Distal peptide elongation by a protease-like ligase and two distinct carrier proteins

Summary

Closthioamide (CTA) is a potent antibiotic with a unique polythioamide scaffold produced by Ruminiclostridium cellulolyticum. Unlike classical non-ribosomal peptide synthetases (NRPSs), which use modular adenylation and condensation domains, CTA biosynthesis proceeds through non-canonical standalone enzymes. Central to this process is the papain-like ligase CtaG, which catalyzes amide bond formation between two distinct peptidyl carrier proteins (PCPs): CtaH, presenting para-hydroxybenzoic acid (PHBA), and CtaE, carrying a tri-β-alanine ((βAla)₃) chain. Using biochemical assays, chemical probes, crystallography, and mutational analysis, we show that CtaG operates via a ping-pong mechanism involving an enzyme-bound intermediate. A single substrate tunnel mediates directional transfer, enabling distal chain elongation that mirrors solid-phase peptide synthesis. Structure-based genome mining revealed homologous enzymes in the biosynthetic pathways of petrobactin, butirosin, and methylolanthanin. Together, our findings uncover a previously overlooked class of thiotemplated ligases and provide a mechanistic blueprint for engineering ribosome-independent peptide assembly lines.

Graphical abstract

Highlights

• •Enzymatic mechanism for distal peptide chain elongation elucidated
• •Characterization of a class of papain-like, PCP-dependent peptide ligases
• •New addition to the catalytic toolbox for ribosome-independent amide bond formation

The bigger picture

Microorganisms synthesize a wealth of ecologically and medically important natural products that rely on ribosome-independent biosynthetic pathways. Most are assembled by modular non-ribosomal peptide synthetases (NRPSs), which drive unidirectional chain elongation via proximal coupling of peptidyl carrier protein (PCP)-bound amino acids. We report a striking deviation from this logic: the protease-like ligase CtaG catalyzes an amide bond between two distinct PCP-linked substrates during the biosynthesis of the antibiotic closthioamide (CTA). Unlike modular NRPSs, CtaG acts via a ping-pong mechanism and extends the peptide chain through distal substrate loading, a process reminiscent of solid-phase peptide synthesis. Structural and biophysical analyses suggest that CtaG adapts to its partners through ligand-induced conformational changes, enabling selective PCP recognition. These findings expand our understanding of enzymatic amide bond formation beyond amino acid substrates, reveal a previously overlooked class of thiotemplated ligases, and open new avenues for the discovery and engineering of biosynthetic platforms.

The enzymatic assembly of the unusual peptide antibiotic closthioamide involves ribosome-independent amide formation by a papain-like ligase, CtaG. Biochemical, mutational, and structural analyses show that CtaG catalyzes a non-canonical distal peptide chain elongation via a ping-pong mechanism using two distinct peptidyl carrier proteins.

Introduction

Non-ribosomal peptides (NRPs) are a structurally diverse class of specialized metabolites with a wide range of biological activities.^1,2 They are typically assembled by modular non-ribosomal peptide synthetases (NRPSs) that operate analogously to molecular assembly lines.^1,2 In these systems, peptidyl carrier proteins (PCPs) present building blocks and intermediates through covalent attachment, making the process thiotemplated.³ While adenylation (A) domains commonly activate amino acids,⁴ many NRPSs incorporate non-proteinogenic monomers such as aryl acids, fatty acids, or modified amino acids. Condensation (C) domains then catalyze peptide bond formation between the PCP-bound units (Figure 1A).^5,6,7 In this canonical model, the aminoacyl group is transferred from the donor PCP to the acceptor PCP and incorporated in a proximal orientation relative to the growing peptide chain.⁸

Figure 1.

Pathways for amide bond formation and model for CTA biosynthesis

(A) Comparative analysis of type I non-ribosomal peptide synthetases (NRPSs), CtaG-dependent biosynthesis, and Merrifield solid-phase peptide synthesis (SPPS). In type I NRPS systems (peptidyl carrier protein [PCP], gray; A, adenylation domain, green; C, condensation domain, purple), compound 2 (orange square) is connected to compound 1 (gray square) in a proximal position. In contrast, CtaG (G, purple) first accepts 2 from CtaH (H, black) and transfers 2 to CtaE (E, black) in a distal position relative to the phosphopantetheine linker. SPPS follows a similar logic: 1 is immobilized on a resin and extended by coupling of 2 in a distal orientation.

(B) Closthioamide (CTA) biosynthesis. p-Hydroxybenzoic acid (PHBA), produced from chorismate by CtaA (chorismate lyase), is loaded onto CtaH by CtaI (acetyl-CoA synthetase). In parallel, CtaE is loaded with three β-alanine units, synthesized from l-aspartate (l-Asp) by CtaD (ATP-grasp enzyme), and CtaF (decarboxylase). PHBA is then transferred to (βAla)₃-CtaE by CtaG, followed by thioamidation by CtaC (Fe-S cluster enzyme). The pathway branches: CtaJ (transglutaminase) cleaves the terminal thioamide, while CtaK (reductase) and CtaB (aminotransferase) generate a product that merges with the intermediate to form CTA.

