
A molecular ruler model has been proposed to explain the substrate preference toward peptides of certain length and 289 sequence for four-domain M1 metallopeptidases ERAP1/ ERAP2 and IRAP. 14, 24, 30 The C-termini of peptide substrates bind to a regulatory site in ERAP1 and induce allosteric activation of ERAP1 by shifting its conformation toward the 291 closed form. 14, 24 In contrast, the presence of peptide substrates has little impact on the hydrolytic efficiency of EryP toward 292 Ala-pNA under assay conditions (Supplementary Fig. 19). In addition, the inter-domain cavity in the closed form of EryP 293 is ~1530 Å 3 , which is not sufficient to fully accommodate the cyclized core peptide of AplA peptide (~2300 Å 3 ). These 294 data implicate that EryP might have distinct mechanisms of substrate recognition and enzyme activation. Although EryP 295 binds to full length EryALP and AplALP peptides (Table 1), it shows no measurable binding with endoproteolytic products 296 EryALP(-15)-(-1) and AplALP(-6)-(-1) and displays no aminopeptidase activity toward them even after long time incubation 297 ( Supplementary Fig. 20). However, EryP could recognize and trim the C-terminal products with cyclized core peptide 298 (AplAcyc-CP(-6)), suggesting that the core peptide plays a critical role to facilitate substrate recognition by EryP during the 299 second phase of leader removal. Incubation of EryP and AplALP(-6) in trans with AplAcyc-CP peptide did not lead to improved 300 aminoproteolytic removal of N-terminal residues from AplALP(-6). Together, these data indicate that the core peptide 301 AplAcyc-CP functions as a substrate recognition unit to deliver the AplALP(-6) segment to EryP, instead of as an allosteric 302 activator. Therefore, we propose that the apparent low aminopeptidase activity of EryP during the second phase of leader 303 removal is primarily due to the low binding affinity to peptide substrates.

Competing interests 433
The authors declare no competing interests.

Additional information 435
Supplementary information is available for this paper online.

436
Correspondence and requests for materials should be addressed to H.W. or R.B.

477
Mutagenesis of AplP and EryP. Mutagenesis of AplP and EryP was carried out by a two-stage protocol, which starts with 478 the generation of primers using two parallel asymmetric PCRs, followed by an exponential whole-plasmid amplification.

537
FITC-labeled EryALP sample (10 μL) was first incubated with EryP (10 μL) of 16 different serial dilutions in the assay 538 buffer for 5 min to allow binding. Samples were then loaded into Monolith NT.115 capillary (Nanotemper Technologies) 539 and measured using 20% (Auto-detect) Nano -BLUE as excitation power and medium MST power. The measurement 540 was repeated for three times. Data analysis was performed using Nanotemper affinity analysis software.

541
Crystallization, data collection of EryP. The purified protein was adjusted to about 9 mg/ml and subjected to screenings    Table S1. The structural figures were drawn with PyMOL (https://pymol.org/2/) and USCF Chimera X 38 .

561
Initial structural preparation for computational studies. The initial structure of EryP with Ca 2+ ion bound was based 562 on the closed structure of EryP (PDB ID: 7V9N). The Ca ion was manually removed from the EryPclosed structure to build 563 an EryP structure without Ca 2+ bound. The starting coordinate of the EryPE802R mutant were based on its crystal structure 564 (PDB ID: 7V9O). The protonation states of charged residues were determined at constant pH 7.5 based on pKa 565 calculations via the PROPKA program 39 and the consideration of the local hydrogen bonding network. In models of 566 EryP with/without Ca ion bound, residues His102, 182, 231, 249, 422, 489, 492, 512, 532, 659, 756, 774 and 797 567 were assigned as HIE; residues His306, 310 and 710 were set as HID; the rest His residues were HIP. In the model 568 of EryPE802R, residues His102, 182, 306, 310, 512, 756, 710, 756 and 797 were assigned as HIE; residues His115, 569 231, 249, 422, 489, 492, 532, 659 and 774 were set as HID; the rest His residues were HIP. In above three models, 570 all Asp and Glu residues were set as deprotonated status except for Asp315, while Lys and Arg resides were all set 571 as protonated status. The force field parameters for the Zn center active site were prepared using the Metal Center 572 Parameter Builder (MCPB.py) 40 as implemented in Amber 16 41 , in which residues H306, H310, E329 and a catalytic 573 water molecule were all treated in coordinate bonded way. Bond and angle force constants were derived using the 574 Seminario method 42 , and point charge parameters for the electrostatic potentials were obtained using the ChgModB 575 method. Each model was neutralized by adding Na + ions and solvated into a truncated octahedron TIP3P 43 water box 576 with a 10 Å buffer distance on each side. These three models consisted of 84549, 84547 and 88052 atoms for the 577 EryP bound with or without Ca 2+ and EryPE802R, respectively. 578 Molecular docking. The GLIDE program 44 was used to perform molecular docking of EryP with the LE peptide. The 579 crystal structure of EryP in the closed state was minimized by using the OPLS3 force field and the ligand was prepared

583
To dock Leu-pNA into the active sites of EryP enzymes, 50000 snapshots uniformly distributed at equal intervals 584 from the last 100 ns MD simulation (with time intervals of 2 ps) were picked up and divided into ten groups using 585 hierarchical agglomerative (bottom-up) approach 45 . Leu-pNA, the favorite aminopeptidase substrate of EryP, was 586 fully optimized at the B3LYP-D3/6-31+G(d) level of Gaussian 16 using the CPCM 46-48 model in water, and then 587 docked into the active site of one representative group snapshot to mimic the native substrate-EryP complex. 588 Molecular docking was performed using the Lamarckian genetic algorithm local search method in the AutoDock 4.2 589 and AutoDockTools-1.5.6 49 . The docking approach was employed on rigid-receptor conformation, while all the 590 rotatable torsional bonds of Leu-pNA were set free. A grid box was centered on the Zn atom and its size was set to 591 40 Å ×40 Å ×40 Å points with a 0.375 Å spacing. A total of 500 independent docking runs were undertaken with a 592 maximum energy evaluation of 2.5 × 10 7 . The obtained 500 docked conformations were clustered with 2.0 Å RMSD 593 and ranked depending on an energy-based scoring function. The possible catalytically active binding modes were 594