Cyclic peptides tend to exhibit improved metabolic stability, membrane permeability, and target specificity as compared to their linear counterparts [1,2,3,4,5]. Consequently, macrocyclization is a significant modification in the development of peptide-based therapeutics. However, despite its importance, efficient construction of macrocyclic scaffolds remains a challenging task, due to inherently competing side reactions like oligomerization and epimerization [6]. Under such circumstances, the use of biocatalysts has garnered keen attention due to their high selectivity and mild reaction conditions. Such peptide-cyclizing biocatalysts with synthetic potential have been identified mainly from secondary metabolism in bacteria and plants [7, 8].

Numerous pharmaceutically important macrocyclic peptides are biosynthesized through non-ribosomal pathways [9]. Representatives of clinically approved non-ribosomal peptides (NRPs) include the immunosuppressive agent cyclosporin, lipopeptide antibiotics like daptomycin and polymyxin, and glycopeptide antibiotics like vancomycin. NRPs are generally biosynthesized by non-ribosomal peptide synthetases (NRPSs). These megasynthetases consist of several functional domains that cooperatively assemble amino acid building blocks to produce grown peptide intermediates that are covalently bound to the synthetase. After assembly, peptide chains are processed by the C-terminal thioesterase domain (TE) and released from the NRPS as linear or cyclic products. Although several enzymatic machineries that mediate cyclization of NRPs have been identified, TEs are the most general biosynthetic mechanisms to generate macrocyclic scaffolds [10,11,12].

TE is a serine protease-type enzyme with a typical α/β hydrolase fold equipped with a Ser/His/Asp catalytic triad. TE-mediated macrocyclization proceeds via two steps. First, the grown linear peptide intermediate is transferred onto the catalytic Ser to form the peptide-O-TE complex. Subsequent attack from intramolecular nucleophiles on the ester carbonyl achieves the cyclization and release of the cyclic products from TEs, while attack from water results in the release of linear products. The hallmark of the TE-catalyzed cyclization is the diversity of nucleophiles used for cyclization: TEs utilize not only N-terminal Nα-amine nucleophiles, but also other nucleophiles such as the β-OH of an N-terminal acyl group, Nω-amine of basic amino acids, β-OH of Ser/Thr/β-OH Phe, phenolic hydroxy of Tyr, and thiol of Cys, resulting in lariat-shaped macrocycles with various linkages at the ring closing site [10, 11]. Moreover, some TEs catalyze ligation along with cyclization, yielding cyclo-dimers and cyclo-trimers [10, 11]. As the functions of TEs have enormous impacts on the molecular shapes of the final products, TEs can be considered as major contributors toward generating the structure diversity of NRPs.

In 2000, Walsh and colleagues reconstituted the cyclization activity of TE excised from tyrocidine synthetase in vitro, and chemoenzymatically synthesized tyrocidine from a linear peptide using a phosphopantetheine surrogate: N-acetylcysteamine (SNAC) [13]. This work set the stage for the exploitation of excised TEs as chemoenzymatic tools for macrocyclization. Since then, TEs from various biosynthetic pathways have been characterized in vitro, with some of them successfully facilitating the chemoenzymatic synthesis to afford access to natural products and libraries of their analogs.

This review focuses on macrocyclizing-TEs in the biosynthesis of bacterial NRPs and their utilization as chemoenzymatic tools, with special emphasis on recent reports. This review also highlights the growing insights into penicillin-binding protein-type thioesterases (PBP-type TEs), a new group of NRP cyclases with remarkable biocatalytic potentials [14,15,16].

Tyrocidine A (1) is a head-to-tail cyclized, amphiphilic non-ribosomal decapeptide with potent membrane disruption activity [17]. In 2000, Trauger et al. reconstituted the macrocyclizing step of tyrocidine biosynthesis in vitro, reporting that an overexpressed excised TE domain (TycC TE) involved in tyrocidine biosynthesis cyclized a seco-tyrocidine N-acetylcysteamine (SNAC) thioester to give 1 [13] (Fig. 1a). This demonstrated for the first time that (i) an excised TE retains macrocyclizing activity in vitro, and (ii) SNAC is an effective low-molecular-weight surrogate of a peptidyl-carrier protein (PCP), a part of the megasynthetase linked at the substrate C-terminus through the phosphopantetheinyl group. This set the stage for detailed analyses of TEs as well as their biocatalytic exploitations. TycC TE requires an aromatic D-aa at the N-terminus and L-Orn at position 9 (Orn9) (Fig. 1a) [13]. The latter was initially thought to play a role as a proton-donor for intramolecular hydrogen-bonding to pre-organize the substrate into a product-like conformation [13]. However, a subsequent study showed that Orn9 can be replaced by non-proton-donating cationic residues like trimethyl-Lys, suggesting that Orn9 may interact with the negatively charged protein surface rather than forming an intramolecular hydrogen bond [18]. TycC TE exhibits remarkable substrate tolerance: an Nα-amine nucleophile can be substituted with a hydroxy group to achieve macrolactonization [19]. It accepts substrates with various ring sizes ranging from 6 to 14 [20], showing its broad tolerance for substrate length. The catalytic ability of TycC TE is not limited to simple cyclization, but also includes cyclo-dimerization, as it converts a pentapeptide-SNAC to gramicidin S by simultaneously catalyzing ligation and cyclization [13]. Influential work led to the proposal of a minimal cyclization substrate for TycC TE, highlighting the importance of the residue near the ring closure site for enzyme recognition [19].

