Peptide 17

Molecular mechanism underlying substrate recognition of the peptide macrocyclase PsnB
Inseok Song, Younghyeon Kim, Jaeseung Yu, Su Yong Go, Hong Geun Lee, Woon Ju Song and Seokhee Kim ✉
Graspetides, also known as ω-ester-containing peptides (OEPs), are a family of ribosomally synthesized and post-translationally modified peptides (RiPPs) bearing side chain-to-side chain macrolactone or macrolactam linkages. Here, we present the molecular details of precursor peptide recognition by the macrocyclase enzyme PsnB in the biosynthesis of plesiocin, a group 2 graspetide. Biochemical analysis revealed that, in contrast to other RiPPs, the core region of the plesiocin precursor peptide noticeably enhanced the enzyme–precursor interaction via the conserved glutamate residues. We obtained four crystal struc- tures of symmetric or asymmetric PsnB dimers, including those with a bound core peptide and a nucleotide, and suggest that the highly conserved Arg213 at the enzyme active site specifically recognizes a ring-forming acidic residue before phosphoryla- tion. Collectively, this study provides insights into the mechanism underlying substrate recognition in graspetide biosynthesis and lays a foundation for engineering new variants.

atural products have been the main source of therapeutic leads owing to their structural diversity and bioactivities1.
RiPPs are a major class of natural products that exhibit a broad scope of chemistry and diverse biological functions2–4. RiPPs are initially synthesized as a precursor peptide by ribosomes and are subsequently modified by enzymes that install a class-defining structural motif2,4. The precursor peptide is frequently composed of two segments, a leader peptide (LP), which is the site for enzyme recognition and is often cleaved during biosynthesis, and a core peptide (CP), which is the site for chemical modification. Because substrate recognition does not solely rely on the direct binding of a modification site to the enzyme active site, modular RiPP biosyn- thesis often leads to the natural evolution of highly diverse CPs and facile engineering of variants with higher chemical diversity.
Many structures of the enzyme–LP complex have provided molecular details of LP recognition5–16. However, our understanding of CP recognition and the following primary modification process has been largely limited because only a few enzyme structures with a bound CP are known. Recent examples of CP-bound structures include the radical S-adenosylmethionine (rSAM) enzyme CteB in ranthipeptide biosynthesis12, backbone N-methyltransferase OphMA in borosin biosynthesis17,18, lantipeptide dehydratase NisB19 and, while not an RiPP enzyme, thioamide-installing YcaO (Mj-YcaO)20. The difficulty of obtaining the native enzyme–CP structure may rise from the transient nature of the enzyme–substrate complex in enzyme reactions and the apparent low affinity of CPs to the cognate enzyme in many RiPP families14,21,22. Therefore, the above examples show only a small part of the CP (CteB)12 or take advantage of the natural covalent linkage to the enzyme (OphMA)17,18, the artificial fusion of a substrate mimic to the enzyme (NisB)19 or the synony- mous activity of the non-RiPP enzyme (Mj-YcaO)20.
Graspetides, formerly known as OEPs, are RiPPs that harbor side chain-to-side chain macrolactone or macrolactam linkages between an acidic residue (aspartate or glutamate) and a hydroxyl- or amine-containing residue (serine, threonine or lysine)4. The prototypic members of this RiPP family are microviridins (group 1 graspetides) that have a cage-like tricyclic structure with two

lactones and one lactam on the highly conserved TXRXPSDXDE core motif23–25. Recent bioinformatic and biochemical analyses revealed 11 new or putative groups of graspetides that have novel core motifs and ring connectivities, including plesiocins (group 2 graspetides; Fig. 1a)26–29. The presence of topologically diverse gras- petides suggests that the formation of multicyclic peptides by side chain-to-side chain crosslinking is a simple but powerful strategy to expand the chemical space of peptides both in natural evolution28,30 and for engineering novel functionalities31–33.
ATP-grasp enzymes are responsible for the biosynthesis of mac- rocyclic linkages in graspetides24,25. Using ATP, they initially activate a carboxyl side chain by phosphorylation and mediate the nucleo- philic substitution of a hydroxyl or amine group to form an ester or amide bond (Fig. 1b). Two crystal structures of the ATP-grasp enzymes in microviridin biosynthesis, an apo form of MdnB and an MdnC–LP complex, revealed a novel LP-binding domain and a distinct conformational change of the enzyme following the leader binding; the movement of the β9β10 hairpin by 25 Å opens the core-binding site14. These structures, however, showed neither the CP nor the nucleotide in the electron density maps, leaving a ques- tion about how the enzyme recognizes the CP and proceeds to the specific phosphorylation and ligation of the acidic residue.
Here, we studied the mechanism underlying precursor recog- nition using plesiocin biosynthesis as a model system (Fig. 1a). Biochemical studies revealed that the binding of the LP and a nucleotide promotes CP binding and that, unexpectedly, CP bind- ing enhances the enzyme–precursor interaction. Four crystal struc- tures of PsnB, including those with the well-resolved nucleotide, LP and CP, revealed residues involved in the binding of nucleotides, LP or CP and showed the conformational changes of the PsnB dimer following substrate recognition. Collectively, these results suggest a molecular mechanism for the precursor binding and activation of graspetide biosynthetic enzymes.
An MP recapitulates an enzyme reaction. To study the molecu- lar details of the macrocyclization reaction in graspetides, we chose

Department of Chemistry, Seoul National University, Seoul, South Korea. ✉e-mail: [email protected]

a b


2-O3PO O







c d e

Precursor of

Minimal precursor (PsnA214–38, MP)


(Kd = 3.1 ± 0.13 µM)

(K = 3.9

µM) 10

d ± 0.20


(K = 40

d ± 1.8 µM)

(Kd = 90 ± 5.9 µM)
0 1 10 100 1,000
PsnB (µM)

0 1 10
MP (µM)


Fig. 1 | Introduction of plesiocin and its minimal precursor peptide. a, Structure of plesiocin, a group 2 graspetide. Plesiocin consists of four core motifs (TTXXXXEE), and each core contains two macrolactone linkages between threonine and glutamate. b, Suggested mechanism underlying macrolactone formation in plesiocin biosynthesis. Using ATP, ATP-grasp macrocyclase (PsnB) phosphorylates the carboxyl side chain of glutamate in the core region of the precursor peptide (PsnA2) and mediates nucleophilic substitution by threonine to form the macrolactone. c, Minimal precursor peptide (MP; PsnA214–38; overlined with purple) was designed from the native sequence of PsnA2. MP contains the key region of the LP (PsnA214–24; red) and one CP (PsnA228–38; blue). Variants of LP used in d are shown as solid lines below PsnA2. d, The affinity of the LP variants was determined by measuring the fluorescence anisotropy of fluorophore-labeled LP variants (0.1 μM) with different concentrations of PsnB. e, ATP consumption by PsnB (0.4 μM) was measured with ATP (5 mM) and different concentrations of the MP. Data are presented as dot plots with mean ± 1 s.d. (n = 3 independent experiments) and are fitted to a hyperbolic equation (d,e).

plesiocin biosynthesis as a model system. Previously, we reported that an ATP-grasp enzyme, PsnB, installs two macrolactones on each of the four conserved core motifs of the precursor PsnA2 (ref. 26) (Fig. 1a–c). To quantitatively analyze the multiple steps of the macrocyclization reaction (precursor binding, ATP consump- tion and macrolactone formation), we started by finding an MP that contains only one core motif and a minimal LP. Our previous report showed that the complete macrocyclization reaction requires the conserved region of the LP, LFIEDL (PsnA214–19; Extended Data Fig. 1a), but not the N-terminal 12 residues (PsnA21–12)26. Here, we tested whether LFIEDL is required for enzyme binding by measur- ing fluorescence anisotropy of fluorophore-labeled LPs of various lengths (Fig. 1c,d). Indeed, the LFIEDL region was essential for tighter enzyme binding, whereas the N-terminal 13 residues were dispensable. Therefore, we chose PsnA214–38 as the MP for biochem- ical and structural studies (Fig. 1c).
We first measured the ATPase activity of PsnB with increasing amounts of MP. The KM and Kcat values were 30 μM and 14 min–1, respectively (Fig. 1e), which are similar to those of AMdnC, an MdnC homolog from Anabaena sp. PCC7120 (27 µM and 12 min–1, respectively)34. ATP titration resulted in KM and Kcat values of 92 μM and 16 min–1 (Extended Data Fig. 1b), of which KM is in the mid- dle of those of non-RiPP ATP-grasp enzymes, glutathione synthe- tase (240 μM)35 and D-alanine-D-alanine ligase (49 μM)36. We also found that 10 mM MgCl2 was sufficient for the maximal activity of PsnB (Extended Data Fig. 1c). By monitoring the loss of water in the MALDI spectra of reaction solutions, we observed that the

modification rate of MP was similar to that of the wild-type PsnA2 (Extended Data Fig. 1d)26.
To obtain evidence for the suggested mechanism of the mac- rocyclization reaction (Fig. 1b), we added to the reaction solution hydroxylamine (NH2OH) for trapping the acyl-phosphate interme- diate37,38. Although 0.5 M NH2OH severely reduced PsnB activity, we detected two trapped peptides: one with the NH2OH-added Glu37 and the other with the NH2OH-added Glu38 and one macrolactone between Thr32 and Glu37 (Extended Data Fig. 2). These results are consistent with the activation of the carboxylate by ATP-grasp enzyme39 (Fig. 1b) and with the reaction order of PsnA2; the Thr32– Glu37 inner ring forms before the Thr31–Glu38 outer ring26.

