Catalytic architecture and cyclase‐mediated dimerization of a newly identified class III‐c lanthipeptide synthetase

Abstract

Lanthipeptide synthetases catalyze dehydration and cyclization reactions that generate structurally complex ribosomally synthesized and post‐translationally modified peptides. Class III LanKC enzymes are distinguished by their tri‐functional kinase, lyase, and cyclase architecture, yet their mechanistic diversity remains incompletely understood due to limited structural and biochemical characterization. Here, we identify a previously unrecognized subgroup of Class III LanKC enzymes, designated Class III‐c, through phylogenetic analysis and conserved substrate peptide features. Using cryo‐electron microscopy, we determined the structure of a representative Class III‐c enzyme, TherKC, in complex with its peptide substrate. We performed a systematic structural and functional characterization of TherKC, encompassing peptide substrate binding, oligomerization, and catalysis, providing a comprehensive mechanistic framework for this new LanKC class. TherKC assembles as a homodimer mediated exclusively by cyclase–cyclase interactions, a mode of oligomerization distinct from all previously characterized LanKC enzymes. Structural analysis identifies Arg685 as a conserved cyclase‐domain hotspot that anchors dimer formation through symmetric electrostatic interactions. Disruption of this interface abolishes dimerization and reduces enzyme stability, while preserving substrate binding and catalytic competence. Structure‐guided mutagenesis and mass spectrometric analyses further define conserved cyclase residues essential for efficient dehydration–cyclization coupling, provide insight into the formation of the conserved labionin‐forming motif, revealing a noncanonical yet conserved catalytic framework. Together, these findings establish Class III‐c LanKC enzymes as a mechanistically coherent subgroup and provide a comprehensive structural and functional understanding of this newly identified class, including its substrate‐recognition, oligomerization, and cyclization mechanisms, thereby expanding the current landscape of lanthipeptide biosynthesis.

1. INTRODUCTION

Ribosomally synthesized and post‐translationally modified peptides (RiPPs) constitute a large class of natural products whose structural diversity arises from enzymatic tailoring of genetically encoded precursor peptides (Arnison et al., 2013; Velásquez & van der Donk, ²⁰¹⁰). Among these, lanthipeptides are defined by the presence of intramolecular thioether cross‐links that confer enhanced stability and diverse biological activities, including antimicrobial, antiviral, and immunomodulatory functions, and have positioned them as promising scaffolds for the development of new therapeutics, particularly in the context of rising antimicrobial resistance (Bierbaum & Sahl, 2009; King et al., ²⁰²³; Knerr & van der Donk, ²⁰¹²; Sahl & Bierbaum, ¹⁹⁹⁸).

Lanthipeptide biosynthesis proceeds through a conserved sequence of post‐translational transformations catalyzed by lanthipeptide synthetases. The precursor peptide (LanA) consists of an N‐terminal leader peptide (LP), which mediates enzyme recognition, and a C‐terminal core peptide (CP), which undergoes chemical modification (Burkhart et al., 2015; Repka et al., ²⁰¹⁷; Viel & Kuipers, ²⁰²²). Biosynthesis involves three principal chemical steps. First, selected serine and threonine residues in the CP are activated through either phosphorylation or glutamylation, depending on the enzyme class, in a nucleatide‐dependent manner (Repka et al., 2017). This activation enables β‐elimination to generate dehydroalanine (Dha) or dehydrobutyrine (Dhb) residues. In the final step, cysteine thiols in the CP undergo stereoselective Michael‐type addition to Dha or Dhb to form thioether cross‐links, yielding lanthionine or methyllanthionine rings. In some Class III lanthipeptides, more complex labionin cross‐links are generated through additional cyclization chemistry, although the underlying mechanisms remain poorly understood (Hernandez Garcia & Nair, 2023). Importantly, this cyclization step is ATP‐independent across all lanthipeptide classes and relies instead on enzyme‐templated chemical reactivity (Habibi et al., 2022; Li et al., ²⁰²¹; Repka et al., ²⁰¹⁷).

Based on the organization and catalytic strategy of their modification enzymes, lanthipeptide biosynthetic systems are currently classified into five major classes (Arnison et al., 2013; Lagedroste et al., ²⁰²⁰; Pei et al., ²⁰²²; Xu et al., ^2020a; Zhang et al., ²⁰¹²). In Class I systems, dehydration and cyclization are catalyzed by separate enzymes, LanB and LanC (Li et al., 2006; Li, Yu, et al., ²⁰⁰⁶; Ortega et al., ²⁰¹⁴; van der Donk & Nair, ²⁰¹⁴). Class II enzymes employ a single bifunctional synthetase, LanM, to perform both steps (Dong et al., 2015; Rahman et al., ²⁰²⁰; Xie et al., ²⁰⁰⁴). More recently identified Class III (LanKC) and Class IV (LanL) enzymes are tri‐functional proteins comprising kinase, lyase, and cyclase domains (Hegemann & Süssmuth, 2020; Hernandez Garcia & Nair, ²⁰²³; Li et al., ²⁰²⁴; Repka et al., ²⁰¹⁷; Sigurdsson et al., ²⁰²³). A fifth class (Class V) has been identified more recently, in which the three catalytic activities are encoded by individual polypeptides (Liang et al., 2022; Pei et al., ²⁰²²; Randall et al., ²⁰²⁴; Xu et al., ^2020b). Among these systems, Class III LanKC enzymes are mechanistically distinctive in that their cyclase domains lack the canonical Zn²⁺‐binding motifs found in classical LanC cyclases, indicating an alternative strategy for thioether bond formation (Hegemann & Süssmuth, 2020; Lee & van der Donk, ²⁰²²; Wang & van der Donk, ²⁰¹²). Structural and biochemical studies of several representative Class III enzymes, including ThurKC (Hernandez Garcia & Nair, 2023), PneKC (Li et al., 2024) and SalKC (Li et al., 2025), have revealed important differences in nucleotide preference, cyclase active‐site organization, and oligomeric state. In particular, some Class III LanKC enzymes form dimers, and in certain systems dimerization has been proposed to influence catalytic behavior through inter‐protomer communication or functional asymmetry (Li et al., 2024).

Despite these advances, current understanding of Class III LanKC enzymes remains limited. Only a small number of representatives have been structurally and biochemically characterized (Hernandez Garcia & Nair, 2023), and many predicted LanKC homologs do not cluster with previously characterized subclasses (Hegemann & Süssmuth, 2020; Hernandez Garcia & Nair, ²⁰²³; Huang et al., ²⁰²²; Li et al., ^{2024, 2025}). As a result, the relationship between sequence divergence, substrate specificity, oligomeric architecture, and catalytic function within Class III LanKC enzymes remains poorly defined. In particular, it is not known whether alternative modes of oligomerization correspond to distinct catalytic mechanisms or primarily serve structural roles.

In this study, we identify and characterize a previously unrecognized subgroup of Class III LanKC enzymes, which we designate Class III‐c. Through phylogenetic analysis and substrate sequence comparison, we show that Class III‐c enzymes form a distinct group associated with conserved leader and core peptide features. Focusing on a representative enzyme, TherKC from Thermoactinomyces sp., we combine cryo‐electron microscopy, biochemical assays, structure‐guided mutagenesis, and mass spectrometric analysis to define its oligomeric organization and catalytic determinants. Our results reveal a cyclase‐mediated dimerization mode that enhances enzyme stability while preserving a conserved catalytic framework for lanthipeptide modification. Together, these findings expand the structural and mechanistic landscape of Class III lanthipeptide synthetases and provide new insight into how architectural diversification can coexist with conserved catalytic logic in lanthipeptide biosynthesis.

2. RESULTS

2.1. Reclassification of Class III LanKC enzymes identify a previously uncharacterized Class III‐c subgroup

Class III lanthipeptide modification enzymes catalyze nucleotide‐dependent dehydration and cyclization reactions but display large sequence diversity, particularly within the cyclase domain (Li et al., 2024; Wiebach et al., ²⁰²⁰). Our previous study has largely grouped Class III LanKC enzymes into only two subclasses based on limited sequence sampling (Li et al., 2024). To establish a more complete and mechanistically informative framework, we performed an expanded phylogenetic and comparative sequence analysis of Class III LanKC enzymes across diverse bacterial lineages. Maximum‐likelihood phylogenetic analysis revealed that Class III LanKC enzymes segregate into four well‐supported subclasses, which we designate Class III‐a, III‐b, III‐c, and III‐d (Figure 1a). Pairwise sequence comparison showed that the overall sequence identity between enzymes belonging to different subclasses is consistently low (<20%), indicating deep evolutionary separation rather than gradual divergence. In contrast, enzymes within each subclass cluster tightly with sequence identity higher than 74.21% for III‐a, 47.49% for III‐b, 55.52% for III‐c and 36.31% for III‐d, supporting their designation as discrete evolutionary groups. Comparison with our previous two‐group framework indicates that the newly defined III‐b subclass largely corresponds to the previously defined III‐b branch, whereas the previously broader III‐a‐related branch can now be further grouped into three distinguishable subclasses: III‐a, III‐c, and III‐d. Among these, III‐a and III‐c form more closely related sister groups, while III‐d represents a more divergent branch with distinct substrate features. Previously characterized LanKC enzymes map cleanly onto this refined classification: ThurKC clusters within Class III‐a, PneKC within Class III‐b, and CurKC within Class III‐d. Notably, a large number of LanKC homologs form a distinct and previously unexplored clade, which we designate Class III‐c, within which TherKC is located (Figure 1a). To date, no representative from this subgroup has been biochemically or structurally characterized, leaving its enzymatic properties and mechanistic features entirely unknown. Domain‐level comparison revealed that while the kinase and lyase domains are broadly conserved across all four subclasses, the cyclase domain shows pronounced divergence, with particularly low sequence identity between subclasses (7.9%–17.6%; Figure S1A). This pattern suggests that subclass‐specific functional adaptations are likely encoded within the cyclase domain, rather than in the upstream modification modules. Within Class III‐c, multiple highly conserved motifs are evident in the cyclase domain that seems are absent from other subclasses. Using TherKC numbering, these include a conserved LIxGDIRQ (AA region 512–520) motif near the N‐terminus, LRSGLAGIGL (AA region 608–617) and GLIDGWSGVS motifs (AA region 663–672) in the central region, and a GILxxNPLYWLP motif (AA region 838–849) near the C‐terminus.

