Structure-based Mechanistic Insights into Terminal Amide Synthase in Nosiheptide-Represented Thiopeptides Biosynthesis

Nosiheptide is a parent compound of thiopeptide family that exhibit potent activities against various bacterial pathogens. Its C-terminal amide formation is catalyzed by NosA, which is an unusual strategy for maturating certain thiopeptides by processing their precursor peptides featuring a serine extension. We here report the crystal structure of truncated NosA1-111 variant, revealing three key elements, including basic lysine 49 (K49), acidic glutamic acid 101 (E101) and flexible C-terminal loop NosA112-151, are crucial to the catalytic terminal amide formation in nosiheptide biosynthesis. The side-chain of residue K49 and the C-terminal loop fasten the substrate through hydrogen bonds and hydrophobic interactions. The side-chain of residue E101 enhances nucleophilic attack of H2O to the methyl imine intermediate, leading to Cα-N bond cleavage and nosiheptide maturation. The sequence alignment of NosA and its homologs NocA, PbtH, TpdK and BerI, and the enzymatic assay suggest that the mechanistic studies on NosA present an intriguing paradigm about how NosA family members function during thiopeptide biosynthesis.

. NosA and its homologs share a common mechanism on the post-modifications of thiopeptides. (A) The catalytic reaction for nosiheptide maturation by NosA. The -OH groups involved in the interactions with NosA are highlighted in red and blue, respectively. The C α -N bond cleavage site is marked in pink and a wavy line. The distances between the oxygen atoms of the -OH groups and the C α -N bond cleavage site in the substrate were measured as 12 Å and 20 Å, indicated by blue dotted lines, respectively. (B-D) the C-terminal amide formation of the thiopeptides similar to nosiheptide, such as nocathiacin (catalyzed by NocA), (C) GE2270A (catalyzed by TpdK or PbtH). (D) berninamycin A (catalyzed by BerI). (E) For comparison, thiostrepton maturation is done through deesterification and amidation by TsrB and TsrC, respectively. In (B-E), the products of the catalytic reactions were highlighted in red dashed boxes. (F) The sequence alignments of NosA and its homologs NocA, TpdK, PbtH and BerI. The conserved residues K49 and E101 were marked with stars on the top of the sequences. On the top of the amino acid sequence, the secondary structures of NosA were displayed, where arrows indicate β -sheets, and coils represent α -helices. The stars indicate the active sites observed in this report for enzymatic reaction.
generates enamide, and subsequent dealkylation requires the activity of NosA, which has recently been characterized as a new terminal amide synthase, leading to a C α -N bond cleavage for nosiheptide maturation with release of the co-product pyruvate. NosA catalyzed reaction is apparently distinct from those of known endogenously C-terminal amide-forming proteins, which typically catalyze an oxidative cleavage of C-terminal Gly-extended peptides accompanying glyoxylate production [5][6][7][8] . In this study, we provide the structural basis of NosA for investigation into its enzymatic mechanism.

Results and Discussion
Derivation of the truncated NosA 1-111 variant and its X-ray crystal structure. The full-length NosA contains 151 amino acids in total. Within ten days, it degrades as a large fragment with a molecular weight about 12 KDa (supporting information, Fig. S1). This instability is likely due to the sequence (residues 106-151) at the C-terminus, which has a potential to form a flexible loop based on the secondary structure analysis (http://bioinf.cs.ucl.ac.uk/psipred/) (supporting information, Fig. S2). Thus, we truncated NosA by cutting off the residues at its C-terminus ten by ten, and constructed the pET28a plasmids containing the genes NosA 1-140 (i.e., residues 1-140), NosA 1-130 (i.e., residues 1-130), NosA 1-120 (i.e., residues 1-120), and NosA 1-111 (i.e., residues 1-111), and carried out their overexpression and purification as we did on wild-type (WT) NosA. By running SDS_PAGE gels (supporting information, Fig.  S1), we found that, except NosA 1-111 , other NosA variants are still unstable (Among them, NosA 1-130 is too unstable to be obtained in the process of purification).
