**4. Microviridin Biosynthesis**

Owing to their atypical conformation, microviridins have been mistakenly labeled as nonribosomal peptides. This concept has been discarded, because numerous studies have failed in the quest for biosynthetic gene clusters with mechanisms linked to NRPS genes and being similar to ribosomally biosynthesized peptides, such as cyanobactins (patellamides, tencyclamides and patellins) and trichamide [15,33]. In addition, NRPS products usually have nonproteinogenic amino acids in their structure and can be paired with hydroxy acids. Furthermore, their amino acids can also be in a D-configuration. These characteristics are not usually present in the family of microviridins [15,33]. Microviridins have recently been identified as ω-ester-containing peptides, along with plesiocins and thuringinins of the ribosomally synthesized and post-translationally modified peptide (RiPP) family [34].

Apart from the fact that microviridins have been isolated and characterized since the 1990s, their biosynthesis started to be elucidated by two groups independently using separate approaches in 2008 [14,15]. Firstly, Ziemert et al. [14] pursued a NRPS gene cluster related to microviridin production in *Anabaena*; however, they detected a gene with similar sequence to microviridin, known as *mdnA*. In the immediate proximity of *mdnA*, two additional genes were discovered, named *mdnB* and *mdnC* [14]. In comparison, Philmus et al. 2008 [15] detected similar genes from *Planktothrix agardhii*. This filamentous cyanobacterium possesses a homologous *mdnA* sequence, named *mvdE*, and homologous genes of *mdnB* and *C* encoding two ATP-grasp ligases (*mvdB* and *mvdC*). In addition, an acetyl transferase (*mvdB*) and an ATP-binding cassette transporter (*mvdA*) were detected, which their homologous genes were identified in *Microcystis* named *mdnD* and *mdnE*, respectively (Table 1) [15].


**Table 1.** Genes involved in microviridins biosynthesis.

These genes have been analyzed by various methods, confirming their roles during the synthesis of microviridins. The heterologous expression of microviridin B *mdnA-C* genes from *Microcystis* in *E. coli* produced a tricyclic microviridin-lacking leader peptide [14]. Concurrently, the in vitro reconstitution of the MvdB-E enzymes from *P. agardhii* also confirmed that these genes were linked to the production of microviridins [15]. These studies were important to demonstrate that the microviridin biosynthetic clusters have different organizations, with or without different genes (Figure 6) [14,15,35].

**Figure 6.** Graphical representation of microviridin biosynthetic clusters. The gene cluster compilation was accomplished through the Gene Graphics application (https://katlabs.cc/genegraphics/app).

Through an extensive bioinformatics study of microviridin biosynthetic gene clusters, a number of variations between them have been identified. The majority of these clusters consisted of *mdnA-C* genes, where *mdnB* and -*C* are normally in strict order. However, *mdnD* is only present in a subset of the clusters found. In comparison, *mdnE* is also absent in microviridin gene clusters or replaced by the C39 peptidase, which is followed by the HlyD3 homolog protein, normally linked to the transport of proteases across membranes. Several other gene clusters carry additional proteins, likely linked to the noncommon post-translational modification of the core sequence, such as *mdnF* and *G* [21,33].

One of the first steps needed to produce a completely tricyclic N-acetylated microviridin is the production of prepeptide. The microviridin precursor gene (*mdnA*) produces an immature peptide that its leader peptide (LP) has preserved among different variants and possessing a highly conserved PFFARFL motif among the microviridin gene clusters, which has a α-helix structure in a solution (Figure 7) [36]. The core sequence frequently contains Asp, Thr, Ser and Lys residues, as well as the TxKxPSD motif, in which both features are related, to form lactone and lactam rings [20]. When evaluating different cyclized peptides with ω-ester and ω-amide bonds, their core sequences have a high frequency of conserved Thr and Glu residues, which are highly related to lactone ring formations. In addition, when contrasting plesiocin, thuringinins and microviridins, the residues involved in both ester and amide bonds are arranged in a similar order: the nucleophilic residues (Lys, Ser and Thr) always precede the acidic residues (Glu or Asp), indicating their relationship to the directionality of the modification enzymes, as described below [34].

