Abstract

   6-thioguanine (6-TG) is a therapeutic medication for childhood acute
   lymphoblastic leukemia (ALL) and a potent antimicrobial agent. Its
   biosynthesis relies on the YcfA-YcfC system, yet the formation of its
   critical thioamide moiety remains incompletely understood. Here, we
   provide a detailed biochemical and structural characterization of YcfA,
   including apo and substrate-bound crystal structures, which reveal that
   substrate adenylation and L-cysteine addition are key initial steps in
   the reaction cascade. Cryo-electron microscopy (cryo-EM) and functional
   analyses highlight YcfA’s assembly into a two-layered heptameric
   structure, essential for the enzymatic function. GTP serves a dual role
   as a substrate and oligomerization enhancer. Additionally, pyridoxal
   5’-phosphate (PLP), a cofactor for YcfC, the partner enzyme in this
   system, promotes YcfA oligomerization but inhibits its activity by
   obstructing GTP binding. Biochemical and structural evidence confirms
   that YcfC acts as a C‒S lyase, which is essential for thioamide
   formation in the presence of PLP. Exploiting substrate flexibility, we
   synthesized a seleno analog with antimicrobial properties. Multi-omics
   analyses of the biosynthetic precursor underscore its potential as an
   antibiotic. Collectively, our findings unravel the distinct
   architecture and functionality of the YcfA-YcfC system, offering an
   evolutionary perspective on noncanonical thioamide biosynthesis and a
   foundation for synthetic biology applications in drug development.

   Subject terms: Biocatalysis, X-ray crystallography, Enzyme mechanisms
     __________________________________________________________________

   Here, the authors structurally and biochemically characterise the
   YcfA-YcfC enzyme system, crucial for biosynthesis of the anticancer and
   antimicrobial agent 6-thioguanine.

Introduction

   Thioamides, characterized by the substitution of sulfur for oxygen in
   the amide group (C(=S)NH), are critical components of numerous
   metabolites, exhibiting antimicrobial and anticancer
   properties^[44]1–[45]3. Their diverse bioactivities make them
   compelling targets for drug discovery. Prominent examples include
   methanobactin, a metal-chelating compound^[46]4–[47]6, and
   closthioamide, an antibiotic agent^[48]7,[49]8. Among these,
   6-thioguanine (6-TG) stands out as a clinically utilized antimetabolite
   for treating acute lymphoblastic leukemia (ALL)^[50]9. Beyond oncology,
   6-TG demonstrates antimicrobial activity against Staphylococcus
   aureus^[51]10 and antiviral effects against
   coronaviruses^[52]11,[53]12. Intriguingly, it also functions as a
   virulence factor in Erwinia amylovora, the causative agent of fire
   blight in rosaceous plants^[54]13,[55]14.

   The biosynthesis of the thioamide group in 6-TG, catalyzed by the
   YcfA-YcfC system^[56]13,[57]15, was initially thought to follow the
   canonical pathways for thioamide formation in nucleobases^[58]16. These
   involve alpha adenine nucleotide hydrolase (AANH), such as ThiI, MnmA,
   TtuA, TtcA, or Ncs6, and a cysteine desulfurase. Together, they
   facilitate the transfer of sulfur from L-cysteine or cystine via a
   persulfide intermediate, often mediated by persulfide carrier proteins
   like TusA, TusBCD, or TusE. The reactive persulfide intermediate then
   reacts with an ATP-adenylated nucleobase, to form the thioamide group,
   requiring multiple steps and complex machinery.

   Emerging evidence suggests that the YcfA‒YcfC system employs a more
   streamlined mechanism^[59]17. Unlike typical AANH enzymes, YcfA
   bypasses the persulfide intermediate, directly utilizing L-cysteine to
   produce a cysteine–guanine nucleotide adduct. YcfC, functioning as a
   C‒S lyase, then cleaves this adduct to generate the thioamide group in
   6-TG. This process eliminates the need for mobile persulfide
   intermediates, presenting a biochemical strategy for thioamidation.

   Despite these advances, key questions remain: (1) How does YcfA
   specifically recognize and utilize L-cysteine for nucleotide
   modification? (2) What are the molecular mechanisms underlying YcfC’s
   interaction with the cysteine-conjugate intermediate? (3) How do YcfA
   and YcfC coordinate their activities to ensure efficient biosynthesis?
   Addressing these questions is crucial for advancing our understanding
   of thioamide biosynthesis and its applications. This research opens
   avenues for the rational design of bioactive compounds, enabling the
   engineering of 6-TG analogs for diverse therapeutic applications, such
   as cancer treatment, antiviral therapies, and antimicrobial
   development.