We thus overexpressed NosA 1-111 variant and its Se-Met form, and purified them to homogeneity for crystallization. The purified proteins were estimated to have a purity of > 95%. The crystals grew in a cubic form. The diffraction of Se-Met NosA 1-111 was extended to 2.40 Å resolution. Its crystals belong to the primitive cubic space group P4 1 32, with unit cell parameters of a = b = c = 143.3 Å. X-ray diffraction data sets of Se-Met NosA 1-111 were processed using data in the resolution ranges 50.0-2.4 Å. The typical Matthews coefficient and solvent content were estimated as 3.94 Å 3 Da −1 and 40.3%, respectively. The three-dimensional (3D) structure of NosA 1-111 was determined using the single-wavelength anomalous-dispersion (SAD) method. The crystal parameters and data-collection statistics were summarized in Table 1.
Three monomers occupy one asymmetric unit ( Fig. 2A,B), two of them form a dimer. One monomer constitutes a dimer with one monomer in an adjacent asymmetric unit. Each monomer is identical to the others with an RMSD value of 0.21 Å for the backbone C α atoms in the secondary structural regions, consisting of four anti-parallel β -sheets (β 1, β 2, β 3, and β 4), three α -helices (α 1, α 2 and α 3) and six loops (L1, L2, L3, L4, L5 and L6), which are arranged in the order of (Fig. 2C). The anti-parallel β -sheets form a semicircular hydrophobic surface, the α -helices and loops are located on the outside of the circle. The first six or fewer residues at the N-terminus are invisible in all monomers, and the residues from 35 to 42 are also invisible in the two monomers that form a dimer conformation. This dimeric structure adopts a global fold, resembling an elliptic β -barrel of 28.9 Å in height, with diameters of 15 Å and 23 Å ( Fig. 2A). The α -helices and loops surround the β -barrel.
The β -barrel is formed through several hydrogen bonds between the residues of the β 4 strand in one monomer and the β 2' strand in another monomer (Fig. 2D), including hydrogen bonds between the E101 carboxyl and S53' -OH group or R29' imide group, between the S99 backbone carboxyl and the S53' backbone nitrogen, and between the S99 backbone nitrogen and the S53' carboxyl; and a hydrogen bond between the Y97 backbone oxygen and the S54' side-chain -OH group. The β -barrel is also stabilized by salt bridge within the β -barrel between the side chains of the R96 and E94' residues ( Fig. 2E). Among these residues, R29 and E101 are conserved in the NosA homologs ( Fig. 1), indicating that they may be important for enzyme dimerization.
The possible key sites for the catalytic reaction. To understand how NosA interacts with its substrate, we searched NosA structural homologs in protein data bank using DALI server 9 . The structures with Z-score higher than 7.0 were selected. They are heme-degrading enzymes [10][11][12][13][14][15] or HapK protein involved in prodigiosin biosynthesis 16 , demonstrating that their ligands bind to the region outside of the β -barrel (supporting information, Fig. S3). Based on these observations, we assumed that the substrate of NosA might also bind to the region out of the β -barrel. To confirm this, we performed alanine-scanning mutagenesis assay on the conserved residues that are located at the entrances of the β -barrel (such as C11 and Q48), within the β -barrel (such as R29 and S53), outside of the β -barrel (such as Q48, K49, W51 and E101), or in the C-terminal loop (such as F122, D123, P124, S126, D128, P129, R132, P133, E135, F136, P138 and P139) in the full-length NosA, respectively. Then, we measured the catalytic activities of these variants by running enzymatic assay on HPLC system ( Fig. 3A and supporting information, Fig.  S4). Only K49A and E101A variants abolish the catalytic activities, indicating that K49 and E101 might be key elements for catalytic reaction. As shown in Fig. 3B, when E101 was replaced by D101, Q101 or K101, and residue K49 was mutated into E49, respectively, the E101K, K49E and E101Q variants abolish or significantly lose catalytic activities, while E101D retains the catalytic activity, indicating that the charged side-chains of E101 and K49 are critical to the catalytic reaction. Finally, we tested the possible effects on the folding and the aggregation state of NosA by these mutations by running the size-exclusion chromatography (SEC) assay, circular dichroism (CD) and nuclear magnetic resonance (NMR) spectroscopies. The results from the SEC assay suggested that the aggregation states of these variants were not affected by these mutations (Fig. 4A-C). Moreover, as shown in Fig. 4D, the retention time of the WT full-length NosA and the truncated NosA 1-111 locates between gel filtration protein markers (3 and 4) with molecular weights of 44KDa and 17KDa, respectively, indicating that they are dimer in solution. Thus, the C-terminal loop is not helpful to NosA dimerization. However, the CD spectrum of the K49A variant looks much different from those of WT protein and other variants (Fig. 4E), and its 2D 1 H-15 N HSQC spectrum did not overlap well with that of WT NosA (Fig. 4F), indicating that the mutation from K49 to A49 might affect the folding of the protein. Thus, the loss of the catalytic activity of K49A might also result from the changes in the folding of the protein.
