Biosynthetic Proteases that Catalyze the Macrocyclization of Ribosomally Synthesized Linear Peptides

Abstract

Circular peptides have long been sought after as scaffolds for drug design as they demonstrate protein-like properties in the context of small, constrained peptides. Traditional routes towards the production of cyclic peptides rely on synthesis or semi-synthetic methods, which restrict their use as platforms for the production of large, structurally diverse chemical libraries. Here, we discuss the biosynthetic routes towards the N-C macrocyclization of linear peptide precursors, specifically, those transformations that are catalyzed by peptidases. While canonical peptidases catalyze the proteolysis of linear peptides, the biosynthetic macrocyclases couple proteolytic cleavage with cyclization to produce a macrocyclic compounds. In this Perspective, we explore the different structural features that impart on each of these biosynthetic proteases the distinct ability carry out macrocyclization, and focus on their potential use in biotechnology.

Graphical Abstract

INTRODUCTION

Ribosomally synthesized and post-translationally modified peptides (RiPPs) form of a class of natural products that are structurally diverse and possess a range of biological activities¹. Unlike non-ribosomally synthesized peptides, RiPPs are derived from peptides synthesized on the ribosome as linear precursor, which are subsequently processed by a cadre of enzymes to produce an elaborate and often conformationally restricted final product². Interest in RiPPs have been spurred by the largely novel nature of chemical transformations, their biosynthetic enzymes can catalyze, the ease in genome mining efforts leading to discovery³, and their use in therapeutic applications^4–5. Here, we review advances in the understanding of the processing enzymes that carry out macrocyclization of linear precursors, and the potential of these enzymes to be used as macrocyclases on a broader panel of substrates.

Given the extent of unique modifications that RiPP biosynthetic enzymes carry out, peptidases are often structurally and biochemically overlooked. The canonical role for peptidases in RiPP biosynthesis, whether exogenous or found within cognate biosynthetic gene clusters, is for the removal of the leader sequence from the processed mature final product. Examples of such proteases include those found in the processing of lanthipeptides⁶, lassopeptides⁷, and thiopeptides. A less frequent utility is the use of proteases, in certain RiPP pathways, to carry out a macrocyclization reaction (i.e. transamidation of the N- and C-termini of a linear peptide precursor) in a manner that is coupled to peptide bond cleavage. Parenthetically, it is of note that this strategy is not strictly limited to RiPP biosynthetic pathways, as the isolated thioesterase domains from non-ribosomal peptide synthases⁸ and the transpeptidase sortase A⁹, have been exploited as biotechnological tools for the production of macrocyclic peptides from linear precursors. Synthetic methods for peptide ligation, although flexible, often require prosthetic groups to be installed on the substrates¹⁰. However, these platforms require either the use of synthetic groups or the retention of recognition sequences in the final macrocyclic product¹¹. In contrast, many RiPP biosynthetic enzymes do not require additional cofactors for ligation, nor do they retain recognition sequences in the final product.

In the context of RiPP pathways, there are several examples of peptidases that catalyze macrocyclization, such as those involved in the biosynthesis of orbitides¹², amatoxins and phallotoxins^13–14, borosins^15–16, cyanobactins¹⁷ and cyclotides¹⁸ (Figure 1A-E). As in all RiPP systems, the biosynthetic enzymes from each of these pathways are contingent on recognition sequences that are excised from the final product, highlighting their portability in biotechnological applications. The lack of signature sequences in the final product, along with the genetic nature of the system (i.e. both the substrate and modification enzyme are genetically encoded) favor the use of RiPP macrocyclases as versatile tools for chemical biology. In contrasts to their linear precursors, macrocyclic peptides are more thermally and chemically stable, resistant protease degradation, and show avid binding to biological targets targets⁴. Moreover, peptide based therapeutics are weakly immunogenic, and their cyclic counterparts can often recapitulate favorably “protein-like” properties. Hence, RiPP-based production of macrocyclic peptide libraries is ideally suited for exploring the scaffold space of macrocyclic peptides.