Recent biosynthetic studies have uncovered a growing number of non-canonical, ribosome-independent peptide synthetases that deviate from the classical NRPS logic.^9,10,11 One prominent example is the biosynthetic pathway of closthioamide (CTA, Figure 1B), produced by the anaerobic bacterium Ruminiclostridium cellulolyticum.^12,13 The polythioamide scaffold of CTA is structurally unique, binds copper(I),¹⁴ and functions as a potent antibiotic targeting bacterial DNA gyrase.¹⁵ Genomic, mutational, and biochemical analyses revealed that CTA is assembled by a non-modular, ribosome-independent system involving nine standalone enzymes and two PCPs.^16,17 Among these, three enzymes catalyze the formation of six amide bonds, resulting in the symmetric peptide backbone that defines the CTA scaffold. Central to this process is CtaG, a papain-like ligase, which catalyzes amide bond formation between para-hydroxybenzoic acid (PHBA) and a tri-β-alanine ((βAla)₃) chain in a ping-pong mechanism involving the carrier proteins CtaH (donor) and CtaE (acceptor), respectively. In contrast to the proximal bond formation observed in canonical NRPS systems, this reaction proceeds through distal peptide extension and represents a fundamentally different strategy of thiotemplated biosynthesis.¹⁸

To better understand this deviation, it is useful to first consider the paradigm of canonical NRPSs. In modular NRPSs, domain architecture ensures unidirectional chain elongation catalyzed by C domains.^5,6,7 Each C domain features a substrate channel that accommodate PCP-bound donor and acceptor intermediates simultaneously, enabling amide bond formation proximal to the phosphopantetheine (PPant) linker. This process relies on precise domain organization and coordination.¹⁹ In contrast, the initial steps of CTA biosynthesis deviate from this paradigm. The end-capping reaction is proposed to be catalyzed by a standalone papain-like enzyme, CtaG,¹⁸ which, unlike canonical C domains, typically lacks dual substrate channels. Furthermore, CtaG operates in conjunction with two PCPs, CtaE and CtaH. Although this setup superficially resembles the NRPS framework, key mechanistic differences exist. CTA biosynthesis proceeds via distal chain extension, reminiscent of solid-phase peptide synthesis (Figure 1A). Moreover, CtaG and its PCP partner proteins are not embedded in a modular architecture, precluding structural control over the reaction sequence. Consequently, the mechanism of thiotemplated amide bond formation in the CTA pathway has remained elusive.

In this study, we provide detailed functional and structural insights into the ligase CtaG, which drives amide bond formation in a ribosome-independent, thiotemplated antibiotic assembly line. We show that CtaG interacts with its substrates exclusively when presented by the PCPs CtaH and CtaE, each of which assumes a defined role as donor or acceptor. Strikingly, CtaG accommodates both substrates through a single tunnel to mediate distal chain extension. In addition, structure-based analyses reveal previously unrecognized CtaG homologs involved in the biosynthesis of environmentally and medically relevant NRPs.

Results

CtaG catalyzes amide bond formation between PCP-conjugated substrates

To investigate the unusual end-capping activity of the papain-like enzyme CtaG in CTA biosynthesis, we first identified its preferred substrates. Specifically, we analyzed whether both reaction partners must be PCP bound. To this end, we synthesized (βAla)₃ and its corresponding PCP conjugate, (βAla)₃-CtaE. The latter was prepared from the pantetheine-linked intermediate (βAla)₃-Pant, which was chemoenzymatically converted into the corresponding coenzyme A (CoA) derivative using CoaA, CoaD, and CoaE.¹⁸ This CoA conjugate was subsequently loaded onto the PCP using the promiscuous phosphopantetheinyl transferase Sfp²⁰ (Figure S1). We then reconstituted the enzymatic reaction using recombinant CtaG, the CtaH-bound donor substrate PHBA,¹⁸ and either free (βAla)₃ or (βAla)₃-CtaE. Reaction products were cleaved from the respective PCP under basic conditions and analyzed by liquid chromatography-high-resolution mass spectrometry (LC-HR-MS). Using a synthetic reference, we demonstrated that the amide-linked dipeptide (βAla)₃-PHBA was formed exclusively when (βAla)₃ was presented as a PCP-bound substrate (Figure 2). These results conclusively demonstrate that CtaG catalyzes amide bond formation strictly between PCP-linked intermediates.

Figure 2.

Specificity of CtaG for PCP-bound (βAla)₃ in PHBA transfer reactions

CtaG is selective for PCP-bound (βAla)₃ as a PHBA acceptor. LC-HR-MS profiles of CtaG reactions following thioester cleavage. Traces correspond to the extracted ion chromatogram of the [M–H]^– ionic species for PHBA-(βAla)₃ and are displayed with m/z values ±5 ppm from the calculated exact mass. Red strikethrough indicates heat-inactivated enzyme.

CtaG selectively transfers PHBA from CtaH to CtaE

The requirement for two PCP-bound substrates prompted us to investigate whether CtaG discriminates between the carrier proteins CtaH and CtaE, which deliver PHBA and (βAla)₃, respectively. Therefore, we prepared defined PCP variants bearing each substrate covalently linked to the phosphopantetheinyl group. Using these conjugates, we reconstituted reactions with recombinant CtaG and tested both native (Figure 3A) and non-natural (Figures 3B–3D) substrate combinations. Matrix-assisted laser desorption/ionization time-of-flight MS (MALDI-TOF-MS) revealed a mass shift of +120 Da, consistent with PHBA transfer, but only when the donor was PHBA-CtaH. No product was observed when PHBA was presented by CtaE, confirming the selectivity of CtaG for its native donor. In contrast, both (βAla)₃-CtaE and non-physiological (βAla)₃-CtaH were accepted as nucleophilic partners. However, transfer remained incomplete when (βAla)₃ was delivered by CtaH. These findings demonstrate that CtaG exclusively acts on PCP-linked ligands and assigns distinct, non-interchangeable roles to its carrier proteins. While CtaH functions as the dedicated PHBA donor, CtaE serves as the preferred acceptor.

Figure 3.

Substrate specificity of CtaG

(A–D) The transacylase CtaG is active only when (βAla)₃ and PHBA are presented on their native PCP carriers. MALDI-TOF-MS spectral overlay of (A) (βAla)₃-CtaE and PHBA-CtaH, as well as artificial PCP-substrate pairs (B) (βAla)₃-CtaH and PHBA-CtaE, (C) (βAla)₃-CtaE and PHBA-CtaE, and (D) (βAla)₃-CtaH and PHBA-CtaH. Expected masses of all PCP species are indicated; red asterisks denote holo-PCP from residual unmodified CoA in reactions containing (βAla)₃-CoA.