TycC TE-catalyzed macrolactamization. Ring closing sites are highlighted by shading. a TycC TE-catalyzed macrolactamization of natural and synthetic substrates. b Examples of macrolactams chemoenzymatically synthesized by TycC TE. c Synthesis of 6 via McyC TE-catalyzed macrolactamization

The biocatalytic potential of TycC TE was then extensively explored [21,22,23,24,25,26,27,28]. TycC TE tolerates oxoester substrates that are tethered to a solid support (PEGA resin) through a biomimetic linker, thus enabling chemoenzymatic synthesis in which combinatorial SPPS (Solid-Phase Peptide Synthesis) chemistry and enzymatic cyclization can be performed seamlessly [21] (Fig. 1a). A library of 192 tyrocidine variants with simultaneous substitutions at positions 1 and 4 was screened for both the minimal concentration of antibacterial activity against Bacillus subtilis (MIC) and hemolysis of human erythrocytes (MHC), which led to the identification of a variant (2) with an improved therapeutic index (MHC over MIC) [21]. The internal variable region of substrates can accommodate diverse structural motifs (Fig. 1b), such as the RGD sequence to yield an integrin-binding macrocycle (3) [24], (E)-alkene dipeptide isosteres to give cyclic peptidomimetics [25], fragments of type B streptogramin to produce chimeric macrocyclic antibiotics [26], and polyketide-like building blocks to generate peptide/polyketide-like hybrid molecules (4) [27]. Lin et al. incorporated a propargyl glycine residue at position 4, enabling post-cyclization structure diversification via copper(I)-catalyzed alkyne/azide cycloaddition (CuAAC) [28]. By coupling with azido sugars, 247 glycopeptidyl variants of tyrocidine A were synthesized, resulting in the identification of a lipoglycopeptide (5) with a sixfold better therapeutic index than natural tyrocidine [28].

Li’s group recently reported that MycC TE in microcystin biosynthesis can cyclize non-native sequences activated by BMT (benzylmercaptan), and utilized it for the chemoenzymatic synthesis of cilengitide (6), a head-to-tail cyclic pentapeptide with potent integrin inhibitory activity (Fig. 1c). Substitution of the catalytic Ser with Cys enhanced the catalytic activity, and 250 µM of benzyl mercaptan thioester was cyclized to 49% within an hour of incubation [29].

Head-to-side chain cyclization, in which the C-terminal carbonyl is linked to side chain nucleophiles at internal residues to give a “lariat-shaped” product, is one of the structural hallmarks of NRPs. Numerous NRPs with pharmaceutical relevance, such as lipopeptide antibiotics, fall into this subgroup. To date, many TEs that catalyze head-to-side chain macrocyclization have been characterized in vitro. Representatives are TEs involved in the biosynthesis of surfactin [20, 30], calcium-dependent antibiotics (CDA) [31], A54145 [32], daptomycin [32], lichelin [33], fengacin [34], lysobactin [35], polymyxin [36] etc. Compared to the Tyc TE that catalyzes head-to-tail macrocyclization, TEs catalyzing head-to-side chain macrocyclization are often less versatile as biocatalysts, since most exhibit narrow substrate scopes in terms of ring size and residues near ring closing sites. A significant hydrolytic flux of peptide-O-TE intermediates is also often reported. Furthermore, some are inactive with a conventional SNAC leaving group and require PCP-loaded substrates [37] or higher activation (i.e., the use of thiophenol as a leaving group) [38]. Ring-opening reverse reactions that hydrolyze the macrocycle products often contribute to low product yields [39]. Nevertheless, the abilities of these TEs to catalyze regio-, stereo-, and chemo-selective cyclizations to yield lariat-shaped products directly from linear substrates provide valuable opportunities to develop efficient chemoenzymatic approaches. Indeed, TEs such as CDA TE and Crp TE were utilized to synthesize variant libraries of daptomycins and cryptophysins, respectively [31, 40]. These TEs are particularly powerful in their abilities to achieve regio-selective macrolactonization, which frequently poses synthetic challenges like low coupling yields and epimerization via chemical methodologies [41].

