Recent advances in the biosynthesis of RiPPs from multicore-containing precursor peptides

Abstract

Ribosomally synthesized and post-translationally modified peptides (RiPPs) compose a large structurally and functionally diverse family of natural products. The biosynthesis system of RiPPs typically involves a precursor peptide comprising of a leader and core motif and nearby processing enzymes that recognize the leader and act on the core for producing modified peptides. Interest in RiPPs has increased substantially in recent years as improvements in genome mining techniques have dramatically improved access to these peptides and biochemical and engineering studies have supported their applications. A less understood, intriguing feature in the RiPPs biosynthesis is the precursor peptides of multiple RiPPs families produced by bacteria, fungi and plants carrying multiple core motifs, which we term “multicore”. Herein, we present the prevalence of the multicore systems, their biosynthesis and engineering for applications.

Natural products have long inspired drug discovery and development due to their intrinsic bioactivity. A major type of natural products, peptides possess diverse biological roles and have received tremendous attention in basic and translational research [43,86]. One specific utility for peptides is to target “undruggable” space, e.g., protein-protein interactions, which is less accessible with other types of small molecules since peptidic compounds demonstrate increased specificity and less off-target effects [115,43,86]. Ribosomally synthesized and post-translationally modified peptides (RiPPs) are one category of peptidic natural peptides that are produced by all major domains of life and contain a wide variety of structural diversity [2]. These peptides demonstrate a wide array of biological activities, e.g., inhibiting nucleotides/proteins synthesis, chelating metal ions, and permeabilizing cell membranes [5,21,41,70,127]. RiPPs also play important roles in internal regulation and host-bacteria interactions [5,70,127]. Of note, the RiPP nisin has been used for food preservation since the late 1960s. As new RiPPs are rapidly discovered, their applications in drug development and other areas will continue to expand.

So far, over 22 RiPP families have been discovered with distinct structural features, e.g., lanthipeptides, thiopeptides and lasso peptides, while their biosynthesis follows a common logic (Fig. 1A) [5]. The RiPP biosynthetic gene clusters (BGCs) minimally consist of one precursor peptide and one or more pathway-specific modification enzymes. The precursor peptide generally includes an N-terminal leader peptide (LP), which often dictates the binding of modification enzymes, and a C-terminal core peptide (CP), which is enzymatically modified and then released to form the advanced biosynthetic intermediate or final product (Fig. 1A). A long list of peptide modification enzymes promotes a wide variety of modifications on the CP, e.g., methylation, oxidation/reduction, dehydration, crosslinks between amino acid residues, and macrocyclization; the chemical diversity of RiPPs rivals that of nonribosomal peptides, another major category of peptidic natural products [3,5,23,87,114,127]. Importantly, capable bioinformatics tools have been developed to predict and discover new RiPP analogs and even novel families from organism genomes [11,20,55–58,64,106–108,113]. The known and newly mined RiPPs gene clusters can be further categorized into three groups based on the status of their precursor peptides. The gene clusters in the first group encode only a single precursor peptide that contains a single CP, while those in the second group carry multiple precursor peptides (Fig. 1B). The single CP of these multiple precursors can range from identical sequences to very low similarity. The RiPP gene clusters in the last group have one or more precursor peptides, each of which consists of multiple CPs after the single LP (Fig. 1B). Compared with the first two RiPP groups, the biosynthetic understanding of the multicore RiPPs remains underexplored. On the other hand, the multicore system holds the promising potential to expand peptide chemical diversity with an economical genetic cost, aiding improved target selectivity and potency in peptide drug research [8,25,44,97,101,125,126]. Herein, we review current knowledge of all known multicore RiPPs systems in bacteria, fungi and plants, highlighting their prevalence, distribution, bioactivity, and biosynthetic logic compared to their respective single-core systems. We will further highlight engineering efforts in developing these RiPPs for biomedical applications.

Fig. 1.

A. General scheme of RiPPs biosynthesis. B. Three groups of precursor peptides found in the RiPP gene clusters. The core peptide of the precursor peptide sometimes carries flanking sequences at its N- and C-termini, named as N-flanking (NF) and C-flanking (CF), respectively.

1. Bacterial multicore RiPPs

The vast majority of RiPPs are of bacterial origins but only a few demonstrate apparent multicore features in the biosynthesis, with repeated CP units within a single precursor peptide, including cyanobactins and microviridins as well as some Ω-ester peptides. We will review each of these RiPPs families along with putative multicore RiPPs discovered through genome mining efforts.

