Molecular insight into the enzymatic macrocyclization of multiply backbone N-methylated peptides

The enzyme OphP is essential for the biosynthesis of the macrocyclic peptide omphalotin A, a dodecamer with 9 backbone N-methylations produced by the wood-degrading fungus Omphalotus olearius. Heterologous expression of OphP and the peptide-precursor protein OphMA in yeast, yields omphalotin A. Thus, Oph P was hypothesized to have a dual function; catalyzing both endoproteolytic release of a peptide intermediate from OphMA, and macrocyclization of the multiply α-N-methylated core peptide with concomitant release of a C-terminal follower peptide. In our in vitro activity assays, OphP showed robust endoproteolytic and macrocyclase activity on α-N-methylated peptides but was unable to cleave OphMA. The enzyme had a strong preference for hydrophobic, highly α-N-methylated peptides and an α-N-methylated glycine residue at the P1 site. OphP adopts a canonical prolyl oligopeptidase (POP) fold with a predominantly hydrophobic substrate binding cleft, and a small and hydrophobic P1 binding pocket. We demonstrate that OphP is a POP-type macrocyclase with a specificity and a substrate route to the active site different from other members of the family. These results could be exploited for the biotechnological production of macrocyclic peptides with multiple backbone N-methylations, which are interesting due to their favorable pharmacological properties.


Introduction 38
Despite violating Lipinski's 'rule of five' (Lipinski et al., 2001), some peptide macrocycles 39 show potent biological activity. This is because the macrocyclization restricts the bond 40 rotations of the peptide backbone and, thus, the favourable conformation of the peptide has a 41 much lower entropic penalty for target engagement (Marsault & Peterson, 2011;Marti-138 In this study, we purified OphP from P. pastoris and assessed its in vitro activity towards a 139 variety of substrates. We found no evidence for OphP being able to process intact OphMA 140 but detected a robust proteolytic and macrocyclase activity of the enzyme towards multiply 141 backbone N-methylated peptides derived from OphMA. In contrast to other macrocyclase 142 systems, the enzyme appears to be insensitive to the sequence of the follower peptide but 143 have a preference for hydrophobic, multiply backbone N-methylated peptides and recognize a 144 backbone N-methylated glycine residue within the OphMA core peptide as P1 site for 145 proteolytic cleavage. These findings are supported by the structural analysis of enzyme-146 substrate complexes. This knowledge may enable the exploitation of the enzyme for the 147 biotechnological production of new-to-nature multiply backbone N-methylated peptide 148 macrocycles for applications in agriculture or medicine. 149 150

151
OphP proteolytically cleaves and macrocyclizes OphMA-derived, multiply backbone N-152 methylated peptides in vitro 153 OphP was purified from P. pastoris as an N-terminal His 8 SUMO* fusion protein (Figure 1-154 figure supplement 1) and confirmed to be catalytically active using a standard chromogenic 155 substrate for POPs, benzyloxycarbonyl-Gly-Pro-p-nitroanilide (Z-Gly-Pro-4-pNA) (Nagatsu 156 et al., 1976) (Figure 1-figure supplement 2). OphP showed optimal activity at 30°C and pH 157 6.0. The reaction was dependent on enzyme and substrate concentration. OphP activity was 158 reduced by Z-Pro-prolinal (ZPP), a common covalent inhibitor for POPs, but the enzyme was 159 not inhibited as strongly as other members of the family (Figure 1-figure supplement 2) 160 (Yoshimoto et al., 1985). Mutation of the putative active site Ser residue at position 580 to 161 Ala (S580A) abolished the detectable enzymatic activity of the fusion protein in vivo. Co-162 incubation of catalytically active OphP with intact OphMA protein showed no detectable 163 peptide product (LC-MS) nor did we observe any mass loss of the protein (ESI-MALDI-TOF) 164 (Figure 1-figure supplement 3). Intact OphMA was expressed as an N-terminal His 8 -fusion 165 protein in E. coli for 72 hours and purified as previously described (Song et  containing 10 methylations (includes one residue in the follower peptide) (Figure 1a). The 169 experiment was repeated by purifying OphMA with a lower degree of methylation 170 (predominantly zero, one or two methylations) and again we detected no turnover by OphP. 171 Addition of SAM in vitro, which led to conversion of OphMA to the completely methylated 172 state over time, to the reaction, did not change this negative result. Based on these data, we 173 conclude that OphP is not able to process the intact OphMA protein.