FIGURE 1.

Phylogenetic reclassification of Class III LanKC enzymes and initial biochemical characterization of TherKC. (a) Maximum‐likelihood phylogenetic analysis of Class III lanthipeptide synthetases identifies four well‐supported subclasses, designated Class III‐a, III‐b, III‐c, and III‐d. TherKC clusters within the previously uncharacterized Class III‐c subgroup. (b) Sequence logo analysis of precursor peptides associated with each subclass. Class III‐a and Class III‐c substrates are enriched in Ser/Thr‐rich core peptides and frequently contain TITTT‐like elements and an STxSNxC motif, whereas Class III‐b and Class III‐d substrates display SxxSxxxC‐type motifs with distinct spacing and context. Conserved features in the LP and CP regions are indicated. (c) SDS‐PAGE gel image of apo TherKC and co‐expressed TherKC‐TherA complex. (d) SEC profiles of TherKC expressed alone or co‐expressed with TherA. A major peak around 11.5 mL corresponds to an apparent molecular mass of ~200 kDa, consistent with a homodimeric TherKC assembly. (e) Nucleotide hydrolysis activity of TherKC in the presence of TherA using different NTPs as substrates. (f) Michaelis–Menten analysis of ATP hydrolysis by TherKC‐TherA. (g) Intact‐mass analysis of TherA before and after modification by TherKC. Unmodified TherA undergoes MMTS labeling (+45.9 Da), whereas TherKC‐modified TherA shows progressive dehydration and no detectable MMTS labeling, consistent with cysteine consumption during thioether‐ring formation. A schematic of the MMTS labeling reaction is shown at lower left. (h) βME trapping analysis of TherKC‐modified TherA. MALDI‐TOF‐MS revealed up to six βME adducts (+78 Da each), indicating six accessible Dha/Dhb residues, while two Dha/Dhb residues remain inaccessible, consistent with participation in ring formation. Data in (e) and (f) are presented as mean ± SD (n = 3); ns, not significant; *p< 0.05; **p< 0.01; ***p< 0.001.

To assess whether this enzyme‐level divergence is reflected at the substrate level, we next analyzed the precursor peptide (LanA) sequences associated with each subclass (Figure 1b). This analysis revealed subclass‐dependent substrate patterns that broadly parallel the phylogenetic separation. Class III‐b substrates contain two prominent core peptide motifs, including an upstream SxxSxxxC‐type motif and a downstream SxxxC‐type motif. Class III‐d substrates also contain two Cys‐associated motif regions, but their spacing and surrounding sequence context differ from those of III‐b, consistent with III‐d forming a separate subclass. In contrast, Class III‐a and III‐c substrates show a more similar organization, with a dominant STxSNxC‐like motif near the core peptide region. Class III‐a substrates additionally display stronger upstream Thr‐rich enrichment, including TITTT‐like features, whereas Class III‐c substrates retain the STxSNxC‐like motif but show a less pronounced upstream Thr‐rich pattern. Thus, the substrate comparison supports the close relationship between III‐a and III‐c while also distinguishing III‐c from the dual‐motif substrate architectures observed in III‐b and III‐d. This relationship is important for interpreting TherKC: although Class III‐c enzymes have evolved distinct features within their cyclase domains, their substrate motif organization remains closer to Class III‐a, suggesting that architectural diversification of the enzyme can occur without a complete change in the core peptide recognition pattern.

Rather than revealing strictly subclass‐exclusive motifs, the analysis highlights shared architectural features with subclass‐dependent enrichment and organization. Class III‐a substrates are enriched in serine‐ and threonine‐rich core peptides and frequently contain TITTT‐like elements together with an STxSNxC motif. Class III‐b substrates are instead characterized by SxxSxxxC‐type motifs that have been linked to distinct cyclization patterns, while Class III‐d substrates also contain SxxSxxxC‐related motifs but with altered spacing and sequence context, consistent with their separate phylogenetic placement. Notably, Class III‐c substrates do not exhibit a fundamentally distinct sequence architecture. Instead, they display a core peptide organization that is more similar to Class III‐a (Hernandez Garcia & Nair, 2023) than to Class III‐b (Li et al., 2024) or III‐d. Class III‐c core peptides are likewise enriched in serine and threonine residues and often contain both TITTT‐like features and an STxSNxC motif, suggesting a shared modification logic involving extensive dehydration followed by cysteine‐mediated thioether bond formation.

In the leader peptide region, a hydrophobic pattern typified by a VLxLxxL motif is prominent in Class III‐c; however, similar hydrophobic motifs are also present in Class III‐a and other subclasses, albeit with different residue identities and positional variability. Together, phylogenetic analysis and cyclase‐domain divergence define Class III‐c as a previously unrecognized subgroup of Class III LanKC enzymes, while revealing that, despite extensive enzyme‐level sequence divergence, Class III‐c enzymes act on substrate peptides that are architecturally similar to those of Class III‐a systems.

2.2. Biochemical characterization of the Class III‐c LanKC enzyme TherKC and its substrate TherA

Having identified Class III‐c as a distinct subgroup of Class III LanKC enzymes and established TherKC as a representative member, we next sought to define its biochemical properties and substrate specificity. To this end, we examined the lanthipeptide biosynthetic gene cluster from Thermoactinomyces sp. DSM 45892, which encodes a canonical Class III lanthipeptide maturation system. The cluster includes the LanKC enzyme TherKC, two nearly identical precursor peptides (TherA), a heterodimeric ABC transporter, and a downstream putative metalloprotease likely responsible for leader peptide removal (Figure S2A). Given the near identity of the two TherA peptides, one representative substrate was selected for subsequent analysis.

TherKC and TherA were heterologously expressed in E. coli BL21 (DE3) (Figure 1c) (Figure S2B). Unmodified TherA was obtained either by denaturation–refolding following inclusion body purification or by co‐expression with TherKC. Size‐exclusion chromatography (SEC) revealed that purified TherKC elutes as a single dominant peak corresponding to an apparent molecular mass of ~200 kDa (Figure 1d), consistent with a dimeric assembly. Notably, this elution profile was unchanged in the presence or absence of the substrate peptide TherA, indicating that dimer formation is an intrinsic property of TherKC rather than a substrate‐induced effect.

We next examined the nucleotide dependence of TherKC activity. To validate the ATPase assay conditions used throughout this study, we first assessed the dependence of ATP hydrolysis on substrate concentration and reaction time. Under the assay conditions employed, ATP hydrolysis increased linearly with increasing TherA concentration over the tested range of 0.5–3.5 μM (Figure S2C). Likewise, time‐course measurements performed using 1.5 μM TherKC and 1.5 μM TherA showed an approximately linear increase in ATP hydrolysis during the first 100 min of the reaction (Figure S2D). Based on these results, all subsequent ATPase assays were performed using 1.5 μM TherKC and 1.5 μM TherA with a reaction time of 1 h, ensuring measurements were obtained within the linear range of the assay. In the absence of substrate, TherKC exhibited low basal nucleotide hydrolysis, with a modest preference for ATP over other NTPs (Figure 1e). Upon addition of equimolar TherA, hydrolysis of all tested NTPs increased significantly, with ATP remaining the preferred substrate. Michaelis–Menten analysis yielded an apparent ATP Km of ~0.40 mM and a kcat of 0.30 s⁻¹ (Figure 1f), indicating that TherKC functions as a substrate‐stimulated ATPase, consistent with its role in lanthipeptide modification. To directly assess peptide modification, TherA was incubated with TherKC in the presence of ATP and Mg²⁺, followed by intact mass spectrometry analysis (Figures 1g and S3). In the absence of TherKC, TherA exhibited a mass consistent with the unmodified peptide (7939.4 Da; theoretical 7939.9 Da). In contrast, TherKC‐treated samples displayed mass shifts corresponding to up to eight dehydration events, consistent with modification of all eight dehydratable Ser/Thr residues present in TherA, indicating extensive post‐translational modification (Figure 1g). To further probe cysteine engagement during modification, we employed S‐methylmethanethiosulfonate (MMTS) labeling to detect free thiols (Figure 1g). Unmodified TherA exhibited the expected mass increase upon MMTS treatment, whereas TherKC‐modified TherA showed no detectable labeling, consistent with cysteine consumption during thioether ring formation.

To gain preliminary insight into the nature of the resulting thioether crosslinks, we performed β‐mercaptoethanol (βME) trapping experiments (Figure 1h), a method previously used to probe accessible Dha and Dhb residues in lanthipeptides (Hernandez Garcia & Nair, 2023). Because βME undergoes Michael‐type addition with accessible Dha and Dhb residues, it provides a convenient means to assess which dehydrated residues remain uncyclized following peptide maturation. Since TherKC can catalyzes up to eight dehydration events, a maximum of eight βME additions would theoretically be possible if all Dha/Dhb residues remained accessible. However, MALDI‐TOF MS analysis of βME‐treated modified TherA revealed a ladder of species corresponding to a maximum of six βME additions (Figure 1h). Thus, at least two dehydrated residues are no longer available for βME trapping and are likely consumed during cyclization. Because formation of a labionin linkage requires two Dha/Dhb residues, whereas a simple lanthionine linkage consumes only one, the observed βME‐trapping pattern is consistent with the formation of at least one labionin‐containing thioether crosslink. Together with the conserved STxSNxC motif characteristic of Class III lanthipeptides, these data support the formation of a labionin‐containing TherA product, although the precise ring connectivity remains to be experimentally determined. Together, these results confirm that TherKC efficiently modifies the conserved STxSNxC motif within the TherA core peptide, in agreement with the Class III‐c substrate signature identified by sequence analysis.