1.1. Cyanobactins:

Cyanobactins were first discovered in 1989 as natural products of marine ascadians, but were later shown to be biosynthesized by the symbiotic cyanobacterium Prochloron didemni [40,47,90,101]. To date, dozens of cyanobactins have been isolated from multiple cyanobacterial species (Fig. 2). Importantly, over 20% of cyanobacterial genomes are predicted to contain at least one cyanobactin cluster [66,105], while the similar clusters have been discovered in other bacterial phyla, such as Actinomycetes [38,53]. The native functions of cyanobactins remain unclear, but these peptides have been shown to execute diverse biological activities, including cancer cytotoxicity, growth and protease inhibition, metal chelation, and reducing antibiotic drug resistance [45,105].

Fig. 2.

Biosynthetic scheme of cyanobactins. The representative gene cluster (pat) of patellamide is shown at the top. Pathway-specific enzymes demonstrate different features toward precursor peptides with a single CP and multiple CPs. Chemical structures of representative cyanobactins, patellamide A and trunkamide, are shown and their thiazole and prenyl moieties are shadowed in green and light blue, respectively. RS: recognition site.

The biosynthesis of cyanobactins obeys the general logic of RiPPs (Fig. 1A) [24,67,73], and demonstrates obvious biosynthetic plasticity in generating chemical diversity. Some cyanobactin gene clusters contain up to 10 precursor peptides, each of which carries a single CP, while others encode a single precursor peptide with up to 4 CP motifs [99]. Our current understanding of cyanobactin biosynthesis mainly comes from the precursor peptides with 2 to 3 CPs, e.g., PatE from the patellamide biosynthetic pathway, while the engineered single-core precursors have often been used for easing enzymatic and engineering studies [33,39]. The cyanobactin CP generally consists of 6 to 8 amino acids with cysteine, serine, threonine residues separated by one nonconserved residue, e.g., xTxCxSxC. Multiple enzymes install modifications on each CP, including heterocyclase (e.g., PatD), N-terminal protease (e.g., PatA), C-terminal protease/cyclase (e.g., PatG), oxidase (e.g., ThcOx), and prenyltransferase (e.g., PatF) (Fig. 2). Some modification enzymes contain a RiPP precursor peptide recognition element (RRE) that is a PqqD domain for recognizing and binding a helix-turn-helix motif typically found in the LP [7]. In addition, some cyanobactin biosynthetic enzymes bind to another type of conserved motifs in the precursor peptide, termed recognition sites (RSs) (Fig. 2). Recognition site 1 (RSI) is N-terminal to the CP and typically has a conserved sequence of LAELSEE. RSI guides the heterocyclization of the conserved cysteine, serine, and threonine in the CP. The cyanobactin heterocyclases contain an ATP-dependent YcaO domain [9]. The heterocyclization within the cyanobactin CP occurs in a C to N direction, with cysteines modified before threonine or serine residues. The majority of cyanobactin CPs contain a C-terminal heterocycle, typically thiazol(in)e [38,67,105]. On the other hand, some cyanobactin BGCs have no heterocyclase and their precursor peptides carry no RSI but often have a C-terminal proline in the CP, e.g., xTxCxSxP [99]. Following RSI, RSII with a conserved sequence of Gx(E/D)xS flanks the CP on the N-terminus, while RSIII as the motif of (A/S)YD is C-terminal to the CP (Fig. 2). RSII interacts with the N-terminal subtilisin-like protease (PatA) to release the heterocyclized CP C-terminal to the conserved Ser of the RSII motif. RSIII in the released modified CP then guides the C-terminal subtilisin-like protease (e.g., PatP) to hydrolyze the amide bond N-terminal to its conserved Ala/Ser residue. In addition, the latter protease often catalyzes head to tail cyclization of the released cyanobactins through a transamidation, but linear cyanobactins have also been found [1,60,61,65,67,76].

The modified CP cleaved from the cyanobactin precursor peptide can undergo additional modifications, commonly oxidation and prenylation, to generate the final product, e.g., patellamides and trunkamides (Fig. 2) [46,77]. The FMN-dependent oxidase can be fused with the C-terminal protease (e.g., PatG) or as a stand-alone enzyme (e.g., ThcOx) [32]. Oxidation on the azolines (e.g., thiazolines) proceeds in an N to C direction on both linear and macrocyclic peptides [32,46]. The oxidase domain contains an RRE that is important to enzyme stability but shows no clear binding to the cyanobactin precursor peptide [32]. The prenylation is often the last possible modification on cyanobactins [75,97]. The prenyltransferase LynF modifies the –OH group of tyrosine residues, while TruF acts on the –OH group of serine and threonine residues. After the O-prenylation, a spontaneous Claisen rearrangement can occur to produce an ortho-C-prenylated cyanobactin. On the other hand, the prenyltransferase is fused with a methyltransferase as a single enzyme AgeMTPT for the biosynthesis of linear cyanobactins aeruginosamides [97]. In this case, after heterocyclization and N-terminal cleavage, the exposed N-terminus is prenylated to prevent the transamidation reaction by the C-terminal protease AgeG that cleaves at RSIII. The AgeMTPT further protects the C-terminus by the methylation [97].