175
To characterize the presumed peptidase and/or macrocyclization activity of OphP, we tested a 176 variety of backbone N-methylated, OphMA-derived C-terminal peptides, consisting of clasp 177 domain, core peptide and follower peptide sequences (Figure 1a), for in vitro processing by 178 OphP. Since, in our hands, the chemical synthesis of methylated or unmethylated OphMA-179 derived C-terminal peptides failed (an issue we attributed to their highly hydrophobic 180 character), the peptides were generated by cleaving purified N-terminally His 8 -tagged 181 recombinant wildtype OphMA using trypsin (Oph-30mer), and variants thereof containing 182 cleavage sites for the tobacco etch virus (TEV) protease in the unmethylated clasp domain at 183 various distances from the core peptide, using TEV protease. Some TEV-cleavable OphMA 184 variants also contained deletions and mutations in the C-terminal core and follower peptide 185 sequences. The OphMA variants were expressed for a sufficient period in E. coli to allow 186 backbone N-methylation to proceed, purified using metal affinity chromatography and 187 processed by trypsin or TEV protease. The proteolytically released peptides were purified by 188 preparative HPLC (Figure 2a). Peptides produced in this way (Oph-30mer, Oph-24mer, 189 Oph-21mer, Oph-18mer, Oph-15mer, Led-21mer and Dbi-21mer), included between three 190 and 12 residues of the unmethylated clasp domain preceding the core peptide. Whereas all 191 other peptides contained the complete 12-residue core and 6-residue follower peptide,  18mer Tr and Oph-15mer contained a truncated core peptide lacking the three C-terminal 193 residues and a truncation of the entire follower peptide, respectively. Led-21mer and  21mer, contained the 12-residue cores and the follower peptides (SVVSSA) of the lentinulin 195 A and dendrothelin A precursor proteins, LedMA and DbiMA1 (Matabaro,Kaspar,et al.,196 2021; Matabaro, Song, et al., 2021), respectively (Figure 2a, Figure 2-figure supplement 1). 197 MS analysis indicated that the Oph-30mer, Oph-24mer, Oph-21mer, Dbi-21mer, and  21mer peptides contained up to ten methylations, both the Oph-18mer Tr had up to 8 199 methylations and the Oph-15mer carried 7 methylations (Figure 2-figure supplements 1, 2). 200 All our attempts to produce and purify non-methylated peptide substrates, using either the 201 catalytically inactive R72A mutant of OphMA or maltose-binding protein fusions of C-202 terminal parts of OphMA, failed due to the insolubility of the peptides.

204
With the exception of Oph-30mer, where no product could be detected, incubation of all 205 these peptides with OphP led to the formation of peptide macrocycles where the 206 unmethylated residues of the clasp domain that precede the methylated core peptide in 207 OphMA, were incorporated into the product (Figure 2a, 2b) backbone N-methylation (Figure 2b, Figure 2-figure supplement 3) (Figure 2c, Figure  215 2-figure supplement 3). The Oph-18mer Tr and -15mer peptides both produced the 12-residue 216 macrocycle cyclo(GFPW me VI me V me V me G me V me I me G) (Figure 2d). These results suggest that 217 OphP is able to in vitro macrocyclize OphMA-derived, multiply backbone N-methylated 218 peptides at two different sites within the methylated core but unable to remove the 219 unmethylated clasp domain residues resulting in peptide macrocycles including these residues. 220 221 In order to test whether OphP is also able to produce macrocycles consisting only of the 222 methylated core peptides, we exploited the fact that the clasp domain ends with a proline 223 residue in OphMA, LedMA and DbiMA1 and treated derived peptides with the proline-224 specific endopeptidases recombinant GmPOPB (Czekster et al., 2017) or ProAlanase 225 (Promega, Madison WI, USA). In detail, these were Oph-21mer (leading to 18mer NL ; NL 226 denotes 'no leader'), Oph-18mer Tr (leading to Oph-15mer Tr ), Oph-15mer (leading to Oph-227 12mer), Led-21mer (leading to Led-18mer) and Dbi-21mer (leading to Dbi-18mer). 228 Treatment of the purified peptides with these enzymes removed the unmethylated residues 229 preceding the methylated core peptides (Figure 2a, Figure 2-figure supplement 4).