2.3. Cryo‐EM analysis reveals a cyclase‐mediated homodimer unique to Class III‐c LanKC enzymes

To elucidate the structural basis of the newly identified Class III‐c LanKC subgroup, we determined the structure of the TherKC–TherA complex purified from co‐expression by single‐particle cryo‐electron microscopy. Purified TherKC was incubated with substrate peptide TherA, Mg²⁺, and the non‐hydrolyzable nucleotide analogue ATPγS prior to grid preparation. Cryo‐EM analysis yielded a reconstruction at an overall resolution of 3.48 Å, revealing a homodimeric assembly of TherKC (Figures 2a, S4, and S5). The density map clearly resolves two TherKC protomers arranged in a symmetric dimer with overall dimensions of approximately 125 Å in length and 80 Å in width (Figure 2a). Only one protomer is observed in a ligand‐bound state, with well‐defined density for both TherA and ATPγS, whereas the second protomer lacks bound substrate and nucleotide. This asymmetry likely reflects differences in substrate occupancy rather than compositional heterogeneity, suggesting that the unliganded protomer adopts a more dynamic or inactive conformation that is less amenable to high‐resolution reconstruction (Figure S6). Each TherKC protomer adopts the characteristic cashew‐shaped architecture of Class III LanKC enzymes (Hernandez Garcia & Nair, 2023; Kobayashi et al., ²⁰²⁴; Li et al., ²⁰²⁴; Li et al., ²⁰²⁵), comprising an N‐terminal lyase domain, a central kinase domain, and a C‐terminal cyclase domain (Figure 2b). The lyase and cyclase domains are positioned at opposite ends of the molecule and are connected by the kinase domain, which forms a wide internal cavity. TherA binds within a groove formed between the lyase and kinase domains, while ATPγS is coordinated within the kinase domain. Importantly, the catalytic sites of all three domains face toward the central cavity, consistent with the established catalytic topology of LanKC enzymes. Critically, analysis of the dimer interface revealed a previously unobserved mode of LanKC oligomerization. In contrast to all reported LanKC dimers (Li et al., 2024, 2025), the TherKC homodimer is formed exclusively through head‐to‐head interactions between the cyclase domains of the two protomers, with no contribution from the lyase or kinase domains (Figure 2a). This architecture is fundamentally distinct from previously characterized Class III‐b LanKC dimers, such as PneKC and SalKC (Li et al., 2024, 2025), in which dimerization is mediated by interactions between the cyclase domain of one protomer and the lyase domain of the other (Figure S6). Moreover, Class III‐a LanKC enzymes have been reported to function as monomers (Hernandez Garcia & Nair, 2023). Together, these comparisons establish that Class III‐c LanKC enzymes employ a unique cyclase‐driven dimerization strategy not observed in any other LanKC subgroup.

FIGURE 2.

Cryo‐EM structure of the TherKC–TherA complex reveals a cyclase‐mediated dimeric assembly. (a) Top and side views of the cryo‐EM density map of the TherKC–TherA complex. Two TherKC protomers form a homodimer through contacts between their cyclase domains. One protomer is substrate‐ and nucleotide‐bound (gray), while the second protomer is ligand‐free (orange). Density corresponding to the bound substrate peptide TherA (green) and ATPγS (red) is observed only in one protomer, highlighting an asymmetric dimer. Domain boundaries of lyase, kinase, and cyclase regions are indicated. (b) Structural organization of a single TherKC protomer. The enzyme adopts a compact, curved architecture in which the lyase (yellow), kinase (pink), and cyclase (blue) domains surround a central cavity. The leader peptide of TherA binds within a groove formed between the lyase and kinase domains, while ATPγS occupies the minor groove in the kinase domain. The N‐terminal loop extends beneath the cyclase domain, and the catalytic centers of all three domains are oriented toward the central cavity.

2.4. Arg685 defines a cyclase‐domain dimerization hotspot essential for TherKC assembly and stability

To define the molecular basis of this unique dimerization mode, we examined the cyclase–cyclase interface in detail. The TherKC dimer is stabilized by two symmetric helical interaction surfaces, primarily involving the α8C helix from each cyclase domain (Figure 3a,b). Interface analysis using PISA (Krissinel & Henrick, 2007) revealed a substantial buried surface area of approximately 980.8 Ų (Table S1). At the center of the interface, Arg685 plays a pivotal anchoring role. The guanidinium group of Arg685 from one protomer forms bidentate salt bridges with Glu749 and Glu750 from the opposing protomer, creating a strong electrostatic clamp (Figure 3b). Surrounding this core interaction, Tyr688 and Phe689 engage in π–π stacking interactions with Phe747 from the opposing protomer, while Arg690 forms a hydrogen bond with Asn700, further reinforcing the interface. Together, these interactions create a tightly interlocked cyclase–cyclase assembly. Consistent with this structural organization, PISA analysis further showed that Arg685 contributes one of the largest residue‐level buried surface areas within the interface and participates in the most extensive electrostatic interaction network among the identified interface residues (Table S1). In contrast, neighboring residues such as Phe689 and Arg690 make smaller individual contributions, suggesting that Arg685 may functions as the primary dimerization hotspot coordinating the cyclase–cyclase assembly.

FIGURE 3.

Identification of Arg685 as a hotspot of TherKC dimerization. (a) Overall structure of the TherKC homodimer, showing two protomers assembled through symmetric interfaces formed by their cyclase domains. The N‐ and C‐termini and the interface‐forming α8C and α10C helices are indicated. (b) Close‐up view of the cyclase–cyclase dimerization interface, which is stabilized by a network of interactions including a salt bridge between Arg685 from one protomer and Glu749/Glu750 from the opposing protomer, π–π stacking interactions involving Tyr688 and Phe689 with Phe747, and a hydrogen bond between Arg690 and Asn700 across the interface. (c) Sequence alignment of residues at the dimerization interface in TherKC and representative Class III LanKC enzymes, showing high conservation of the Arg685‐centered dimerization motif within the Class III‐c subgroup. (d) SEC analysis of wild‐type TherKC and interface mutants shows that substitution of Arg685 with alanine (R685A) shifts the elution profile from a dimeric to a monomeric species, whereas mutations at neighboring interface residues have smaller effects. SEC analysis of the Class III‐c homolog SoyaKC also indicates a predominantly dimeric assembly. (e) Representative cryo‐EM 2D class averages showing predominantly dimeric particles for WT TherKC and SoyaKC, whereas the R685A mutant is primarily monomeric.

Sequence alignment of interface residues across representative Class III LanKC enzymes revealed that Arg685 and its surrounding interaction network are highly conserved within Class III‐c, but are absent or poorly conserved in Class III‐a, III‐b, and III‐d enzymes (Figure 3c). This conservation pattern strongly supports the idea that cyclase‐mediated dimerization is a defining and subclass‐specific feature of most Class III‐c LanKC enzymes. To experimentally assess whether this dimeric assembly is conserved beyond TherKC, we additionally characterized a second representative Class III‐c enzyme, SoyaKC. Similar to TherKC, co‐expressed SoyaKC–SoyaA purified as a stable complex and eluted as a single dominant SEC peak corresponding to an apparent molecular mass of ~200 kDa, consistent with a homodimeric assembly (Figures 3d and S7A). Cryo‐EM analysis further revealed predominantly dimeric particles in both raw micrographs and 2D class averages, closely resembling the overall assembly observed for TherKC (Figures 3e and S7B,C). Together, these observations provide experimental support that cyclase‐mediated dimerization is likely a conserved structural feature shared among multiple Class III‐c LanKC enzymes rather than a unique property of TherKC alone.

To directly test the functional importance of these interactions, we generated a series of point mutants targeting key interface residues. SEC demonstrated that mutation of Arg685 to alanine (R685A) completely abolished dimer formation, yielding a monomeric species, whereas mutations of neighboring residues (e.g., F689A, R690A) had more modest effects (Figure 3d). Two‐dimensional cryo‐EM classification further confirmed that wild‐type TherKC predominantly forms dimers, while the R685A mutant exists almost exclusively as monomers (Figure 3e). These results identify Arg685 as the dominant dimerization hotspot in TherKC.

2.5. Dimerization enhances catalytic efficiency and thermal stability in Class III‐c LanKC enzymes

In previously characterized Class III‐b LanKC enzymes, such as SalKC (Li et al., 2025) and PneKC (Li et al., 2024), dimerization is mediated by extensive interactions between the lyase domain of one protomer and the cyclase domain of the other. In those systems, substrate binding to one protomer induces conformational changes that are transmitted across the dimer interface, resulting in functional asymmetry and suppression of activity in the opposing protomer (Li et al., 2024). In contrast, the cyclase–cyclase–only dimerization architecture observed for Class III‐c TherKC lacks direct lyase or kinase domain contacts, raising the question of whether dimerization in this subgroup serves an allosteric regulatory role or instead fulfills a distinct function. To directly probe the functional consequences of dimer disruption, we determined the structure and activity of the TherKC^R685A monomer mutant, in which the principal cyclase‐domain dimerization hotspot is eliminated. Cryo‐EM analysis of TherKC^R685A yielded a reconstruction at 4.25 Å resolution, revealing a monomeric architecture with well‐resolved density for the bound substrate LP (Figures 4a and S8). The presence of clear LP density demonstrates that dimerization is not required for substrate engagement. Structural superposition of the monomeric TherKC^R685A with the ligand‐bound protomer of wild‐type TherKC showed a high degree of similarity, with no discernible rearrangements in the lyase, kinase, or cyclase active sites (Figure 4b). This observation indicates that disruption of the cyclase–cyclase interface does not perturb the intrinsic fold of the enzyme or its substrate‐binding configuration.