(E–G) (E) CtaG interacts with both CtaE and CtaH via a shared tunnel leading to the active site. AlphaFold3²¹ models depict (F) CtaG with CtaH (dark gray) and (G) CtaG with CtaE. The catalytic cysteine in CtaG as well as the serine residues coordinating the phosphopantetheinyl residue in CtaH and CtaE are shown in red.

CtaG recruits donor and acceptor PCPs in a defined order

To further investigate the directionality of CtaG:PCP interactions, we applied a covalent crosslinking approach using the electrophilic surrogate chloroacrylamide (Figure S2). The ligand was converted into its CoA derivative²² and enzymatically loaded onto apo-CtaH or apo-CtaE using Sfp (see the supplemental information). The resulting PCP conjugates were then incubated with CtaG, and covalent adduct formation was monitored by SDS-PAGE (Figure S3). Intriguingly, a CtaG:CtaH complex formed within 1 min, while the corresponding adduct with CtaE appeared only after 1 h. This marked difference might correlate with the nature of the PCPs and their reaction order. While the compact and aromatic PHBA moiety on CtaH facilitates rapid conjugation, the same ligand installed on CtaE reacts slower. To confirm that covalent complex formation requires active site engagement, we substituted the catalytic cysteine (Cys11) of CtaG with alanine (see below). As expected, the CtaG_C11A variant did not form a crosslink with chloroacrylamide-CtaH (Figure S3B). In conclusion, the crosslinking results recapitulate the native reaction sequence and support a directional mechanism in which CtaG first engages PHBA-CtaH as the donor, followed by (βAla)₃-CtaE as the nucleophilic acceptor.

Structural plasticity of CtaG drives ligand-dependent selectivity for carrier proteins

To understand how CtaG differentiates between its two carrier proteins, we combined isothermal titration calorimetry (ITC), MS, and structural analysis (see the supplemental information). ITC revealed that CtaH interacts with CtaG in multiple ligand states, while CtaE binds only when CtaG is preloaded with PHBA (Figure S4). Although these interactions were too weak and transient for quantitative fitting, they provided qualitative insights into a pronounced asymmetry in binding behavior. This selectivity correlates with our MALDI-TOF-MS data (Figures 3A–3D), which show that PHBA can be transferred from CtaH to CtaG and subsequently to CtaE, but not vice versa.

To explore the structural basis of this ligand-dependent selectivity, we examined the electrostatic surface potentials of four CtaG conformers obtained from crystallographic data (see below). Two of these, apo-CtaG (PDB: 9QUZ) and the H128A variant (PDB: 9QVO), share compact entrance loops and conserved surface features, consistent with broad CtaH compatibility. Strikingly, the PHBA-bound H128A mutant crystallized in two distinct forms (PDB: 9QVR, 9QVQ). While the first closely resembles the apo state, the second shows an enlarged substrate tunnel and a marked increase in negative surface potential near the active site, which likely promotes selective interaction with CtaE (Figure S5). These experimentally determined conformations could not be replicated by AlphaFold3²¹ (Figures 3F and 3G), whose predictions failed to capture the loop dynamics and charge redistribution required for modeling such transient, ligand-specific interactions. To overcome these limitations and define the molecular basis of selectivity, we now pursued high-resolution crystallographic analysis.

Crystal structure of CtaG reveals a bipartite architecture and catalytic triad

To elucidate how conformational dynamics translate into molecular recognition and catalysis, we performed a series of in-depth structural and mutational analyses. Notably, a comprehensive primary sequence search in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB) for homologs of CtaG did not yield any significant hit, and the Foldseek database²³ predicted just two entry codes, K0EUI2_NOCB7 and A0A175VN02_9PEZI, with Z scores of 6.35 × 10⁻⁶ and 4.93 × 10⁻⁶, respectively. Given that both proteins exhibit only 11% sequence identity with CtaG, it is evident that structural analyses were required to elucidate mechanistic insights into CtaG. Therefore, we expressed CtaG in Escherichia coli BL21(DE3) and purified the recombinant enzyme by Ni²⁺ affinity and size-exclusion chromatography. Subsequently, we crystallized CtaG and solved its structure by single-wavelength anomalous dispersion (SAD) techniques to resolutions up to 1.4 Å (R_free = 18.9%; PDB: 9QUZ, 9QVJ; see Table S5).

The crystal structure of monomeric CtaG reveals a bipartite architecture linked by a short sequence (Figure 4A). The N-domain, comprising residues 1–181, exhibits a mixed α/β-fold structure. This includes an α-helical bundle composed of helices α1–α4, and a twisted, six-stranded antiparallel β sheet (strands S1–S6), which is flanked by helices α5 and α6 (Figure 4B). The C domain consists of residues 188–309 and adopts a compact coiled-coil configuration formed by five helices (α7–α11) that interact with helices α2 and α3 from the N-domain. In conclusion, CtaG forms a globular particle, with its N- and C-terminal domains creating a spacious intramolecular cavity of approximately 17 × 13 × 8 ų in size (Figure 4C). Notably, access to this cavity is restricted by a single entry pore of about 8 Å in diameter. While the AlphaFold3²¹ model of CtaG shows overall agreement with the crystal structure (Figure S6), pronounced deviations are observed in the region spanning residues 105–127. As this segment contributes to the formation of the substrate entry pore, the structural differences likely reflect its role in mediating specific interactions with CtaH and CtaE and may be critical for substrate recognition.

Figure 4.