**Figure 7.** Leader peptide sequences from different microviridins. The PFFARFL motif is highly conversed among them. This sequence and some of its flanking amino acids are structured as an α-helix, responsible for recognition by ATP-grasp ligases. Multiple alignment was obtained by Clustal Omega (https://www.ebi.ac.uk) and visualized using JalView software (https://www.jalview.org).

The PFFARFL motif and its α-helix structure is crucial as a recognition motif for the ATP grasp-type ligases (MdnB and -C), as can be visualized in Figure 8A, considering that both enzymes do not modify the core microviridin peptide when the leader peptide is absent, and lactonization and lactamization occur with the PFFARFL motif presence [36–38]. The PFFARFL motif is also present in the leader peptide of marinostatin, a double-cyclic peptide with serine protease inhibitor activity, which, by a phylogenetic analysis, suggests that this bicyclic peptide is derived from microviridins [20,37]. Nevertheless, the N-terminal ten-residues sequence of MdnA is not relevant for MdnB and C activity, as this modified prepeptide still containing a PFFARFL motif can also be cyclized and processed. However, a N-His6-tagged MdnA with an integral LP fused to three consecutives core peptides was not able to be processed by MdnC from *Anabaena* sp. PCC7120 [36,39].

**Figure 8.** Microviridin biosynthesis. (**A**) Ester bond formations by MvdD/MdnC. (**B**) Amide bond formations by MvdC/MdnB. (**C**) Removal of a peptide leader by a proteolytic enzyme. (**D**) N-acetylation by MvdB/MdnD.

In addition to this motif, a proline-rich segment is present in the C-terminal region of the leader peptide in a variety of microviridins from *Microcystis* organisms, close to those eukaryotic signal peptides normally associated with cleavage sites. In contrast, microviridin K obtained from *Planktothrix aghardii* CYA128/8 possesses only one proline at the same region [15,35,40]. However, the substitution of these prolines in *Microcystis* did

not affect the removal of the peptide leader but resulted in the cessation of microviridin production [37], suggesting the necessity of the β-turn of peptide leaders MdnB and C.

Ahmed et al [20] analyzed several *mdnA* sequences among their biosynthetic clusters and divided this gene into three different classes. Class I precursor peptides contain the LP fused to only one core sequence and are often associated with the presence of *mdnE*, which occurs in a majority of the strains. Class II precursor peptides present a single leader peptide for up to five core peptides in tandem, separated or not by double-glycine cleavage sites, and these clusters normally encode a C39 peptidase membrane protein. Finally, class III is identical in length to class II, but the former has its core sequence only at the C-terminal of the prepeptide [20]. This indicates a number of pathways for the genetic organization of *mdnA*.

After *mdnA* has been expressed, prepeptides should be submitted for cycling by the sequential catalysis of the enzymes. Thus, in order to understand the mechanisms related to this stage, Philmus et al. [15,37] were the first to define, through biochemical methods, the steps taken by MvdC and D of *P. agardhii* CYA126/8. Both enzymes are carboxylateamine/thiol ligases that belong to the ATP-grasp superfamily and act by requiring ATP and Mg2+ [38,40] to form a carboxylate–phosphate intermediate, which is then susceptible to nucleophilic attacks to form ester, amide or thioester bonds [17,36]. MvdD/MdnC are responsible for the first step in the formation of both ester bonds (Figure 8A) and, subsequently, lactone rings in the linear prepeptide, while MvdC/MdnB are responsible for the formation of lactam rings by amide bonds [15,35].

Both enzymes are homodimers with related assemblies, similar to most proteins of this family, having three subdomains: N-domain, central domain and C-domain. Besides their overall similarities, there are differences comparing their central and C-domains. MdnC/MvdD possess a two-stranded antiparallel β-sheet forming a hairpin structure, followed by a reasonably ordered α-helix that anchors the leader peptide. Meanwhile, this hairpin region is located at the C-domain of MdnB/MvdC, followed by a flexible loop in the α-helix region. The MdnB has a closed conformation, compared to MdnC, because the antiparallel β-sheet hairpin blocks the pocket site where MdnA interacts. Those differences can be related to their specificity and mode of action, as can be seen below. Regarding the ATP-binding pocket, it is structurally conserved, as confirmed by mutagenesis, where substitution of the key amino acids completely abolished the MdnC reaction [36].