Figure 1.

Representative members of different classes of RiPPs produced by N-C macrocyclization catalyzed by peptidases. (A) Segetallin A, a representative orbitide. (B) α-Amanitin, which belongs to the amatoxin class. (C) Omphalotin A, a representative borosin. (D) Patellamide A, a member of the cyanobactin family. (E) Solution NMR structure of the cyclotide kalata B1 (PDB ID: 4TTM).

Despite the overall similarities in the chemical strategies used for N-C transamidation, the biosynthetic enzymes from the different RiPP pathways differ both in structure, as well as in the mechanistic details. In this Perspective, we consider the structural characteristics that impart upon these RiPP proteases the N-C transamidative activity. This activity is particularly striking given their sequence and structural conservation with orthologous enzymes that are not all capable of conducting the same reaction¹. A detailed structural and mechanistic biochemical understanding of how these unique peptidases function may facilitate greater uses of these biocatalysts as biotechnological tools.

I. Mechanistic Insights into the Formation of Macrocycles by Peptidases

Studies on the natural product biosynthesis have enumerated multiple strategies towards macrocycle formation. In many cases, adenosine triphosphate (ATP) is required as a co-factor in order to modify either the α-carboxylate or the β-hydroxyl groups of Ser/Thr residues, either through phosphorylation¹⁹ or adenylation²⁰. ATP-dependent activation generates a suitable leaving group that can be displaced by the incoming free amine, which can include the α-amino of the peptide. Activation of the otherwise inert α-carboxylate may also be carried out by formation of an oxo-ester with the terminal adenine 76 of tRNA, as evidenced for cyclodipeptide biosynthesis²¹, or through thioesterification via the phosphopantheinenylated peptidyl carrier protein (PCP) as observed in nonribosomal peptide synthases³. In the latter case, the thioester is transferred from the PCP to the active site Ser hydroxyl of the thioesterase (TE) domain utilized in NRPS chain termination. In brief, TE domains catalyze chain termination of non-ribosomal peptides through formation of an acyl-enzyme intermediate with the nascent peptide. This intermediate may be resolved either hydrolytically, to yield a linear product, or by an intramolecular transamidaton, resulting in a macrocylic product⁸. Proteases from RiPP biosynthetic pathways that carry out macrocyclization utilize a similar chemical strategy of capturing a linear substrate as an acyl-enzyme intermediate. Specifically, the catalytic nucleophile attacks the carbonyl carbon of the amino acid at the P1 position in the intact linear precursor²². Collapse of the resultant tetrahedral intermediate results in the scission of the C-terminal peptide and yields an acyl-enzyme intermediate. Typically, residues that are necessary for substrate recognition are encoded in the C-terminus of the precursor peptide and the removal of these residues during formation of the acyl-enzyme intermediate results in a final product that lacks any recognition elements²³.

In order to facilitate N-C transamidation, the α-amine of the linear peptide must be directed back toward the active site and be favorably positioned to attack the acyl-enzyme intermediate to form a cyclic product. Moreover, water molecules must be precluded from the active site in order to avert liberation of the intermediate as a linear hydrolytic product (as would be typical for canonical proteases). Moreover, in vitro reconstitution studies suggest that depending on the reaction conditions used, linear peptides may be the more thermodynamically favored product over their cyclic counterparts²⁴. Hence, these enzymes must efficiently achieve the permissibility of the free amine to reach the acyl-enzyme adduct, along with the exclusion of solvent. Each of the macrocycle-forming peptidases utilizes different mechanisms towards achieving this goal.