FIGURE 4.

Structural and functional characterization of the monomeric TherKC^R685A mutant. (a) Cryo‐EM reconstruction of the TherKC^R685A–TherA complex reveals clear density for the bound leader peptide, indicating that the monomeric mutant retains substrate‐binding capability. (b) Structural superposition of a wild‐type TherKC protomer and the TherKC^R685A monomer shows close overall agreement (RMSD = 0.686 Å), demonstrating that disruption of dimerization does not alter the global fold or substrate‐binding configuration. (c) ATPase activity of wild‐type TherKC and the R685A mutant measured in the absence and presence of TherA (enzyme: Substrate ratio 1:1), showing reduced but detectable substrate‐stimulated activity for the monomeric mutant. Data are shown as mean ± SD (n = 3); ns, not significant; *p< 0.05; **p< 0.01; ***p< 0.001. (d) Intact‐mass analysis of MMTS‐treated TherKC^R685A reaction products. Samples were treated with MMTS prior to mass spectrometric analysis. The absence of an MMTS‐derived MSH adduct indicates that the monomeric mutant retains the ability to consume the cysteine thiol during substrate modification, consistent with preservation of cyclization activity despite disruption of the dimer interface. (e) Differential scanning fluorimetry (DSF) reveals a reduced melting temperature for the R685A mutant relative to wild‐type TherKC, indicating that cyclase‐mediated dimerization contributes to thermal stability. (f) Temperature dependence of ATPase activity. Activities are normalized to the maximum activity observed for each protein, showing a lower optimal temperature for the R685A mutant relative to wild‐type TherKC.

To further assess local map quality, we compared the cryo‐EM densities of the dimeric wild‐type TherKC and monomeric TherKC^R685A at identical contour levels (Figure S9). In the WT dimer, one protomer displays weaker density than the other, while both cyclase domains remain well resolved with clear side‐chain densities visible for most residues. In contrast, although the overall architecture of the R685A monomer is well defined, the cyclase‐domain density is noticeably less resolved, and side‐chain features are largely missing. By comparison, the kinase and lyase domains exhibit similar density quality in both structures, with densities primarily sufficient for tracing the polypeptide backbone but containing limited side‐chain information. These observations suggest that cyclase‐mediated dimerization preferentially stabilizes the cyclase domain, resulting in improved local structural order, whereas the kinase and lyase domains are comparatively unaffected by disruption of the dimer interface.

We next examined whether loss of dimerization affects enzymatic activity. In nucleotide hydrolysis assays performed at a 1:1 molar ratio of enzyme to TherA, the TherKC^R685A mutant exhibited significantly reduced ATPase activity compared with wild‐type TherKC, both in the absence and presence of substrate (Figure 4c). Nevertheless, measurable ATP hydrolysis was still observed upon substrate addition, indicating that individual TherKC protomers remain catalytically competent. Consistent with this conclusion, intact‐mass analysis of MMTS‐treated reaction products confirmed that TherKC^R685A retains the ability to modify TherA in vitro. The absence of an MMTS‐derived MSH adduct indicates consumption of the cysteine thiol during substrate processing, demonstrating that cyclization activity is retained despite disruption of the dimer interface. However, substrate modification was less efficient than that observed for the dimeric wild‐type enzyme (Figure 4d). Interestingly, whereas wild‐type TherKC produced a continuous distribution of partially dehydrated intermediates, the monomeric R685A mutant predominantly accumulated species corresponding to 1‐, 3‐, 5‐, and 8‐fold dehydration. Although the mechanistic basis of this pattern remains unclear, it suggests that disruption of the cyclase‐mediated dimer interface alters the efficiency or coordination of substrate processing. Given that the dimer places two complete catalytic assemblies in close proximity, dimerization may facilitate more efficient progression through multiple dehydration events, potentially by enhancing substrate processivity or increasing access of the flexible core peptide to catalytic domains within the dimeric assembly. The ability of R685A to still generate the fully dehydrated product indicates that dimerization is not strictly required for catalysis, but may instead promotes efficient and coordinated substrate modification.

Finally, we assessed the functional consequence of dimer disruption by measuring thermal stability. Differential scanning fluorescence revealed that wild‐type TherKC exhibits a melting temperature of approximately 42°C, whereas the monomeric R685A mutant unfolds at a significantly lower temperature (~36°C) (Figure 4e). This reduction in thermal stability indicates that cyclase‐mediated dimerization substantially stabilizes TherKC, likely contributing to its enzymatic robustness, particularly under the thermophilic growth conditions of Thermoactinomyces. Consistent with this interpretation, temperature‐dependent ATPase assays further showed that wild‐type TherKC retained substantially higher activity at elevated temperatures, with maximal activity near 40°C, whereas the monomeric R685A mutant exhibited a lower optimal temperature (~35°C) together with a pronounced loss of activity at higher temperatures (Figure 4f). Together, these results demonstrate that cyclase‐mediated dimerization in Class III‐c LanKC enzymes is not strictly required for substrate recognition or basal catalytic competence under standard conditions, but instead plays an important role in enhancing catalytic robustness, thermal stability, and efficient enzymatic activity under elevated temperature conditions. Importantly, the absence of strong inter‐protomer communication differs from Class III‐b systems and suggests that Class III‐c LanKC protomers function largely independently, even within a dimeric assembly.

2.6. Leader peptide recognition by TherKC is mediated by a conserved hydrophobic groove between the lyase and kinase domains

Recognition of the leader peptide is the initiating step in lanthipeptide biosynthesis and plays a central role in substrate specificity. Previous studies of Class III LanKC enzymes have shown that the leader peptide adopts an amphipathic α‐helical conformation that docks into a hydrophobic surface formed between the lyase and kinase domains, despite the absence of a canonical RiPP recognition element (RRE) (Burkhart et al., 2015; Huang et al., ²⁰²²; Kobayashi et al., ²⁰²⁴; Ortega et al., ²⁰¹⁴; Wiebach et al., ²⁰²⁰). The cryo‐EM structure of the TherKC–TherA complex is fully consistent with this model. Clear density corresponding to the LP of TherA is observed bound to TherKC, adopting a helical conformation (Figure 5a). Based on sequence conservation and density quality, residues Val4‐Leu10, encompassing the conserved VLxLxxL motif, were modeled into the density. The remainder of the LP and the downstream core peptide exhibit weaker density, likely reflecting conformational flexibility during active processing. The LP helix inserts into a groove formed at the interface between the lyase and kinase domains, where it is oriented approximately perpendicular to the antiparallel β‐sheet of the kinase N‐lobe (Figure 5b). This binding mode closely resembles that reported for ThurKC (Hernandez Garcia & Nair, 2023) and PneKC (Li et al., 2024), indicating conservation of the overall recognition strategy across Class III LanKC enzymes. At the residue level, LP binding is stabilized predominantly through hydrophobic interactions, supplemented by specific hydrogen bonds. Val4 of the LP packs against Tyr148 in the lyase domain, while Leu10 is stabilized by hydrophobic contact with Tyr242 in the kinase domain. Additional polar interactions include hydrogen bonding between Asp6 of the LP and Asn222 of the lyase domain, as well as interactions involving Gln8 and Gln12 with Tyr194 (Figure 5b). Notably, Phe235 in the kinase domain disrupts the LP helical conformation, promoting a transition toward a more flexible coil. This residue occupies a position analogous to Phe236 in ThurKC and is thought to facilitate delivery of the core peptide into the central catalytic cavity (Hernandez Garcia & Nair, 2023). Surface analysis confirms that both the LP helix and its binding groove are strongly hydrophobic, underscoring the importance of shape and chemical complementarity in substrate recognition (Figure 5c). Alanine substitution of key groove‐lining residues, including Tyr194, Tyr242, Trp301, and Phe235, resulted in significant reductions in ATPase activity, confirming their functional importance in LP engagement (Figure 5d). Consistent with the endpoint ATPase measurements (Figure 5d), time‐course assays showed that mutations at the LP‐binding interface reduce the overall level of substrate‐stimulated ATP hydrolysis in the presence of TherA (Figure 5e).

FIGURE 5.

Leader peptide recognition by TherKC is mediated by a conserved hydrophobic groove between the lyase and kinase domains. (a) Cryo‐EM structure of the TherKC–TherA complex showing that the LP of TherA adopts an α‐helical conformation upon binding. Different domains and the N‐ and C‐termini are indicated. (b) Close‐up view of the LP bound within a hydrophobic groove formed at the interface between the lyase and kinase domains, with interacting residues indicated. (c) Surface representation colored by hydrophobicity highlights that LP binding is dominated by hydrophobic interactions. (d) Effects of alanine substitutions of groove‐lining residues on ATPase activity of TherKC, indicating their functional importance in LP recognition; data are shown as mean ± SD (n = 3); ns, not significant; *p< 0.05; **p< 0.01; ***p< 0.001. (e) Time‐course measurements of ATP hydrolysis by wild‐type TherKC and LP‐binding mutants in the presence of TherA (1:1 molar ratio) show reduced maximal activity for interface mutants relative to wild type, consistent with impaired substrate engagement. (f) Sequence alignment of representative Class III LanKC enzymes shows that residues forming the LP‐binding groove are broadly conserved within the Class III‐c subgroup and are variably conserved across other Class III subclasses.

Sequence comparison across Class III LanKC subgroups revealed that many LP‐interacting residues are conserved within Class III‐c enzymes, supporting a shared recognition mechanism within this subgroup (Figure 5f). Unlike certain Class IV systems, such as CurL23, in which the leader peptide wraps around the kinase base, TherKC lacks such extended LP interactions and does not employ a dedicated RRE, indicating that Class III‐c LanKC enzymes rely on direct LP–enzyme contacts for substrate discrimination.