   In this work, we determine a series of crystal structures of YcfA and
   YcfC, complemented by molecular docking and biochemical assays.
   Together with the cryo-EM analysis, we reveal YcfA’s two-layered
   heptameric architecture, which is essential for its catalytic activity.
   We also examine the regulatory interplay within the YcfA-YcfC system,
   discovering that YcfA oligomerization is promoted by its nucleotide
   substrates and the YcfC cofactor pyridoxal 5’-phosphate (PLP).
   Interestingly, PLP exhibits dual effects, enhancing oligomerization
   while inhibiting enzymatic activity, suggesting a sophisticated
   feedback mechanism. Furthermore, we synthesize a seleno analog,
   6-seleno-GTP, which exhibits potent antimicrobial activity against
   multidrug-resistant S. aureus. Multi-omics analyses further elucidate
   its antibacterial mechanism. These findings offer insights into this
   biosynthetic pathway and its potential applications in drug discovery.

Results

In vitro reconstitution of the YcfA-YcfC system

   To elucidate the catalytic mechanism of the YcfA-YcfC system in
   synthesizing 6-thioguanine (6-TG) nucleotides, we reconstituted the
   system in vitro using various substrates, including sulfur and selenium
   sources—L-cysteine (L-Cys) and L-selenocysteine (L-SeCys)—and guanine
   nucleotides (GTP, GDP, and GMP). Consistent with previous
   studies^[60]15,[61]17, combining YcfA, YcfC, ATP, Mg²⁺, and L-Cys with
   each guanine nucleotide yielded products with absorption peaks at
   340 nm (Fig. [62]1a; Supplementary Fig. [63]1a). High-performance
   liquid chromatography (HPLC) and high-resolution mass spectrometry
   (HRMS) identified these products as 6-thio-GTP, 6-thio-GDP, and
   6-thio-GMP (Fig. [64]1a; Supplementary Fig. [65]1b, c). With more
   6-thio-GTP produced, GTP was selected as the optimal substrate for
   further studies. Substituting L-Cys with L-SeCys resulted in
   corresponding 6-selenoguanine nucleotides with a redshifted absorption
   peak at 360 nm (Fig. [66]1b; Supplementary Fig. [67]1d).

Fig. 1. In vitro reconstitution of the YcfA-YcfC system.

   [68]Fig. 1
   [69]Open in a new tab

   a UV–Vis spectra detection of the thioamide formation catalyzed by the
   YcfA-YcfC system using L-Cys and GTP as substrates. A characteristic
   absorption peak at 340 nm was observed. The molecular weight (as
   determined by MS) and structure of the product are shown. b UV–Vis
   spectra detection of the selenoamide formation catalyzed by the
   YcfA-YcfC system using L-SeCys and GTP as substrates. The resulting
   product exhibits an absorption peak at 360 nm. The molecular weight (as
   determined by MS) and structure of the product are shown. c, d
   Continuous monitoring of the reaction process of the YcfA-YcfC system.
   c Incubation of YcfA with ATP, GTP, and L-Cys results in increased
   absorbance at 315 nm over 200 s. Subsequent addition of YcfC leads to a
   decrease in absorbance at 315 nm, concomitant with the emergence of a
   new peak at 340 nm. d YcfC alone showed no activity with ATP, GTP, and
   L-Cys. Addition of YcfA produced simultaneous absorption peaks at
   315 nm and 340 nm. e The formation of the 315 nm intermediate requires
   the adenylation activity of YcfA. f Adenylation activity assay of YcfA
   with diverse amino acids and 3MP. Only L-Cys and 3MP yielded the 315 nm
   peak. g Proposed catalytic steps of the YcfA-YcfC system in
   6-Thioguanine nucleotides formation, however, the detailed regulatory
   and catalytic mechanisms of the two enzymes remain elusive. Source data
   are provided as a Source Data file.

   To delineate the individual roles of YcfA and YcfC, we monitored the
   reaction kinetics using high-frequency UV‒visible absorption scans.
   Sequential enzyme addition experiments revealed that YcfA catalyzes the
   initial formation of a 315 nm intermediate, which YcfC subsequently
   converts into the final product, 6-thio-GTP (340 nm) (Fig. [70]1c).
   Starting with YcfC produced no detectable product; however, adding YcfA
   afterward led to the sequential formation of both the 315 nm
   intermediate and the 340 nm final product (Fig. [71]1d). These results
   underscore YcfA’s role in initiating the reaction and YcfC’s role in
   processing the intermediate.