II. PCY1, a Prolyl Oligopeptidase, and the Macrocyclization of Orbitides

Orbitides, or Caryophyllaceae-type homodetic cyclic peptides (CPs) are plant-produced RiPPs that are N-C cyclized products of five to twelve amino acids. Putative orbitide gene clusters have been identified in Annonaceae, Caryophyllaceae, Euphorbiaceae, Lamiaceae, Linaceae, Phytolaccaceae, Rutaceae, Schizandraceae, and Verbenaceae¹. Furthermore, orbitides have been shown to possess antimalarial²⁵, immunomodulating²⁶, vasodilatory²⁷, and xenoestrogenic²⁸ activities. Segetalians A through L are a family of orbitides produced by S. vaccaria (syn. Vaccaria hispanica)²⁹. Each segetalin has a corresponding precursor peptide that consists of a variable central region that contains residues found in the final cyclized product, flanked on either side by sequences that are conserved. During segetalin maturation, a generic oligopeptidase (OLP 1) cleaves the flanking residues N-terminal to the variable region to produce two linear products. The C-terminal peptide, consisting of the variable region and a conserved “follower” sequence is processed by the macrocyclase PCY1 (peptidase cyclase 1), which utilizes a serine protease mechanism to remove the follower sequence and carry out the N-C cyclization of the hypervariable sequence²⁴.

The primary sequence of PCY1 reveals that it is part of the S9A protease family, which includes prolyl oligopeptidases (POPs). Canonical POPs are serine proteases that cleave peptides after a Pro residue, but cannot hydrolyze substrates longer than 30 amino acids³⁰, with noted exceptions^31–32. POPs have been studied in the context of human physiology and disease^33–34, although homologs are found in all domains of life³⁵. These enzymes possess a α/β hydrolase fold at the C-terminus, a seven-bladed propeller domain at the N-terminus, with the active site sitting at the interface of these two domains³⁶. There are no reports of mammalian POPs that are capable of cyclizing linear peptide substrates.

Biochemical and structural analyses have provided insight into how PCY1 has been adapted to carry out the macrocyclization, rather than hydrolysis of peptide substrates. The structure of PCY1 bound to presegetalin A1 (residues 14–32) demonstrates a canonical S9A protease α/β hydrolase fold, linked to a β-propeller domain connected by a hinge region (Figure 2A)¹². Presumably, flexibility along the hinge allows the enzyme to adopt a closed position when the substrate is bound³⁷. The PCY1 cocrystal structure reveals that the hinge region engages the terminal six residues in the follower sequence (N₂₇ASAPV₃₂) providing a means for recognition of substrate peptides (Figure 2B). Binding of the follower peptide to the enzyme forces a closed state, which could preclude solvent and facilitate the N-C transamidation reaction. Notably, in PCY1 the catalytic His695 is displaced away from the catalytic Ser562 as a result of a single residue deletion preceding the His (Figure 2C). A PCY1 variant wherein a Gly is inserted before His695 demonstrated a decrease in cyclization activity, and favored the production of linear products¹². Sequence alignment of PCY1 to other POP-ligases, including borosin and amatoxin producing POPs, as well as to non-cyclase POPs, reveal that the Gly residue deletion is unique to PCY1. Thus, the protease-ligase functionality can be attributed to this deletion only for PCY1. Hence, it was proposed that in PCY1 His695 possesses a range of motion to serve the dual function of both deprotonating the catalytic Ser562, as well as deprotonating the N-terminus of the core peptide to create a nucleophile that can resolve and cyclize the acyl-enzyme intermediate¹⁰. Interestingly, despite possessing the conserved Gly residue, the catalytic His and Ser residues in GmPOPB are also displaced. This suggests that a sub-optimal arrangement of the catalytic triad may impart the ability to carry out macrocyclization, but this has yet to be determined experimentally.

Figure 2.

Cocrystal structures of PCY1 bound to the six terminal residues of the presegetalin A1 follower peptide and Z-Proprolinal (ZPP). (A) Overall fold of PCY1 (in ribbons) bound to presegetalin A1 follower peptide (in CPK spheres) shows the enzyme in a closed conformation, with the peptide situated at the hinge region between the two domains (PDB ID: 5UW7). (B) Residues on PCY1 (orange) that are closely interacting with the six terminal residues of presegetalin A1 (white). (C) Active site residues of PCY1 (orange) in the presence of covalent inhibitor ZPP (cyan). Note the displacement between Ser562 and His695, which has been shown to be essential for macrocyclization potential of PCY1 (PDB ID: 5UW6).