NosA may function as a dimer. It was reported that the removal of two -OH groups on the substrate (highlighted in red and in blue in Fig. 1, respectively) (by knockouting the genes nosC and nosB in nosiheptide biosynthesis responsible for these -OH groups formation, respectively) made NosA completely lose the catalytic activities 17 , suggesting that these two -OH groups in the substrate are important for the catalytic reaction. Interestingly, as shown in supporting information, Fig. S5, in the current crystal structure of NosA 1-111 , the intra-molecule distance between the oxygen atom in the side-chain of E101 and the nitrogen atom in the side-chains of K49 is approximately 26 Å, much longer than the corresponding inter-molecular distance (13.6 Å) between these two atoms. The latter is close to that (~12 Å) between the -OH group in the pyridine ring (highlighted in red, Fig. 1A) and the C α -N bond cleavage site in the substrate (measured from the crystal structure of nosiheptide in complex with ribosomal subunit, pdb code 2ZJP, in which the distance between -OH group (highlighted in blue, Fig. 1A) and the C α -N bond cleavage site is 20 Å) 18 . Thus, we suggest that NosA may function as a dimer to catalyze the maturation  NosA 112-151 is crucial to the catalytic reaction. As we mentioned above, to get a stable form of NosA, we prepared several truncated NosA variants. Among them, NosA 1-111 is the most stable. However, the enzymatic assay performed on the HPLC system indicates that NosA 1-111 has no catalytic activity ( 15 N-labeled NosA 112-151 in its free state (black), in complex with NosA 1-111 (green), in complex with the substrate (blue), in complex with NosA 1-111 and substrate (red), respectively. (E) Two hydrogen bonds between NosA and the substrate are formed, supported by the distance changes during MD simulation trajectory. The change in the distance between nitrogen atom of the side-chain of K49 and oxygen atom in -OH group in pyridine of the substrate is shown in black, and the change in the distance between the backbone oxygen of S126 and oxygen atom in -OH group in the substrate highlighted in blue in Fig. 1 is shown in red, respectively. (F) The last snapshot of the MD simulation trajectory, where the two monomers of NosA were displayed in green and yellown ribbon modes, respectively. The substrate was displayed in cyan-sticks mode. The main residues contributing to the protein-substrate interactions, including K49 and E101 in NosA 1-111 and residues in NosA 1-120 , were also shown in sticks mode. The two hydrogen-bonds between NosA and substrate were displayed in dotted lines. To probe whether the whole sequence of NosA 112-151 affects the catalytic reaction, we measured the catalytic activities of the truncated NosA variants with different length, and found that the NosA 1-140 variant maintained the catalytic activity almost similar to the full-length NosA, whereas NosA 1-120 significantly lost the catalytic activity (Fig. 6A), revealing that the residues from 120 to 140 (i.e., A 121 FDPASPEPLTRPQEFVPPG 140 ) of NosA 112-151 are important to the catalytic reaction. To investigate whether the sequence and the coiled-coil conformation of NosA 112-151 are specific to the catalytic reaction, we replaced NosA 112-151 in the mixture by three randomly-selected peptides (Did2, Vps60 and JmjN) available in the lab [19][20][21][22] , respectively, and tested the catalytic activities of these mixtures, all displaying no catalytic activities at all (Fig. 6B). These observations suggest that the sequence and the flexible loop conformation of the NosA 112-151 variant is crucial to the catalytic reaction. (2) The hydrogen bond between the backbone carbonyl oxygen atom of residue S126 and the oxygen atom in the -OH group next to the carbonyl group on the substrate (highlighted in blue in Fig. 1A) was formed around 15 ns, and was maintained during the next 10 ns of simulation (Fig. 5E). We assumed that the formation of this hydrogen bond might induce the C-terminal loop to bend and wrap the substrate. The second hydrogen bond was formed between K49 side-chain in NosA and the -OH group in pyridine ring of the substrate, because the distance between them was kept less than 4 Å (Fig. 5E). These observations are supported by the fact that the catalytic power of NosA was completely abolished after removing these two hydroxyl groups in the substrate 17 . Moreover, the -OH group in pyridine ring is obviously more acidic than -OH group next to the carbonyl group in the substrate, which could interpret why the side-chain of residue K49 of NosA easily interacts with the -OH group in pyridine ring. (3) Residues involved in the protein-substrate interaction mainly locate at residues 120-140 on the C-terminal loop (Fig. 5F), consistent with the results from the enzymatic assay above. (4) The distance between the side-chain carboxylic oxygen atoms of E101 and the carbon atom at the C α -N bond cleavage site on the substrate is kept larger than 5 Å, smaller than 12 Å during the MD simulation (data not shown).