   Further analysis revealed that ATP, GTP, and L-Cys are essential for
   forming the 315 nm intermediate (Fig. [72]1e). As an AANH-like
   enzyme^[73]15, YcfA likely catalyzes the adenylation of GTP, activating
   its amide carbonyl group. The resulting adenylated GTP, though unstable
   and undetectable via UV‒visible absorption^[74]18, reacts with the
   thiol group of L-Cys to form the 315 nm intermediate. Testing
   alternative amino acids confirmed the critical role of the thiol group,
   as only L-Cys produced the 315 nm peak (Fig. [75]1f). Similarly,
   replacing L-Cys with mercaptopyruvate (3MP), which lacks only an amine
   group, also yielded a product absorbing at 315 nm (Fig. [76]1f). These
   findings suggest that the thiol group of L-Cys interacts with the
   activated amide carbonyl of guanine in adenylated GTP, forming a
   6-Cys-GTP S-adduct. Tandem mass spectrometry confirmed this adduct
   (Supplementary Fig. [77]1e, f).

   YcfA’s dual catalytic roles are thus evident: it catalyzes the
   adenylation of guanine nucleotides and facilitates the formation of
   6-Cys-guanine or 6-SeCys-guanine nucleotide S-adducts. YcfC, acting as
   a C–S or C–Se lyase, cleaves these intermediates to produce
   6-thio-guanine or 6-seleno-guanine nucleotides, as depicted in
   Fig. [78]1g. However, the inability to isolate and independently
   measure the adenylation step precluded precise differentiation of
   YcfA’s contributions to each step. These results align with recent
   studies highlighting YcfA’s dual catalytic functions and YcfC’s
   specific role as a C–S lyase^[79]17.

The dual roles of YcfA in adenylation and L-Cys adduct formation

   To elucidate the dual catalytic roles of YcfA, we determined the
   crystal structures of apo-YcfA, YcfA in complexes with ATP (YcfA–ATP)
   or GTP (YcfA–GTP), and an inactive variant, YcfA^D19A complexed with
   ATP and GTP (YcfA^D19A–ATP–GTP) (Supplementary Table [80]1).
   Unexpectedly, YcfA assembles into a two-layered heptameric ring,
   a distinct arrangement in this enzyme family (Fig. [81]2a). Structural
   analyses revealed that catalytic cavities are located at the interface
   of adjacent monomers across the two heptameric layers, forming a
   catalytic dimer unit (Fig. [82]2b). Notably, a single monomer lacks a
   complete cavity for guanine substrates, highlighting the
   interdependence of monomers for substrate binding.

Fig. 2. The dual roles of YcfA in adenylation and L-Cys adduct formation.

   [83]Fig. 2
   [84]Open in a new tab

   a A two-layered heptameric architecture of YcfA, with each monomer
   distinctly colored. b The catalytic cavities of adjacent monomers
   across the two heptameric layers. c ATP binding mode in the YcfA-ATP
   complex structure. Interaction networks formed between ATP and
   surrounding residues of YcfA, with hydrogen bonds represented as blue
   sticks. d Adenylation activity assay of YcfA mutants targeting ATP
   binding. The WT YcfA is regarded as possessing 100% enzyme activity.
   The mean is displayed ±SD for n = 3 biological replicates. e GTP
   binding mode in the YcfA-GTP complex structure. The polar interactions
   formed between GTP and surrounding residues are represented as blue
   sticks, and the π-π stacking interaction is shown as yellow stick. f
   Adenylation activity assay of YcfA mutants targeting GTP binding. The
   WT YcfA is regarded as possessing 100% enzyme activity. The mean is
   displayed ±SD for n = 3 biological replicates. g Surface representation
   of YcfA^D19A-ATP-GTP complex. ATP and GTP are shown as sticks, Mg^2+ is
   shown as green sphere. h The cooperative binding of ATP and GTP in the
   complex structure of YcfA^D19A-ATP-GTP. Interaction networks in the
   active site involving ATP, GTP, Mg^2+, and surrounding residues are
   presented, with polar contacts represented as blue sticks. i Docking
   simulations of YcfA with adenylated-GTP and L-Cys. The surrounding
   residues of L-Cys binding are presented, with potential polar contacts
   represented as blue sticks. j Mutagenesis analysis of the L-Cys binding
   site shows YcfA’s relative activity in forming the 6-Cys-GTP S-adduct.
   The WT YcfA is regarded as possessing 100% enzyme activity. The mean is
   displayed ±SD for n = 3 biological replicates. k Proposed catalytic
   mechanism of YcfA. Key residues involved in adenylation and L-Cys
   nucleophilic addition are indicated. Source data are provided as a
   Source Data file.