LP activates the enzyme. We next investigated the role of LP in enzyme behavior. By monitoring the fluorescence anisotropy of the fluorophore-labeled CP (Fl_PsnA228–38 or Fl_CP) and ATP con- sumption of PsnB, we initially observed that the LP (PsnA214–24) enhances the enzyme–CP interaction and the ATPase activity by PsnB (Extended Data Fig. 3a,b). To stably attach the LP to the enzyme, we used the leader-fused PsnB (LP_PsnB), which was previously shown to install two macrolactones in CP33. The fluo- rescence anisotropy of the fluorophore-labeled LP (Fl_PsnA214–24 or Fl_LP) indicated that the LP in LP_PsnB successfully competes for the cognate binding site in PsnB (Extended Data Fig. 3c). We then measured the fluorescence anisotropy of Fl_CP with PsnB or LP_PsnB and found that Fl_CP binds more tightly to LP_PsnB (Kd = 11 μM) than to PsnB (Kd > 100 μM; Fig. 2a). The LP-mediated

a b





3 LP_PsnB



Kd = 11 ± 0.87 µM


Kd > 100 µM 0


0 1 10 100
Enzyme (µM)

0 100
CP (µM)


1,050 1,100 1,150 1,200

d e Relative f




Kd = 1.2 ± 0.078 µM
Kd = 7.3 ± 0.67 µM

Sequence Affinity to PsnB

reaction rate 20




0 0.1

1 10 100
PsnB (µM)

0 5 10 15

0 10
Precursor (µM)


Kd (µM)

Fig. 2 | Roles of LPs and CPs in enzyme behavior. a, The affinity of the CP to PsnB or to the leader-fused PsnB (LP_PsnB) was determined by measuring the fluorescence anisotropy of the fluorophore-labeled CP (Fl_CP; 0.1 μM). b, The ATPase activity of PsnB or LP_PsnB (2 μM; 37 °C) was measured with different amounts of CP. c, Macrocyclization of CP (200 μM) by PsnB or LP_PsnB (10 μM; 4 h, 37 °C) was monitored by the loss of water in MALDI spectra of reaction solutions. d, The affinity of MP or LP to PsnB was determined by fluorescence anisotropy of fluorophore-labeled MP or LP (0.1 μM).
e, MP variants with one or two threonine-to-alanine or glutamate-to-alanine mutations were tested for affinity to PsnB (bar graphs) and macrocyclization reaction (circles). Affinity was determined as in d, and bar graphs represent the determined Kd and the error of fitting (see Supplementary Fig. 1a for data and fitting). The relative reaction rates were estimated by using MALDI spectra of the samples taken at different time points (see Supplementary Fig. 1b for MALDI spectra). f, ATPase activity of PsnB with different concentrations of the MP variants shown in e. Data are presented as dot plots with mean ± 1 s.d. (n = 3 independent experiments; a,b,d,f) and fitted to a hyperbolic equation (a,d,f).

enhancement of the enzyme–CP interaction was also observed in the class II lantipeptide synthetase HalM2 (ref. 21). LP_PsnB had much higher ATPase activity, which further increased with more CP, whereas PsnB had minimal ATPase activity regardless of the CP (Fig. 2b). As previously reported33, LP_PsnB, not PsnB alone, modi- fied CP (Fig. 2c). Collectively, the LP of PsnA2 promotes later steps of the enzyme reaction (enzyme–core interaction, ATP consump- tion and macrolactone formation). These results are consistent with enzyme activation by LP in the biosynthesis of microviridins and other families of RiPPs6,14,21,32,40–43.

CP enhances enzyme binding via conserved glutamate. To study the effect of the CP on enzyme–precursor interactions, we per- formed fluorescence anisotropy experiments with either Fl_LP or the fluorophore-labeled MP (Fl_PsnA214–38 or Fl_MP). Surprisingly, the MP binds to the enzyme approximately sixfold tighter than the LP (Fig. 2d), suggesting that the CP of PsnA2 also contributes to the enzyme–precursor interaction. By contrast, the CP of MdnA (a microviridin precursor) did not enhance the interaction with MdnB or MdnC14. To determine which residues in the CP enhance the enzyme–precursor interaction, we determined the affinity of MP variants in which the conserved threonine or glutamate residue of the CP is mutated to alanine. Alanine mutations in glutamate, but not in threonine, substantially reduced the affinity (four- to tenfold), indicating that the conserved glutamate residues are critical for the affinity-enhancing effect of the CP (Fig. 2e and Supplementary Fig. 1a). Interestingly, the rates of ATP consumption and macrocycli- zation of all these variants were considerably reduced (>fourfold),

although some of these variants still had the ring-forming threo- nine–glutamate pair (Fig. 2e,f and Supplementary Fig. 1b). In partic- ular, the T31A or T32A variants that retain two conserved glutamate residues also showed much lower ATPase activity than the MP, sug- gesting that threonine may help the phosphorylation of the acidic residue, or the phosphorylation step may not be independent of the subsequent macrocyclization. Collectively, these results suggest that the PsnB enzyme binds to the CP primarily by the conserved gluta- mate residues but efficiently mediates phosphorylation and macro- cyclization only with the native CP sequence TTXXXXEE.

Binding of ADP or AMPPNP enhances enzyme–core interaction. Next, we probed the effect of nucleotides on the enzyme–precursor interaction. The fluorescence anisotropy of Fl_MP or Fl_LP with ADP or AMPPNP, a nonhydrolyzable analog of ATP, showed that two nucleotides enhance the enzyme–MP interaction by ninefold and twofold, respectively (Fig. 3a), but not as much the enzyme–LP interaction (Fig. 3b), indicating that ADP and AMPPNP strengthen the enzyme–CP interaction. In agreement with this result, higher amounts of AMPPNP at a single concentration of PsnB (1 µM for Fl_MP and 5 µM for Fl_LP), at which only 30–40% of the peptide binds to PsnB without AMPPNP, increased the bound fraction of Fl_MP but not of Fl_LP (Fig. 3c). These results suggest cooperative interactions among the enzyme, the precursor, including the CP, and a nucleotide. By contrast, the MdnC–MdnA interaction was not altered by the addition of ATP, ADP or AMPPNP14.
We also determined the affinity of the MP, the single-ring inter- mediate (MP-1H2O) and the double-ring product (MP-2H2O) with

a b c







Fl_MP, 1 µM PsnB

0 0.1

PsnB (µM)

10 100

0 0.1

1 10
PsnB (µM)


0 100

Fl_LP, 5 µM PsnB

1,000 10,000

Fig. 3 | ADP and AMPPNP enhance the affinity of the CP to PsnB. a,b, The affinity of MP (a) or LP (b) to PsnB in the presence of ADP or AMPPNP (1 mM) was determined by fluorescence anisotropy. The experiments were performed together with those in Fig. 2d, and the data with no nucleotide (gray) represent those from Fig. 2d. c, The fractions of PsnB-bound MP or LP were determined by fluorescence anisotropy with different amounts of AMPPNP. Either 1 µM PsnB or 5 µM PsnB was used for MP or LP, respectively, to start with the bound fraction of 0.3–0.4 in the absence of AMPPNP. Data are presented as dot plots with mean ± 1 s.d. (n = 3 independent experiments; a–c) and fitted to a hyperbolic equation (a,b).