The proposed mechanism for NosA catalytic amidation reaction. Taken all results above together, we proposed the following catalytic mechanism (Fig. 7): (1) The terminal dehydroalanine unit is tautomerized to methyl imine intermediate A in basic buffer condition, supported by the previous studies on thiostrepton synthesis 23,24 , where similar reaction is initiated by Et 2 NH; (2) The substrate is fixed into the active sites by hydrogen-bond and hydrophobic interactions between the substrate and residue K49, the C-terminus of NosA, generating intermediate B, supported by the structural and MD studies above; (3) The negatively charged side-chain of E101' interacts with one molecule H 2 O, supported by the findings that several water molecules exist close to E101' in the crystal structure; (4) The nucleophilic attack by H 2 O to methyl imine produces intermediates C and D, leading to the final Cα -N bond cleavage to yield nosiheptide and pyruvate.

NosA family members catalyze the amidation reaction through a similar way. Most impor-
tantly, it was reported that NosA homologs could also catalyze the terminal amide formation of some thiopeptides (Fig. 1). For example, NocA (64% identity to NosA) catalyzes the formation of nocathiacin 25 , TpdK (34% identity to NosA) and PbtH (44% identity to NosA) catalyzes the final formation step in GE2270A biosynthesis 26,27 , BerI (48% identity to NosA) catalyzes the final step of berninamycin A in its biosynthesis 28 . Among these enzymes, the residues K49 and E101 are conserved, corresponding to residues K41 and E93 in NocA. Thus, we made the NocA K41A and E93A variants, and measured their catalytic activities on the same substrate (Fig. 3D). The results indicate that either the NocA K41A or E93A variant loses catalytic activities, revealing that NosA and its family members might share a common mechanism to catalyze the terminal amidation reaction.
In conclusion, we characterized three key elements (basic lysine, acidic glutamic acid and flexible C-terminal loop) for the terminal amide formation in nosiheptide-represented thiopeptide biosynthesis, these mechanistic studies on how NosA works present an intriguing paradigm about how NosA family members function during thiopeptide biosynthesis. NosA (151 amino acids in total), truncated NosA variants, including the truncated NosA 1-111 (i.e., residues 1-111), NosA 1-120 (i.e., residues 1-120), NosA 1-130 (i.e., residues 1-130), NosA 1-140 (i.e., residues 1-140), and NocA (151 amino acids in total) were engineered into a pET28a vector with a His 6 -tag using NheI and HindIII restriction sites. Site-directed mutagenesis was performed using a QuikChange site-directed mutagenesis kit (Stratagene Inc.). The ORF of the C-terminal NosA 112-151 (i.e., residues 112-151) was engineered into a pSMT3 vector with a SUMO tag, which can be removed using ULP1 protease. The constructs were verified by DNA sequencing, and the plasmids were transformed into Escherichia coli BL21(DE3) competent cells. The transformed cells were grown at 310 K in a Luria Broth (LB) medium containing 50 μ g ml −1 kanamycin and were induced (24 h, 291 K) by the addition of 0.1 mM isopropyl-d-thio-b-D-galactopyranoside (IPTG) when the OD 600 value was measured as 0.5 -0.6. The cells were harvested and resuspended in lysis buffer (50 mM phosphate, pH 7.5, 500 mM NaCl), lysed with 10 μ g ml −1 PMSF by sonication on ice. The lysates were clarified by centrifugation (30 min, 16,000 rpm), and the soluble proteins were purified by nickel-affinity chromatography (GE Healthcare) though a linear gradient of 0-500 mM imidazole in the lysis buffer. Protein fractions were collected and dialyzed twice at 277 K against the lysis buffer. To remove the His 6 -tag, the fusion proteins were digested with thrombin protease (ULP1 protease for NosA 112-151 ) overnight at 4 °C, followed by running a second nickel column.