Unlike other POP enzymes, PCY1 does not require a Pro residue preceding the cleavage site in the substrate, which would otherwise be retained in the final sequence of the cyclized peptide. Therefore, PCY1 could be perceived as a “traceless” macrocyclase. PCY1 has been successfully used to cyclize synthetic linear peptides between five and nine amino acids in length, including the precursors for segetalins A-J, as well two other orbitides from Dianthus caryophyllus²⁴.

Substrate scope analysis using Ala substitutions and D-amino acid variants of synthetic presegetalin A, demonstrates that PCY1 is able to tolerate single amino acid substitutions at each residue in the cyclized core, except at Val19²⁴. To assess PCY1’s ability to be used as an general cyclase in vitro, PCY1 has been used to cyclize presegetalin F1 core variants containing a minimal three residue follower sequence³⁸. Notably, PCY1 is reported to be 11–23 times more active than the cyanobactin macrocyclase PatG, using a linear peptide resembling the native PatG substrate³⁸. PCY1 has been shown to accept the amino octanoic acid, N-methylated alanine, and a thiazole ring, inserted into the core peptide of presegetalin F1. Kinetic parameters have also been established for PCY1 using presegetalin A1 (14–32) as a substrate, yielding a k_cat of 0.06 s⁻¹, and a K_M of 0.77 µM¹². The k_cat values for PCY1 fares less favorably compared to other macrocyclases^39–40. Despite the ease of using their biosynthetic gene clusters in combinatorial engineering, a factor limiting biotechnological use is the yield of cyclic products from heterologous production. To our knowledge, there are no reports of segetalin production in Escherichia coli, or other heterologous hosts.

III. Macrocycle-forming POPs in Fungal RiPP Biosynthesis

To date, there are four known classes of RiPPs that are found only in fungi: amatoxins/phallotoxins⁴¹, borosins^15–16 dikaritins⁴², and epichocyclins^43–45. Based on sequence similarities and homology modeling, the biosynthetic gene clusters for both amatoxins/phallotoxins and borosins were predicted to encode enzymes that are similar to POPs. Consequently, each of these enzymes may be considered as members of the S9A protease family. Macrocyclization of the amatoxin precursor utilizes a single enzyme (GmPOPB), which carries out sequential cleavage events: hydrolytic removal of the leader sequence, followed by proteolysis and transamidation of the hypervariable core. Biochemical characterization of amatoxin producing POPs suggests that they are among the most catalytically efficient of the macrocylases. Thus, utilization of these enzymes may provide a more efficient platform for the in vitro production of cyclic peptides.

The amatoxins and phallotoxins are actually fungal toxins that only differ in the size of the final macrocycle with amatoxins consisting of octapeptide cycles, and phallotoxins composed on heptapeptide cycles. Both of these toxins contain a characteristic Trp-Cys (tryptathionine) cross bridge⁴⁶, as well as hydroxylated Trp, Pro, and Ile residues. In addition, phallotoxins can also contain hydroxylated Asp, and Leu residues ⁴¹. Amanita bisporigera and other members of the Amanita species are capable of synthesizing both amatoxins and phallotoxins. Examples fungal toxins that fall within this RiPP class include the eukaryotic RNA polymerase II/III inhibitor α-amanitin^{13, 47}, and the actin stabilizer, phalloidin⁴⁸. As with other ribosomally synthesized macrocyclic peptides, the precursor peptides contain an N-terminal leader sequence, a hypervariable core, and a C-terminal follower sequence. The hypervariable core is flanked by Pro residues, suggesting that a prolyl oligopeptidase is involved in its processing. A candidate POP was isolated from the phallotoxin producing fungus Conocybe albepies, and was shown to process synthetic precursor peptide, but the enzyme generated only linear products and did not carry out cyclization¹⁴.