Structural insights into CtaG catalysis

(A) Ribbon diagram of the CtaG_H128A variant (cyan) in complex with PHBA (carbon atoms in green; PDB: 9QVQ). The entrance pore, formed by residues 105–127 (orange), differs significantly from the AlphaFold3²¹ prediction (Figure S6A).

(B) Topology of CtaG shown from N to the C terminus with a blue-to-red color gradient. The catalytic chamber is indicated in purple; amino acids involved in catalysis and PHBA binding are labeled by the one letter code.

(C) Sliced surface representation of CtaG_H128A:PHBA with A128 in cyan. The structure reveals a prominent reaction chamber and entrance pore likely serving as docking site for crypto-CtaH and crypto-CtaE. Trp50 (black), distorted in apo-CtaG (PDB: 9QUZ), sequesters the PHBA thioester from bulk solvent and is positioned opposite the catalytic triad.

(D) Structural superposition of CtaG_H128A:PHBA (Ala128 in cyan) with apo-CtaG (carbons in gray) displays a perfect match of side chains. The F_o − F_c electron density map for PHBA is presented in gray mesh and contoured to 3σ (with the ligand omitted prior to phasing). The indole residue of Trp50 is unresolved in apo-CtaG (W^∗50).

(E) Proposed trajectory of the ping-pong reaction catalyzed by CtaG.

A more detailed structural analysis shows that the α/β-fold of CtaG is a motif commonly found in hydrolases. Moreover, the N-domain of CtaG comprises a catalytic cascade, consisting of Cys11, His128, and Asp144, which constitute the active site. This triad is related to papain-like proteases (Cys-His-Asn) (Figure S6B), where cysteine and histidine serve their typical functions, and Asp144 acts as a basic residue to enhance the nucleophilicity of Cys11. Within the central chamber, aromatic residues such as Trp50, Phe105, Tyr146, and Tyr147 play a crucial role in stabilizing the complex with the CtaH cargo to facilitate the transfer of PHBA to Cys11. These residues participate in π-π stacking with PHBA, forming an aromatic patch that enhances substrate specificity and catalytic efficiency, consistent with a ping-pong reaction mechanism (see below). Despite the high resolution of the dataset, Trp50 was structurally distorted in wild-type (WT) CtaG, indicating its flexibility and suggesting that the dynamic nature of this residue plays a prominent role in selectively binding the PHBA moiety (Figures 4C and 4D).

Aromatic gating enables directional catalysis by CtaG

To investigate the impact of catalytic residues in CtaG, we generated a series of variants that were subsequently analyzed by crystallography. The CtaG_C11A mutant (PDB: 9QVK, 9QVL; Figure S7; Tables S5 and S6) shows only minor structural deviations from WT. Notably, the indole ring of Trp50, now well defined in the F_oF_c map, was positioned at the base of the substrate pocket, increasing accessibility to the active site. In stark contrast, the conservative D144N replacement (PDB: 9QVS, Figure S7B; Table S7) resulted in significant structural rearrangements within the central cavity. Here, Trp50 divides the catalytic chamber into two equal halves. Moreover, the side chain of His128 is flipped by 120° and involved in π-π stacking with Phe105 and Trp50. Thus, Asp144 is crucial to maintain the structural integrity and functionality of CtaG. Conversely, the major substitution in H128A (PDB: 9QVO and 9QVP; Figure S7A; Table S6) display no structural alterations of residues in the substrate binding pocket, with Trp50 remaining disordered as seen in WT.

Next, we analyzed the CtaG-catalyzed ligation using PHBA bound to N-acetylcysteamine (SNAC-PHBA)²⁴ as a substitute for crypto-CtaH. Co-crystallization trials with WT resulted only in the apo form of CtaG, likely because the protonation of His128 during the transfer reaction promotes the gradual hydrolysis of the thioester intermediate over time. As expected, the C11A and D144N variants also failed to bind the surrogate, due to the absence of the nucleophilic sulfur atom and significant rearrangements in the active site, respectively. However, PHBA-complex structures were successfully obtained with the CtaG_H128A mutant in two different space groups with resolutions up to 1.6 Å (PDB: 9QVQ, PDB: 9QVR, Table S7). In these structures, PHBA is covalently bound to Cys11 via a thioester bond, and both datasets exhibit the same binding profile (Figure 4D). The orientation of the hydroxyl group on the benzoate toward the entrance pore indicates that PHBA has undergone a 180° rotation through an sp³-hybridized (O,S,S)-dithio-orthoester-oxyanion transition state. While the thioester between PHBA and CtaG is not stabilized by protein interactions, the aromatic portion of the ligand is anchored in the substrate binding pocket via intensive π-π stacking involving Trp50, Phe105, Tyr146, and Tyr147 (Figure 4D). This integrated aromatic interplay explains why PHBA derived from crypto-CtaH is preferred over the (Ala)₃-group of crypto-CtaE in the first step of the ping-pong reaction.

Notably, cargo delivery does not displace solvent from the active site. In contrast to cysteine proteases of the papain family,²⁵ no water molecules are positioned along a Bürgi-Dunitz trajectory, which prevents thioester hydrolysis from competing with the subsequent peptide coupling step via crypto-CtaE. Instead, the surrounding aromatic network, coordinated by Trp50, shields the thioester from solvent and stabilizes the intermediate. Since PHBA is thioester-bound to both CtaH and CtaG, this arrangement also prevents reverse transfer and suppresses hydrolysis at Cys11, thereby ensuring forward progression of the reaction.