or without ADP (1 mM). The single-ring intermediate had the high- est affinity in both conditions, although its maximal ATPase activity was approximately half that of the MP (Extended Data Fig. 4).
Structures of PsnB reveal core-bound asymmetric dimers. To understand the molecular mechanism underlying the enzyme– substrate interaction, we crystallized PsnB with a nucleotide (ADP or AMPPNP) and an precursor peptide (MP or its vari- ant with the phosphomimetic Glu37 (MP(pE37)), in which the unstable acyl-phosphate intermediate (CO-OPO 2−) is replaced with a more stable acyl-difluoro-phosphonate (CO-CF2PO 2−; see Supplementary Fig. 2 and Supplementary Note for chemical syn- thesis of Fmoc-pE37 (6) and MP(pE37)). We determined four crystal structures containing different components (Extended Data Fig. 5 and Supplementary Tables 1 and 2). Basically, PsnB has a common architecture and the conserved ATP-binding site of ATP-grasp enzymes, including MdnC and non-RiPP ATP-grasp enzymes with known structures (RimK, GshB, LysX, PurT and DdlB) (Supplementary Fig. 3)14,44–48. Compared to MdnC, PsnB has a longer β6α3 loop region (Arg72–Gln90) and a shorter β9β10 loop (Lys172–Arg181; Supplementary Fig. 3)14. More notable differences of these PsnB structures are, however, that some PsnB subunits have well-resolved nucleotides and the CP as well as the LP, and that the dimers show different levels of asymmetry (Fig. 4a, Extended Data Figs. 5 and 6 and Supplementary Table 2). Nine independent PsnB dimers are classified into four states based on their components (Extended Data Fig. 5 and Supplementary Table 2). In the two most asymmetric states (ENLC–E and ENLC–EN; root mean square deviation of the Cα atoms in the two superposed subunits is >1), one PsnB subunit is a full enzyme–nucleotide–leader–core complex (ENLC), while the other subunit is an apo form (E) or is complexed with ADP (EN). In the other two states (EL–E and EL–EL) that are slightly asymmetric or almost symmetric (root mean square devia- tion of the Cα atoms is <1), the PsnB subunit has either one LP (EL) or nothing (E). Although nucleotides and LPs are sometimes found in both subunits of dimers, the CP is found only in one subunit of dimers (Supplementary Table 2), suggesting that the binding of the CP rather than of the LP or nucleotide contributes more to the dimeric asymmetry. The β13β14 loop is visible only in the LP-bound PsnB subunit, but the β6α3 loop is shown only in the CP-bound PsnB subunit (Supplementary Table 2).
To examine whether PsnB behaves as an asymmetric dimer in solution, we monitored the fraction of PsnB-bound Fl_MP with increasing amounts of unlabeled MP in the presence of saturating amounts of PsnB (10 µM) and ADP (1 mM). The simulation of two opposite models, in which the precursor–enzyme stoichiometry is

strictly 1:2 (complete asymmetry; only one PsnB subunit of dimer can bind to the precursor with a Kd of 0.13 µM) or 2:2 (complete symmetry; two subunits independently bind to the precursor with the same Kd), shows that the bound fraction falls from 1 at the approximate MP:PsnB ratio of 0.5 and 1 and reaches 0.25 and 0.5 at the MP:PsnB ratio of 2, respectively (Fig. 4b). However, the experi- mental bound fraction decreased slightly at a ratio of 0.5, like the complete asymmetry model, but converged to the curve of the com- plete symmetry model at high MP concentrations (Fig. 4b). This result can be explained by an intermediate model in which the two subunits of the dimer can bind to the precursor with different affini- ties. The two different affinities would result in negative cooperativ- ity of the enzyme–ligand interaction. Indeed, the data in Fig. 2d fit much better to the Hill curves with Hill coefficients of 0.75 and 0.79 for MP and LP, respectively (Supplementary Fig. 4).

A conserved DFR motif is critical for enzyme activity. Although our structures show that a dimer is the fundamental unit, we found that the LP-bound PsnB subunits (ENLC and EL) have distinct inter- dimeric interactions; the β13β14 loop in one subunit (A*) of the neigh- boring dimer, including the DFR233–235 residues, interacts with the β9β10 loop and a nucleotide in the synonymous subunit (A; Fig. 4c) or, if no nucleotide is present, occupies the nucleotide-binding site of subunit A (Supplementary Fig. 5a). In particular, Arg235 inter- acts with the phosphates of the nucleotide, Asp233 forms hydrogen bonds with nitrogens of Arg235 and Phe234 makes hydrophobic contacts with Ile227/Ile241/Phe292. Interestingly, the D(F/W)R motif is largely conserved in the ATP-grasp enzymes for graspetide biosynthesis (Fig. 4d and Supplementary Fig. 3c). RimK, GshB and PurT also have similar sequences (Supplementary Fig. 3c), but they do not bind to the nucleotide44,45,47. The alanine mutation on these residues resulted in the loss of macrocyclization activity, indicating that the DFR motif is critical for enzyme activity (Fig. 4e). We tested whether these interdimeric interactions lead to the formation of higher oligomers by using size-exclusion chromatography and fluo- rescence resonance energy transfer (FRET), but found no evidence for stable higher oligomers in solution (Supplementary Fig. 5b–d and Extended Data Fig. 7a–e). Therefore, we believe that the PsnB dimers do not form higher oligomers, and the functional impor- tance of the DFR233–235 motif can be explained by putative intra- molecular interactions (Extended Data Fig. 7f), as shown in other ATP-grasp enzymes45,49. Alternatively, the PsnB dimers may only transiently associate together for the reaction via the DFR motif.

LP Phe15 is important for enzyme binding. The enzyme–LP inter- face revealed a hydrophobic interaction between the hydrophobic


Leader (L)

Subunit A (E)

Subunit B (E)

Subunit B (E)




Core (C)

ENLC–E chain A, B and E of 7DRM


b c


Subunit B*

Subunit A*


Subunit A

Subunit B


0 0.5 1.0 1.5 2.0

Chain A, B, C and D of 7DRM (precursor is not shown)

Loop β9β10
of subunit A

d e –2 H O –1 H O –0 H O

2 2 2
DFR 0.5 h

DFR 5 h



3.0 Å



2.8 Å
3.1 Å

2.5 Å
2.9 Å

AFR 5 h

DAR 5 h

DFA 5 h

of subunit A*


3.0 Å

2.3 Å R235

AAA 5 h



I227 I241


Fig. 4 | Crystal structures of PsnB reveal asymmetric dimers with bound nucleotide and CP. a, Overall structure of a PsnB dimer (ENLC–E) in which PsnB (E; yellow or pale cyan cartoons) is complexed with nucleotide (N; cyan sticks), leader (L; magenta sticks) and core (C; green sticks). Loops β6α3, β9β10 and β13β14 are colored in blue. Polder OMIT maps (gray mesh) for the ADP, leader and core are contoured at 4σ. In subunit B, loops β6α3 and β13β14 are disordered. b, PsnB-bound MP fractions were measured by fluorescence anisotropy of the fluorophore-labeled MP (0.1 μM) with PsnB (10 µM) and increasing amounts of unlabeled MP. Data are presented as dot plots with mean ± 1 s.d. (n = 3 independent experiments). The simulation curves of two models, in which the MP:PsnB stoichiometry is either 1:2 or 2:2, are shown as red and blue solid lines, respectively. c, Loop β13β14 shows an interdimeric interaction with the neighboring subunit. Loop β13β14 from the subunit A* in the neighboring dimer (red sticks/cartoon) interacts with a nucleotide (cyan sticks), loop β9β10 and hydrophobic residues (yellow sticks) of subunit A. The atoms in sticks are colored in red (O), blue (N), orange (P) and light green (Mg2+). Hydrogen bonds are shown as red dashed lines. d, Sequence alignment of the DFR motif in representative enzymes that synthesize 12 groups of graspetides. e, MALDI spectra of the reactions with PsnB variants (0.4 µM; 40 µM MP) that contain alanine mutation(s) in the DFR motif.

pocket of PsnB (Tyr171/Ile193/Leu196/Val201/Phe203) and LP-Phe15, a putative cation-π interaction between PsnB-Arg192 and LP-Phe15 and an ionic interaction between PsnB-Arg181 and LP-Asp18 (Fig. 5a). In support of the divergence of the leader sequences among different groups of graspetides, many of these

residues are largely conserved only in the enzymes for group 2a graspetides (Supplementary Figs. 3c and 6a)28. A similar hydropho- bic interaction was also observed in the MdnC–MdnA structure (Leu196/Leu199/Phe206–Phe14)14. To test their contributions to the enzyme–leader interaction, we characterized the leader affinity,

a c

d e
WT R101A R213A

10 32 h
30 min
5 0.1 min

–1 –0 –2

–1 –0

–2 –1 –0


0 10
MP (µM)





Fig. 5 | Molecular interactions between PsnB and its precursor peptide substrate. a, Interaction scheme of LP (magenta sticks/cartoon) and PsnB (chains A and E of 7DRM (Protein Data Bank (PDB)). Five hydrophobic residues of PsnB (green sticks) form a hydrophobic pocket for Phe15 of LP. Arg192 (red sticks) of PsnB interacts with Phe15 of LP by hydrogen bond and cation-π interaction. Arg181 (blue sticks) of PsnB shows auxiliary electrostatic interaction with Asp18 of LP. b, The affinity to LP (bar graphs) and enzymatic activity (circles) of leader-binding site mutants. Affinity was determined by fluorescence anisotropy, and bar graphs represent the determined Kd and the error of fitting (see Supplementary Fig. 6b for data and fitting). The relative activities were estimated by MALDI spectra (see Supplementary Fig. 6d for MALDI spectra). c, Interaction scheme of CP (green sticks/cartoon) and PsnB (chains A and E of 7DRN). Glu37 and Glu38 of CP interact with Arg213 of PsnB and Arg101 from the neighboring subunit, respectively. The backbone amide of Gly36 of the core peptide interacts with Arg72 of PsnB. The carboxylate oxygen of Glu37, which forms the first ring, is located only 5.2 Å away from the γ-phosphate of AMPPNP. d,e, ATPase activity (d) and macrocyclization reaction (e) of the wild-type PsnB, R101A mutant or R213A mutant. Data are presented as dot plots with mean ± 1 s.d. (n = 3 independent experiments) and are fitted to a hyperbolic equation (d).