Interestingly, A. bisporigera, and Galerina marginata possess two genes that encode POP enzymes, but only one (POPB) is only found in amatoxin producing species¹¹. The other ortholog (POPA) is a general protease that catalyzes the hydrolytic cleavage of the full-length precursor but does not carry out cyclization. In contrast, the other ortholog POPB carries out both the hydrolytic removal of the leader, as well as the transamidative N-C cyclization of the hypervariable region⁴⁹. Structural and biochemical characterization of POPB from G. marginata (GmPOPB) have allowed for better understanding of how thess macrocyclases function. The crystal structure of ligand-free GmPOPB reveals that it possesses the same folds and domains as members of the S9A protease family, and that the enzyme is in the open state in the absence of bound substrate (Figure 3A)³⁹. Reconstitution studies reveal that the full-length 35-residue precursor peptide binds to POPB, whereupon the N-terminal 10 residue leader is excised to yield a 25 residue intermediate. This intermediate is released and then binds again to the enzyme, which carries out removal of the C-terminal follower peptide with the concomitant macrocyclization of the hypervariable region. The cocrystal structure of GmPOPB bound to 25- and 35-residue peptides reveals the enzyme adopts a closed state upon substrate binding (Figure 3B). As in the structure of PCY1, there is a notable displacement between the catalytic His and Ser residues (Figure 3C). Additionally, the crystal structure and biochemical data reaffirms that GmPOPB mainly recognizes its substrate through interactions with a six-residue region linker that follows the core peptide. Surprisingly, there are minimal hydrogen bonding contacts in the follower peptide-binding region (Figure 3D). GmPOPB completely lacks affinity for the N-terminal leader sequence³⁹.

Figure 3.

Structure of ligand-free GmPOPB (WT) and ligand bound GmPOPB (S577A mutant) (A) GmPOPB without substrate bound adopts an open conformation. (PDB ID: 5N4F). (B) GmPOPB with its 25-mer substrate adopts a closed conformation (PDB ID: 5N4B). (C) GmPOPB utilizes a Glu-His-Ser catalytic triad (purple) to conduct proteolysis. The His-Ser dyad is separated by 12Å. The 25 residues colored in pink indicate the indispensable linker recognition sequence, while residues colored in orange denote the remainder of the follower tail. Electron density is not observed for the core peptide. (PDB ID: 5N4B). (D) Complete view of GmPOPB interactions with the follower peptide, emphasizing the lack of contacts within hydrogen bonding distance to the follower peptide. Residues in yellow associate with linker region, while the residues in cyan associate with the remainder of the tail.

As with other RiPP macrocyclases, GmPOPB can tolerate a number of substitutions in the hypervariable region that is retained in the final product. The substrate tolerance of GmPOPB was probed using variants of its native core peptide (mature α-amanitin, I₁₁WGIGCRP₁₈), demonstrating that cyclization efficiency decreased the most in the W2G, G3L, and G5S variants, suggesting that these residues in core peptide may be involved in interactions with the enzyme⁴¹. GmPOPB accepts D-amino acids, as well as hydroxylated substrates, which are natural occurring post-translational modifications found in amatoxins⁵⁰. Interestingly, GmPOPB can be used to generate rings between 8 and 16 residues in size; the lower limit of eight residues is the size of the native hypervariable sequence of the core peptide. The cyclization efficiency of GmPOPB is greater than other macrocyclases, with K_M and k_cat values of 25.5 µM and 5.6 sec⁻¹ for the 35-residue native precursor⁵¹. While the requirement of a Pro residue at the P1 position is restrictive, GmPOPB is an otherwise efficient and flexible macrocyclase.