CtaG catalyzes distal chain extension via a ping-pong mechanism

In the following cascade, the (βAla)₃-NH₂ group of crypto-CtaE forms an amide bond by attacking the thioester of the CtaG:PHBA intermediate via its amine group. Attempts to co-crystallize CtaG with CtaE were unsuccessful, and the AlphaFold3²¹ structure predictions for complex formation with CtaH and CtaE remain inconclusive. Notably, both models suggest interactions with residues 105–127, a region that differs significantly between the model and our experimental structure (Figures 3F, 3G, and S6A). Furthermore, crystallographic analysis indicates that productive binding of (βAla)₃-NH₂ requires repositioning of Trp50 to permit access to the active site. In the ligand-bound state, the indole nitrogen of Trp50 participates in a polar interaction network with Glu124, Asp126, and Lys167 (Figure S7C), which must be disrupted to enable catalysis. Binding of crypto-CtaE likely induces a conformational switch in Trp50 and opens a channel for nucleophilic attack by the (βAla)₃-NH₂ chain on the CtaG:PHBA thioester intermediate.

Notably, PHBA transfer is only effective when (βAla)₃ is loaded on CtaE, highlighting the functional specificity of the acceptor PCP. Thus, the CtaG_H128A:PHBA complex provides the first molecular snapshot of this reaction sequence and reveals a specificity pocket near the PHBA thioester (Figure 4C). This prechamber coordinates both peptide bonds of the (βAla)₃ chain and orients the terminal amine from the Si face for nucleophilic attack. Amide bond formation with PHBA proceeds via an (O⁻,N,S)-orthoester-oxyanion transition state, rendering the reverse reaction to the thioester energetically disfavored. Furthermore, this ping-pong mechanism enables CtaG to extend the N terminus of crypto-CtaE and release the distal, elongated metabolite, which remains covalently attached to CtaE (Figures 3E and 4E). Thus, the unique sequence and architecture of CtaG allow interaction with two distinct PCPs, representing a striking evolutionary adaptation to the biosynthetic framework of CTA assembly.

Structural homology search reveals related biocatalysts

The unusual mode of peptide bond formation catalyzed by the protease-like ligase CtaG prompted us to search for structural homologs potentially involved in related biosynthetic pathways. A distance matrix alignment (DALI) search²⁶ using the CtaG crystal structure revealed only two proteins with comparable folds: AziU2/U3, a structurally similar but functionally unrelated aziridine-forming enzyme²⁷ (Z score = 22.8; PDB: 8GS1; Figure S6C), and CcbD, a protein of unknown function (Z score = 22.0; PDB: 7YN2; Figure S6D). To move beyond solved structures, we employed a structure-based genome mining, which uncovered a set of CtaG-like enzymes embedded in diverse biosynthetic gene clusters (Figures 5 and S8; Table S4). These candidates define a previously unrecognized group of catalysts that likely function in non-canonical, ribosome-independent peptide assembly, underscoring the broader relevance of the CtaG fold in natural product biosynthesis. Among the most prominent homologs are AsbE and BtrH. AsbE is involved in petrobactin biosynthesis, while BtrH participates in the production of the aminoglycoside antibiotic butirosin.^28,29,30,31 Despite sharing less than 14% sequence identity with CtaG, both proteins exhibit strong structural conservation, including the catalytic triad and bipartite architecture with an N-terminal papain-like domain and a C-terminal helix-rich region (Figures S6C and S6D). While AsbE and BtrH accept PCP-bound donor substrates, they differ mechanistically from CtaG by acting on free acceptor substrates. In contrast, the fusion protein MllDE, comprising a carrier protein and a transacylase domain, is implicated in the biosynthesis of the lanthanide-binding siderophore methylolanthanin.³² Its AlphaFold3²¹ structure reveals both the conserved catalytic triad and the overall fold of CtaG. In addition, the carrier domain occupies a position analogous to the predicted PCP-binding interface in CtaG models (Figures 3F, 3G, and 5C), supporting a shared catalytic strategy based on single-tunnel substrate coordination. Altogether, the structural conservation across distant homologs expands the functional scope of CtaG-like enzymes and reveals a previously unrecognized class of thiotemplated peptide ligases relevant to the evolution and engineering of ribosome-independent peptide bond formation.

Figure 5.

CtaG structural homologs function in diverse biosynthetic pathways

AlphaFold3²¹ models of CtaG homologs involved in distinct biosynthetic pathways are shown: (A) AsbE from petrobactin biosynthesis, (B) BtrH from butirosin biosynthesis, and (C) MllDE from methylolanthanin biosynthesis. The respective reactions or final products are indicated. The conserved catalytic triad (Cys, His, and Asp), also present in CtaG, is highlighted in red. Transacylase domains are shown in gray; the carrier protein domain of MllDE is depicted in black.

Discussion

Modular NRPSs have long served as a blueprint for the biosynthesis of structurally diverse peptides. However, their large size, rigid architecture, and reliance on co-localized modules pose significant challenges for reprogramming and synthetic applications.³³ In contrast, non-modular peptide assembly lines such as the CTA biosynthetic pathway offer a more flexible platform. While earlier studies have primarily focused on substrate-activating ligases such as ATP-grasp or adenylate-forming enzymes,^10,34 protease-like ligases remain largely unexplored as catalysts for amide bond formation. Our biochemical and structural investigation of CtaG reveals a distinct mechanism for peptide extension that diverges fundamentally from NRPS logic. CtaG catalyzes amide bond formation via a ping-pong mechanism, operating through a single substrate tunnel. In this sequence, PHBA is first transferred from the donor PCP (crypto-CtaH) to the catalytic cysteine (Cys11) of CtaG, generating a covalent intermediate. This intermediate is subsequently resolved through nucleophilic attack by the acceptor PCP (crypto-CtaE), enabling distal substrate loading and facilitating extension of the polythioamide backbone. Unlike canonical NRPS C domains, which accommodate both PCPs simultaneously, CtaG is structurally restricted to engage with one PCP at a time. This directional mode of chain extension is conceptually reminiscent of solid-phase peptide synthesis. Interestingly, our functional and structural data indicate that this selectivity is dynamically regulated. CtaH binds to CtaG in both the apo and PHBA-bound states, indicating broader compatibility. In contrast, CtaE interacts only when CtaG has undergone PHBA-induced remodeling, which reshapes the substrate tunnel and alters the electrostatic surface. This asymmetric recognition explains the observed substrate flow and highlights the role of ligand-triggered conformational change in directing selective protein-protein interactions. Notably, AlphaFold3²¹ failed to predict these dynamic features. The structural models lacked the loop rearrangements and charge redistribution observed in our crystallographic data, illustrating current limitations in structure prediction of transient, ligand-sensitive protein complexes.