   In the YcfA–ATP complex, ATP is deeply embedded in the catalytic cavity
   and forms critical hydrogen bonds with Asp19, Gln129, and Glu183,
   stabilizing the transition state during ATP cleavage (Fig. [85]2c;
   Supplementary Fig. [86]2a). Mutations at any of these residues
   abolished 6-Cys-GTP S-adduct production, underscoring their essential
   roles in adenylation. Pyrophosphate is the leaving group in this
   reaction, facilitated via these residues by promoting the
   phosphodiester bond cleavage (Fig. [87]2d).

   In the YcfA–GTP structure, GTP binds near the catalytic cavity
   entrance, with its guanine ring stabilized by forming interactions with
   Phe132 and Glu134 (Fig. [88]2e; Supplementary Fig. [89]2b). Glu134
   serves as a general base, deprotonating the guanine ring to activate
   the amide carbonyl for adenylation (Fig. [90]2f). The triphosphate tail
   of GTP is anchored by Arg18, Lys169, and Lys180 from the adjacent
   monomer. Mutation of Arg18 significantly reduced S-adduct formation,
   suggesting its dual roles in GTP binding and facilitating subsequent
   steps in catalysis (Fig. [91]2f).

   The YcfA^D19A–ATP–GTP complex revealed the interactions between ATP and
   GTP, with Gln135 aligning the substrates by interacting with the
   α-phosphate of ATP and the guanine base of GTP (Fig. [92]2g, h).
   Mutation of Gln135 reduced S-adduct production (Fig. [93]2f). In this
   inactive mutant, Glu183 coordinates with all three phosphate groups of
   ATP via Mg²⁺, but the absence of interaction with Asp19 and Gln129
   prevents the stabilization of ATP transition state, underscoring the
   cooperative roles of these residues in adenylation (Fig. [94]2c, h).

   Following adenylation, YcfA catalyzes the incorporation of L-Cys into
   adenylated GTP. Substituting L-Cys with inorganic sulfides failed to
   produce detectable thiolated GTP (Supplementary Fig. [95]3),
   highlighting YcfA’s specificity for L-Cys. Structural data from the
   YcfA^D19A–ATP–GTP complex suggest that L-Cys binding is sterically
   hindered until pyrophosphate release creates sufficient space.
   Molecular docking simulations revealed that adenylated GTP undergoes a
   conformational change upon pyrophosphate release, enabling L-Cys
   binding (Fig. [96]2h, i). The carboxylate group of L-Cys is anchored by
   Arg18, while Tyr128 and Gln129 stabilize its orientation (Fig. [97]2i).
   Substitution of Tyr128 with alanine abolished S-adduct formation,
   whereas phenylalanine substitution had minimal effect, underscoring the
   aromatic ring’s role in L-Cys positioning. Gln129 likely facilitates
   thiol nucleophilic attack by deprotonating L-Cys (Fig. [98]2j) and is
   unable to catalyze this step with L-serine (Fig. [99]1f), further
   confirming substrate specificity.

   YcfA’s dual catalytic steps, adenylation and L-Cys adduct formation,
   are mediated by distinct interdependent residues. Asp19, Gln129, and
   Glu183 stabilize the ATP transition state, while Glu134 deprotonates
   the guanine ring to drive adenylation (Fig. [100]2k). Pyrophosphate
   release creates space for L-Cys accommodation, where Gln129 facilitate
   nucleophilic addition to adenylated GTP. Disruption of any key residue
   compromises one or both catalytic steps, highlighting the intricate
   coordination required for YcfA’s enzymatic function. These findings
   provide mechanistic insights into YcfA’s dual roles and broader
   implications for this enzyme family.

Tetradecamer assembly is essential for YcfA activity

   Beyond its dual catalytic functions, YcfA adopts a rare tetradecamer
   arrangement, comprising a two-layered heptameric architecture as
   revealed by crystal structures and cryo-EM (Figs. [101]2a, [102]3a;
   Supplementary Fig. [103]4; Supplementary Table [104]2). This
   physiologically relevant structure could not be predicted by
   computational methods like AlphaFold3^[105]19 without experimental
   confirmation. Retrospective size-exclusion chromatography (SEC) data
   revealed that increasing protein concentration decreased the apparent
   elution volume (Fig. [106]3b), suggesting concentration-dependent
   oligomerization.