The fractions were collected and further purified by running gel-filtration chromatography on a Superdex 75 column (GE Healthcare) with a buffer containing 25 mM phosphate, pH 7.5, and 50 mM NaCl. Finally, the peak fractions containing proteins were concentrated to 50 mg ml −1 using an ultra-centrifugal filter tube (Millipore) and were used for crystallization (50 mM Tris-HCl, pH 7.5 50 mM NaCl for crystallization), NMR experiments, stability testing experiment by running SDS-PAGE gels, or kinetic experiments. The pure protein fractions were further verified by running an SDS-PAGE gel and electrospray mass spectrometry. The protein concentrations were estimated from the absorbance at 280 nm with their corresponding absorption coefficients.
The truncated NosA 1-111 variant has one Met residue at its N-terminal sequence. Thus, for crystallization, the selenomethionine NosA 1-111 (Se-Met NosA 1-111 ) was successfully expressed in M9 medium using the reported methionine-pathway inhibition protocol 29,30 and was purified as performed on native NosA 1-111 above.
For NMR experiments, the 15 N-labeled NosA 112-151 , the full-length NosA and its K49A variant, and the 13 C-and 15 N-labeled NosA 112-151 were expressed in M9 medium containing 15 NH 4 Cl and/or 13 C-glucose as the nitrogen and carbon source, respectively. NosA 1-111 crystallization and its X-ray data collection. Initial crystallization trials were performed at 293 K with Crystal Screen HT and Index HT kits in 96-well plates using the sitting-drop vapor-diffusion method (Hampton Research). For refinement of the crystallization conditions, 1 μ l of protein solution was mixed with an equal volume of reservoir solution and equilibrated against 0.5 ml of the reservoir solution at 293 K in 24-well plates using the sitting-drop vapor-diffusion method. Crystals of Se-Met NosA suitable for X-ray analysis were obtained under the following conditions: 0.02 M magnesium chloride hexahydrate, 0.1 M HEPES pH 6.0 -9.0, and 22% w/v polyacrylic acid 5100.
The crystals were picked up in a nylon loop (Hampton Research) and mounted in liquid nitrogen for flash-cooling. X-ray diffraction data were collected at beamline BL17U of the Shanghai Synchrotron Radiation Facility (SSRF, China) using a MAR CCD MX-225 detector. The wavelength of the radiation was 0.9794 Å, and the distance between the crystal and the detector was 400 mm. The exposure time for each frame was 1 s with a 1 o oscillation, and 360 frames were collected. The data were indexed, integrated and scaled using the HKL-2000 program suite 31 . The structure was solved using the single-wavelength anomalous dispersion method implemented in Phenix 32 , which identified 3 distinct Se atoms and yielded a figure of merit of 0.22. Model building was performed using Coot 33 . Structure refinement was carried out using Phenix and Refmac5 32,34 . Structure analysis was carried out using programs in CCP4 35 . The figures were generated using Pymol (http://www.pymol.org). The statistics of the structure refinement and the quality of the final structure models are summarized in Table 1.
The enzymatic assay on HPLC system. To assay the activities of the full-length NosA, NocA and their variants, the truncated N-terminal NosA 1-111 variant, the mixture of NosA 1-111 with the C-terminal NosA 112-151 peptide, each protein solution (including the full-length NosA, NocA or their mutants, or the mixture of NosA 1-111 with NosA 112-151 at mole ratio of 1:1.) was diluted to 1 mg/ml in buffer A (50 mM Glycine buffer, pH 9.0). To avoid degrading, all samples are used after purification was finished. The substrate was first dissolved in DMSO, and then diluted to 1 mg/ml by buffer A. For simple comparison, 5 μ L of diluted protein solution and 5 μ L of substrate solution were added into 40 μ L of buffer A, respectively, and then was incubated at 303 K for 60 min. To quench the reaction, 5 μ L of methanol were added. To determine whether the product nosiheptide was generated or not, HPLC analysis was performed by running C-18 reverse phase column (Agilent 1100) with a MeCN/H 2 O gradient mobile phase. The wavelength of nosiheptide detection was specified as 254 nm.