Prompted by these mechanistic insights, we next explored whether the structural and functional logic of CtaG extends to other biosynthetic systems. While protease-like enzymes are classically associated with peptide bond hydrolysis, they are increasingly recognized as catalysts for amide bond formation in specialized biosynthetic pathways.^35,36 One example is CcbD, a C enzyme with remote similarity to cysteine proteases, which catalyzes amino sugar acylation during lincosamide biosynthesis.³⁷ Although structurally unrelated to CtaG, CcbD likewise acts on carrier protein-bound substrates, supporting a broader role for protease-like folds in peptide assembly. To assess whether this principle applies more widely, we performed structure-guided genome mining and identified distant CtaG homologs in the biosynthetic gene clusters of petrobactin, butirosin, and methylolanthanin.^29,31,32 Despite low sequence identity, these enzymes share the conserved bipartite architecture and catalytic triad of CtaG. At least two candidates also act on PCP-linked substrates, and their colocalization with ATP-grasp, decarboxylase, and CoA-ligase domains suggests a shared biosynthetic logic. Taken together, our data define a previously unrecognized class of thiotemplated peptide ligases. These enzymes expand the catalytic toolbox for ribosome-independent amide bond formation and provide a mechanistic foundation for the engineering of modular and non-modular peptide assembly lines.

Methods

General materials and methods

All enzymes used were purchased from either Thermo Fisher Scientific (USA) or New England Biolabs (NEB, USA). Sequencing and oligonucleotide primer synthesis was performed by Eurofins Genomics (Germany) and media components were purchased from Sigma-Aldrich (USA) and Roth (Germany).

Bacterial strains and culturing conditions

Escherichia coli liquid cell cultures were grown in lysogeny broth (LB) with agitation or on LB agar plates at 37°C containing the appropriate antibiotic (25 μg mL⁻¹ chloramphenicol, 50 μg mL⁻¹ kanamycin). Plasmid construction and storage was performed with E. coli TOP10, while E. coli BL21 (DE3) and E. coli Rosetta (DE3) was used for heterologous protein overproduction (Table S1).

Plasmid construction for heterologous production of CoaA, CoaD, CoaE, and Sfp in E. coli

The target genes coaA, coaD and coaE were amplified by PCR from a colony of E. coli TOP10, whereas the gene sfp was amplified from genomic DNA of Bacillus subtilis strain 168, using the oligonucleotide primers listed in Table S3. PCR amplifications were performed using the Phusion High-Fidelity DNA Polymerase (NEB) and amplicons were purified using a Monarch PCR & DNA Cleanup Kit (NEB). The purified PCR products were assembled with NcoI- and XhoI-digested pET28a vector using NEBuilder HiFi DNA Assembly Master Mix (NEB) according to the manufacturer’s protocol. Assembled expression vectors were introduced into E. coli TOP10 through electroporation and transformants were selected on LB agar plates supplemented with the relevant antibiotic (Table S2). Plasmids were isolated from the transformants using a Monarch Plasmid Miniprep Kit (NEB) and verified by sequencing using the primers T7 Seq F and T7 Seq R (Table S3).

Site-directed mutagenesis

Plasmids were amplified by PCR using the mutagenesis oligonucleotide primers listed in Table S3 and Phusion High-Fidelity DNA Polymerase (NEB). Template DNA was removed by digestion with DpnI (1 U; NEB) for 4 h at 37°C. Digested samples were used to transform E. coli TOP10 cells and transformants were selected on LB agar supplemented with the relevant antibiotic (Table S2). Plasmids were isolated from transformants and verified by sequencing using the primers T7 Seq F and T7 Seq R (Table S3).

Heterologous production and purification of proteins

E. coli BL21 (DE3) (for Sfp-His, CoaA-His, CoaD-His and CoaE-His) or E. coli Rosetta (DE3) (for His-CtaE-His-CtaH) cells were transformed with pET28a-derived expression vectors and transformants were selected on LB agar plates supplemented with the relevant antibiotic. A 500 mL volume of LB medium supplemented with the relevant antibiotic was inoculated with 5 mL of an overnight pre-culture of E. coli BL21 (DE3) harboring recombinant plasmid and grown at 37°C with agitation, until an OD₆₀₀ of 0.5–0.75 was reached. Cultures were then iced for 20 min before protein overproduction was induced by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 0.4 mM. Following induction, cells were grown at 18°C with agitation overnight and afterward harvested by centrifugation at 4,000 × g and 4°C for 10 min. The resulting pellet was washed with Tris buffered saline (50 mM Tris pH 7.5, 150 mM NaCl) and stored at −20°C.