230
Subsequent incubation of Oph-18mer NL , Led-18mer and Dbi-18mer with OphP led to the 231 formation of omphalotin A, lentinulin A and dendrothelin A, respectively (Figure 2e). 232 Noteworthy, in all reactions where peptide macrocycles were detected, also linear peptide 233 products were found (e.g. Figure 2-figure supplement 2, 4 and 6). This is typical for 234 peptide macrocyclases and a consequence of the inherent endopeptidase activity of these 235 enzymes (Czekster et al., 2017). Incubation of Oph-15mer Tr and Oph-12mer with OphP only 236 produced linear peptides (W me VI me V me V me G me V me I me G), suggesting OphP cannot form a 9-237 residue macrocycle. These results are in accordance with the detection of linear 9-residue 238 core peptides in a recent in vivo study .

240
To test whether OphP could process known peptide substrates for POPs, GmPOPB peptide 241 substrates, namely the 25mer amatoxin precursor (AMA1) and the 17mer phalloidin 242 precursor peptide (PHA1) were chemically synthesized. Consistent with the previous report supplement 5). In contrast, neither OphP nor its homolog LedP was able to process these 246 non-native peptides. These results are in agreement with the specificity of OphP (and LedP) 247 for α-N-methylated peptide substrates.

249
Taken together, our results show that OphP is able to proteolytically cleave and macrocyclize 250 OphMA-derived peptides of 12 to 24 residues containing a multiply backbone N-methylated 251 core of at least nine residues. The OphP-mediated macrocyclization reaction involves the 252 proteolytic clipping of three to six unmethylated or partially methylated residues following 253 the methylated core peptide and produces peptide macrocycles of 12 to 18 residues that can 254 include stretches of up to six unmethylated residues preceding the methylated core peptide. 255 On the other hand, OphP appears to be unable to remove unmethylated residues preceding the 256 methylated core peptide or process intact OphMA.

258
OphP peptide cleavage and macrocyclization activity has a strong preference for an α-N-259 methylated glycine residue at the P1 site 260 A closer analysis of the data revealed that peptide processing by OphP depended on the 261 presence of an α-N-methylated glycine ( me Gly) residue at the P1 site and an unmethylated 262 residue at the P1' site. This preference is particularly evident on the processing of Oph-263 18mer NL and Oph-15mer by OphP (Figure 3a). MS analysis of purified Oph-18mer NL 264 showed predominantly 10 methylations with 3-to 11-fold methylated species being present in 265 lower amounts (11 th methylation of the Met414 residue in the follower peptide) whereby the 266 methylation pattern of the various species can easily be inferred since methylation of the core 267 peptide by OphMA occurs sequentially from the N-to the C-terminus (van der Velden et al., 268 2017). Incubation of Oph-18mer NL with OphP showed that OphP consumed all species but 269 the 3-, 5-and 6-fold methylated ones. The peptide species consumed by OphP, all have a 270 me Gly followed by an non-methylated residue while the three non-processed ones do not.

271
These results were confirmed by the in vitro processing of a 1:1 mixture of 6-and 7-fold 272 Oph-15mer. It should be noted that 7-fold methylation is the maximal degree of methylation 273 that is achievable for a OphMA lacking the 6-residue follower peptide .

274
While the 7-fold methylated peptide species was readily consumed by OphP, the 6-fold 275 methylated one was hardly diminished (Figure 3a and Figure 3-figure supplement 1).

276
These results were confirmed by analyzing the processing of another batch of Oph-15mer 277 where purified OphMA was incubated with SAM in vitro before cleavage by TEV protease 278 and HPLC-purification of the peptide (Figure 3-figure supplement 2). This procedure is 279 known to increase the degree of core peptide methylation (Song et al., 2018) and, accordingly, 280 yielded Oph-15mer with a 1:9 ratio of 6-fold:7-fold methylated peptide. MS analysis of 281 consumption of this batch of Oph-15mer by OphP and the formation of the respective 282 macrocyclic products, cyclo(GFPW me VI me V me V me G me V me I me G) and 283 cyclo(GFPW me VI me V me V me G me V me IG), demonstrated that the 7-fold methylated species was 284 strongly preferred over the 6-fold methylated one (Figure 3b,c and Figure 3- figure  285 supplement 2). The only difference between the two peptide substrates is the α-N-286 methylation of the glycine residue at the P1 site of the proteolytic cleavage.