2.7. A conserved catalytic framework within the structurally distinct Class III‐c LanKC

To define the catalytic basis of the newly identified Class III‐c LanKC enzymes, we next combined sequence comparison, structural analysis, and functional validation to identify residues essential for catalysis in TherKC. Although Class III‐c enzymes adopt a distinct quaternary organization and cyclase‐mediated dimerization mode, it remained unclear whether their catalytic machinery diverges from that of previously characterized Class III systems.

We first compared the overall architecture of TherKC with representative tri‐functional LanKC enzymes from other Class III subclasses (Figure S10). Global structural alignment revealed that TherKC is most similar to the Class III‐a enzyme ThurKC, with an overall RMSD of 2.44 Å, whereas substantially larger deviations were observed when compared with Class III‐b enzymes (PneKC and SalKC; RMSD >5.5 Å) and the kinase domain of Class III‐d enzyme CurKC (RMSD 4.31 Å) (Figure S10). Domain‐wise comparisons further clarified this relationship. The kinase domains of TherKC and ThurKC overlay closely (RMSD 0.926 Å), as do those of TherKC and PneKC (RMSD 1.218 Å), and the lyase domains likewise show high structural similarity across subclasses (Figure S11). These observations indicate that, despite divergence in higher‐order assembly and cyclase organization, the core catalytic scaffold of the kinase and lyase domains is strongly conserved in Class III‐c enzymes.

Consistent with these structural observations, sequence alignment of the kinase domains revealed broad conservation of residues involved in nucleotide binding and phosphoryl transfer (Figure 6a). The ATP‐binding cleft remains positioned between the N‐ and C‐lobes of the kinase domain, and the surrounding catalytic motifs align closely with those in previously characterized Class III enzymes. Subclass‐specific differences were nevertheless apparent. For example, Arg234 is conserved in Class III‐c and Class III‐b but absent from Class III‐a, whereas the DFE motif is conserved in Class III‐c and Class III‐a but diverges in Class III‐b. These variations do not alter the overall fold of the kinase domain but may reflect evolutionary tuning of nucleotide coupling or substrate handling among subclasses.

FIGURE 6.

Conserved catalytic framework of the Class III‐c LanKC enzyme TherKC. (a) Sequence alignment of the TherKC kinase domain with kinase domains from structurally characterized Class III LanKC enzymes shows strong conservation of residues involved in nucleotide binding and phosphoryl transfer. (b) Sequence alignment of lyase domains reveals broad conservation of residues implicated in phosphate elimination during dehydration. (c) Sequence alignment of the cyclase domain demonstrates that residues previously identified as essential for catalysis in the Class III‐a enzyme ThurKC are conserved in TherKC and across the Class III‐c subgroup. (d) Structural alignment of the cyclase domains of TherKC and ThurKC reveals a conserved barrel‐like architecture composed of 14 α‐helices; comparison highlights the absence of canonical Zn²⁺‐binding motifs and the presence of a shared non‐canonical active‐site configuration. (e) Alanine substitution of conserved cyclase residues significantly reduces substrate‐stimulated ATPase activity of TherKC, indicating their importance for efficient pathway coupling; data are shown as mean ± SD (n = 3); ns, not significant; *p< 0.05; **p< 0.01; ***p< 0.001. (f) LC–MS/MS analysis of the conserved STVSNNCK motif in TherA modified by cyclase‐domain mutants. Reaction products were treated with IAA prior to digestion and LC–MS/MS analysis to monitor the presence of free cysteine thiols through carbamidomethylation. The D666A, R609A, and D770A mutants retain a carbamidomethylated cysteine and show little or no modification within the motif, indicating defects in both dehydration and cyclization. In contrast, the F769A mutant lacks carbamidomethylated cysteine but retains a Dha residue corresponding to the predicted cyclization partner of the cysteine, consistent with formation of a partially modified intermediate. These distinct modification patterns suggest that R609, D666, and D770 are required for productive maturation of the motif, whereas F769 may contribute to a later stage of cyclization following the initial cysteine addition reaction.

Analysis of the lyase domain similarly revealed a high degree of conservation. The lyase domain of TherKC retains the defining features of tri‐functional LanKC enzymes, including the extended N‐terminal loop, a six‐stranded antiparallel β‐sheet core, and a surrounding helical bundle (Figure S8). The putative catalytic pocket is formed by a cluster of conserved positively charged residues, including His61, Lys65, Lys94, Lys117, and Arg163, which generate an electropositive environment well suited for engaging phosphorylated Ser/Thr intermediates generated by the kinase domain (Figure 6b). Additional conserved residues, including Lys117, Tyr122, and Asp152, are positioned near this pocket and have been implicated in phosphate elimination in other Class III systems, supporting a conserved dehydration pathway in Class III‐c LanKC enzymes.

We next focused on the cyclase domain, where Class III‐c enzymes show the greatest sequence divergence and where TherKC additionally supports dimerization. Sequence alignment across Class III‐c homologs revealed a set of highly conserved residues, including Arg609, Asp666, Ser725, Phe769, and Asp770, that cluster around the predicted cyclase active site and overlap with residues previously shown to be essential for cyclization in Class III‐a ThurKC (Figure 6c). Structural alignment of the TherKC and ThurKC cyclase domains revealed a conserved barrel‐like architecture composed of 14 α‐helices arranged as repeating helix‐turn‐helix units (Figure 6d). Thus, despite differences in quaternary assembly, the overall cyclase fold is preserved.

Importantly, comparison with canonical Zn²⁺‐dependent lanthipeptide cyclases highlights a defining feature of Class III enzymes. In classical cyclases, a conserved Cys‐His‐Cys motif coordinates Zn²⁺ and plays a central catalytic role24. In TherKC, as in ThurKC, these positions are instead occupied by Ser725, Phe769, and Asp770, forming a noncanonical cyclase motif (Figure 6d). This configuration preserves the architecture of the catalytic cavity while replacing the classical metal‐binding triad, indicating that Class III‐c cyclases operate through a conserved yet distinct chemical framework.

To validate the functional importance of these conserved residues, we generated alanine substitutions and assessed their effects on catalysis. Mutations at all five positions significantly reduced substrate‐stimulated ATPase activity, with the most severe defects observed for R609A, D666A, and F769A (Figure 6e). While ATPase measurements reflect overall pathway coupling, they do not distinguish defects in dehydration from failures in cyclization. We therefore analyzed the products of these mutants by fragmentation mass spectrometry.

To directly assess modification of the predicted labionin‐forming region, we focused on the conserved STVSNNCK motif within the TherA core peptide (Figures 6f, S12 and Table S2). This analysis simultaneously reports on both dehydration and cyclization. Dehydration of Ser/Thr residues can be detected by the appearance of Dha or Dhb, whereas cyclization was monitored by treatment with iodoacetamide (IAA), which selectively labels free cysteine thiols to generate a carbamidomethylated cysteine adduct. Thus, retention of IAA labeling indicates that the cysteine residue remains uncyclized, whereas loss of IAA labeling is consistent with cysteine consumption during thioether ring formation. The LC–MS/MS analyses (Figure 6f) were performed on peptides derived from in vitro modification reactions using purified TherA and purified TherKC variants. Following the reaction, TherA was isolated by SDS–PAGE, subjected to in‐gel digestion, and analyzed by LC–MS/MS. MS analysis revealed clear mechanistic distinctions among the mutants (Figures 6f and S12). In the R609A, D666A, and D770A mutants, the STVSNNCK motif remained largely unmodified, and the cysteine residue retained IAA labeling, indicating that both dehydration and cyclization within this motif were severely impaired. These observations identify the conserved R609–D666–D770 network as a critical catalytic element required for productive modification of the labionin‐forming region. Interestingly, a similar phenomenon was reported for the Class III‐a enzyme ThurKC, where mutation of corresponding cyclase‐domain residues unexpectedly affected dehydration in addition to cyclization. As dehydration is formally catalyzed by the kinase and lyase domains, these findings suggest that modification of the labionin‐forming motif may proceed through tightly coordinated interdomain processes rather than fully independent catalytic steps.

The F769A mutant exhibited a distinct phenotype (Figures 6f and S12). In this mutant, a Dha residue corresponding to the predicted cyclization partner of cysteine within the STVSNNCK motif remained detectable, indicating that dehydration of the motif could still occur. In addition, the cysteine residue was no longer detected as IAA‐labeled, demonstrating loss of the free thiol group and indicating that the cysteine had participated in a cyclization reaction. Notably, the remaining Dha corresponds to the predicted cyclization partner of cysteine in the mature product, making this intermediate consistent with formation of an initial lanthionine‐like linkage while preventing completion of the fully matured labionin scaffold. This phenotype differs obviously from those of the R609A, D666A, and D770A mutants, in which both dehydration and cyclization were impaired. The coexistence of a residual Dha residue and the absence of IAA‐reactive cysteine suggests that modification has progressed beyond the earliest stages of maturation. One possible interpretation is that an initial lanthionine‐like cyclization event can still occur in F769A, whereas subsequent reactions required for formation of the final labionin scaffold are impaired. Although the precise sequence of bond‐forming events cannot be assigned unambiguously from the current data, the accumulation of this partially modified intermediate suggests that F769 contributes to a later stage of productive cyclization. Together, these observations support a catalytic model in which the conserved R609–D666 ion pair helps organize the cyclase active site, while D770 functions as a key catalytic residue required for productive cyclization (Figure S13). Similar to the mechanism proposed for the Class III‐a enzyme ThurKC, cyclization is likely initiated by activation of the cysteine thiol followed by Michael‐type addition to a dehydrated residue (Figure S13). Interestingly, mutation of cyclase‐domain residues affects dehydration within the motif despite dehydration being formally catalyzed by the kinase and lyase domains, mirroring observations previously reported for ThurKC. In contrast, the unique phenotype of F769A suggests that modification can proceed beyond the earliest dehydration and cyclization steps but fails to generate the fully matured product. These findings are consistent with a model in which formation of the labionin scaffold occurs through sequential reactions and in which F769 contributes to productive completion of the cyclization process residue (Figure S13). While the detailed chemical mechanism remains unresolved, the overall catalytic logic appears highly conserved between Class III‐c TherKC and Class III‐a ThurKC despite their markedly different oligomeric assemblies. Notably, the partially modified intermediate observed for F769A was not resolved in previous studies of ThurKC and may therefore provide additional insight into the sequential nature of labionin formation.