Multicore cyanobactin precursors are often found in the patellamide/trunkamide-like BGCs and contain up to 4 CP motifs. Each CP is flanked by an RSII and RSIII with only one RSI site before the first CP, if heterocyclization enzymes are present (Fig. 2) [39,99]. The modifications on these CPs demonstrate distinct features, although the overall modifications on each CP are the same as the single-core system (Fig. 2) [98]. In contrast to the engineered single-core precursor peptide, the heterocyclase iteratively binds to, releases, and rebinds to the parental substrate and modified intermediates during the reaction, and displays no preferred order in modifying the multiple CPs. Indeed, cyanobactin intermediates have been detected with cyclization on either CP [39]. The N-terminal protease prefers the cleavage at the first CP under non-reductive conditions, but distributive catalysis still occurs with initial cleavage at the second CP as a minor product (Fig. 2). Interestingly, the order of cyclization and cleavage is redox-state dependent [39]. Overall, in the multicore processing, the majority of cyanobactin modification enzymes are distributive and tolerant to core sequence variations, leading to the generation of multiple intermediates depending on the processing order [39].

The inherent biosynthetic plasticity of cyanobactins can be leveraged to produce unnatural analogs. Several libraries of unnatural cyanobactins have been synthesized, with most proteinogenic residues being accepted at several positions within the CP, without affecting enzyme activity [95]. Furthermore, cyanobactin biosynthetic enzymes successfully processed the precursor peptides carrying non-proteinogenic residues in the CP [46,100,112]. For example, the macrocyclase PatG has been shown to cyclize compounds with no natural residues or the presence of heterocycles [87]. Remarkably, the use of modification enzymes from different cyanobactin pathways further expanded the chemical diversity, generating hundreds of new compounds [46,95,98]. Importantly, the engineering of the multicore system, in conjunction with combinatorial biosynthesis, is expected to lead to new or improved activities for a wide range of applications.

1.2. Microviridins:

Microviridins were discovered in 1990 from an extraction of the cyanobacterium Microcystis viridis [49]. Since then, tens of microviridins have been isolated from a variety of cyanobacterial strains [2,128]. These tricyclic peptides possess potent and selective serine protease inhibitory activity (Fig. 3A) [80,93,104,128]. In addition, microviridin J has demonstrated the toxicity to the water flea Daphnia pulicaria that is the predator of the producing cyanobacterium [93,94].

Fig. 3.

A. A representative gene cluster (mdn) of microviridin. The annotated functions of biosynthetic proteins are shown at the top. MdnE encodes a precursor peptide. B. General scheme of microviridin biosynthesis. The interaction of LP/enzyme is shown on the top right. C. Structure of the most abundant modified AMdnA in the reaction of AMdnC. Each CP in AMdnA is shadowed in light blue and the first and second macrolactones are shown as red and brown lines, respectively. The blue dashed lines indicate the predicted macrolactams.

Microviridin BGCs have a minimal gene set of a precursor peptide and two ATP-grasp ligases but can contain additional genes, including an acyltransferase and an ABC transporter. The ATP-grasp ligases contain an atypical ATP-binding site, named the ATP-grasp fold, and typically catalyze the intermolecular formation of a peptide or ester bond. The microviridin precursor peptide (e.g., MdnA) carries an N-terminal conserved leader motif of PFFARFL and a conserved 10-amino acid core motif of 1-TxKYPSDx(D/E)(D/E)-10 (Fig. 3A) [88]. The two ATP-grasp ligases (e.g., MdnB and MdnC) interact with the conserved leader motif and install three macrocycles on the CP. The first ATP-grasp ligase MdnC lactonizes the CP between the side chains of its Thr1 and Asp7. Upon the formation of the first macrolactone, MdnC further forms the second macrolactone between the side chains of Ser6 and Asp/Glu9. The bicyclic microviridin CP is then modified by the second ATP-grasp ligase MdnB to install one macrolactam between the side chains of Lys3 and Asp/Glu10 (Fig. 3A) [69,88,121]. The order of these three cyclizations is strict [88], while natural bicyclic microviridin analogs have been isolated when their core Lys3 is mutated to Arg [109,111]. Interestingly, recent structural studies revealed that neither MdnC nor MdnB carries the RRE domain [69]. Instead, MdnC and MdnB share the standard ATP-grasp ligase structural motifs comprising of N-domain, central domain and C-domain, and their central domain interacts with an α-helical element formed from the conserved leader motif (Fig. 3A) [69], representing a new strategy of substrate/enzyme recognition in the RiPPs biosynthesis. The LP/enzyme interaction opens the catalytic pocket formed by residues from the N- and C-domains for cyclizing the CP [69]. Furthermore, the leader motif can activate the ATP-grasp ligases for the cyclizations when it is fused to the enzymes or added in trans [2,69,91,121]. The remaining biosynthetic steps of microviridin include the proteolytic cleavage of the LP and the acetylation of N-terminus of the released CP to produce mature microviridins that may be subjected to a pathway-specific ABC transporter [2,121,128].