288
OphP adopts a typical POP-fold with an unusual hydrophobic tunnel 289 OphP shares 35% and 38% sequence identity to PCY1 from Saponaria vaccaria and 290 GmPOPB from Galerina marginata, respectively, including conservation of the putative 291 catalytic triad residues Ser580, Asp665 and His701 (Figure 4-figure supplement 1). Despite 292 these similarities, OphP has two remarkable features that distinguish it from these and other 293 members of the POP family: its preference for multiply backbone N-methylated substrates 294 including a me Gly residue (rather than a proline residue) at the P1 site, and the apparent lack 295 of a site for the recognition of the substrate C-terminus (follower peptide) (Figures 2 and 3; 296 Figure 4-figure supplement 2). In order to elucidate the structural basis of these differences, 297 we determined the crystal structures of OphP alone (apo-structure) and in complex with 298 substrates and inhibitors. 299 300 We first determined the structure of the putatively catalytically inactive OphP variant S580A. 301 The structure of the apo-form was determined in space group P1 (eight monomers in the 302 asymmetric unit) to 1.9 Å by molecular replacement using GmPOPB structure (PDB entry 303 5N4C) as the search model. Visual analysis and the PISA server suggest OphP is a monomer 304 in the crystal and other POP enzymes. The eight monomers differ only in flexible loops some 305 of which are involved in crystal contacts. Similar to other POPs, the monomer has two 306 domains, an α/β hydrolase domain (residues 1-82, 453-738) and the seven-bladed β-307 propeller domain (residues 83-452) (Figure 4a). All the monomers in the asymmetric unit 308 adopt a "closed" conformation with the two domains packed against each other which closes 309 off the direct route to the active site. OphP is most similar to GmPOPB (5N4B) with a root-310 mean-square-deviation (RMSD) of 1.5 Å over 686 residues and closely related to PCY1 311 (5O3W, RMSD of 1.7 Å over 650 residues) with a more distant relation to porcine muscle both subunits His701 is around 15 Å away from Ser580 and, thus, a structural rearrangement 318 is required to form the canonical catalytic triad. 319 320 The structure of the covalent complex between OphP and ZPP (Figure 4-figure supplement  321 3) was obtained by soaking wild-type OphP crystals with 5 mM ZPP overnight. The structure 322 of the complex was determined at 2.0 Å resolution. Additional electron density, which we 323 fitted as ZPP covalently linked to Ser580 was seen in two subunits (Figure 4- figure  324 supplement 3). Crystal packing would appear to limit the ability of the domains to have 325 separated sufficiently to allow ZPP to access the active site. The calculated solvent content of 326 the crystals is below 40 % consistent with visual observation that there are no large solvent 327 channels that would allow large domain movements that do not disrupt crystal packing. The 328 benzene ring of ZPP is located in an aromatic hydrophobic pocket. The other two subunits 329 had weak electron density suggesting low occupancy. The overall structure of the OphP:ZPP 330 complex is largely unchanged from the apo-structure of OphP(S580A) (0.5 Å over 716 Cα 331 atoms. There are differences in loops Leu139-Ala146, Ser164-Met171, Ser179-Met195, 332 Pro222-Gly230, Leu696-Ser706 and Ala622-Tyr664. As a result, His701 has moved closer 333 (7.6 Å to serine) to both Asp665 and Ser580 to form the catalytic triad (Figure 4- figure  334 supplement 3). These results confirm Ser580 as the key catalytic residue and support our 335 suggestion of structural rearrangement of loops is required to form the catalytic triad. 336 337 me

Gly at the P1 site of the peptide substrate is coordinated by a specific binding pocket 338
To characterize the interactions of OphP with its presumed natural substrate, crystals of 339 OphP(S580A) were soaked with mainly 10-fold methylated Oph-18mer NL  (Figure 5a-c and Figure 5- figure  343 supplement 1, respectively) (both substrates yield dodecameric peptide macrocycles with 344 the underlined residues; the numbering of the residues eases the visualization of the structural 345 arrangement of the peptide in the active site of the protein displayed below). The Oph-346 18mer NL complex was determined to 2.0 Å and the Oph-15mer complex at 2.5 Å resolution. 347 The protein structures superimpose well with the apo structure with an RMSD of 0.4 Å over 348 715 Cα atoms (Oph-18mer NL ) and 0.8 Å over 714 Cα atoms (Oph-15mer). The peptide 349 substrates are located at the interface between the two domains with its N-terminus extending 350 into the β-propeller domain and its C-terminus contacting the hydrolase domain (Figure 5a; 351 Figure 5-figure supplement 1). Substrate binding displaced both glycerol and water 352 molecules found in the apo-structure.