Frozen cell pellets were resuspended in 40 mL of lysis-wash buffer (50 mM Tris pH 8.0, 300 mM NaCl, 25 mM imidazole, 5% glycerol v/v). Cells were sonicated at 4°C using a SONOPLUS HD 4000 ultrasonic homogenizer with a TS109 microtip (Bandelin) and the following parameters: 70% amplitude, four 2 min cycles with 0.5 s pulses and 2 min breaks between cycles. The lysate was then centrifuged for 30 min at 12,000 × g and 4°C, to pellet insoluble debris. The supernatant was loaded onto 1–2 mL HisPur Ni-NTA resin (Thermo Fisher Scientific) equilibrated with lysis-wash buffer. Unbound proteins were removed by washing with 100 mL lysis-wash buffer. Bound proteins were eluted with 15 mL elution buffer (50 mM Tris pH 8.0, 300 mM NaCl, 250 mM imidazole, 5% glycerol v/v). The eluate was concentrated using Amicon Ultra-3kDa molecular weight cut-off centrifugal filters (Merck Millipore) and buffer exchanged into storage buffer (25 mM HEPES pH 7.5, 200 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine [TCEP]) using PD MidiTrap G-25 columns (Cytiva). Purified proteins were flash frozen in liquid N₂ and stored at −70°C. Protein concentration was determined by absorbance at 280 nm, and purity was assessed by SDS-PAGE (Figure S1).

In vitro phosphopantetheinylation of PCPs

Generation of PHBA- or (βAla)₃-loaded PCPs was performed with 30 μM apo-PCP, 1 μM Sfp-His, 0.75 μM CoaA-His, 0.75 μM CoaD-His, 0.75 μM CoaE-His, and 400 μM pantetheine conjugate, (βAla)₃- (Figure S9), PHBA- (Figure S10), or chloroacrylamide-pantetheine, and 2 mM ATP in reaction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP, and 20 mM MgCl₂). For the generation of holo-PCPs, reactions were performed with 400 μM CoA instead of pantetheine conjugates, and without ATP and CoA biosynthetic enzymes. Reactions were allowed to proceed for 1–2 h at 30°C and the obtained PHBA-, (βAla)₃- or holo-PCPs were used for enzymatic assays without further purification.

CtaG reconstitution assays with CtaE-bound (βAla)₃ and non-CtaE-bound (βAla)₃

Reactions with CtaE-bound (βAla)₃ were performed with 0.5 μM CtaD, 30 μM holo-CtaE, 0.5 μM CtaF, 1 μM CtaG, 2 μM holo-CtaH, 0.5 μM CtaI, 1 mM l-aspartate, 1 mM ATP, 0.5 mM PHBA, and 5 μM PLP in reaction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP and 20 mM MgCl₂). Reactions with non-CtaE-bound (βAla)₃ were performed with 1 μM CtaG, 2 μM holo-CtaH, 0.5 μM CtaI, 30 μM (βAla)₃, 1 mM ATP and 0.5 mM PHBA in reaction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP and 20 mM MgCl₂). Control reactions were performed with heat denatured CtaG. Reactions were allowed to proceed for 2 h at 30°C and for identification of the CtaG reaction product by LC-HR-MS, samples were added with 10 mM potassium hydroxide and incubated at 55°C for 1 h. The samples were diluted 1:1 with methanol and incubated at room temperature (RT) for 5 min prior to analysis by LC-HR-MS. Enzymatically produced PHBA-(βAla)₃ was compared with an authentic synthetic standard.

CtaG backbone assembly reactions with (βAla)₃-/PHBA-CtaE and (βAla)₃-/PHBA-CtaH

Reactions were performed with 1 μM CtaG and 15 μM of two of the following PCP species: (βAla)₃-CtaE, PHBA-CtaE, (βAla)₃-CtaH, and PHBA-CtaH in reaction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP, and 20 mM MgCl₂). Assays were allowed to proceed for 4 h at 30°C and subsequently analyzed by MALDI-TOF-MS.

Chloroacrylamide-pantetheine crosslinking of CtaG and CtaH

Reactions were performed with 15 μM CtaG or CtaG_C11A, and 15 μM chloroacrylamide-CtaH in reaction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP, and 20 mM MgCl₂). Assays were allowed to proceed for 1 min to 24 h at 30°C and subsequently analyzed by SDS-PAGE.

LC-HR-MS

LC-HR-MS measurements were performed with an Accela high-performance liquid chromatography system coupled to an Exactive Hybrid Quadrupole Orbitrap (Thermo Fisher Scientific) mass spectrometer equipped with an electrospray ion source. Separation was performed with an Betasil C18 column (2.1 × 150 mm, 3 μm, Thermo Fisher Scientific) operating at a flow rate of 200 μL min⁻¹, with 0.1% formic acid (solvent A) and acetonitrile + 0.1% formic acid (solvent B) and the following gradient: 5% solvent B for 0.1 min, 5%–98% solvent B over 10 min, hold 98% solvent B for 4 min, 98%–5% solvent B over 0.1 min, hold 5% solvent B for 6 min.

MALDI-TOF-MS

MALDI-TOF-MS analysis of protein samples was performed with an ultrafleXtreme TOF/TOF (Bruker). To prepare protein samples for analysis, 2 μL of protein sample was mixed with 2 μL 100 mM 2′-5′-dihydroxyacetophenone (DHAP) and 2 μL 2% trifluoroacetic acid (TFA). Afterward, 2 μL of this mixture was spotted onto a MTP AnchorChip 384 T F (Bruker), crystallized at 20°C for at least 5 h, and washed three times with 2 μL 0.1% TFA to remove residual salts. The instrument was operated in positive linear mode. Data acquisition was performed using flexControl 3.3 (Bruker) in a mass range of 5–25 kDa, with 1 kHz laser frequency, 50% laser power, and 5,000 shots per spot. The spectrometer was calibrated to commercially available standards (Protein Calibration Standard I and II, Bruker) prior to each measurement. For data analysis the program flexAnalysis 3.3 (Bruker) was used.

Protein structure predictions

The amino acid sequences of CtaE, CtaG, and CtaH were used to predict their monomeric structures and the dimeric structures of CtaE-CtaG and CtaG-CtaH using AlphaFold3.²¹ Structure predictions were performed using the default parameters.