Phylogenetic studies and the study of preserved sequences of different classes of prepeptides forming cyclic structures by the action of ATP-grasp ligases (plesiocins, microviridins and thuringinins) suggest that the enzymes coevolved with their respective precursor peptides due to the specificity of the preserved residues present in the core sequence [36]. Consequently, the association between microviridin production and ATPgrasp enzymes indicates that cyanobacteria recycled primary metabolic enzymes for the production of natural products, such as ribosomal peptides, as most ATP-grasp ligases are engaged to primary metabolism [15,17,35,36]. In addition, MdnC is well-conserved among *Microcystis* species, suggesting its derivation from a common ancestor, as well as its dependence on the core motif KYPSD and threonine and aspartate conservation sites of microviridins, as seen by the mutagenesis and phylogenetic analysis [14,15].

As described by Li et al. (2016) [36], the reaction of the bond formation by MdnC (Protein Data Bank (PDB) code 5IG9) is driven by the interaction with the leader peptide. Thus, the PFFARFL motif structured as a α-helix and its flanking amino acids interact with the MdnC hairpin, inducing its movement towards the linear prepeptide bound to the enzyme, then acting as an allosteric region. Considering the ATP-grasp ligase from *Microcystis aeruginosa*, these interactions occur between the amino acids Arg17 (MdnA) and Glu191, Asp192 and Asn195 (MdnC) and Ser20 (MdnA) and Val182 (MdnC). However, Glu191 is mostly replaced by an aspartate residue among all microviridin macrocyclases but still bears the negative charge required to recognize the LP [36].

After binding to the peptide leader through the PFFARFL motif, the ester bond formations are strictly required to occur in a specific order in microviridins: MvdD catalyze the lactone ring between Asp44 and Thr38, then Glu46 and Ser43 into the prepeptide, by phosphorylating the carboxyl side chain of Asp and Glu with ATP, thus forming the large then small lactone rings, respectively [37]. These residues participating in the amide bond and ring formation are highly conserved among the cyclic peptides, suggesting their requirement for the correct cyclization and similar catalysis between ATP-grasp ligases from different groups [34].

When a site-directed mutagenesis was applied to produce different variants of MvdE/ MdnA, S43A and T38A, it has been noticed that MvdD catalyzes a reaction following a N-terminal-to-C-terminal direction, as the S43A variant is still lactonized, producing a monocyclic microviridin. In addition, the amino acid bearing the hydroxyl group is crucial for the reaction, as it seems that MvdD cannot react when it is moved one position in either the N- or C-terminal direction [15,37,41]. It seems that both ATP-grasp ligases are highly tolerant for nonconserved residues, then being able to catalyze different microviridins. However, they are not flexible to conserved residues that are involved in cyclization [34,35].

For a better understanding of the different MdnC/MvdD enzymes, Zhang et al. [39] characterized a homolog of these enzymes from *Anabaena* sp. PCC7120, AMdnC, which belonged to a biosynthetic cluster with a prepeptide of class II, with a LP followed by three consecutives core sequences (AMdnA). The mode of action of AMdnC indicates a distributive catalysis, where the ATP-grasp ligase dissociates from the processed peptide after each monocyclization, until achieving all lactone ring formations. This feature has been also described in other modification proteins from RiPP pathways, such as the NisB, LctM, LabKC and HalM2 enzymes from lanthipeptides processing; microcin B17 synthetases; ATP-grasp enzyme PsnB and N-methylation enzyme OphA of omphalotin. Additionally, AMdnC also demonstrates a preferential N-to-C directionality when catalyzing the reaction but not unstrict. Thus, this homolog of MdnC can process each core peptide independently from AMdnA. Moreover, the calculated Km from AMdnC when catalyzing AMdnA or MdnA is comparable to MdnC values when processing MdnA; however, the *k*cat of the ATP hydrolysis of AMdnC were up to 60 times faster, suggesting a different mechanism for processing a prepeptide with multiple core sequences.

MdnB has a similar structure and mechanism of activation as MdnC, where the PFFARFL motif interacts with the hairpin, resulting in the activation of the enzyme [36]. Then, the bicyclic prepeptide produced by MvdD/MdnC is catalyzed by MndB/MvdC, and the lactam ring is formed through the amide bond formation between the ε-amino group of Lys40 and δ-carboxyl group of Glu47 (Figure 8B). The omega-amide bond is similar to those present in microcin J25 and capistruin; however, the enzymatic mechanism is different, because microviridin K synthesis occurs via an acyl-adenylated intermediate [17].