To investigate the importance of the sequence and the coiled-coil conformation of the C-terminal NosA 112-151 in the catalytic reaction, three randomly selected peptides (Did2, Vps60 and JmjN from Scientific RepoRts | 5:12744 | DOi: 10.1038/srep12744 KDM5B) available in the lab were used to replace NosA 112-151 , and to mix with the N-terminal NosA 1-111 variant. Both Did2 (176-204 aa, with an amino acid sequence as follows: NVPEIKAKEV NVDDEKEDKL AQRLRALRG) and Vps60 (128-186 aa, with an amino acid sequence as follows: INIDKLQDMQ DEMLDLIEQG DELQEVLAMN NNSGELDDIS DAELDAELDA LAQEDFTLP) peptides were reported in ESCRT-III system involved in multivesicular bodies (MVB) pathway 21,22 , with a helix conformation and a random-coiled conformation in their free states, respectively. The JmjN peptide (with an amino acid sequence as follows: ECPVFEPSWA EFRDPLGYIA KIRPIAEKSG ICKIRPPAD) locates at the N-terminal histone H3K4me3/2 demethylase KDM5B 19,20 , with a regular conformation. NosA 1-111 was first mixed with each peptide at mole ratio 1:1, then the assay was performed as we did on the mixture of NosA 1-111 and NosA 112-151 .
For k cat and K m measurements of the full-length NosA, a time course was carried out to determine the initial rate in 50 mM Glycine-NaOH buffer (20 μ L, pH = 9.0) that contained 10 nM NosA and 200 μ M substrate. The reactions were initiated by the addition of NosA into diluted substrate solution, incubated at 303 K, and then terminated by adding 50 μ L of methanol into the solution at 2, 5, 10, 15, 30, 60, 120, and 240 min, respectively. The samples were subjected to the same workups and HPLC analysis as described above. The production of product (i.e., the intensity of the peak was used to quantify the nosiheptide), linear with respect to time during 0-5 min, was fitted into a linear equation to obtain the initial velocity. To determine the kinetic parameters for the conversion of substrate to product, the reactions were carried out at 303 K for 5 min in each 20 μ l of the mixture that contained 10 nM NosA, 50 mM Glycine-NaOH (20 μ L, pH All assays were performed in duplicate. Each conversion was analyzed by HPLC as described above. The resulting initial velocities were then fitted to the Michaelis-Menten equation using GraphPad Prism 5 to extract K m and k cat parameters.

Molecular modeling of complexation between full-length NosA with its substrate.
To understand how the substrate interacts with NosA, as the first step, the substrate molecule was docked onto the crystal structure of NosA 1-111 . A distance-restrained docking was performed based on two observations. Firstly, the results of enzymatic assay indicate that the C-terminal NosA 112-151 loop is important for catalytic reaction, which may be due to the stabilizing effect of the C-terminus on the complex structure. Due to the absence of C-terminus, it is probably difficult to generate a rational binding mode for the substrate and NosA 1-111 by an unrestrained docking. Secondly, based on the information that the residues K49 and E101 play a vital role in maintaining the catalytic activity of NosA, the distances between the side-chain amino group of K49 and the oxygen atom of the -OH group in the pyridine ring of the substrate and between the side-chain carboxylic oxygen atoms of E101 and the nitrogen atom at the C α -N bond cleavage site on the substrate were restricted within 2.5-3.5 Å during molecular docking. All ligand docking jobs were performed using the GOLD software suite 36 .
As the second step, a structural model of C-terminal NosA 112-151 was derived through homology modeling. Our homology search found that the C-terminus of the HTLV-II matrix protein (PDB code 1JVR) 37 , which is also an irregular loop, had a sequence identity of 37.5% to the C-terminal NosA 112-151 peptide. Thus, it was used as the template to build a model of the NosA C-terminus. The homology modeling job was completed by using the MOE software (version 2013) (Chemical Computing Group Inc., Montreal, Quebec, Canada). The resulting structure was adjusted manually to avoid steric hindrance between C-terminus and the rest part of the complex. Simultaneously, the main chain of C-terminus should stay as close as possible to the -OH group next to the carbonyl group of the substrate (Fig. 1A, highlighted in blue) according to the fact that the catalytic activity was abolished due to the deletion of this -OH group 17 . Finally, the adjusted structure was assembled onto the crystal structure of the N-terminal NosA  .