Bioinformatics

For the CtaG multiple sequence alignment, potential structural homologs were identified with Foldseek²³ using the apo-CtaG structure as input. Structural homologs were manually curated, and the corresponding protein sequences used for a multiple sequence alignment with ClustalW³⁸ using the default parameters.

Protein crystallization and structure determination

Crystals of CtaG (20 mg mL⁻¹) were grown using sitting drop vapor diffusion methods at 20°C. For co-crystallization experiments, the respective ligand (100 mM stock solution in DMSO) was added to CtaG to a final concentration of 2 mM. Crystallization droplets had a maximum volume of 0.4 μL with a protein to reservoir ratio of either 1:1, 2:1 or 3:1. The distinct crystallization parameters for each of the recorded datasets are shown in Tables S5–S7.

Crystals were cryoprotected with a 7:3 mixture of mother liquor and 100% (v/v) glycerol prior vitrification in liquid nitrogen. High-resolution datasets of CtaG and its variants were collected using synchrotron radiation (λ = 1.0 Å) at beamline X06SA, Swiss Light Source (SLS), Paul Scherrer Institute, Villigen, Switzerland. Reflection intensities for each dataset were processed with the XDS program package (Tables S5–S7).³⁹ Experimental phases were obtained by soaking native CtaG crystals in KAuCl₄ (10 μM dissolved in mother liquor) for 4 h to allow for heavy metal decoration. A dataset of the derivative was recorded at 1.5 Å in space group P2₁ (a = 47.0 Å, b = 61.1 Å, c = 106.1 Å, β = 106.1°), resulting in the identification of six Au sites applying SHELXD.⁴⁰ Subsequent SAD phasing with SHARP⁴¹ combined with solvent flattening, provided adequate phases at 2.5 Å resolution. PHENIX AutoBuild⁴² was then used to generate a high-quality model, which was positioned in the native dataset (Tables S5–S7). Model building was carried out with Coot⁴³ through iterative rounds. Water molecules were automatically added by ARP/wARP⁴⁴ and refinement cycles were conducted with REFMAC5⁴⁵ yielding excellent statistics and model parameters (Tables S5–S7). Eventually, the structure of CtaG was solved in space group P2₁ at 1.4 Å resolution. In the following, the structures of CtaG mutants and the CtaG_H128A:PHBA complex were determined in various space groups using the apo-CtaG coordinates as a starting model for Molecular Replacement. All crystal structures have been deposited in the RCSB Protein Data Bank with their respective accession codes (Tables S5–S7).

ITC

All ITC experiments were carried out using a MicroCal PEAQ-ITC microcalorimeter (Malvern Panalytical) at 25°C. To minimize heat of dilution effects, protein and ligand solutions were prepared in identical buffer: 20 mM MES (pH 6.0) with 200 mM NaCl. Proteins were buffer-exchanged using HiTrap Desalting columns packed with Sephadex G-25 resin (Cytiva). Prior to use, all solutions were filtered through 0.22 μm membranes and degassed under vacuum for 15 min to remove air bubbles that could interfere with measurements. Protein and ligand concentrations were determined by Bradford assay. The sample cell (0.2 mL) was filled with protein solution, while MilliQ water was used in the reference cell. The injection syringe (60 μL) was loaded with the PCP proteins. For titrations, CtaH was used at 400 μM and CtaE at 1.2 mM. CtaG was tested at 30 μM (for CtaH titrations) and 120 μM (for CtaE). Each titration started with a 0.4 μL pre-injection for equilibration (data discarded), followed by 18 injections of 2 μL each. Injections were spaced 200–250 s apart to allow baseline stabilization. All experiments were conducted at constant temperature (25°C) with continuous stirring at 750 rpm to ensure homogeneous mixing. Control experiments were performed by titrating CtaE or CtaH into buffer alone or buffer into protein solution to determine heats of dilution. Raw thermograms were baseline-corrected using AFFINImeter ITC software (AFFINImeter). Due to the low-affinity, highly transient nature of the interactions between CtaG and its PCP partners CtaH and CtaE, combined with minimal enthalpic contributions, reliable fitting of the ITC data was not feasible. Consequently, the experiments provided only qualitative insight into the ligand- and state-dependent binding behavior of these protein-protein complexes.

Further details regarding the methods can be found in the supplemental information.

Resource availability

Lead contact

Requests for further information and resources should be directed to Christian Hertweck (christian.hertweck@leibniz-hki.de).

Materials availability

All unique reagents generated in this study are available from the lead contact upon request.

Data and code availability

All data needed to evaluate the conclusion of this work are presented in the paper and the supplemental information. Other data related to this work are available from the lead contact upon reasonable request. Full experimental procedures are provided in the supplemental information.

Acknowledgments

We are grateful to H. Heinecke and A. Perner for NMR and LC-MS measurements. We thank A. König for assistance in protein crystallization experiments and the staff of the beamline X06SA at the Paul Scherrer Institute, Swiss Light Source, Villigen, Switzerland, for assistance during data collection. The research leading to these results has received funding from the European Community’s Seventh Framework Programme (FP7/2007-2013) under BioStruct-X (grant agreement no. 283570). Financial support from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), Project-ID 239748522-SFB 1127, and Leibniz Award (to C.H.) as well as the European Regional Development Fund (ERDF) (MassNat and WiVi) is gratefully acknowledged.

Author contributions

Conceptualization, M.G. and C.H.; methodology, F.G., A.B., M.D., J.F., K.L.D., and M.G.; investigation, F.G., A.B., M.D., J.F., and M.G.; writing – original draft, F.G., A.B., and M.G.; writing – review and editing, M.G. and C.H.; funding acquisition, M.G. and C.H.; supervision, K.L.D., M.G., and C.H.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

The authors used DeepL and ChatGPT to improve language clarity and style. All scientific content was generated by the authors, who take full responsibility for the final manuscript.

Published: September 16, 2025