Borosins are a relatively newly characterized family of fungal RiPPs, previously believed to have been of non-ribosomal origin. In addition to the N-C macrocycle, borosins are additionally distinguished by the presence of N-methylation along the peptide backbone. Notably, the “substrate” peptide for the N-methylation and macrocyclization is produced as a fusion with the N-methyltransferase, and is excised and cyclized by a POP macrocyclase. Biosynthetic gene clusters for members of the borosin class also possess tailoring enzymes that can generate different congeners¹⁵. Heterologous expression of the N-methyltransferase and the POP enzyme in a fungal host is sufficient for the in vivo production of product¹³. It can be inferred that borosin POPs will work in a similar fashion to GmPOPB, as the borosin precursor peptide contains a leader, hypervariable core, and follower region. The omphalotin A precursor peptide contains a single Pro upstream of the hypervariable core. Although the order of modifications that the peptide can undergo is currently not known, the POP enzyme is expected to process the peptide in two sequential binding steps as with GmPOPB.

IV. PatG, a Promiscuous Subtilisin-like Protease in Cyanobactin biosynthesis

Cyanobactins are a class of N-C macrocyclized peptides that often, but not always, contain additional modifications, including azole moieties derived from β-nucleophile-containing amino acids (Ser/Thr/Cys), and/or isoprene groups⁵². Cyanobactins are produced by cyanobacterial symbionts found in marine environments, and one of the earliest cyanobactins to have been characterized is patellamide. The cyanobactin biosynthetic gene cluster contains two proteases, PatA and PatG⁵³, which process the PatE precursor peptide to yield the cyclic natural product. The five-membered (methyl)oxazole and thiazole groups are installed by the concerted actions of a heterocyclase/dehydrogenase⁵⁴, and isoprenylation is catalyzed by unique prenyltransferases that preferentially modify the cyclic compound⁵⁵. Like other RiPP enzymes, PatG is an extremely promiscuous macrocylase that can process a number of substrates, yielding products of different ring sizes and composition. The precursor peptides for cyanobactins are encoded as cassettes, and a single polypeptide contains sequences that elaborate multiple final products.

PatA and PatG proteases both fall under the family of subtilisin-like (peptidase S8 family) proteases, which utilize the Asp-His-Ser catalytic triad in order to process their substrates. The hypervariable regions within the precursor PatE cassette are flanked by N- and C-terminal recognition motifs. The first protease PatA removes the N-terminal recognition motif and produces a linear product. The second protease excises the C-terminal recognition motif and produces an N-C macrocyclized product⁵³. Based on primary sequence, PatA and PatG are strongly related and share 45% sequence identity. This raised the question as to why PatA produces only linear products but PatG was able to carry out a macrocyclization reaction. The crystal structures of these two enzymes are virtually superimposable, but PatG contains an insertion of two alpha helices, dubbed “capping helices” or “macrocyclase domain”, that is situated directly above the active site of the enzyme (Figure 4A)^{17, 56}. It was initially thought that the “capping helices” serve to both direct the α-amine of the bound peptide towards the oxo-esterified Ser, and to exclude solvent from the active site. Solvent exclusion is additionally mediated by the follower peptide^56–57. Recent identification and characterization of AgeG, a “capping domain” containing homolog of PatG, demonstrates that this homolog is only capable of generating linear peptide products⁵⁸. Thus, the specificity for transamidation over hydrolysis is established by the inherent chemistry of the respective enzymes, supporting what is termed the kinetic liability hypothesis⁵⁹.

Figure 4.

Crystal structure of PatG, H618A mutant, bound to a peptide substrate. Cocrystal structure with the PatG enzyme shown as ribbons and the substrate as CPK spheres. The capping helices (cyan) may serve as a substrate recognition domain (PDB ID: 4AKT). Catalytic residues are highlighted in the inset.