Early investigations revealed that TE-mediated head-to-side chain cyclization is generally sensitive to alterations in the environment near the nucleophiles. For example, Srf TE in surfactin biosynthesis does not tolerate the stereochemical inversion of the β-OH nucleophile in the acyl chain [20]. Similarly, Syr TE in syringomycin biosynthesis [38] and CDA TE in CDA biosynthesis [42] do not tolerate the inverted Cα-stereochemistry of their nucleophilic residues, L-Ser and L-Thr, respectively. Cyclization is generally chemo-selective, while A15145 TE exceptionally accepts not only the natural nucleophile β-OH L-Thr (macrolactonization) but also β-NH2 L-Dap (macrolactamization) [42].

Substrates with long acyl chains are poorly soluble in the aqueous reaction conditions of TEs, making in vitro characterizations and biocatalytic applications of lipopeptide TEs challenging. Shortening the acyl chains to improve the solubility reportedly compromises the regio-specificity in cyclization [42]. To tackle the solubility issue, Marahiel’s group performed Srf TE cyclization in 100% DMF, which dramatically enhanced the catalytic activity and suppressed hydrolysis to afford surfactin (7) in a quantitative conversion (Fig. 2a) [43]. Similarly, the non-polar detergent Brij57 profoundly improved the catalytic efficiency and catalyst lifetime of Tyc TE [18]. These observations highlight the significance of solvent engineering for optimizing the biocatalytic applications of TEs.

TE-catalyzed head-to-side chain macrocyclizations. Ring closing sites are highlighted by shading. a Srf TE-catalyzed macrolactonization in DMF. b SnbDE TE-catalyzed stereoselective macrolactonization of a racemization-sensitive thioester. c Skyxy TE-catalyzed epimerization/macrolactonization

The stereoselective nature of TE was combined with dynamic kinetic resolution (DKR) to achieve the stereoselective cyclization of a highly racemizable substrate [44]. The SNAC substrate of SnbDE TE, involved in streptogramin biosynthesis, possesses L-Phg at its C-terminus, which is prone to racemization. However, when starting from a 1:1 racemic mixture of the SNAC substrate, SnbDE TE gave a single stereoisomeric macrolactone (8) due to the strict stereospecificity of SnbDE TE for the C-terminal L-Phg. The yield was greater than 50%, due to the in situ stereochemical inversion of the C-terminal D-Phg to L-Phg (Fig. 2b).

Recently, Ma’s group characterized Skyxy-TE in skyllamycin biosynthesis. Skyxy-TE is a bifunctional TE that actively catalyzes the epimerization of a non-racemizable C-terminal L-aa of a linear substrate and performs regioselective macrolactonization to generate skyllamycin analog (9) (Fig. 2c) [45]. Epimerization is suggested to occur at the point of peptide-O-TE complex formation, and then regioselective macrolactonization affords the epimerized macrolactone. The crystal structure and mutagenesis of Skyxy-TE identified a key Gln residue near the catalytic Ser, and the substitution of this Gln affords racemic products.

Teixobactin, a depsipeptide with promising antibiotic activity, is produced by the β-proteobacterium Eleftheria terrae, and was isolated using the in situ cultivation method iChip [46]. The teixobactin NRPS possesses duplicated TE domains, sometimes referred to as tandem TEs, and similar architectures are also observed in the NRPSs of lysobactin [35] and arthrofactin [47]. While only the N-terminal TE in the lysobactin system is responsible for macrolactonization, Zhang’s group reported that both TEs can catalyze macrolactonization in the teixobactin system [48]. Although accompanied by substantial amounts of hydrolytic products, the teixobactin TE (Txo TE) afforded teixobactin analogs (10) from linear methyl ester substrates, which are much easier to synthesize than conventional SNAC thioesters (Fig. 3a). Txo TE accepted D-Ser as a nucleophile alongside the native D-Thr, but not L-Thr.

Chemoenzymatically synthesized macrolactones. Ring closing sites are highlighted by shading. a Chemical structures of teixobactin analog (10), seongsanamide E (11), and rhizomide A (12). b Chemical structures of cryptophycin analogs and a low-picomolar potency analog (13)

Examples of other recent additions include Sgd TE in the biosynthesis of the lipodepsipeptide seongsanamide [39]. Boddy’s group reported that seongsanamide E (11), a lariat-shaped peptide cyclized through the β-OH of L-Thr, could not be chemically synthesized from a linear free acid via macrolactonization [39]. However, Sgd TE catalyzed the macrolactonization of a linear SNAC substrate to give seongsanamide E in a 27% yield, without suffering from the epimerization typically occurring in chemical strategies, demonstrating the advantage of enzymatic macrolactonization [39]. Similarly, RzmA TE in the biosynthesis of the lipodepsipeptide rhizomide catalyzes a chemically challenging macrolactonization through the β-OH of L-Thr to give rhizomide A (12) from a linear SNAC substrate in a 15% yield, without substantial flux of hydrolysis [49].