Recent bioinformatics studies identified hundreds of microviridin BGCs from the genomes of cyanobacteria, as well as those of proteobacteria, bacteriodetes, and acidobacteria phyla [2]. These mined clusters demonstrate tremendous diversity as some encode up to 10 precursor peptides, while the precursor peptides of other clusters contain up to five CPs [2,127]. Of note, the microviridin multicore precursors appear in the clusters found in all four bacterial phyla [2], presumably reflecting the biological significance of the multicore design. The first understanding of the enzymatic processing of the microviridin multicore systems comes from the research of the cluster in the filamentous cyanobacterium Anabaena sp. PCC 7120 [127]. In this cluster, the precursor peptide AMdnA has three predicted CPs and the lactone-forming ATP-grasp ligase AMdnC cyclizes these cores with a favored N to C directionality (Fig. 3B) [127]. On the other hand, the first CP of the most abundant product is macrocyclized only once, while two macrolactones are installed on each of the remaining CPs (Fig. 3B). The order of two cyclization reactions on the two CPs of AMdnA is the same as the single-core containing MdnA. These results together suggest that the modification status on one CP of AMdnA has no, to minimal, influence on the enzymatic macrocyclization of others, highlighting the enzyme’s substrate promiscuity. The broad substrate scope of AMdnC is further supported by its reactions on engineered AMdnA analogs with one to four CPs and the noncognate substrate MdnA. Similar to the biosynthetic enzymes of cyanobactins, AMdnC is distributive in cyclizing CPs within AMdnA [39,127]. The distributive catalysis may be one hallmark of the processing of multicore precursors in the RiPPs biosynthesis. A conserved GG motif is often available between two microviridin CPs that can be cleaved by a C39 protease [2]. However, individual microviridins have not been isolated from any multicore systems yet [2,127].

The plasticity of microviridin biosynthetic systems has allowed engineering efforts to expand the chemical diversity of these peptides for the discovery of potent and selective serine protease inhibitors [2,128,122]. Almost all proteinogenic residues are tolerant in the nonconserved sites of the CP to generate diverse peptide libraries [91,122]. It is expected that the multicore systems can further facilitate the synthesis of new analogs that can be used to screen for inhibitors of serine proteases essential to all forms of life.

1.3. Ω-Ester peptides:

In the continued effort to uncover microviridin-like peptides, the Kim laboratory predicted a total of 12 putative peptide groups whose precursor peptides demonstrate distinct features but contain multiple potential pairs of Thr/Ser/Lys and Asp/Glu residues for the ATP-grasp ligase mediated macrocyclizations [63]. Remarkably, similar to microviridins (Group 1), some precursor peptides from Groups 2 to 6 also carry multiple CP repeats [63]. Among them, the biosynthesis of plesiocin (Group 2) and thuringinins (Group 3) have been studied in good detail, as discussed below.

The plesiocin gene cluster was first mined from the genome of a marine myxobacterium Plesiocystis pacifica in 2017, and its putative clusters were later found to be widely distributed in actinobacteria, cyanobacteria, firmicutes, proteobacteria, acidobacteria, euryarchaeota, and plantomycetes [63,64]. The understanding of the plesiocin biosynthesis came from the in vitro biochemical characterization of the plesiocin precursor PsnA2 processed by the pathway specific ATP-grasp ligase PsnB (Fig. 4A). PsnA2 carries four repeats of 1-TTLxxGEE-8 separated by 2 to 4 residues, while its homologs can have up to 15 repeats. Recombinant PsnB prepared from E. coli successfully processed all four repeats of PsnA2 by forming two ester linkages between the side chains of Thr1 and Glu8 and of Thr2 and Glu7 in each repeat, yielding four hairpin-like bicyclic units. PsnB was distributive in modifying these four repeats but demonstrated low activity toward engineered repeats with more than four residues between the inner pair of Thr and Glu [62]. Furthermore, the plesiocin precursor does not share the same conserved leader motif as microviridins, suggesting a new way for the interactions of leader/ATP-grasp ligase in the peptide macrocyclization. The minimal leader peptide for the effective processing was determined with serial truncated PsnA2 analogs (Fig. 4A). Although it is still unclear if one or more repeats of plesiocin are proteolytically released in the native producers, both plesiocin and its engineered single repeat inhibit serine proteases elastase and chymotrypsin, but not trypsin, and plesiocin is 10 to 20-fold more potent. On the other hand, the replacement of Leu3 with Arg introduced potent trypsin activity. In this regard, plesiocins can be engineered to target one or multiple proteases at once potently and selectively.