The crystal structure of PatG also rationalizes the strict requirement for a Pro or thiazoline at the P1 site, as this residue is necessary to direct the α-amino group of the substrate back towards the active site. The lack of additional contacts between the “capping helices” and the hypervariable region of the substrate explains the substrate promiscuity⁵⁶. Given this substrate permissivity, PatG has been manipulated to generate compounds that contain unusual features including D-amino acids in centered positions in the peptide chain⁶⁰, non-proteinogenic amino acids⁶⁰, aryl rings, polyethers, and alkyl chains^61–62. Lastly, the strict Pro/thiazoline requirement at the P1 site may be abrogated by using peptide bond mimetics such as using 1,4-anti-1,2,3-triazole-alanine and vicinal cysteine disulfide bonds⁶¹. Cyanobactin gene clusters have been successfully refactored into E. coli, with yields ranging from ~20–100 µg/L to ~2.0 mg/L^63–65. With optimized culture conditions, heterologous production of novel cyclic compounds using macrocyclase variants may be feasible.

V. Asparaginyl Endopeptidases in the Biosynthesis of Cyclic Plant-Derived Peptides

Another class of protease identified to function in the macrocyclization of linear substrates is the asparaginyl endopeptidases (AEPs). Plant-derived AEPs are involved in the biosynthesis of a number of cyclic plant products, including cyclotides, a class of plant cyclic peptides that are resistant to chemical, thermal, and enzymatic degradation⁶⁶. Other than the macrocycle, the defining feature of these natural products is a cyclic cysteine knot (CCK) motif, which is formed as the result of internal disulfide linkages. Other plant-based cyclic peptides are formed by AEPs, but lack the classical CCK motif. The primary sequences of plant-based AEPs identify these enzymes as members of the C13 family of cysteine proteases (clan CD), which are homologous to mammalian CD clan members such as legumain. A deepened understanding of AEPs involved in the production of cyclic plant products can provide multiple platforms for production of novel cyclic peptides.

There are several examples of characterized plant AEPs that produce cyclic peptide products, including sunflower seed AEP involved in the biosynthesis of sunflower trypsin inhibitors (SFTIs)⁶⁷, butelase 1 from Clitoria ternatea⁶⁸, and OaAEP1 isolated from Oldenlandia affinis, which is involved in the biosynthesis of kalata B1⁶⁹. The mechanism of AEP-mediated macrocyclization has been hypothesized to involve peptide bond cleavage coupled to hydrolysis-independent transpeptidation, as O¹⁸ could not be detected in the mature cyclic product⁶⁷. However, structural and biochemical characterization of human AEP suggests that the catalytic Cys is dispensable, and that ligation is accomplished via pH-dependent activation of a catalytic Asp to succinimide⁷⁰.

The crystal structure of the OaAEP1 proenzyme provides structural insights into substrates recognition⁷¹. Superimposition of OaAEP1 structure with that of human prolegumain shows that the two enzymes share nearly identical folds. Both enzymes contain a prodomain, termed the legumain stabilization and activity modulation (LSAM) domain, which is poised above the catalytic domain to block active site⁷². Additionally, the activation peptide region (AP) may also serve as a substrate mimic to block substrate entry⁷². It was proposed that the active site of OaAEP1 becomes accessible to the substrate only upon autocatalytic cleavage of the sterically hindering LSAM and AP regions under acidic conditions (Figures 5A, B). However, AEPs can exist as a two-chain intermediate in which the LSAM remains noncovalently bound to the catalytic domain, and hence can serve as a gatekeeper to modulate ligase activity⁷³. The crystal structure of OaAEP1 also identifies a cavity adjacent to the catalytic Cys nucleophile, which may accommodate the cyclic product⁷¹. The substrate specificity for each AEP is reflected by subtle amino acid differences in this cavity⁷⁴.

Figure 5.

Structural elements found in OaAEP1. (A) Individual domains within OaAEP1 are color coded as indicated. The activation peptide and the LSAM domain must to be cleaved off for the active site to become accessible. (B) Superimposition of the proenzymatic forms of OaAEP1 (PDB ID: 5H0I) and human legumain (PDB ID: 4NOK).