Both preformed lactone rings are required by MvdC/MdnB, as the linear and monocyclic peptides are not modified. A single mutation in the PFFARFL pattern in the leader peptide prevents the formation of amide bonds, as well as the proper conformation of the β-turn by the proline-rich region at the C-terminal of microviridin from *Microcystis*, suggesting a lower flexibility compared to MndC [39]. In addition, the amino acid sequence of the core peptide can also influence the correct cyclization, even those not well-conserved, requiring the TxKxPSD motif and Lys and Glu residues [41].

MdnC/MvdD is less rigid than MdnB/MvdC, as it can still catalyze both lactone rings besides single and double mutations in the PFFARFL motif and proline-rich region of the leader peptide (in *Microcystis*) but could result in producing different microviridin variants differing at the N-terminal [37]. This versatility is possibly due to the more open conformation of the hairpin structure and to the binding interaction between MdnC and the prepeptide relative to MdnB. As seen in vitro, the binding interaction between LP and MdnC is approximately tenfold higher compared to the LP and MdnB, resulting in a rapid processing of the linear prepeptide compared to the bicyclic modification. It is also believed that this is due to the fact that linear MdnA is less stable, requiring prompt modification [36]. However, this post-translational modification does not seem to be strictly sufficient for further steps, as bicyclic microviridins can still be cleaved and N-acetylated

and possessing inhibitory activity against proteases [41]. As seen by bioinformatic analyses, the absence of MdnB is normal and is likely to lead to the formation of marinostatin, a peptide that lacks an amide bond and is closely related to microviridins [20]. Firstly, it was suggested that MdnE, an ABC transporter, could be related to the removal of the leader peptide through peptide cleavage due to the presence of a N-terminal C39 peptidase domain from *Anabaena* PCC7120 [14,37]. However, not all MdnE carry that domain, and the heterologous expression of the microviridin cluster lacking this protein still produces this tricyclic peptide, indicating other roles [14,37]. Comparing the microviridin expression with the presence and absence of MdnE, it was noted that these peptides were not correctly processed at the N-terminal and were incompletely cyclized due to a lack of lactam rings, as was the amount of MdnB observed in the cytoplasmic fraction when the ABC transporter was absent. This pattern then suggests the hypothesis that MdnE is a scaffolding protein, anchoring and stabilizing the microviridin biosynthesis complex on the cytosolic side of the membrane [37]. In addition to its similarity to transporter proteins, its function in exporting microviridin from the cell has not been demonstrated.

Knowledge on the removal of a peptide leader has so far been scarce in the literature. However, the heterologous expression of microviridin suggests that this step can be mediated by a nonspecific proteolytic enzyme (Figure 8C), as *E. coli* expressing only MdnA-C was capable of producing microviridin lacking a LP [14]. Moreover, GluC endoprotease is capable of cleaving peptide bond C-terminals to glutamic acid residues, and during the in vitro production of class II microviridin with three core sequences, this enzyme released all three mono-, bi- and tricyclized microviridins [40]. Finally, another hypothesis related to MdnE was raised. As described above, this enzyme may have a peptidase domain, typically present in class II clusters. This function may be linked to the presence of interspaced regions between the core sequences of MdnA and the release of each individual microviridin [14,20,37,39].

Acetylation is one of the last steps for the development of a fully matured tricyclic microviridin. Microviridin synthesis in vitro has shown that MvdB from *P. aghardii* CYA126/8 does not require the presence of the peptide leader to acetylate the microviridin N-terminal. In addition, MdnD/MvdB can react with mono-, bi- and tricyclic, being more flexible than MdnB and C. Thus, it can be assumed that this step occurs after the peptide leader removal, or this enzyme does not interact with this region (Figure 8D) [38]. In addition, a 12 amino acid-long tricyclic peptide is not N-acetylated by MdnD from *P. aghardii* CYA126/8, but those microviridin with 13/14 amino acids are acetylated, indicating that there a specific size requirement by this enzyme, and it is flexible regarding the core sequence [20], thus suggesting that N-acetylation occurs only after leader peptide removal [15,20,35,38].