The third step was to refine the binding mode of the substrate to NosA, which was generated by molecular docking as described above, through molecular dynamics (MD) simulation. Our MD simulations were performed by using the AMBER software (version 12) 38 . The docking model was supplied as the initial configuration for MD simulation. The AMBER FF12SB force field 39 was applied. The partial charges were calculated with the Gaussian 09 software 40 at the HF/6-31G(d) level and were further processed by the RESP model 41 . The complex structure was soaked in a TIP3P 42 water box with a margin of 10 Å at each direction. The whole system was neutralized by addition of a proper number of counterions (Na + ). The system was gradually heated to 300 K over 100 ps. Then, a total of 25 ns simulation was performed at a 2-fs interval under a constant temperature of 300 K and a constant pressure of 1 atm. During simulation, all covalent bonds containing hydrogen atoms were constrained with the SHAKE algorithm 43 . Long-range electrostatic effects were modeled using the particle-mesh-Ewald method 44 with a cutoff of 12 Å. The Langevin equilibration scheme 45 was used to control and equalize the temperature. Snapshots were saved every 500 fs during simulation. The resulting MD trajectories were analyzed using the PTRAJ module and visualized with the VMD software 46 . Circular dichroism spectra of NosA 112-151 , full-length NosA and its variants. To probe the folding of full-length NosA and its variants, and the C-terminal NosA 112-151 peptide, circular dichroism (CD) experiments were performed at 298 K on a JASCO-715 spectropolarimeter (Jasco International Co., Tokyo, Japan). Data were collected at 0.1-nm intervals at a scan speed of 20 nm/min, a 1-nm bandwidth, and a 0.25-s response time from 250 to 190 nm. Circular quartz cells of 1-and 0.1-cm path lengths were used for the far-UV regions. The CD intensities are expressed as the molar residue ellipticities given in units of degrees cm 2 mol −1 . The concentration of protein or peptide was about 20 μ m. The buffer conditions used for running the CD spectra were 50 mM sodium phosphate (pH 7.0).
To probe the effects on the folding of NosA by the mutation from K49 to A49, we first run circular dichroism (CD) spectra on both wild-type (WT) NosA and its K49A variant. The results from CD spectra indicate that the folding of K49A looks much different from that of WT full-length protein. To double check these changes on the folding of the K49A variant, 1 H-15 N HSQC spectra were acquired on WT full-length NosA and its K49A variant at 293 K in NMR buffer (50 mM Na 2 HPO 4 , 50 mM NaCl, 0.01% NaN 3 , pH 7.0 and 10% D 2 O).
To determine the flexibility of the NosA 112-151 in the presence of NosA 1-111 and the substrate, the measurements of the relaxation times, T 1 and T 2 and 15 N-1 H NOEs of the backbone atoms of the NosA 112-151 were performed at 293 K in NMR buffer (50 mM phosphate, 50 mM NaCl, 0.01% NaN 3 , pH 7.0 and 10% D 2 O) at 293 K for NosA 112-151 on a Varian INOVA 600-MHz spectrometer. Ten different values for the relaxation delay were used for the T 1 (delays 60, 80, 100, 140, 240, 360, 540 and 760 ms) and T 2 (Carr-Purcell-Meiboom-Gill mixing times of 10, 30, 50, 70, 90, 110, 150, 170 and 190 ms) relaxation experiments. The T 1 and T 2 values were obtained by nonlinear squares fits using the program Prism 5. The 15 N-1 H NOE values were calculated as the ratio of the intensities of the paired 15 N-1 H correlation peaks from interleaved spectra acquired with and without 1 H pre-saturation during a recycle time of 3 s.

Size-exclusion chromatography (SEC) assay.
To probe effects on the aggregation state of NosA by the mutations, the size-exclusion chromatography assay was performed on a Superdex 75 column (10/300 GL) (GE Healthcare), which was previously equilibrated with buffer B (50 mM Tris-HCl, pH 7.5, 50 mM NaCl).

Isothermal titration calorimetry (ITC) binding assay.
To investigate the binding affinity of the N-terminal NosA 1-111 with the C-terminal NosA 112-151 , the isothermal titration calorimetry (ITC) binding assay was performed. An ITC-200 microcalorimeter (GE Healthcare) was used with a buffer containing 20 mM Tris-HCl, 150 mM NaCl, pH 7.0 at 298 K. The reference titration of small molecules in the buffer was subtracted from the experimental data, and the data were fitted using the Origin 7.0 (OriginLab Corporation) software.