ATPase activity and macrocyclization activity of the alanine mutants in these residues; Leu196 and Phe203 in PsnB were critical for all these functions (fourfold to fivefold reduction in binding affinity and a >64-fold reduction in ATPase and macrocyclization activity for the alanine mutants), whereas other residues only mod- estly contribute (twofold to threefold reduction in binding affinity and intermediate levels of ATPase and macrocyclization activity for the alanine mutants; Fig. 5b and Supplementary Figs. 6b–d). The affinity of the precursor variants with a mutation (F15A, D18A or a Phe15 mutation to the pentafluorinated Phe15, F15(F5)F, which can substantially reduce the cation-π interaction50) indicates that Phe15 in the LP has an important role in the enzyme–LP interaction and that the cation-π and ionic interactions contribute only modestly (Supplementary Fig. 6e).

Conserved Arg213 in PsnB recognizes ring-forming glutamate. Inspection of the enzyme–core interface revealed distinct ionic and hydrogen bond interactions between the enzyme–CP residue pairs, Arg213–Glu37, Arg101–Glu38 and Arg72–Gly36 (Fig. 5c). Interestingly, Arg213 interacts with the carboxyl side chain of Glu37, which is the first residue to be phosphorylated and is located near the γ-phosphate of AMPPNP (5.2 Å), and Arg101 from the neighboring subunit binds to Glu38, which forms the second ring. Arg72 and Arg213 are highly conserved in ATP-grasp enzymes for 12 groups of graspetides, whereas Arg101 is conserved only in the enzymes for group 2a graspetides (Supplementary Fig. 3a), suggest- ing their different roles.

To examine the functional importance of these residues, we pre- pared leader-fused PsnB variants with the R72A, R101A or R213A mutations and found that all three arginine residues are required to tightly bind to the CP (Extended Data Fig. 8). We also purified the PsnB variants with the R101A or R213A mutation (we were unable to obtain the R72A variant) and found that Arg213 is critical for ATP consumption and the macrocyclization reaction, whereas Arg101 is largely dispensable for both activities (Fig. 5d,e). These results suggest that Arg213 has an important role in specifically recognizing the ring-forming carboxyl side chain. In agreement with this hypothesis, the crystal structure of PsnB with the precur- sor carrying the phosphomimetic Glu37 (MP(pE37)) revealed that Arg213 binds to Glu38 instead of pE37 (Extended Data Fig. 9a). In this structure, we could not observe any direct interaction with pE37. This partial interaction of the core peptide of MP(pE37) is supported by additional results that the affinity of MP(pE37) to PsnB is between those of the MP and the LP, and that MP(pE37) inhibits the macrocyclization reaction better than the LP (Extended Data Fig. 9b,c). It remains elusive, however, how the hydroxyl nucleophile of Thr32 substitutes the phosphate to form an ester linkage.

Substrate binding induces a conformational change in PsnB. Comparison of different PsnB structures revealed conformational changes in PsnB following binding of the precursor or a nucleotide (Fig. 6a and Supplementary Video 1). Leader binding moves the α8α9 helices of the leader-binding domain to the dimeric center by
2.6 Å, generates a hydrophobic pocket to accommodate Phe15 in the

a b

2.4 Å

Fig. 6 | Conformational changes in the PsnB dimer during substrate binding. a, Superposition of five PsnB subunits (E, empty PsnB, pale cyan; EN, enzyme–nucleotide complex, teal; EL, enzyme–leader complex, pink (from EL–E dimer) or orange (from EL–EL dimer); ENLC, enzyme–nucleotide–leader– core holo-complex, magenta). Leader binding moves the α8α9 helices by 2.6 Å, nucleotide binding shifts the β9β10 loop by 2.3 Å and core binding attracts the α3α4 helices by 2.4 Å. The β6α3 loop is ordered only when the CP is bound to PsnB. b, Conformational change in PsnB in the dimeric context. Two pairs of α3α4 helices in the PsnB dimer move together during the 2.4-Å shift from symmetric dimer (EL from EL–EL; orange) to asymmetric dimer (ENLC and
E; magenta and teal, respectively). c, Suggested model for substrate recognition of PsnB. The binding of leader and ATP induces distinct conformational changes in PsnB, resulting in enhancement of the core binding. The core binding via conserved Arg213 generates the asymmetric dimer with a compact active site in core-bound PsnB.

LP and relocates Arg195 out of the entry of the pocket (Extended Data Fig. 10a). Nucleotide binding moves the β9β10 loop toward the nucleotide by 2.3 Å via Lys172 and Thr180 (Extended Data Fig. 10b). The binding of both LP and CP shifts the α3α4 helices toward the active site by 2.4 Å (Fig. 6a,b). The α3α4 helices extensively interact with the same helices in the neighboring subunit, and, there- fore, the movement of the rigid body of the two α3α4 pairs appears to induce dimeric asymmetry (Fig. 6b and Extended Data Fig. 10c). The β6α3 loop becomes ordered only after the CP binds to PsnB and tightly packs the active site (Extended Data Fig. 10d). The overall structures of the CPs from multiple subunits overlap well with each other (Extended Data Fig. 10e). Collectively, the overall conforma- tional change of PsnB following precursor and nucleotide binding establishes the space for the CP and makes the active site compact.
By using quantitative biochemical analyses and core-bound three-dimensional structures, we obtained a glimpse of how the ATP-grasp enzyme in plesiocin biosynthesis specifically recognizes CP, coordinates all relevant molecular interactions and changes its

conformation before the macrocyclization reaction. We suggest a model for substrate recognition (Fig. 6c), where LP binds to PsnB mainly by hydrophobic interactions using LP Phe15. PsnB binds to ATP via several conserved residues. LP binding stabilizes the con- served DFR motif, which binds to ATP. The conformational changes associated with the binding of LP and ATP generate the proper space for the CP, in which the ring-forming side chain carboxylate is specifically recognized and relocated to the site near ATP by the highly conserved Arg213 in the enzyme. CP binding shifts the two α3α4 pairs at the dimeric center toward the CP and stabilizes the β6α3 loop, which results in close packing of the active site and the forma- tion of the asymmetric dimer, in which only one subunit forms the enzyme–ATP–LP–CP holo-complex. The Arg213-bound carboxyl- ate is then ready to move to ATP for phosphorylation. The presence of these features in enzymes for other graspetide groups remains unknown, but the high conservation of Arg213 and the D(F/W)R motif in graspetide macrocyclases indicates that their functions are conserved in graspetide biosynthesis.
The molecular details of the precursor recognition in plesiocin biosynthesis differ at several points from those in microviridin

biosynthesis: (1) the CP and nucleotide also contribute to the enzyme–precursor interaction, and (2) the enzyme forms an asym- metric dimer as an active state. The enhanced enzyme–precursor interaction by CP is most likely due to the interactions between the conserved arginine residues in the enzyme and the glutamate residues in the CP. Although Arg213 is also conserved in enzymes for microviridins, this affinity-enhancing effect was not observed, suggesting the presence of another affinity-lowering factor in this enzyme. Because we observed a symmetric PsnB dimer with two LPs (EL–EL), we suggest that the reported symmetric MdnC–MdnA structure presents only one possible conformation. Collectively, plesiocin biosynthesis appears to be a better model system for understanding substrate recognition in graspetide biosynthesis.
Previously, we reported that ATP-grasp enzymes have high spec- ificity only to the cognate core motif and thus are not cross-reactive to CPs in other graspetide groups28. Our structures support this high substrate specificity for the following reasons. (1) The core-bound structure shows a highly compact active site with well-overlapped conformations of the CPs. The β6α3 loop, which is much shorter in MdnC, is ordered only in the CP-bound PsnB and therefore generates a highly compact active site of PsnB. (2) Arg101, which binds to Glu38 that forms the second ring, is highly conserved only in enzymes for group 2a graspetides. It is likely that enzymes in other graspetide groups have different active-site environments to recognize their cognate core motif. Therefore, the construction of graspetide-like libraries with different multicyclic architectures may require diverse enzymes that produce different groups of graspetides.
There are still several unsolved questions regarding the macro- cyclization reaction: what residues are involved in the phosphoryla- tion of glutamate and the nucleophilic addition of the ring-forming threonine to generate the macrolactone linkage? Although our structure with the phosphomimetic precursor did not reveal resi- dues that help the nucleophilic addition, our results suggest that two threonine residues are also involved in the phosphorylation of glutamate or that phosphorylation is coupled to nucleophilic addi- tion. Additional structures that present snapshots of phosphoryla- tion or nucleophilic addition may help to understand the molecular mechanisms of these steps.
Our data present a rare example of the CP-bound structure of enzymes involved in RiPP biosynthesis. We believe that this is pos- sible because of the well-defined MP and the contribution of CP to enzyme binding. Because the MP, PsnA214–38, is short enough for solid-phase peptide synthesis, we believe that chemical syn- thesis of intermediate mimics as well as mutant enzymes that may stabilize the enzyme–intermediate complex may help to further dissect the macrocyclization process. The molecular details of the macrocyclization reaction will not only improve our understand- ing of graspetide biosynthesis but will also help the construction of graspetide-like peptide libraries to explore diverse biological func- tions for therapeutics and biotechnology.
Online content
Any methods, additional references, Nature Research report- ing summaries, source data, extended data, supplementary infor- mation, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at s41589-021-00855-x.
Received: 1 March 2021; Accepted: 8 July 2021;