Fig. 3. Nucleotide substrates drive functional YcfA assembly.

   [107]Fig. 3
   [108]Open in a new tab

   a Cryo-EM density map of YcfA with the fitted atomic model. Side view
   (top) shows the two layers, and the top view (bottom) reveals the
   heptameric arrangement in each layer, forming a two-layered heptameric
   (tetradecameric) assembly. b SEC analysis of YcfA at varying
   concentrations (0.5 mg/mL, 1 mg/mL, 2 mg/mL and 4 mg/mL). Inset
   presents a calibration curve for SEC using proteins of known sizes. c
   SEC analysis of YcfA^L173D and YcfA^D166R/Y170N. d The adenylation
   activity assays of YcfA^L173D and YcfA^D166R/Y170N using ATP, GTP, and
   L-Cys as substrates. The WT YcfA is regarded as possessing 100% enzyme
   activity. The mean is displayed ± SD for n = 3 biological replicates. e
   The effects of different substrates (Mg^2+, L-Cys, ATP, and GTP) in
   promoting YcfA oligomerization were analyzed by SEC. Both ATP and GTP
   can promote the assembly of YcfA. f SEC analysis of YcfA with GMP, GDP,
   and GTP showed that GDP had a weaker effect than GTP on
   oligomerization, whereas GMP had a negligible effect. Source data are
   provided as a Source Data file.

   To probe the functional significance of the tetradecamer, we introduced
   mutations at the oligomeric interfaces based on structural analysis
   (Supplementary Fig. [109]5a–f). The K180A mutation promoted
   polymerization but severely reduced enzymatic activity (Supplementary
   Fig. [110]5e, f). As Lys180 is critical for guanine substrate binding
   at the interface, this mutation likely disrupts optimal substrate
   alignment (Fig. [111]2e, f). To mitigate substrate-binding confounders,
   we further introduced the L173D mutation, disrupting a hydrophobic
   interface region. This resulted in a stable dimer (confirmed by
   analytical ultracentrifugation, AUC) but without enzymatic activity
   (Fig. [112]3c, d; Supplementary Fig. [113]5g–i). Conversely, the
   D166R/Y170N double mutant enhanced interfacial stability through
   synergistic modifications: The Y170N substitution was designed to
   introduce a more favorable hydrogen-bonding network with Asp190 and
   potentially introduce a new polar interaction with Arg78. This change
   was expected to increase the number of contacts between the two layers
   of YcfA. As for Asp166, while it is located near Interface III, it is
   positioned at the interlayer boundary. The D166R mutation changes the
   charge from negative to positive, which could enhance interfacial
   stability by forming a polar interaction with Asp190. (Fig. [114]3c, d;
   Supplementary Fig. [115]5d, j). These results indicate that YcfA’s
   tetradecameric assembly is crucial for activity, as it forms a
   continuous ring of catalytic sites, improving substrate access and
   residue positioning. Notably, the D166R/Y170N mutant exhibited stable
   oligomerization compared to the variable forms of the wild type,
   presenting the potential for industrial applications, particularly in
   enhancing 6-TG production.

Nucleotide substrates drive functional YcfA assembly

   The active tetradecamer form of YcfA contrasts with its presence in
   multiple oligomeric states in solution, suggesting a substrate-induced
   mechanism for functional assembly. To test this, we analyzed the
   effects of ATP, GTP, Mg²⁺, and L-Cys on oligomerization. SEC showed
   that ATP and GTP facilitated assembly, with GTP exhibiting a stronger
   effect (Fig. [116]3e). AUC further confirmed that both nucleotides
   predominantly promoted tetradecamer formation, with increased
   resistance to trypsin digestion supporting the enhanced oligomerization
   (Supplementary Figs. [117]6a–c).

   The stronger effect of GTP arises from its binding characteristics.
   While ATP binds deeply within a monomer cavity, GTP binds at the
   catalytic interface, with its guanine base interacting with one monomer
   and its phosphate tail engaging Lys169, Lys180, and Arg18 of an
   adjacent monomer (Fig. [118]2e). This inter-monomer linkage is
   critical, as mutations at Lys169 or Arg18 disrupted GTP-induced
   assembly and binding (Supplementary Fig. [119]6d). Surprisingly, the
   K180A mutant promoted stable oligomerization independent of GTP, though
   the mechanism remains unclear, highlighting the regulatory role of
   subtle interface modifications (Supplementary Fig. [120]6d).