Fig. 4.

The processing of multiple CP repeats within the plesiocins (A) and thuringinin (B) precursor peptides PsnA2 and TgnA by their cognate ATP-grasp ligases PsnB and TgnB, respectively. Conserved key residues in each repeat that is shadowed in light blue are highlighted in bold, while the minimal leader peptides potentially for the interactions with ATP-grasp ligases are shadowed in light grey. The ester bonds formed on each repeat are shown with red lines. In the thuringinin biosynthesis, the inner macrolactone is formed first and the outer cycles are shown with brown lines.

Similar to plesiocins, the thuringinin (Tgn) gene cluster was first mined from the genome of one Bacillus species in 2019 and appeared in multiple species of Firmicutes and Proteobacteria [92]. All clusters consist of at least one precursor peptide (e.g., TgnA), a transmembrane protein (e.g., TgnC), and an ATP-grasp ligase (TgnB) (Fig. 4B). The thuringinin precursor peptides have an N-terminal leader and 2 to 7 repeats of the conserved motif of 1-TxxTxxxExxD-11 at the C-terminus. TgnA from the Bacillus species carries three repeats and the co-expression with its cognate TgnB in E. coli generated the recombinant TgnA that receives six macrolactones, two on each repeat (Fig. 4B) [92]. The first ester linkage is formed between the side chains of Thr4 and Glu8 and the other from Thr1 and Asp11. Remarkably, the conserved leader motif of TgnA for the binding of TgnB is discovered by substrate engineering and distinct from those of microviridin and plesiocin precursor peptides (Fig. 4B), indicating that ATP-grasp ligases can evolve multiple modes of protein/protein interactions in the RiPPs biosynthesis. Processed thuringinins with the attached leader showed no inhibitory activity against several serine proteases. Furthermore, it is unclear if any repeats of thuringinins are proteolytically released.

2. Fungal multicore RiPPs

So far, only several families of fungal RiPPs have been isolated, e.g., amatoxins and phallotoxins [4,5,72,123]. On the other hand, multicore precursor peptides are increasingly being discovered due to improvements of genome mining approaches. The biosynthetic understanding of multicore fungal RiPPs has been elucidated, in part, for dikaritin-type peptides.

Dikaritins are a family of cyclic peptides produced by fungal species of the Dikarya subkingdom [28]. Within this family are currently ustiloxins, phomopsins, and asperipins (Fig. 5) [117]. The characteristic cycle of the dikaritin family is formed between the aryl hydroxyl group of Tyr and ß-carbon of another residue, e.g., Ile for ustiloxins and phomopsins. Ustiloxins were first isolated from parasitic false smut balls growing on rice plants and were shown to inhibit microtubule assembly and mitosis of human tumor cell lines in the low μM range and are extremely toxic against mice [89,90]. Similarly, phomopsins are a group of hexapeptide mycotoxins that are the major cause of lupinosis disease and inhibit microtubule assembly [19,117]. In contrast, asperipins were recently identified from Aspergillus flavus primarily driven by searching biosynthetic enzyme homologs of ustiloxins from fungal genomes [82].

Fig. 5.

A. Structures of representative phomopsins and asperipins. B. A schematic representation of the ustiloxin B (ust) gene cluster. The precursor peptide gene is shown in dark red, while the ustQ, ustYa and ustYB genes are in orange. Two peptidase genes ustP1 and ustP2 are shown in blue. C. General scheme of ustiloxin biosynthesis. The leader peptide region is in grey. The tetrapeptide core in light blue is flanked with conserved N and C-terminal sequences in green and orange, respectively. CP* represents the modified CP. The details of the majority of biosynthetic steps remain unavailable. Ustiloxin F and B are used as representative examples.