3. CONCLUSION AND DISCUSSION

A key outcome of this work is the establishment of a four‐subclass framework for Class III LanKC enzymes that is supported by multiple independent lines of evidence. Phylogenetic separation, low inter‐subclass sequence identity, and divergence concentrated in the cyclase domain together indicate that Class III‐c enzymes form a coherent and distinct group rather than a marginal extension of previously defined subclasses. The structural analysis of TherKC reveals that Class III‐c enzymes adopt a mode of dimerization that has not been observed in other LanKC systems. In contrast to Class III‐b LanKC enzymes, where dimerization involves extensive cross‐protomer interactions between lyase and cyclase domains and is associated with functional asymmetry, TherKC dimerization is mediated exclusively by cyclase–cyclase contacts. The absence of lyase or kinase involvement in this interface is supported by both structural analysis and mutational validation. These findings establish cyclase‐mediated homodimerization as a defining architectural feature of Class III‐c LanKC enzymes.

Functional characterization of the dimer interface provides insight into its role. Disruption of dimerization through mutation of Arg685 abolishes dimer formation and reduces thermal stability, yet does not prevent substrate binding or abolish catalytic activity. However, the monomeric R685A mutant consistently exhibits reduced ATPase activity, lower modification efficiency, and impaired activity at elevated temperatures. Notably, MMTS‐labeling experiments indicate that the monomeric enzyme retains the ability to consume the substrate cysteine during modification, suggesting that cyclization can still occur in the absence of dimerization. Thus, dimerization primarily enhances catalytic robustness and stability rather than serving as an absolute requirement for catalysis. This behavior differs from that reported for Class III‐b systems, where dimerization has been shown to influence activity through inter‐protomer coupling. Together, these observations suggest that different Class III subclasses employ distinct oligomeric strategies, even while carrying out the same overall chemical transformations.

Despite these differences in higher‐order assembly, the catalytic framework of Class III‐c LanKC enzymes remains strongly conserved. Structural comparisons show that the kinase and lyase domains of TherKC closely resemble those of previously characterized Class III enzymes, and sequence analysis confirms conservation of residues involved in nucleotide binding, phosphoryl transfer, and phosphate elimination. These findings are consistent with earlier biochemical studies demonstrating that dehydration chemistry is mechanistically conserved across Class III LanKC systems (Asamizu, 2025; Hegemann & Süssmuth, ²⁰²⁰; Lagedroste et al., ²⁰²⁰; Xie & van der Donk, ²⁰⁰⁴).

The cyclase domain, although structurally divergent and responsible for dimerization in Class III‐c enzymes, also retains key catalytic features (Hernandez Garcia & Nair, 2023; Zhang et al., ²⁰¹²). The barrel‐like fold observed in TherKC closely matches that of the Class III‐a enzyme ThurKC, and conserved residues identified by sequence alignment are positioned around the predicted active site. Notably, the non‐canonical cyclase motif that replaces the classical Zn²⁺‐binding triad is preserved, consistent with previous reports that Class III LanKC enzymes employ a Zn²⁺‐independent cyclization mechanism. Mutational and mass spectrometric analyses further demonstrate that individual conserved residues within this motif play distinct roles in coordinating dehydration and cyclization steps. In particular, mutation of R609, D666, or D770 blocks both dehydration and cyclization within the conserved STVSNNCK motif, whereas F769A produces a distinct partially modified intermediate in which cysteine consumption occurs despite incomplete maturation of the motif. These findings suggest that cyclization of the labionin‐forming region proceeds through tightly coordinated reaction steps and identify F769 as a residue potentially involved in a later stage of productive cyclization.

Taken together, these results indicate that Class III‐c LanKC enzymes combine a distinct structural organization with a relatively conserved catalytic mechanism. Rather than representing a fundamentally different mode of lanthipeptide biosynthesis, Class III‐c enzymes appear to implement established chemical steps within a modified architectural context. This finding aligns with broader observations in RiPP biosynthetic enzymes, where variation in domain arrangement, oligomerization, or substrate engagement often coexists with conservation of core catalytic chemistry.

In summary, this study defines Class III‐c LanKC enzymes as a distinct subgroup within the Class III family and provides the first detailed structural and functional analysis of a representative enzyme. By integrating classification, cryo‐EM structure determination, biochemical assays, and mutational analysis, we show that Class III‐c enzymes employ a unique cyclase‐mediated dimerization mode that enhances stability while preserving a conserved three‐domain catalytic architecture. These findings refine our understanding of LanKC diversity and provide a foundation for future studies aimed at exploring additional Class III‐c systems and their associated lanthipeptide products.

4. METHODS

4.1. Sequence analysis and phylogenetic reconstruction

Class III LanKC amino acid sequences were retrieved from the NCBI non‐redundant protein database using previously characterized Class III LanKC enzymes as query sequences (Schoch et al., 2020). Redundant and truncated sequences were removed manually. The resulting dataset was aligned using ClustaW implemented in MEGA 11 with default parameters (Kumari & Sidhu, 2024). Phylogenetic reconstruction was performed using the maximum‐likelihood method based on the Jones–Taylor–Thornton (JTT) substitution model, with 1000 bootstrap replicates to assess branch support (Little, 2024). Phylogenetic trees were visualized and annotated in MEGA 11.

Multiple sequence alignments used for motif and residue conservation analysis were generated using ClustalW and rendered with ESPript 3.0, with secondary‐structure elements overlaid where structural information was available (Robert & Gouet, 2014).

4.2. Cloning and site‐directed mutagenesis

The coding sequences of TherKC (WP_093291005.1) and TherA (WP_218122178.1) were amplified by PCR from genomic DNA of Thermoactinomyces sp. DSM 45892 using gene‐specific primers (Table S3). And the coding sequences of SoyaKC (WP_131847794.1) and SoyaA (WP_207896784.1) were amplified by PCR from genomic DNA of Baia_soyae using gene‐specific primers. PCR products were cloned into the pQLinkN vector (Scheich et al., 2007), which introduces an N‐terminal hexahistidine tag for affinity purification. Constructs encoding TherKC and TherA individually were generated as pQLinkN‐TherKC and pQLinkN‐TherA, respectively.

For co‐expression, the two pQLinkN plasmids were assembled using the pQLink multigene expression system to generate a single plasmid encoding both His‐tagged TherKC and TherA (Yang et al., 2023). Site‐directed mutagenesis of TherKC was performed using PCR‐based mutagenesis with mutagenic primers (Table S3). Following PCR, template plasmids were digested with DpnI at 37°C for 1 h. All constructs were transformed into Escherichia coli DH5α for plasmid propagation. Correct insertion and all mutations were verified by Sanger DNA sequencing.

4.3. Protein expression and purification of TherKC and the TherKC–TherA complex

Expression plasmids were transformed into E. coli BL21 (DE3). Cells were grown in LB medium supplemented with appropriate antibiotics at 37°C until OD₆₀₀ reached ~0.8. Protein expression was induced by addition of 0.5 mM IPTG, followed by incubation at 18°C for 18 h.

Cells were harvested by centrifugation and resuspended in lysis buffer containing 50 mM Tris–HCl (pH 7.5), 300 mM NaCl, and 5 mM imidazole. Cells were lysed using a high‐pressure homogenizer, and insoluble material was removed by centrifugation at 20,000 × g for 30 min at 4°C.

The clarified lysate was incubated with Co‐NTA affinity resin (Thermo Fisher Scientific) at 4°C for 1 h with gentle mixing. The resin was washed extensively with lysis buffer, and bound proteins were eluted using buffer containing 250 mM imidazole. Eluted proteins were concentrated using a 100 kDa molecular‐weight cutoff centrifugal concentrator and further purified by size‐exclusion chromatography on a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated in 25 mM Tris–HCl (pH 7.5) and 150 mM NaCl. Protein purity was assessed by SDS–PAGE, and concentrations were determined by absorbance at 280 nm using a NanoDrop spectrophotometer.

For cryo‐EM studies, all structures reported in this work were determined using the co‐expressed TherKC–TherA complex purified directly from E. coli. Co‐expression was employed to preserve the native enzyme–substrate interaction and facilitate structural characterization of the complex. In contrast, for all biochemical assays, including ATPase measurements, peptide modification assays, and mass spectrometry analyses, TherKC and TherA were expressed and purified separately before being mixed and incubated in vitro. This experimental design ensured that enzyme concentration, substrate concentration, reaction time, and buffer conditions could be precisely controlled and kept consistent across experiments, thereby enabling direct comparison and reproducible quantification of enzymatic activity and peptide modification.

4.4. Expression and purification of TherA from inclusion bodies

For preparation of unmodified TherA, the TherA‐pQLinkN construct was expressed in E. coli BL21 (DE3). Cultures were grown at 37°C to OD₆₀₀ ~0.8 and induced with 1 mM IPTG for 4 h. Cells were harvested and lysed as described above. The resulting inclusion body pellet was collected by centrifugation at 20,000 × g.

Inclusion bodies were solubilized in denaturation buffer containing 8 M urea in PBS (pH 8.0) and clarified by centrifugation. The supernatant was incubated with Co‐NTA resin at 4°C overnight. The resin was washed with urea‐containing buffer and eluted with 250 mM imidazole in denaturation buffer. Eluted protein was refolded by stepwise dialysis against refolding buffer (25 mM Tris–HCl, pH 7.5, 150 mM NaCl), with buffer exchanges every 12 h. Precipitated material was removed by centrifugation, and soluble TherA was concentrated and quantified for downstream assays.