The 1.8 Å. crystal structure of the catalytic domain of sunflower AEP1 further clarified some mechanistic ambiguities by which plant AEPs ligate their products⁷⁴. In this structure, the substrate forms a tetrahedral intermediate with the enzyme, and mutational analyses shows that the active site succinimide is dispensable for ligation⁷⁴. Most importantly, ligation is shown to be a pH-dependent process. At near neutral pH, the catalytic Cys could be deprotonated to function as a nucleophile; additionally in this pH range, the N-terminal amine may be deprotonated, and primed for nucleophilic attack of the acyl-enzyme intermediate⁷⁴. Conversely, proteolysis will be favored at a lower pH. The ligation functionality may be additionally modulated by a α-helix bundle that controls substrate and solvent access to the active site⁷³.A plausible mechanism of cyclization initiates with proteolytic cleavage of the linear propeptide to elaborate an exposed N-terminal amine. Next, the AEP recognizes an Asp/Asn at the C-terminus of the substrate, resulting in formation of an acyl-enzyme adduct with the Cys nucleophile. Due to proximity effects, the exposed N-terminal amine can attack the thioester to regenerate the active site thiolate, and releasing a cyclized product. Interestingly, molecular simulations have supported experimental data that the catalytic Cys is dispensable, and that His alone can facilitate ligation⁷⁵.

Out of the ligases discussed, AEPs may be the most versatile class of proteolytic ligases. Butelase 1 has been used to cyclize non-native peptides substrates of varying chain length, and even protein substrates (including GFP), in which its preferred N/D-HV recognition motif at the C-terminus is appended at the end of the region to be cyclized^{40, 68, 76}. Similarly, cyclization has also been demonstrated with OaAEP1 using an anti-malarial peptide R1 with it’s preferred N-GLPS recognition sequence appended to its C-terminus. As a class, AEPs are catalytically robust, for example butelase1 has a catalytic efficiency of 10,700 M⁻¹ s⁻¹ for its native substrate kalata B1, and 11,700 M⁻¹ s⁻¹ for the non-native substrates, with >95% conversion in both cases⁶⁸. However, AEPs require substrates with an N/D at the P1 position (in addition to their respective recognition motifs) that is retained in the final macrocycle. Therefore, AEPs do not strictly function as “traceless” ligases.

CONCLUSION

Over the course of the past decades, research efforts from multiple groups have identified macrocyclic natural products that originate as linear peptides synthesized on the ribosome, and are subsequently processed to yield cyclic products. Biochemical studies of the corresponding biosynthetic enzymes, many of which are recapitulated in this Perspective, provide detailed understanding for the promiscuity of these catalysts, which allow for the production of sequence divergent molecules, with varying ring sizes. While the individual catalysts are all different, at both the sequence, structural, and mechanistic levels, there is a common theme to the strategies employed by these disparate pathways. Specifically, the peptide precursors for nearly all of these macrocyclic products contain N-terminal leader and/or C-terminal follower sequences that are highly conserved, within a given class of products. These sequences flank regions that are hypervariable with regards to both sequence length and composition. Residues in the N-terminal leader recruit a generic peptidase, resulting in excision of the leader and a resultant linear product consisting of the hypervariable region and the follower peptide. The enzyme that catalyzes macrocyclization is recruited to this substrate by virtue of the follower sequence, which is cleaved to yield an acyl-enzyme intermediate with the hypervariable sequence. The enzyme enforces solvent exclusion and proximity, thus facilitating transamidation of the N- and C-termini to yield the macrocycle.

Such RiPP macrocyclases are ideally suited for the in vivo production of large libraries of cyclic products as a result of the genetic nature of the substrate, as well as of the enzyme, the hypervariability of the sequences retained in the macrocycle and the lack of requirements for any additional cofactors. Hence, large compound libraries in an expedient manner without the need for purification of any single product. Discovery of additional catalysts with even greater substrate range provides further impetus for driving RiPP enzyme-based platforms for the production and screening of peptide-based therapeutics.