The biosynthesis of dikaritins has been characterized to a certain degree in the biochemical and genetic studies of ustiloxins and phomopsins. The ustiloxin BGCs contain 11 to 17 genes, encoding a precursor peptide (e.g., UstA), a copper-dependent tyrosinase (e.g., UstQ), and a variety of other tailoring enzymes, e.g., flavin-dependent monooxygenases (e.g., UstF1-F2), P450 (e.g., UstC), methyltransferase (e.g., UstM), and pyridoxamine 5′-phosphate (PLP) oxidase (e.g., UstO) [82,116,125]. The precursor peptide UstA from Aspergillus flavus contains 16 repeats with a highly conserved sequence motif of NSVEDYAIGIDKR, of which the tetrapeptide YAIG fragment appears in the final structures of ustiloxins (Fig. 5B). A combination of functional prediction, gene knockouts, and biochemical characterization suggests key modifications on the conserved repeats, but the details of these modifications (e.g., order and distributive catalysis) are still unknown. Briefly, the processing requires multiple proteases to release the conserved repeats and the modified or unmodified tetrapeptide. A subtilisin-like endoprotease Kex2 is expected to release the repeat fragments at the C-terminus of the conserved KR motif but this protease is not pathway-specific. The ustiloxin BGC encodes two S41 family peptidases UstP1 and UstP2 that may remove the N and C-terminal sequences of the tetrapeptide core in the released repeat fragments. The conversion of the tetrapeptide YAIG into ustiloxin F requires a series of reactions, including β-hydroxylation and 3-hydroxylation on the phenol of Tyr, crosslinking between 3-OH of Tyr and β-carbon of Ile and N-methylation of the primary amine of the core motif [82,116,125]. The tyrosinase UstQ is likely responsible for the phenol hydroxylation, while UstM catalyzes the N-methylation (Fig. 5B). Two essential uncharacterized proteins UstYa and UstYb may catalyze the β-hydroxylation of Tyr and the oxidative cyclization. Of note, except UstM, the order and other details of the proteolysis and tailoring reactions in processing the multicore UstA are completely unknown. Furthermore, it is unclear if and how the processing enzymes interact with the LP in the modifications. The synthesis of other ustiloxin analogs requires additional modifications. For example, a norvaline residue is added through a sulfur linkage at the tyrosine benzyl group of ustiloxin F to generate ustiloxin B, which likely requires a cytochrome P450 UstC, a cysteine desulfurase UstD, a glutathione S-transferase UstS, two flavin-dependent enzymes UstF1 and UstF2, and a PLP-dependent oxidase UstO (Fig. 5B) [116,125]. The functions of UstF1, UstF2 and UstD in the biosynthesis of ustiloxin B have been confirmed in both genetic and biochemical studies [116,125]. The biosynthesis of phomopsins is highly similar to ustiloxins [117]. The precursor peptides of phomopsins carry multiple repeats with a highly conserved sequence motif of VEDY(V/A/F)I(G/P)(I/V/A)DKR. The central hexapeptide fragment appears in the final structures of phomopsins. The proteolytic cleavages of the native or modified repeats are presumably mediated by Kex2 and one pathway-specific S41 family peptidease PhomP1, while the modifications on the hexapeptide can be catalyzed by the tyrosinase PhomQ1, one or more of five uncharacterized enzymes PhomYa-e, a hydroxylase PhomE, an oxidoreductase PhomF, and a methyltransferase PhomM [117]. Again, the order and details of these potential processing steps in the phomopsin biosynthesis have not been studied yet. Collectively, the known dikaritin precursor peptides contain multiple repeats of conserved sequence motifs and the biosynthesis requires multiple proteolysis steps and enzymatic modifications, particularly aromatic hydroxylation and oxidative crosslinking, which await further characterization. The similar biosynthetic logic can be employed in the biosynthesis of epichloëcyclins that are produced by Epichloë fungi associate with grasses in mutualistic to parasitic manners [54,55]. Furthermore, the relatively conserved modifications and the variations on several positions of the dikaritin CPs indicate great potential in the creation of diverse libraries and combinatorial biosynthesis [22,82,116,125].

3. Plant multicore RiPPs

Plant RiPPs are stable cyclic peptides of significant pharmacological interests. These peptides are very diverse in the size and residue components. Although they can be produced from the precursor peptides containing a single CP, recent genome mining studies revealed the prevalence of multicore precursor peptides. Here we will review the biosynthesis and engineering of three known families, cyclotides, orbitides and lyciumins.

3.1. Cyclotides:

Cyclotides were first discovered from Oldelandia affinis DC, a plant used by African women to accelerate contractions and childbirth [35,36,96], and hundreds of analogs have been identified from many other plants [15,18,79,119]. These compounds have a common characteristic structural feature, a cysteine knot in which a ring formed by two disulfide bonds is threaded by a third disulfide linkage (Fig. 6A). This structural fold is further divided into three subfamilies, bracelet-like in which the backbone is not twisted (e.g., Cycloviolacin O2), Möbius where the peptide is cyclized through a backbone twist (e.g., Kalata B12), resembling a Möbius strip, and the trypsin inhibitor (e.g., MCoTI-II), which has a longer sequence resulting in a looser cysteine knot (Fig. 6A) [34,77]. The bracelet family is the most common, accounting for approximately two-thirds of known cyclotides, followed by the Möbius family [120]. On the other hand, while many cyclotides are also head to tail cyclized, there are known linear cyclotides [18,48,96]. Cyclotides often contain 28 to 37 amino acid residues, including six conserved cysteines at specific positions for forming the characteristic fold. Furthermore, the first and last residues of cyclotides are generally A/G and N/D, respectively [17,52], while the rest are highly variable. The sequence diversity may be related to their potential biological role as a host defense factor as some plants produce large quantities of these compounds (e.g., 1g compound/kg leaf extract) [18], while cyclotides also demonstrate diverse other activities, including antiviral, protease inhibition, insecticidal, and cytotoxic activities.