1. Newman, D. J. & Cragg, G. M. Natural products as sources of new drugs from 1981 to 2014. J. Nat. Prod. 79, 629–661 (2016).

2. Arnison, P. G. et al. Ribosomally synthesized and post-translationally modified peptide natural products: overview and recommendations for a universal nomenclature. Nat. Prod. Rep. 30, 108–160 (2013).
3. Cimermancic, P. et al. Insights into secondary metabolism from a global analysis of prokaryotic biosynthetic gene clusters. Cell 158, 412–421 (2014).
4. Montalban-Lopez, M. et al. New developments in RiPP discovery, enzymology and engineering. Nat. Prod. Rep. 38, 130–239 (2021).
5. Ortega, M. A. et al. Structure and mechanism of the tRNA-dependent lantibiotic dehydratase NisB. Nature 517, 509–512 (2015).
6. Koehnke, J. et al. Structural analysis of leader peptide binding enables leader-free cyanobactin processing. Nat. Chem. Biol. 11, 558–563 (2015).
7. Regni, C. A. et al. How the MccB bacterial ancestor of ubiquitin
E1 initiates biosynthesis of the microcin C7 antibiotic. EMBO J. 28, 1953–1964 (2009).
8. Evans, R. L. III, Latham, J. A., Xia, Y., Klinman, J. P. & Wilmot, C. M. Nuclear magnetic resonance structure and binding studies of PqqD, a chaperone required in the biosynthesis of the bacterial dehydrogenase cofactor pyrroloquinoline quinone. Biochemistry 56, 2735–2746 (2017).
9. Sumida, T., Dubiley, S., Wilcox, B., Severinov, K. & Tagami, S. Structural basis of leader peptide recognition in lasso peptide biosynthesis pathway. ACS Chem. Biol. 14, 1619–1627 (2019).
10. Chekan, J. R., Ongpipattanakul, C. & Nair, S. K. Steric complementarity directs sequence promiscuous leader binding in RiPP biosynthesis. Proc. Natl Acad. Sci. USA 116, 24049–24055 (2019).
11. Ghilarov, D. et al. Architecture of microcin B17 synthetase: an octameric protein complex converting a ribosomally synthesized peptide into a DNA gyrase poison. Mol. Cell 73, 749–762 (2019).
12. Grove, T. L. et al. Structural insights into thioether bond formation in the biosynthesis of sactipeptides. J. Am. Chem. Soc. 139, 11734–11744 (2017).
13. Davis, K. M. et al. Structures of the peptide-modifying radical SAM enzyme SuiB elucidate the basis of substrate recognition. Proc. Natl Acad. Sci. USA 114, 10420–10425 (2017).
14. Li, K., Condurso, H. L., Li, G., Ding, Y. & Bruner, S. D. Structural basis for precursor protein-directed ribosomal peptide macrocyclization. Nat. Chem. Biol. 12, 973–979 (2016).
15. Chekan, J. R., Estrada, P., Covello, P. S. & Nair, S. K. Characterization of the macrocyclase involved in the biosynthesis of RiPP cyclic peptides in plants. Proc. Natl Acad. Sci. USA 114, 6551–6556 (2017).
16. Ludewig, H. et al. Characterization of the fast and promiscuous macrocyclase from plant PCY1 enables the use of simple substrates. ACS Chem. Biol. 13, 801–811 (2018).
17. Song, H. et al. A molecular mechanism for the enzymatic methylation of nitrogen atoms within peptide bonds. Sci. Adv. 4, eaat2720 (2018).
18. Ongpipattanakul, C. & Nair, S. K. Molecular basis for autocatalytic backbone N-methylation in RiPP natural product biosynthesis. ACS Chem. Biol. 13, 2989–2999 (2018).
19. Bothwell, I. R. et al. Characterization of glutamyl-tRNA-dependent dehydratases using nonreactive substrate mimics. Proc. Natl Acad. Sci. USA 116, 17245–17250 (2019).
20. Dong, S. H., Liu, A., Mahanta, N., Mitchell, D. A. & Nair, S. K. Mechanistic basis for ribosomal peptide backbone modifications. ACS Cent. Sci. 5, 842–851 (2019).
21. Thibodeaux, G. N., McClerren, A. L., Ma, Y., Gancayco, M. R. &
van der Donk, W. A. Synergistic binding of the leader and core peptides by the lantibiotic synthetase HalM2. ACS Chem. Biol. 10, 970–977 (2015).
22. Cogan, D. P. et al. Structural insights into enzymatic [4+2] aza-cycloaddition in thiopeptide antibiotic biosynthesis. Proc. Natl Acad. Sci. USA 114, 12928–12933 (2017).
23. Ishitsuka, M. O., Kusumi, T., Kakisawa, H., Kaya, K. & Watanabe, M. M. Microviridin—a novel tricyclic depsipeptide from the toxic cyanobacterium Microcystis viridis. JACS 112, 8180–8182 (1990).
24. Ziemert, N., Ishida, K., Liaimer, A., Hertweck, C. & Dittmann, E. Ribosomal synthesis of tricyclic depsipeptides in bloom-forming cyanobacteria. Angew. Chem. Int. Ed. Engl. 47, 7756–7759 (2008).
25. Philmus, B., Christiansen, G., Yoshida, W. Y. & Hemscheidt, T. K.
Post-translational modification in microviridin biosynthesis. Chembiochem 9, 3066–3073 (2008).
26. Lee, H., Park, Y. & Kim, S. Enzymatic cross-linking of side chains generates a modified peptide with four hairpin-like bicyclic repeats. Biochemistry 56, 4927–4930 (2017).
27. Roh, H., Han, Y., Lee, H. & Kim, S. A topologically distinct modified peptide with multiple bicyclic core motifs expands the diversity of microviridin-like peptides. Chembiochem 20, 1051–1059 (2019).
28. Lee, H., Choi, M., Park, J. U., Roh, H. & Kim, S. Genome mining reveals high topological diversity of ω-ester-containing peptides and divergent evolution of ATP-grasp macrocyclases. J. Am. Chem. Soc. 142, 3013–3023 (2020).
29. Unno, K. & Kodani, S. Heterologous expression of cryptic biosynthetic gene cluster from Streptomyces prunicolor yields novel bicyclic peptide prunipeptin. Microbiol. Res. 244, 126669 (2021).

30. Ahmed, M. N. et al. Phylogenomic analysis of the microviridin biosynthetic pathway coupled with targeted chemo-enzymatic synthesis yields potent protease inhibitors. ACS Chem. Biol. 12, 1538–1546 (2017).
31. Weiz, A. R. et al. Harnessing the evolvability of tricyclic microviridins to dissect protease–inhibitor interactions. Angew. Chem. Int. Ed. Engl. 53, 3735–3738 (2014).
32. Reyna-Gonzalez, E., Schmid, B., Petras, D., Sussmuth, R. D. & Dittmann, E. Leader peptide-free in vitro reconstitution of microviridin biosynthesis enables design of synthetic protease-targeted libraries. Angew. Chem. Int. Ed. Engl. 55, 9398–9401 (2016).
33. Lee, C., Lee, H., Park, J. U. & Kim, S. Introduction of bifunctionality into the multidomain architecture of the ω-ester-containing peptide plesiocin. Biochemistry 59, 285–289 (2020).
34. Zhang, Y. et al. A distributive peptide cyclase processes multiple microviridin core peptides within a single polypeptide substrate. Nat. Commun. 9, 1780 (2018).
35. Tanaka, T., Nishioka, T. & Oda, J. Nicked multifunctional loop of glutathione synthetase still protects the catalytic intermediate. Arch. Biochem. Biophys. 339, 151–156 (1997).
36. Shi, Y. & Walsh, C. T. Active site mapping of Escherichia coli D-Ala-D-Ala ligase by structure-based mutagenesis. Biochemistry 34, 2768–2776 (1995).
37. Lipmann, F. & Tuttle, L. C. A specific micromethod for the determination of acyl phosphates. J. Biol. Chem. 159, 21–28 (1945).
38. Jencks, W. P. The mechanism of the reaction of hydroxylamine with activated acyl groups. Biochim. Biophys. Acta 27, 417–418 (1958).
39. Fawaz, M. V., Topper, M. E. & Firestine, S. M. The ATP-grasp enzymes.
Bioorg. Chem. 39, 185–191 (2011).
40. Weiz, A. R. et al. Leader peptide and a membrane protein scaffold guide the biosynthesis of the tricyclic peptide microviridin. Chem. Biol. 18, 1413–1421 (2011).