4.5. Biochemical characterization of TherKC

Nucleotide hydrolysis activity of TherKC was measured using a malachite green phosphate assay, as described previously (Xu et al., 2023). Reactions (20 μL) contained 1.5 μM TherKC, 1 mM NTP (ATP, GTP, CTP, or UTP), and 10 mM MgCl₂ in the presence or absence of 1.5 μM TherA. To establish assay conditions for endpoint measurements, preliminary experiments were performed to evaluate the dependence of phosphate production on substrate concentration and reaction time. ATP hydrolysis increased linearly with increasing TherA concentration from 0.5 to 3.5 μM and remained approximately linear during the first 100 min of the reaction under the assay conditions employed (Figure S2C,D). Based on these results, all subsequent ATPase assays were performed using 1.5 μM TherKC and 1.5 μM TherA with an incubation time of 1 h to ensure measurements were obtained within the linear range of the assay.

Reactions were incubated at room temperature for 1 h and quenched by addition of 80 μL malachite green–molybdate reagent. After 2 min, 10 μL of 34% (w/v) sodium citrate was added to stop color development. Absorbance was measured at 650 nm. The optimal temperature was determined by assaying activity between 20 and 60°C at pH 8.0. Released inorganic phosphate was quantified using a standard curve generated with Na₂HPO₄. Kinetic data were analyzed using GraphPad Prism, and Michaelis–Menten parameters were obtained by nonlinear regression.

4.6. Differential scanning fluorimetry (DSF)

Thermal stability of TherKC and mutants was assessed by DSF using a Bio‐Rad CFX384 Touch real‐time PCR system, following established protocols (Niesen et al., 2007). Each reaction (10 μL) contained 10 μM protein and 10× SYPRO Orange dye (Sigma‐Aldrich) in buffer comprising 25 mM Tris–HCl (pH 7.5) and 150 mM NaCl.

Fluorescence was monitored using the SYBR Green channel while temperature was increased from 25 to 95°C at 1°C min⁻¹. Melting curves were analyzed using Bio‐Rad CFX Manager 3.0, and melting temperatures (T _m) were determined from the first derivative of the fluorescence signal (dF/dT).

4.7. Cryo‐EM sample preparation

For structural analysis of the dimeric TherKC–TherA complex, purified TherKC–TherA was incubated with 1 mM ATPγS and 1 mM MgCl₂ for 20 min at room temperature prior to grid preparation. Protein samples were adjusted to a final concentration of ~1.5 mg mL⁻¹. A 3.5 μL aliquot was applied to glow‐discharged Quantifoil R1.2/1.3400‐mesh copper grids, blotted for 3.5 s at 100% humidity, and plunge‐frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific).

For the monomeric TherKC mutant (R685A) dataset, the same procedure was followed using purified R685A TherKC co‐expressed with TherA, under identical buffer and nucleotide conditions, ensuring direct comparability between datasets.

4.8. Cryo‐EM data acquisition

Cryo‐EM data for both datasets were collected at liquid nitrogen temperature on a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV, equipped with a Gatan K3 Summit direct electron detector and a GIF Quantum energy filter with a 20 eV slit width. Movies were recorded in counting mode using SerialEM (Eisenstein et al., 2024).

Data were acquired at a nominal magnification of 81,000×, corresponding to a calibrated pixel size of 0.858 Å at the specimen level. The dose rate was maintained at ~7.6 electrons pixel⁻¹ s⁻¹, with a total exposure time of 6 s, resulting in a cumulative dose of ~33.5 e⁻ Å⁻², fractionated into 32 frames (150 ms per frame). Detailed acquisition parameters are summarized in Table S4.

4.9. Cryo‐EM image processing

A total of 3619 movies were collected for the wild‐type TherKC–TherA complex. Image processing was performed using cryoSPARC (Punjani et al., 2017). Movies were subjected to patch motion correction, followed by patch CTF estimation (Zhang et al., 2024), with manual inspection and rejection of low‐quality micrographs, yielding 3514 micrographs for further analysis. Initial particle picking was performed using blob‐based picking, resulting in ~3.75 million particles, which were subsequently subjected to multiple rounds of 2D classification to remove noise and contaminants. A subset of 377,076 particles was selected for ab initio 3D reconstruction into four initial classes. These classes were refined through heterogeneous refinement under C1 symmetry, followed by homogeneous refinement and non‐uniform (NU) refinement (Punjani et al., 2020). The final reconstruction converged to a dimeric TherKC–TherA structure at an overall resolution of 3.48 Å, as determined by the gold‐standard Fourier shell correlation (FSC = 0.143) criterion (Figure S4D). Angular distribution analysis indicated broad particle orientation coverage (Figure S4C), and local resolution estimation revealed well‐resolved core regions with reduced resolution in peripheral and flexible regions (Figure S4E).

For the R685A monomer mutant, 2379 movies were collected and processed in parallel using the same cryoSPARC workflow. After motion correction and CTF estimation, 2279 micrographs were retained for analysis. Blob‐based particle picking yielded ~4.34 million particles, which were subjected to iterative 2D classification. A subset of 618,961 particles was selected for ab initio reconstruction, followed by heterogeneous refinement and non‐uniform refinement under C1 symmetry. The final map corresponds to a monomeric TherKC–TherA complex, resolved to an overall resolution of 4.25 Å based on the FSC = 0.143 criterion (Figure S8D). Angular distribution and local resolution analyses are shown in Figure S8C,E, respectively.

4.10. Model building and optimization

An initial atomic model of TherKC was generated using AlphaFold3 (Abramson et al., 2024). The predicted model was rigid‐body fitted into the cryo‐EM density using UCSF Chimera followed by iterative manual adjustment in COOT (Bond et al., 2020). Real‐space refinement was carried out using PHENIX (Poon et al., 2024), with secondary structure restraints and geometry optimization applied throughout refinement (Singharoy et al., 2015).

Although ATPγS and TherA were not explicitly included in the AlphaFold prediction, clear corresponding densities were observed in the cryo‐EM maps. These ligands were manually built into the density based on shape complementarity and known interaction geometries, followed by additional rounds of real‐space refinement in PHENIX. Model quality was assessed using MolProbity (Williams et al., 2017), Ramachandran statistics (Park et al., 2023) and clash score analysis within PHENIX (Headd et al., 2012). The final models showed good stereochemistry and agreement with the experimental maps.

To further validate the observed quaternary organization, AlphaFold3 was also used to predict the TherKC–TherA complex. The predicted assembly closely matched the experimentally determined cryo‐EM structures, including the dimeric arrangement of wild‐type TherKC and the leader peptide α‐helical binding mode, providing independent support for the structural interpretation.

4.11. LC‐TOF‐MS characterization for distinguishing modified and unmodified TherA

To evaluate the enzyme's modification activity on the substrate peptide, we used S‐methylmethanethiosulfonate (MMTS) cross‐linking to assess whether TherA was modified (Chen et al., 2018). Specifically, the Cys in the unmodified TherA is in an exposed state and can be cross‐linked with MMTS, while the modified TherA cannot be cross‐linked. The increase in mass number detected by MS analysis can determine whether the substrate peptide is modified.

For the MMTS cross‐linking experiment, we first incubated TherKC (1.5 μM) and TherA (1.5 μM) with the same molar concentration with 1 mM ATP, 10 mM MgCl₂ for 30 min, then added 5 mM TCEP for 30 min at 37°C, and then added 10 mM MMTS to each sample and incubated at room temperature for another 1 h for loading. Alkylated and non‐alkylated samples of TherA alone, and alkylated and non‐alkylated samples of TherKC and TherA complex were examined. Each sample was reacted once without replication (n = 1). The mass spectrometer used an Acquity Premier BEH C4 protein analysis column (1.7 μm, 2.1 mm × 50 mm), and the mass spectrometry system was water: acetonitrile (ACN): formic acid (FA), and the solvent system consisted of 0.1% FA in water (solvent A) and 0.1% FA in ACN (solvent B). After injecting 10 μL of sample, solvent B was used to elute and separate at a flow rate of 0.3 mL/mL within 8 min, and then detected by SCIEX TripleTOF 5600 equipped with DuoSpray ion source, and TOF mass spectra were collected from 400 to 2500 m/z m/z range, with an accumulation time of 1 s. Deconvolution was performed using the Biotool Kit in SCIEX OS Analytics, with an output range of 2000–120,000 Da.

4.12. Chemical modification of the Dha/Dhb residues

β‐Mercaptoethanol (βME) trapping of dehydroalanine (Dha) and dehydrobutyrine (Dhb) residues was performed as previously described with minor modifications (Meyer et al., 1994). Briefly, 1.5 μM TherA was first incubated with 1.5 μM TherKC in the presence of ATP and Mg²⁺ under standard modification conditions to generate dehydrated and cyclized products. Following completion of the modification reaction, the modified peptide was isolated by SDS‐PAGE and recovered for subsequent βME treatment at a final concentration of 10 μM in 50 mM Tris–HCl buffer (pH 8.5). βME was then added to a final concentration of 5 mM, and the reaction mixture was incubated at 30°C for 2 h.

Under alkaline conditions, βME undergoes Michael‐type addition to accessible Dha and Dhb residues, generating stable thioether adducts with a characteristic mass increase of approximately 78 Da per modification. Dha/Dhb residues that have already participated in lanthionine or labionin formation are no longer accessible for βME addition and therefore do not exhibit the corresponding mass shift. After incubation, reaction products were directly analyzed by MALDI‐TOF mass spectrometry. The number of βME additions was determined from the observed mass increases relative to the untreated modified peptide, providing an estimate of the number of accessible Dha/Dhb residues remaining after peptide maturation.