354
For Oph-18mer NL , the clearest density for the substrate was found in subunit A. There is 355 some weak electron density for the loop (698-704) which contains the catalytic histidine 356 His701, but we did not model it. Only residues me I8 me G9V10 me I11 me G12S13 me V14 were 357 fitted to the electron density whilst the remainder were presumed to be disordered ( Figure  358 5b). The side chain of me I8 packs against Arg316 whilst me G9 makes almost no contact; the 359 side chain of V10 interacts strongly with the aromatic ring of Phe617, me I11 makes a similar 360 interaction with Phe502 whilst its N-methyl group points to solvent (Figure 5c). The P1 361 binding pocket that accommodates me G12 can be decomposed into two components (amide 362 and side chain) for ease of discussion. The N-methyl group sits in a pocket formed by Phe502, 363 Ile606 and Trp621 (indole makes van der Waal contact) (Figure 5d). The Cα of me G12 is 364 surrounded by Ile606, Val668 and Ser580 (< 5 Å). Further away from the Cα of me G12 are 365 the indole of Trp621 (6 Å) and the hydroxyl of Tyr625 (7 Å) which complete the pocket. The 366 peptide bond between me G12 and S13 is positioned for nucleophilic attack by Ser580. The Arg667 are found in many other members of the POP family where they play a similar role 52 .

373
In the Oph-15mer complex, residues G1F2P3W4 me V5I6 me V7 me V8 me G9 me V10 me I11 were 374 located in the electron density in subunit D (Figure 5-figure supplement 1). The peptide 375 bond of me V8 is closest to Ser580 but its orientation is not consistent with catalysis ( Figure 5-376 figure supplement 1). We conclude this complex is not directly relevant to understanding 377 catalysis but rather is evidence for substrate binding plasticity.

379
To further investigate the role of Ile606 and Trp621 in substrate recognition and catalysis, we 380 produced OphP variants I606A and W621A. Incubation with Oph-15mer (in almost 381 exclusively 7-fold methylated form) showed that I606A is only slightly less active than the 382 native enzyme while W621A had no activity (Figure 5e). We hypothesised that Ile606 might 383 play a role in limiting the volume of the P1 pocket thus introducing selectivity. To test this, 384 we explored whether bulkier side chains at P1 could be processed by the I606A variant of 385 OphP. Unfortunately, we were unable to purify G12V and G12L substrate variants of Oph-386 15mer and the variant G12A resulted in a complex mixture that prevented further analysis. 387 388

389
The multiply backbone N-methylated peptide macrocycle omphalotin A from Omphalotus 390 olearius originates from a RiPP pathway involving the peptide auto-α-N-methyltransferase 391 OphMA and the serine peptidase OphP ( (Figure 1a). The simplest model of the 396 biosynthetic pathway is that OphP represents a dual function macrocyclase, that releases a 397 peptide intermediate (core plus follower peptide) from OphMA (function 1) and cleaves off 398 the follower peptide and concomitantly macrocyclizes the core peptide to yield the final 399 product omphalotin A (function 2). 400 401 OphP is not bifunctional 402 Our in vitro experiments with various backbone N-methylated peptide substrates showed 403 oligopeptidase and macrocyclase activity of purified OphP (Figures 2 and 3 including 404 respective figure supplements). In contrast, OphP showed no proteolytic activity towards 405 full-length OphMA, irrespective of its methylation state (Figure 1-figure supplement 3).