Fig. 6.

A. Representative cyclotides showing structural variations. The PDB IDs of Cycloviolacin O2, Kalata 12 and MCoTI-II are 2KCG, 2KVX, and 1HA9, respectively. The figures were created by Pymol. The disulfide bond is shown in brown and the brown sphere represents sulfur atom. B. General scheme of cyclotide and orbitide biosynthesis. The orange line represents a disulfide bond. C. Representative orbitide structures. The peptide bond for the cyclization is shadowed in orange.

Cyclotide BGCs have been uncovered with up to 10 precursor peptides, while each precursor peptide can contain up to eight CP motifs. In the multicore systems, these CPs can encode both identical and unique cyclotides, i.e., the cyclotide precursor peptide from Cucurbitacaeae can produce four distinct products from six CP units [17,18,52,81]. Each CP of the cyclotide precursor peptide is flanked by two conserved motifs, termed terminal repeats (NTR/CTR) (Fig. 6B) [41]. On the other hand, the NTR/CTR sequences of the precursor peptides from different plants are highly variable. Furthermore, the cyclotide CPs have been found to be embedded in the albumin gene, which normally encode hormone-like peptides with insecticidal properties [37,89]. The albumin BGC can generate the active cyclotide, without clear cyclotide NTR/CTR [90]. The same “hijacking” mechanism has also been found in the biosynthesis of cyclic plant peptides PawS and PawL [27,30,31,50].

The cyclotide precursor peptide is ribosomally translated and targeted to endoplasmic reticulum (ER) by its N-terminal ER signal peptide, where the signal peptide is removed (Fig. 6B). The details of the remaining steps, particularly the order, in the cyclotide biosynthesis have not been clearly defined. Briefly, a protein-disulfide isomerase (PDI) is expected to catalyze the formation of the three disulfide bonds on each CP in the ER [42]. It is unclear if and how the leader peptide interacts with PDI and other processing enzymes in the cyclotide biosynthesis. The modified precursor peptide is then transported to the vacuole, where a yet-identified enzyme first cleaves the NTR of each modified CP. The exposed amino group of A/G of each modified CP then acts as a nucleophile for intramolecular transpeptidation mediated by an asparaginyl endopeptidase (AEP) to produce cyclic cyclotides (Fig. 6B) [14,17,83,84]. On the other hand, the head-to-tail cyclization mediated by AEP is independent of the presence of disulfide bonds and the enzyme is promiscuous. Additional modifications to the cyclic cyclotides are present, e.g., reductions and alkylations [102]. Furthermore, the current understanding of the cyclotide biosynthesis has mainly relied on the use of natural or engineered single-core precursor peptides, and the processing details of multicore systems remains unknown.

Given the availability of multiple nonconserved sites, cyclotides have been used for molecular grafting, in which epitopes are engineered into cyclotides, for improving the stability and efficacy of the epitope and introducing new bioactivities for the recombinant cyclotides [16,118]. Unfortunately, some epitopes prevent enzymatic cyclization. Recently, an enzymatic method for improved cyclization of epitope-linked cyclotides has been reported [108]. In this case, an alternative cyclization site was introduced, allowing for efficient ligation and efficacy of the bioactive epitope, without affecting normal cyclization of the peptide. This improved method opens new opportunities to create large cyclotide libraries in a rapid and low-cost manner.

3.2. Orbitides:

Orbitides are another family of plant cyclic peptides that contain 5–16 residues (Fig. 6C) [6,28]. Over 130 orbitides have so far been discovered and some possess anti-inflammatory, antiplatelet, antimalarial, immunosuppressive, and cytotoxic activities [12,13,79,85,110,119,124]. Compared with cyclotides, orbitides are also head-to-tail cyclized through the peptide backbone, but have no disulfide bridges. On the other hand, the biosynthesis of cyclotides and orbitides share multiple common features (Fig. 6B). The orbitide precursors contain up to nine CP motifs whose sequence components and sizes can be varied [29,86,103]. Each orbitide CP is flanked by conserved NTR/CTR repeats whose sequences often differ among plants. Similar to cyclotides, the orbitide CPs have been found in the albumin gene cluster [30,50]. The processing of the CP is believed to start by the removal of NTR by a serine protease OLP1, followed by the cleavage of CTR and the intramolecular transpeptidation by a peptide cyclase of the S9A protease family, PCY1 (Fig. 6B) [6,10,13,71]. Additional modifications on the cyclized orbitides have been observed [68,109]. Similar to AEP in the biosynthesis of cyclotides, PCY1 is promiscuous in modifying a variety of non-native substrates [106], suggesting its biotechnological applications for the synthesis of the orbitide libraries. Similar to the cyclotide biosynthesis, the details of the enzymatic processing of the orbitide multicore system, including the LP/protein interactions, have yet been deduced.