Cryptophycin TE (Crp TE) is one of the most successfully applied TEs in the chemoenzymatic synthesis of bioactive macrocycles. The cryptophycins are a class of highly potent microtubule-binding agents isolated from Nostoc cyanobacteria [50, 51] and marine sponge [52]. They consist of four unusual building blocks, including phenyl-octenoic acid (unit A), 3-Cl-O-methyl-D-Tyr (unit B), methyl β-Ala (unit C), and L-leucic acid (unit D), which are assembled via a PKS/NRPS hybrid system (Fig. 3b) [53]. In the final step of assembly, Crp TE catalyzes the macrolactonization between the δ-hydroxy group in unit A and unit D to construct a 16-membered cyclic depsipeptide scaffold, followed by the stereospecific epoxidation of unit A mediated by CrpE P450 [53]. The biocatalytic potential of Crp TE was demonstrated by Sherman’s group in 2005. Crp TE from Nostoc sp. ATCC53789 efficiently catalyzed the macrolactonization of seco-cryptophycin-I SNAC thioester with a cyclization: hydrolysis (cyc:hyd) ratio of 10:1 [54]. Crp TE tolerated the variation of the methyl group at β-alaninyl unit C, while it requires an aromatic moiety at unit A [54]. Crp TE was further demonstrated to be tolerant of an activated acylsulfonamide substrate on solid support (PEGA resin), potentially accelerating the preparation of cryptophycin and its analogs [55]. Moreover, Crp TE-mediated macrolactonization was coupled with CrpE P450-mediated stereospecific epoxidation of unit A to generate natural and unnatural cyclic structures with an epoxide moiety, in a one pot manner [53, 56]. Recently, the same group developed a scalable synthesis of linear SNAC substrates with various five- or six-membered heterocycles at unit A and dimethyl derivatives at unit C [40]. Crp TE tolerated heterocycles at unit A and cyclized the SNAC substrates in 69% to 97% conversion, with a cyc:hyd ratio greater than 10:1. Crp TE also cyclized dimethyl derivatives at unit C in 68% to 71% conversions, with a cyc:hyd ratio greater than 7:1. Reactions were conducted on a semi-preparative scale and twelve cyclic products were isolated and tested for their cytotoxicity, which led to the identification of a new derivative (13) with single-digit picomolar potency against the HCT-116 human colorectal cancer cell line. Notably, 13 deviated from the requirement of an epoxide group in unit A, which was previously thought to be invaluable for high potency. This study not only provided one of the most potent cryptophycins to date, but also demonstrated the viability of the TE-mediated cyclization strategy in medicinal chemistry [40].

Oligomeric cyclic scaffolds, in which relatively short fragments are oligomerized and cyclized, are another structural hallmark of macrocycles biosynthesized by thio-templated machineries. Ligation and successive cyclization are solely mediated by TEs, which are often referred to as ‘iterative TEs’. The linkages formed by iterative TEs include amide, ester, and thioester, with two to three monomers being oligomerized.

The cyclo-oligomerization activity of a TE was initially characterized with EntF TE, involved in enterobactin biosynthesis [57], but later intensively characterized with GrsB TE, which catalyzes dimerization and cyclization of pentapeptidyl precursors to generate gramicidin S (14) [58]. Gramicidin S is a head-to-tail cyclic membrane-active decapeptide antibiotic homologous to tyrocidine A [17]. Generation of the decapeptide SNAC from the pentapeptide SNAC substrate indicated the so-called ‘reverse transfer mechanism’ for TE-catalyzed oligomerization, where the peptide-O-TE is attacked by a second monomer (SNAC substrate in vitro or PCP-tethered intermediate in vivo) to form a dimerized linear intermediate, then reloaded onto TE to form the peptide-O-TE complex [58] (Fig. 4a). GrsB TE tolerates various substrate sizes to catalyze dimerization/trimerization and cyclization to give 6–15 residue macrocycles [58]. The fate of the peptide-O-TE (ligation with next monomer or intramolecular cyclization) is thought to be affected by both the pre-folding of the linear intermediate and the catalytic pocket capacity of TE. Notably, TycC TE catalyzes the cyclo-dimerization of a pentapeptide SNAC to generate 14, indicating that the ligase activity is not confined to a specific subgroup of TEs [