406
OphP was unable to remove unmethylated residues preceding the methylated core peptide, 407 rather in some cases these additional residues were incorporated into the resulting 408 macrocycles (Figure 2 for the requirement of separate enzymes. If the cellular endoproteinase responsible for the 422 cleavage of OphMA does not cut after Pro399, additional aminopeptidases will be needed to 423 remove eventual unmethylated residues preceding the omphalotin core peptide and expose 424 Trp400 for macrocyclization of the core peptide by OphP. Based on the sensitivity of the in 425 vivo but not in vitro production of omphalotin A to proteasomal inhibitors (  (Figure 4a). We obtained ZPP, Oph-18mer NL and Oph-15mer co-complexes by soaking the 442 peptide substrates into crystals, with only some small changes in active site loops but no large 443 domain movements. Crystal packing would seem to prevent the large closed to open to closed 444 conformational changes required by the "hinge" mechanism. We cannot exclude the 445 possibility of such gross movements and annealing in the crystal, but it would be unusual. 446 Thus, we rather considered two alternative routes for large peptides to reach the active site. 447 The first is an entrance between the two / and -propeller domains (the side cavity) 448 (Figure 4a). Such side entry has been observed previously, in dipeptidyl peptidase IV, a However, this side cavity was much larger than that seen in apo OphP. In OphP, the 451 disordered loops Pro222-Gly230 and Leu697-Gly704 would be expected to further constrict 452 the side entry. Furthermore, the protein surface for this route is polar which does not match 453 the nature of the substrate (Figure 4b). We cannot exclude this means of entry but do not 454 consider it as very likely. A second and more plausible route is via the central hydrophobic 455 tunnel (Figure 4b) that leads to the cleft. Compared to the other POP enzymes (1NU6, 5O3U 456 and 5N4F), the tunnel in OphP is wider (Figure 4-figure supplement 4). The apparent lack 457 of conformational change, the tunnel width match to substrate size, and the low polarity of 458 this route make the tunnel an attractive route of substrate entry (Figure 4b). Entry by this 459 route that has not been observed in any member of the POP family. Interestingly, in a study 460 of porcine muscle prolyl oligopeptidase (Fülöp et al., 1998;Fülöp et al., 2000), the authors 461 proposed the opening of the central tunnel of the β-propeller domain with blades 1 and 7 462 acting as the gate to allow substrate to gain access to the active site.

464
The tight P1 pocket controls selectivity 465 Detailed analysis of the products derived from the various substrates revealed a me Gly residue 466 as the point of macrocyclization (P1 site) (Figures 2 and 3 including respective figure  467 supplements). The requirement for me Gly contrasts the requirement for Pro at P1 which 468 defines the POP family ( me Gly resembles Pro in the sense that both are tertiary amides).

469
Structural analysis shows no specific interaction of OphP with residues C-terminal to the core 470 peptide (Figure 5 including respective figure supplements). Accordingly, the six C-471 terminal (follower) residues show no binding to OphP (Figure 4-figure supplement 2) and 472 there is no conservation in length or amino acid sequence of the follower residue of peptides 473 processed by OphP (Figure 2a). We conclude that OphP does not utilize the C-terminal 474 follower residues as a recognition sequence, in contrast to GmPOPB and PCY1 (Chekan et al., residues at the N-terminus (Figure 2a). This plasticity is consistent with the observation that 481 the interactions between OphP and the peptide substrate side chains (Figure 5c In the Oph-18mer NL complex, me G12 is located at the P1 site consistent with the expected 489 product. The active site configuration creates an appropriately positioned oxy-anion hole for 490 catalysis. me Gly12 binds to a hydrophobic pocket formed by Trp621, Tyr625, Val668, 491 Asn581, Phe502 and Ile606 (Figure 5d). The N-methyl group interacts with Trp621 and the 492 mutant W621A is inactive (Figure 5e) thus, the entropic penalty for binding ( me Gly12 has Φ and Ψ dihedral angles that are similar to 496 the prolines (P1 site) in structures of PCY1 (PDB 5O3U) and GmPOPB (5N4C)). We suggest 497 this combination of enthalpy and entropic factors drives the preference of OphP (and its 498 relatives) for tertiary amide substrates. 499 500 Except for Ile606, the other residues that could interact with a side chain at P1 (Phe502, 501 Val668, Trp621, Tyr625) are strictly conserved across all POP enzymes to date (Figure 4-502  figure supplement 1, Figure 5-figure supplement 2). Isoleucine is almost exclusively found 503 in OphP and closely related methylated peptide macrocyclases (Figure 4- in a clash (distance < 2.5 Å) with Ile606 ( Figure 5-figure supplement 2), We propose that 507 Ile606 selects against recognition of Pro. In support of this argument, the covalent OphP:ZPP 508 complex, which has a proline group at the P1 site, results in Ile606 adopting an unfavorable 509 conformation (Figure 5-figure supplement 2). This correlates with the lower-than-expected 510 inhibitory potency of ZPP against OphP (Figure 2-figure supplement 6). The ZPP result 511 does show the selection is not absolute and the pocket can undergo conformational change. 512 Gymnopeptides are members of the borosin family that includes the omphalotins (Quijano et  513 al., 2019) and they possess me Val at the P1 position. The valine side chain would clash not 514 only with Ile606 but also with Val668. The sequences of the OphP homologues of the 515 producing fungus Gymnopus fusipes are not known but we would predict their P1 pocket has 516 to be altered from OphP to to accomodate the larger side chain at the P1 position.