4.13. In‐gel protein extraction for MS analysis

For analysis of TherKC‐catalyzed peptide modifications, purified TherKC (1.5 μM) was incubated with purified TherA (1.5 μM) in the presence of ATP and Mg²⁺ under the standard modification conditions described above. Following the reaction, samples were separated by SDS–PAGE and visualized by Coomassie Blue staining. The TherA band was excised to separate the substrate peptide from TherKC and other reaction components prior to downstream LC–MS/MS analysis. First, 50% ACN/50 mM NH₄HCO₃ solution was added to the cut gel for washing, repeated 2–3 times, and then the gel was dehydrated with 100% ACN, and then 10 mM DTT/50 mM NH₄HCO₃ was added, and incubated at 56°C for 45 min for reduction. After discarding the solution, 55 mM iodoacetamide (IAA)/50 mM NH₄HCO₃ was added, and incubated in the dark for 30 min for alkylation. Finally, the dried gel block was soaked in Chymotrypsin digestion solution and placed at 37°C for enzymatic digestion for 12–16 h, and then centrifuged to collect the supernatant to obtain the peptide solution.

4.14. Nano‐HPLC‐MS/MS analysis

The enzymatically digested protein samples were injected into the Nano‐HPLC system (Jia et al., 2023). The separation column used was Thermo Scientific Acclaim PepMap C18, 100 μm × 2 cm, mobile phase A was 0.1% FA in H₂O, mobile phase B was 0.1% FA in ACN (acetonitrile), the flow rate was set to 10ul/min, for 3 min, and then a linear gradient elution was performed in the analytical column with mobile phase B from 5 to 35 for 60 min, the flow rate of the chromatographic column was maintained at 300 nL/min, and the column temperature was maintained at 60°C. Subsequently, the mass spectrometer inlet was equipped with an electrospray voltage of 2.3 kV, and the Orbitrap Exploris 480 mass spectrometer automatically switched between MS and MS/MS acquisition, and the full scan MS spectrum (m/z 350–1600) was acquired and measured at a mass resolution of 60,000, and then the top 20 strongest ions were selected for MS/MS fragmentation (HCD mode), and continuous energy collisions were performed at a resolution of 15,000, and dynamic exclusion lasted for 30 s. The RAW files were compared to the Swissprot target protein database using Proteome Discoverer software (Thermo Fisher Scientific, version 2.4). We specified a maximum of two allowed deletion cleavage sites. Oxidation of methionine, dehydration of serine and threonine were specified as potential modifications, and the whole process ensured that the false discovery rate (FDR) of proteins and peptides remained below 1%. Figure S12 and Table S2 in the Supplementary Information present the raw mass spectra and detailed mass analysis of the fragment ions.

AUTHOR CONTRIBUTIONS

Haojie Zhang: Investigation; methodology; writing – original draft; visualization. Jian Shi: Data curation. Min Luo: Conceptualization; supervision; project administration; writing – review and editing; formal analysis; resources; funding acquisition. Bochu Wang: Formal analysis; validation. Yifan Li: Investigation. Yi Shi: Investigation; validation.

FUNDING INFORMATION

This work was supported by the National University of Singapore (start‐up grant) and the Ministry of Education, Singapore, through Tier 1 (A‐8002655‐00‐00) and Tier 2 (MOE‐T2EP30222‐0015, T2EP30223‐0029, T2EP30123‐0017) grants.

CONFLICT OF INTEREST STATEMENT

The authors declare no competing financial interest.

Supporting information

Figure S1. Sequence alignment of class III cyclase domains reveals multiple conserved motifs in class III‐c LanKC enzymes. Sequence alignment of the cyclase domains in Class III modifying enzymes. The sequence identity between class III‐c and class III‐a is 17.64%, the identity between class III‐c and class III‐b is 7.90%, and the identity between class III‐c and class III‐d is 12.50%.

Figure S2. Genomic context and expression of the Class III‐c lanthipeptide TherA from Thermoactinomyces sp. (A) Organization of the putative lanthipeptide biosynthetic gene cluster encoding the Class III‐c LanKC enzyme TherKC, the precursor peptide TherA, an ABC transporter system, and a downstream M16B family peptidase likely involved in leader peptide processing. (B) SDS‐PAGE analysis of purified His‐tagged TherA, showing a band corresponding to the expected molecular weight (~7.9 kDa). (C) Dependence of ATP hydrolysis activity on TherA concentration. ATP hydrolysis increased linearly with increasing TherA concentration over the tested range (0.5–3.5 μM) in the presence of 1.5 μM TherKC. (D) Time‐course analysis of ATP hydrolysis. Reactions containing 1.5 μM TherKC and 1.5 μM TherA showed an approximately linear increase in phosphate production during the first 100 min. Data are presented as mean ± SD from independent replicates.

Figure S3. Chromatographic and mass spectrometry information of substrate peptides.

Figure S4. Single‐particle Cryo‐EM analysis of TherKC in complex with TherA. (A) Representative Cryo‐EM images showing several particles. (B) Image processing flow chart with final map showing the main conformations of the TherKC‐TherA complex at 3.48 Å. (C) Angular distribution of Cryo‐EM particles included in the final 3D reconstruction. (D) Fourier shell correlation (FSC) curve: gold standard FSC between two half‐maps, indicating resolution FSC = 0.143. (E) Full Cryo‐EM map and slices of surface representation, filtered to estimated global resolution and colored according to local resolution.

Figure S5. Fitting of cryo‐EM images of different domains of TherKC and TherKC bound to TherA. Density display of the cyclase domain and the kinase domain that binds TherA to nucleotides.

Figure S6. Assembly comparison of TherKC with other members in Class III lanKC. PneKC and SalKC in Class III‐b are homodimers of TherKC in Class III‐c. However, the dimer of Class III‐b is formed by the interaction of the cyclase domain of one protomer with the lyase domain of another protomer, while TherKC is a dimer formed by the cyclase domains of two protomers. In comparison, the class III‐a enzyme ThurKC is observed as a monomer in the crystal structure, indicating that it likely functions in a monomeric state.

Figure S7. Characterization of class III‐c SoyaKC. (A) SDS‐PAGE gel image of co‐expressed SoyaKC‐SoyaA complex. (B) Representative Cryo‐EM images showing several particles. (C) Representative 2D class averages from cryo‐EM analysis show predominantly dimeric particles for wild‐type SoyaKC.

Figure S8. Cryo‐EM workflow of TherKC monomer mutant (R685A) complexed with TherA. (A) Representative Cryo‐EM images showing several particles. (B) Image processing flow chart with final map showing the main conformations of the TherKC‐TherA complex at 4.25 Å. (C) Angular distribution of Cryo‐EM particles included in the final 3D reconstruction. (D) Fourier shell correlation (FSC) curve: gold standard FSC between two half‐maps, indicating resolution FSC = 0.143. (E) Full Cryo‐EM map and slices of surface representation, filtered to estimated global resolution and colored according to local resolution.

Figure S9. Local resolution estimates of TherKC structures. The cryo‐EM maps of (A) wt TherKC dimer and (B) TherKCR685A monomer. Compared to the monomer, the cyclization domain within the dimer exhibits significantly higher resolution and better side‐chain density.

Figure S10. Protomer‐level structural comparison of the Class III‐c enzyme TherKC with representative Class III LanKC enzymes. The overall RMSD between TherKC (this study) and ThurKC (PDB: 8SAM) protomers is 2.44 Å, showing some similarity, while the overall RMSD compared with the structure of PneKC (PDB: 8WGO) and SalKC (PDB: 9UGQ) are 5.90 Å and 5.57 Å, and compared with the structure of the kinase domain of CurKC (PDB: 7WJ6), the RMSD was 4.31 Å, showing obvious structural differences.

Figure S11. Kinase‐domain structural comparison of the Class III‐c enzyme TherKC with other Class III LanKC enzymes. (A) Structure overlay of the kinase domains of TherKC and ThurKC (overall RMSD: 0.926 Å). (B) Structure overlay of the kinase domains of TherKC and PneKC (overall RMSD: 1.218 Å). (C) Comparison of the lyase domain structures of Class III‐c enzyme TherKC and Class III‐a enzyme ThurKC (overall RMSD: 0.780 Å). (D) Comparison of the lyase domain structures of Class III‐c enzyme TherKC and Class III‐b enzyme PneKC (overall RMSD: 1.018 Å).

Figure S12. The raw MS spectrum of precursor ions for MS/MS analysis.

Figure S13. Proposed catalytic model for cyclization by TherKC. The conserved R609–D666 ion pair is proposed to organize the cyclase active site, while D770 may assist cysteine thiol activation for Michael‐type addition to a dehydrated residue, forming a lanthionine‐like intermediate. The F769A mutant produces a distinct partially modified product, suggesting that F769 may contribute to a later stage of cyclization, potentially by positioning the intermediate for completion of the labionin scaffold. The final bond‐forming step remains unresolved and is therefore shown as a proposed mechanism.

Table S1. Summarize the contribution of the PISA interface and the residual horizontal buried surface area (BSA) and change in solvation free energy upon interface formation (Δ ⁱ G).

Table S2. Summary of theoretical and observed m/z values of diagnostic b‐ and y‐type fragment ions identified by LC–MS/MS analysis of IAA‐labeled TherA products generated 106 by wild‐type and mutant TherKC enzymes.

Table S3. List of oligonucleotides used in this study.

Table S4. Cryo‐EM data collection, refinement and validation statistics.

Zhang H, Li Y, Shi Y, Shi J, Wang B, Luo M. Catalytic architecture and cyclase‐mediated dimerization of a newly identified class III‐c lanthipeptide synthetase. Protein Science. 2026;35(9):e70776. 10.1002/pro.70776

DATA AVAILABILITY STATEMENT

Cryo‐EM density maps for the TherA‐bound TherKC dimer and the TherKC (R685A) mutant monomer have been deposited in the Electron Microscopy Data under accession codes EMD‐66012 (homodimer) and EMD‐66127 (mutant monomer), respectively. The atomic coordinates for these structures have been deposited in the Protein Data Bank under accession codes 9WJ0 and 9WOP, respectively. The data that support the findings of this study are available from the corresponding author upon reasonable request.