41. Oman, T. J., Knerr, P. J., Bindman, N. A., Velasquez, J. E. & van der Donk, W. A. An engineered lantibiotic synthetase that does not require a leader peptide on its substrate. J. Am. Chem. Soc. 134, 6952–6955 (2012).
42. Dunbar, K. L. & Mitchell, D. A. Insights into the mechanism of peptide cyclodehydrations achieved through the chemoenzymatic generation of amide derivatives. J. Am. Chem. Soc. 135, 8692–8701 (2013).
43. Wang, H. & van der Donk, W. A. Biosynthesis of the class III lantipeptide catenulipeptin. ACS Chem. Biol. 7, 1529–1535 (2012).
44. Zhao, G. et al. Structure and function of Escherichia coli RimK, an ATP-grasp fold, L-glutamyl ligase enzyme. Proteins 81, 1847–1854 (2013).
45. Hara, T., Kato, H., Katsube, Y. & Oda, J. A pseudo-Michaelis quaternary complex in the reverse reaction of a ligase: structure of Escherichia coli
B glutathione synthetase complexed with ADP, glutathione, and sulfate at
2.0 Å resolution. Biochemistry 35, 11967–11974 (1996).
46. Ouchi, T. et al. Lysine and arginine biosyntheses mediated by a common carrier protein in Sulfolobus. Nat. Chem. Biol. 9, 277–283 (2013).
47. Thoden, J. B., Firestine, S., Nixon, A., Benkovic, S. J. & Holden, H. M. Molecular structure of Escherichia coli PurT-encoded glycinamide ribonucleotide transformylase. Biochemistry 39, 8791–8802 (2000).
48. Batson, S. et al. Inhibition of D-Ala:D-Ala ligase through a phosphorylated form of the antibiotic D-cycloserine. Nat. Commun. 8, 1939 (2017).
49. Fan, C., Moews, P. C., Walsh, C. T. & Knox, J. R. Vancomycin resistance: structure of D-alanine:D-alanine ligase at 2.3 Å resolution. Science 266, 439–443 (1994).
50. Salwiczek, M., Nyakatura, E. K., Gerling, U. I., Ye, S. & Koksch, B. Fluorinated amino acids: compatibility with native protein structures and effects on protein–protein interactions. Chem. Soc. Rev. 41, 2135–2171 (2012).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature America, Inc. 2021

General materials and methods. Reagents for cloning were purchased from Enzynomics or Toyobo. Oligonucleotides were purchased from Macrogen. Escherichia coli DH10β and BL21(DE3) strains were used for cloning and protein overexpression, respectively. Protein and peptide concentrations were determined by ultraviolet absorbance at 280 nm. Amino acids, coupling reagents and resins for

MgCl2 (10 mM) and KCl (50 mM). Reaction mixtures (47.5 μl) without ATP were transferred into a 384-well microplate (SPL life science). Then, 2.5 μl of ATP (100 mM) was supplied just before measurement. After ATP was added,
absorbance at 340 nm was measured using an Infinite M200pro (Tecan) equipped with i-Control software. The rate of ATP hydrolysis was calculated from the following equation:

peptide synthesis were purchased from GL-Biochem. For protein purification, Ni

( )
−1 −1

dA340 OD −1 −1

sepharose 6 FastFlow beads were obtained from GE Healthcare. Anion exchange
chromatography was performed using an ÄKTA pure system (GE Healthcare)

ATP consumption rate enz min = − dt min × Kpath × [enz]

on a MonoQ 5/50 GL column (GE Healthcare). Sample purification and analysis with HPLC were performed using an Agilent 1260 Infinity on a ZORBAX
SB-C18 semipreparative column (9.4 × 250 mm, 5-μm particle size; Agilent) and a ZORBAX SB-C18 analytical column (4.6 × 250 mm, 5-μm particle size; Agilent), respectively. Trifluoroacetic acid (TFA; 0.05% (vol/vol)) in water (Solvent A) and 0.05% (vol/vol) TFA in acetonitrile (CH3CN; Solvent B) were used as mobile phase for HPLC. Mass analysis was performed using an Ultraflextreme MALDI–TOF/ TOF mass spectrometer (Bruker Daltonics).
Cloning, overexpression and purification of PsnB variants. Plasmids and oligonucleotides used in this study are listed in Supplementary Tables 3 and 4, respectively. Plasmids expressing PsnB variants were constructed using an inverse PCR method51. PsnB and its variants were expressed and purified as previously described33.

Peptide synthesis. MP (PsnA214–38), LP (PsnA214–24), CP (PsnA228–38) and its variants were synthesized by solid-phase peptide synthesis. Wang resin
(200 μmol, 208 mg) was washed and soaked with 5 ml of 1:1 dimethylformamide (DMF) and dichloromethane (DCM) for 20 min in a reaction vessel for
resin swelling. Attachment of the first amino acid was performed by N,N′- diisopropylcarbodiimide (DIC) coupling. The solution of Fmoc-amino acid-OH (5 equiv.), 4-dimethylaminopyridine (DMAP; 0.1 equiv.) and DIC (4 equiv.) in
2–3 ml of DMF was added to the reaction vessel carrying the resin and incubated at room temperature for 2–3 d. Resins were washed three times with DMF and DCM and were capped by acetic anhydride; acetic anhydride (10 equiv.) and DMAP
(0.1 equiv) in 2 ml of DCM was incubated with resins for 1 h. After capping, resins were washed with DMF and DCM again and mixed with 20% (vol/vol) piperidine in DMF for 30 min to cleave Fmoc. After the DMF and DCM wash, the first amino acid-bound resins went through sequential coupling–capping–deprotection.
Coupling was performed by incubating the resins with reaction solution that contained Fmoc-amino acid-OH (5 equiv.) and HATU/N,N-diisopropylethylamine (DIEA; 10/5 equiv.) dissolved in DMF. Then, resins were treated with capping solution (DMF/acetic anhydride/DIEA in a 9:1:0.05 ratio) for 7 min. For the deprotection of Fmoc, resins were mixed with 20% piperidine in DMF for 30 min.
Resins were washed with DMF and DCM three times between each step. All the remaining amino acids were sequentially attached by repeating coupling, capping and deprotection. The completion of each step was monitored by tests with ninhydrine or 2,4,6-trinitrobenzenesulfonic acid. After completion of the synthesis, peptides were detached from the resins by incubating the resins with cleavage cocktail (95% TFA, 2.5% deionized water, 2.5% triisopropyl silane) for 3 h. Cleavage solutions were then evaporated and mixed with tenfold excess of ether:hexane solution (1:1) for peptide precipitation. The precipitated peptides
were collected by centrifugation at 4,000g for 15 min, dried, dissolved in DMSO, purified by HPLC and lyophilized. For the synthesis of fluorophore-labeled peptides, 5(6)-carboxyfluorescein (Acros) and serine (as a linker) were attached to the N termini of the peptides.