518
In summary, we show that OphP is not a dual function proteinase/macrocyclase and that in 519 vivo a third enzyme is almost certainly required for the production of ribosomally synthesized, 520 backbone N-methylated peptide macrocycles from auto-α-N-methylating precursor proteins 521 like OphMA. OphP recognises an N-methylated glycine rather than a proline residue at the 522 P1 site of macrocyclization using a hydrophobic pocket that selects against residues larger 523 than alanine whilst selecting for methylated amides. The hydrophobic and spacious nature of 524 the substrate-binding pocket results in few specific contacts endowing OphP with some 525 substrate promiscuity and the unique ability to cyclize multiply methylated substrates. These 526 unique properties of OphP may turn this enzyme into a valuable tool for producing highly α-527 N-methylated cyclic peptides for pharmaceutical applications. 528 529 530

650
Protein sequences (the portion used after TEV protease cleavage site (cs) is underlined, 651 the TEV protease cleavage site is in bold) 652 653 His  For the experiment testing for preferred production of 7-fold over 6-fold methylated species 753 (Figure 3-figure supplement 2), the reactions were performed as described above, but with 754 varying 6.0-6.5) and 28% PEG 3350). Crystals appeared within eight days and grew to full size 827 within three weeks at room temperature. Crystals were fished, transferred to cryo-protectant 828 (reservoir solution supplemented with 40% PEG 3350) and flash-frozen in liquid nitrogen.

829
The structure of the OphP-ZPP complex was determined by soaking OphP crystals with 5 830 Sequences and chemical structures of omphalotin A and homologous peptide natural products 876 dendrothelin A and lentinulin A shown with methylation sites in orange and the cyclization 877 site in green. These peptide macrocycles were produced in Pichia pastoris by expressing 878 OphP together with OphMA variants with respective C-termini. Sequences of peptides produced in this work, with α-N-methylated residues highlighted in 882 orange. Note that methylation by OphMA proceeds from N-to C-terminus and that only the 883 completely methylated forms of the peptides are displayed. The P1 residues recognized by 884 OphP are highlighted in purple and the P1' residues in grey. Leader peptide removal was 885 achieved by incubating TEV protease-cleavable OphMA variants with TEV protease and 886 GmPOPB or ProAlanase. Macrocyclization of the peptides by OphP is indicated by green 887 brackets above the peptide sequence. Note that only the macrocyclic products are shown. orthologues lentinulin A and dendrothelin A produced from Oph-18mer NL , Led-18mer NL and 893 Dbi-18mer NL using OphP. The MS chromatograms show masses of both protonated cyclic 894 peptides and corresponding sodium and ammonium adducts. species was drawn at the same level (6-fold methylated species of Oph-15mer and Oph-901 18mer NL ). These species were marked with blue asterisks. The sequence of the various 902 peptide species and their processing by OphP is indicated on the right. Green ticks stand for 903 processing and red crosses for lack of processing by OphP. (b,c) Time course of in vitro 904 reaction between 10µM OphP and 100µM total Oph-15mer Substrate (mixture of 6Me and 905 7Me species) in HEPES buffer at pH 7.0. EIC peak areas of 6-and 7-fold methylated peptide 906 substrate (b) and product (c) species were measured at 7 time points over 4 hours. Left panels 907 show the absolute EIC peak areas and the right panels the ratio between the 7-to 6-fold 908 methylated peptide species over the 7 time points. electrostatic potential is set at ± 5 kT/e. There is no access to the active site (black circle) 919 from the side of the structure in the closed state. (b) Rotation of 90° around the X-axis shows 920 there is a tunnel through the β-propellor domain (circled in red in cartoon). On the right is the 921 is electrostatic surface calculated as above, the tunnel allows access to the active site (black 922 circle). The yellow carbon atoms of ZPP can be seen through the pore in both representations. 923 924 Figure 5. me Gly12 recognition site of OphP. (a) OphP S580A complex to peptide substrate 925 Oph-18mer NL is shown in spheres. (The color schemes is as Figure 4a)