3.3. Lyciumins:

Lyciumins are a family of branched cyclic peptides produced by Lyceum barbarum (Fig. 7) and are potent inhibitors of proteases, angiotensin-converting enzymes and renin. The biosynthetic logic of lyciumins was uncovered in 2018 through genome-mining efforts [59]. The first precursor gene discovered, LbaLycA, contains 12 CP motifs coding for four different compounds, while the precursor peptides mined from other plants can carry 1 to 18 CPs [59]. In addition to the CPs, the precursor peptides have an N-terminal ER signal and a C-terminal BURP domain that is specific in terrestrial plants [42,59]. The CPs can reside within the BURP domain or be N-terminal to it. The 12 CPs in the LbaLycA show a conserved motif of 1-QP(W/Y)GVG(I/S)W-8, with a kexin-like cleavage site N-terminal to each core. However, the LP portion of LbaLycA and its role in the CP processing have not been defined. Biochemical studies of the lyciumin biosynthesis have not been available, while the functional prediction and heterologous expression suggested that the processing is started by cyclizing the CP by a C-N bond between the α-carbon Gly4 and the indole nitrogen of Trp8, likely through a radical mechanism (Fig. 7). Subsequently, the N-terminus of each modified CP is exposed after the cleavage by yet-identified endopeptidases. The Gln1 is then converted into a pyroglutamate moiety by a pathway-specific glutamine cyclotransferase. The final products are then generated by proteolytic cleavage C-terminal to each modified CP (Fig. 7). The processing enzymes of lyciumins are tolerant alanine replacement of the CP at every position other than the residues involved in cyclization [59]. Furthermore, the CP with a Gly4Thr mutant was converted into a cyclized product, Lyciumin I, as well as a dehydrothreonine derivative. On the other hand, the details in processing the lyciumin multicore system have not been studied. Collectively, the enzymes involved in the lyciumin biosynthesis show different biotransformations with those of cyclotides and orbitide (Fig. 6).

Fig. 7.

General scheme of lyciumin biosynthesis. Key chemical transformations are shadowed in orange. R1 and R2 represent the side chains of Trp/Tyr and Ile/Ser, respectively.

4. Conclusion and future perspectives

The wide-spread presence of multiple core motifs within the precursors of different RiPPs families is a recently discovered phenomenon. The multicore biosynthesis allows for increased peptide production and diversity, without the added cost of translating multiple precursors or enzymes. This review highlights multicore systems in both eukaryotic and prokaryotic organisms, outlining the current biosynthetic understanding of these systems and differences between multicore and single-core, as well as the engineering. As a bonus for pharmacological research, enzymes in the multicore systems are naturally promiscuous to account for changes in size or residues within the various CPs. This inherent promiscuity enables the incorporation of canonical and nonproteinogenic residues, as well as medically relevant epitopes, to enhance the potency and selectivity. Indeed, the multicore system has been utilized to generate diversity in cyanobactin libraries, clearly demonstrating the potential of these systems for improving peptide production efficiency. However, a major limitation to broad utilization of these multicore systems is our limited understanding of how the enzymes act on multiple core-containing substrates. Indeed, the biosynthetic logic of the multicore systems has often been characterized with engineered single-core substrates that significantly ease the monitoring of CP modifications [47]. However, the simplified substrates lead to the missing of critical biochemical and mechanistic insights into the enzymatic processing of multicore systems, e.g., the timing and order of different modifications on the multiple CPs and the potential cooperation of different biosynthetic enzymes in the processing [39, 127]. In this regard, the actual promiscuity and combinatorial biosynthesis potential of the enzymes of the multicore systems have not been assessed in detail. On the other hand, as the understanding of the multicore biosynthetic logic improves with increased research efforts, we believe that there is tremendous potential for the generation of large non-natural peptide libraries with demonstrated biological activity. In addition, different sets of enzymes used in the biosynthesis of the same RiPP family members and proven enzyme promiscuity enable combinatorial biosynthesis, further diversifying these peptides. While many single-core RiPPs have potent bioactivities, their multicore counterparts can demonstrate increased potency and specificity [64]. Importantly, these multicore peptides can execute different bioactivities by engineering individual CPs [62]. We are optimistic that through their natural diversity-generating pathways in conjunction with substrate and enzyme engineering, multicore RiPPs systems can greatly improve basic and translational studies and expand industrial applications.