Chemical synthesis of the phosphomimetic glutamate. The phosphomimetic glutamate (pGlu or pE) contains acyl-difluoro-phosphonate (CO-CF2PO 2−) instead of acyl-phosphate (CO-OPO 2-). Synthesis of the phosphomimic glutamate is described in Supplementary Fig. 2 and in the Supplementary Note. Briefly, benzyl lithiodifluoromethylphosphonate was added to Fmoc-protected glutamic anhydride to furnish phosphomimetic glutamate52. Peptide synthesis with
the phosphomimetic glutamate was performed in the same fashion as normal peptides.
Fluorescence anisotropy. Fluorescence anisotropy was performed as follows: a dye-labeled peptide (0.1 μM) and twofold serial dilutions of PsnB variants
(starting at 100–500 μM) were mixed in a buffer containing Tris-HCl (100 mM, pH 7.3), MgCl2 (10 mM) and KCl (50 mM) at 37 °C. Fluorescence (excitation,
485 nm; emission, 535 nm; parallel or perpendicular) was monitored after a 1-min
incubation by using an Infinite F200pro (Tecan). Data analysis was performed using GraphPad Prism.
ATPase assay. ATPase assays were performed as previously described53. Briefly, precursor peptide variants (0–200 μM) and PsnB variants (single concentrations at 0.4–2 μM) were co-incubated at 37 °C in a buffer containing ATP (5 mM; Sigma-Aldrich), NADH (400 μM; Sigma-Aldrich), pyruvate kinase (20 U ml–1; Sigma-Aldrich), L-lactic dehydrogenase (20 U ml–1; Sigma-Aldrich), phosphoenolpyruvate (3 mM; Sigma-Aldrich), Tris-HCl (100 mM, pH 7.3),

Here, Kpath is the molar absorption coefficient for NADH for a given optical path length, OD is the optical density and Kpath is equal to 6.67 × 102 for a 50-μl-well fill volume. The rates were normally corrected for background NADH decomposition of controls containing neither enzyme nor peptide. Data analysis was performed using GraphPad Prism.
In vitro reaction and analysis of PsnA2 variants. In vitro reactions were performed at 37 °C in solutions containing precursor peptide variants (40–100 μM),
PsnB variants (0.5–5 μM), ATP (5 mM), Tris-HCl (100 mM, pH 7.3), MgCl2
(10 mM) and KCl (50 mM). At the designated time points, reaction solutions were quenched by adding the same volume of 10% (vol/vol) TFA. After quenching, peptides were desalted using a C18 zip-tip (Millipore) and were analyzed by MALDI–MS. The MS profiles are provided in Supplementary Table 5.
Trapping the acyl-phosphate intermediate. Reaction conditions were the same as the in vitro reaction conditions except for the addition of hydroxylamine (0.5 M, final concentration) at the start of the reaction. The reaction was quenched after
a 4-h incubation. Hydroxylamine-trapped peptides were monitored by MALDI, purified by HPLC and further analyzed by MALDI–MS/MS. The MS/MS profiles are provided in Supplementary Table 5. Methanolysis to confirm the connectivity of the hydroxylamine-trapped peptides with one ester bond (2-c) was performed as previously described26.

Crystallization, data collection and crystallographic analysis. For PsnB–MP–ADP (PDB: 7DRM), PsnB–MP–AMPPNP (PDB: 7DRN) and PsnB–MP(pE37)–ADP
(PDB: 7DRP), purified PsnB (3.8 mg ml–1) was mixed with nucleotide (5 mM), MgCl2 (5 mM) and 1.2 equiv. of MP (PsnA214–38) or phosphomimetic precursor (PsnA214–38(pE37)). Initial screening was performed in a vapor diffusion and sitting drop format. Small crystals were identified in a reservoir condition of 0.1 M sodium cacodylate, pH 6.5, 0.1 M calcium acetate and 12% PEG8000. pH, salt and precipitant optimization were performed in hanging drop screens. Final reservoir conditions were as follows: PsnB–MP–ADP, 0.1 M sodium acetate pH 5.4 and 7% PEG3350; PsnB–MP–AMPPNP, 0.1 M sodium acetate pH 5.2 and 6% PEG3350; PsnB–MP(pE37)–ADP, 0.1 M sodium acetate pH 5.0 and 3% PEG3350. Crystals of the PsnB complex were grown at 20 °C. For PsnB–MP (PDB: 7DRO), purified PsnB (7.6 mg ml–1) was mixed with ADP (1 mM), MgCl2 (1 mM) and 1.2 equiv. of MP (PsnA214–38). Initial screening was performed in the same format, and small crystals were identified in a reservoir condition of 0.2 M potassium citrate tribasic monohydrate and 20% (wt/vol) PEG3350. The same optimization processes were done to obtain optimized protein crystals. The final reservoir condition was 10% tacsimate pH 7.6 and 14% PEG3350. Crystals were frozen in liquid nitrogen, and diffraction data were collected on the beamline 7A of the Pohang Accelerator Laboratory at a wavelength of 0.97934 Å. Data were processed by using HKL2000 (ref. 54). Molecular replacement was performed using Phaser55, whereby the structure of MdnC (PDB: 5IG9) was used as a model. Refinement with Phenix56
along with manual model rebuilding and ligand placement with COOT57 produced the final models. Polder OMIT maps were calculated for LPs, CPs and nucleotides in Phenix58. Structural figures were prepared with PyMol (
Binding stoichiometry. Binding stoichiometry between PsnB and MP was measured by fluorescence anisotropy of fluorophore-labeled MP (Fl_MP). Fl_MP (0.1 μM) and PsnB (10 μM) were co-incubated with 0–20 μM of non-labeled MP in buffer containing 100 mM Tris-HCl (pH 7.3), 50 mM KCl, 10 mM MgCl2 and 1 mM ADP at 37 °C. The binding fraction of Fl_MP was measured with the different concentrations of non-labeled MP. Totally symmetric and totally asymmetric binding models were simulated with the Kd of 0.13 μM.
Size-exclusion chromatography. Size-exclusion chromatography was performed using an ÄKTA pure system (GE Healthcare) on a Superdex 200 column (GE Healthcare). Three types of solution were used as an eluent: (1) 20 mM Tris-HCl (pH 8.0), 100 mM NaCl solution; (2) 20 mM Tris-HCl (pH 8.0), 100 mM NaCl,
10 mM MgCl2 solution; and (3) 20 mM Tris-HCl (pH 8.0), 100 mM NaCl, 10 mM MgCl2, 1 mM ATP solution.
Dye labeling and FRET. Purified PsnB was incubated with 10 mM Tris and 10 mM DTT for 1 h and then purified by gel filtration using fresh 10 mM Tris
buffer. At first, labeling with phenylmaleimide was performed to monitor which cysteine residues of PsnB were preferentially labeled. Reduced PsnB (50–100 μM) and 5 equiv. of phenylmaleimide were mixed in 10 mM Tris, pH 8.0. After a 2-h incubation at room temperature, the reaction was quenched with excess DTT.

Labeled PsnB was cleaved by adding trypsin (trypsin/PsnB 1:100 (wt/wt)) and analyzed by MALDI–MS/MS. Alexa Fluor 555 and Alexa Fluor 647 (Thermo Fisher Scientific) were covalently linked to native cysteine residues of PsnB in the same condition of phenylmaleimide labeling. After dye labeling, dye-labeled PsnB was purified by size-exclusion chromatography. FRET assays for enzyme assembly were performed with 0.2 μM donor- and acceptor-labeled PsnB using a microplate reader (excitation, 520 nm; emission, 570 or 670 nm; Infinite F200pro, Tecan).

Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Coordinates and structure factors for the reported crystal structures in this work were deposited in the RCSB PDB under accession numbers 7DRM (MP- and ADP-bound PsnB), 7DRN (MP- and AMPPNP-bound PsnB), 7DRP (MP(pE37)- and ADP-bound PsnB) and 7DRO (MP-bound PsnB). Source data are provided with this paper.
51. Hemsley, A., Arnheim, N., Toney, M. D., Cortopassi, G. & Galas, D. J. A simple method for site-directed mutagenesis using the polymerase chain reaction. Nucleic Acids Res. 17, 6545–6551 (1989).
52. Bouwman, S., Orru, R. V. & Ruijter, E. Stereoselective synthesis of fluorinated aminoglycosyl phosphonates. Org. Biomol. Chem. 13, 1317–1321 (2015).
53. Sausen, C. W., Rogers, C. M. & Bochman, M. L. Thin-layer chromatography and real-time coupled assays to measure ATP hydrolysis. Methods Mol. Biol. 1999, 245–253 (2019).
54. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326 (1997).
55. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).
56. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213–221 (2010).
57. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics.
Acta Crystallogr. D 60, 2126–2132 (2004).
58. Liebschner, D. et al. Polder maps: improving OMIT maps by excluding bulk solvent. Acta Crystallogr. D 73, 148–157 (2017).

59. Ko, J. et al. The FALC-Loop web server for protein loop modeling.
Nucleic Acids Res. 39, W210–W214 (2011).
60. Park, H., Lee, G. R., Heo, L. & Seok, C. Protein loop modeling using a new hybrid energy function and its application to modeling in inaccurate structural environments. PLoS ONE 9, e113811 (2014).

We thank H. Lee, G. Eom, C. Lee, H. Roh, H. Cho and H. Park for helpful discussions and technical assistance, H. Chung, Y.J. Lee and T.Y. Im for assistance in synthesizing compounds, Y.T. Kim, W.J. Jeong and Y. Choi for advice in crystallization and crystallographic analysis and H. Woo for help in loop modeling. We also thank the staff of the 5C and 7A beamlines at the Pohang Light Source. This research was supported by the Basic Science Research Program through the National Research Foundation
of Korea (NRF) funded by the Ministry of Education (NRF-2020R1F1A1054191 and 2021R1A2C1008730) to S.K.

Author contributions
I.S. and S.K. conceived the project, designed experiments, analyzed data and wrote the paper. S.K. supervised the project. I.S. performed the majority of experiments.
Y.K. performed a part of the mutant analysis. I.S. and J.Y. conducted crystallographic experiments under the supervision of W.J.S. and S.K. I.S. and S.Y.G. synthesized the phosphomimetic glutamate under the supervision of H.G.L. and S.K.

Competing interests
The authors declare no competing interests.

Additional information
Extended data is available for this paper at
Supplementary information The online version contains supplementary material available at
Correspondence and requests for materials should be addressed to S.K.
Peer review information Nature Chemical Biology thanks Jesko Koehnke and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Reprints and permissions information is available at

Extended Data Fig. 1 | Designed minimal precursor recapitulates the reactivity of PsnB. a, Sequence logo of precursors of Group 2a graspetides including the leader peptide (red region) and one core motif (blue region). The LFIEDL region is highly conserved in Group 2a graspetides (yellow box). b,c, ATPase assays titrating with ATP (b) in solutions containing 0.4 μM PsnB, 200 μM minimal precursor (MP), 100 mM Tris pH 7.3, 50 mM KCl, and 10 mM MgCl2,
or titrating with MgCl2 (c) in solutions containing 0.4 μM PsnB, 200 μM MP, 100 mM Tris pH 7.3, 50 mM KCl, and 5 mM ATP. Data are presented as dot plots with mean ± 1 SD (n = 3 independent experiments) and fitted to a hyperbolic equation. d, Minimal precursor was successfully modified by PsnB with a modification rate similar to that of wild-type PsnA2. 50 μM MP (top panel; the product contains two ester bonds) or wild-type PsnA2 (bottom panel; the product contains eight ester bonds) was co-incubated with 0.5 μM PsnB in buffer A (100 mM Tris pH 7.3, 50 mM KCl, 5 mM ATP, and 10 mM MgCl2) at
37 °C, and the reaction solutions at designated time points were analyzed by MALDI-MS.

Extended Data Fig. 2 | Acyl-phosphate intermediates were trapped by hydroxylamine. a, Scheme of the acyl-phosphate trapping during the macrocyclization reaction. Nucleophilic attack of hydroxylamine rather than the core threonine generates the hydroxylamine adducts of the precursor.
b, MALDI analysis of the reaction solution with (left) or without (right) hydroxylamine. 0.5 M hydroxylamine was added to the reaction mixture containing 100 μM MP (1-a) and 6 μM PsnB in buffer A. The reaction mixtures were analyzed by MALDI after 4 hour incubation at 37 °C. Co-incubation of 0.5 M hydroxylamine generated NH2OH-added precursor peptides (1-b and 2-b), which are the result of nucleophilic attack of hydroxylamine to acyl-phosphate intermediates. c, 2-b was purified by HPLC and methanolysis was performed with purified 2-b as previously reported26. The methanolysis product (2-c) was detected by MALDI. d, MALDI-MS/MS analysis of three hydroxylamine adducts. The connectivities of ester bonds were determined by MS/MS analysis with NH2OH-added precursor peptides (1-b and 2-b) and methanolysis product (2-c).

Extended Data Fig. 3 | Binding of the leader peptide activates PsnB. a, LP (PsnA214-24) enhances the affinity of the CP (PsnA228-38) to PsnB. Fraction bound of Fl_CP (0.1 µM) to PsnB (50 µM) was determined by fluorescence anisotropy. b, LP enhances the ATPase activity of PsnB. Basal ATP consumption rate of PsnB was 0.14 min-1enz-1, whereas the addition of 200 μM LP increased the ATP consumption rate to 0.89 min-1enz-1. P value < 0.01 by two-sided Student’s t-test. c, Fluorescence anisotropy of Fl_LP (0.1 µM) to wild-type PsnB or leader-fused PsnB (LP_PsnB). Data are presented as dot plots with mean ± 1 SD
(n = 3 independent experiments; a-c) and fitted to a hyperbolic equation (c).

Extended Data Fig. 4 | Binding and modification of ring-containing precursors. a,b, Affinity of ring-containing MPs (MP-1H2O, the single-ring intermediate; MP-2H2O, the double-ring product) to PsnB was determined by fluorescence anisotropy without nucleotide (a) or with 1 mM ADP (b).
c, ATPase activity of PsnB was measured with different concentrations of MP or the single-ring intermediate (MP-1H2O). Data are presented as dot plots with mean ± 1 SD (n = 3 independent experiments; a-c) and fitted to a hyperbolic equation (a-c).

Extended Data Fig. 5 | Four crystal structures of PsnB complexes. Structure of four different states of PsnB dimers (E, enzyme, yellow or pale cyan cartoon; N, nucleotide, cyan sticks; L, leader, magenta sticks; C, core, green sticks). Polder OMIT map (gray mesh) of each peptide or nucleotide is contoured at 4.0 σ.

Extended Data Fig. 6 | Polder OMIT maps that are calculated without the model for LP, CP, or nucleotide (contour level at 4.0 σ). For the LP and CP, resolved residues are listed on Supplementary Table 2. Phospho-mimetic side-chain in CP is only resolved well in chain E of 7DRP.

Extended Data Fig. 7 | No higher oligomers were observed from PsnB. a, Gel-filtration chromatogram of PsnB without MgCl2 (black solid line), with MgCl2 (red solid line), or with both MgCl2 and ATP (blue solid line). A chromatogram of molecular weight standard (black dashed line) is also shown with known molecular weights. Addition of nucleotide did not induce the formation of the stable higher oligomer of PsnB dimer. b‒e, FRET of a solution containing both donor- and acceptor-labeled PsnB was measured with different amounts of MgCl2 (b), AMPPNP (c), LP (d), or MP (e). Additional components in solutions are listed above the plots. Neither nucleotide nor precursor induced stable intermolecular interaction of PsnB. f, Loop modeling revealed that β13β14 loop is long and flexible enough for intramolecular interaction. Residues between Ile227 and Ile247, the β13β14 loop region, were modeled by FALC59,60 and overall complex structure was optimized by relaxation. In the modeled structure, the conformation of the β13β14 loop was flipped and the DFR motif moved toward the enzyme active site. Also, Arg235 had hydrogen bonding with nucleotide which is similar to the intermolecular interaction scheme shown in crystal structures (Fig. 4c).

Extended Data Fig. 9 | Molecular interaction, binding affinity, and inhibition effect of MP(pE37). a, Detailed interaction scheme of MP(pE37) and PsnB. pGlu37 makes no interaction, whereas Glu38 interacts with Arg213 (yellow sticks) of PsnB. b, Binding affinity of Fl-MP(pE37) to PsnB was measured
by fluorescence anisotropy. MP(pE37) binds less tightly to PsnB than the MP no matter whether ADP is present. Data are presented as dot plots with mean ± 1 SD (n = 3 independent experiments) and fitted to a hyperbolic equation. c, Inhibition assay was performed with MP(pE37) or LP. 10 μM MP and 0.4 μM PsnB was co-incubated with various concentrations of MP(pE37) (left) or LP (right) in a buffer containing Tris-HCl (100 mM; pH 7.3), MgCl2 (10 mM), KCl (50 mM) and ATP (5 mM) at 37 °C for 10 min.

Extended Data Fig. 10 | Precursor and nucleotide binding induces conformational change of PsnB. a, Surface models of the leader-binding domain without (left) or with (right) the bound Phe15 in the LP (magenta sticks). Hydrophobic pocket for Phe15 is shown as magenta dashed lines. b, The nucleotide (cyan sticks) binding shifts the β9β10 loop (from gray to yellow cartoon). Lys172 and Thr180 that interact with a nucleotide are shown as green sticks. c, Two α3α4 pairs of a PsnB dimer show extensive interactions to each other to form a rigid body. Two PsnB subunits are shown as yellow and pale cyan cartoons, and their interacting residues are presented as orange and cyan sticks, respectively. d, Surface model of the ENLC-E complex (yellow and pale cyan for two PsnB subunits). Binding of LP (magenta) and nucleotide (cyan spheres) induces the conformational change of PsnB dimer to generate a compact core (green) binding site. e, Superposition of four LP and CP pairs from PsnB-MP-ADP (7DRM, green and cyan ribbons) and PsnB-MP-AMPPNP (7DRN, magenta and yellow ribbons). All LPs and CPs are well overlapped. PsnB is shown as gray cartoon and half-transparent surface. Nucleotide and Arg213 are shown as gray sticks. Glu37 in CP is shown as sticks.

Extended Data Fig. 8 | Binding property of core-binding site mutants of leader-fused PsnB. Fluorescence anisotropy of Fl_CP (0.1 µM) with leader-fused PsnB mutants. Mutation of core-binding residues reduced the affinity between CP and the enzyme. Data are presented as dot plots with mean ± 1 SD
(n = 3 independent experiments) and fitted to a hyperbolic equation.Peptide 17