   Our biochemical assays demonstrated that the YcfA-YcfC system exhibits
   a strong preference for GTP over GDP and GMP, with GTP being the
   optimal substrate. This preference is linked to their phosphate group
   compositions (Supplementary Fig. [121]6e). SEC analysis confirmed that
   GTP strongly facilitated oligomerization, GDP had a weaker effect, and
   GMP had negligible influence (Fig. [122]3f). These findings suggest
   that GTP’s interactions at the catalytic interface drive the transition
   from diverse oligomeric states to a stable tetradecamer, modulating
   catalytic activity.

PLP promotes YcfA assembly by competing with GTP

   Building on our investigation of YcfA substrates, we examined whether
   molecules associated with YcfC could modulate YcfA assembly. Our
   results identified three basic residues at the catalytic interface,
   which are essential for GTP binding and pivotal to YcfA’s functional
   assembly. Given that PLP, the cofactor of YcfC, is known to covalently
   bind to lysine residues^[123]20–[124]24, we explored its potential
   influence on YcfA oligomerization. Remarkably, PLP facilitated YcfA
   assembly in a manner comparable to GTP. SEC analysis revealed that PLP
   coeluted with YcfA, demonstrated by overlapping absorbance peaks at
   420 nm (characteristic of Lys-PLP) and 280 nm (protein), suggesting a
   covalent attachment (Fig. [125]4a; Supplementary Fig. [126]6f). This
   attachment, likely via a protonated Schiff base^[127]20,[128]24, was
   localized to Lys169 of YcfA by mass spectrometry (Supplementary
   Fig. [129]6g, h). A Lys169-to-alanine mutation abolished PLP binding
   and its effect on oligomerization (Fig. [130]4b). As Lys169 mediates
   GTP interactions, PLP binding at this residue significantly disrupts
   GTP recruitment (Fig. [131]4c, d).

Fig. 4. PLP promotes YcfA assembly by competing with GTP.

   [132]Fig. 4
   [133]Open in a new tab

   a SEC analysis of YcfA (1 mg/mL) with or without the presence of PLP
   (200 μM). b SEC analysis of the YcfA^K169A (1 mg/mL) mutation with or
   without the presence of PLP (200 μM). c, d ITC measurement determined
   the binding affinity of GTP to YcfA alone and in the presence of PLP,
   respectively. The outer panel shows the raw titration traces; the inner
   panel shows the fitted titration curve. e Adenylation activity assay of
   YcfA with different concentrations of PLP over 150 s. The absorption
   peak at 315 nm gradually increased with decreasing concentration,
   showing that PLP inhibits the enzymatic activity of YcfA in a
   dose-dependent manner. Source data are provided as a Source Data file.

   Although PLP induces oligomeric organization similar to GTP, it blocks
   substrate binding and thereby inhibits YcfA’s enzymatic activity
   (Fig. [134]4e). This inhibition, while seemingly detrimental, likely
   functions as a feedback mechanism within the YcfA-YcfC system.
   Specifically, YcfC converts the unstable 6-Cys-GTP S-adduct produced by
   YcfA into 6-thio-GTP. In the absence of YcfC, PLP stabilizes YcfA’s
   oligomeric structure but suppresses its catalytic function, thereby
   preventing the wastage of ATP, guanine substrates, sulfur sources, and
   unstable intermediates. When conditions improve, the reversible
   PLP-lysine linkage can be cleaved, restoring YcfA’s active state and
   enabling its catalytic activity alongside YcfC.

Structural basis for YcfC C–S lyase

   To elucidate our understanding of the catalytic mechanism of YcfC, we
   resolved its crystal structure in complex with PLP at a resolution of
   2.09 Å (Supplementary Table [135]3). The structure revealed that YcfC
   forms a homodimer, with each monomer comprising small and large domains
   that create two identical active sites. These sites accommodate both
   the PLP cofactor and the substrates (Fig. [136]5a; Supplementary
   Fig. [137]7a). PLP forms a covalent Schiff base bond with Lys148,
   stabilized by π stacking interactions with Tyr76 and Ser124, and
   hydrogen bonding with His125, and Asp122 (Fig. [138]5b; Supplementary
   Fig. [139]7b). Mutations at Lys148, which covalently binds PLP, and
   His125, which interacts with PLP’s hydroxyl group, drastically reduced
   YcfC activity. In contrast, mutations affecting other stabilizing
   interactions had minimal impact, suggesting the covalent PLP-Lys148
   bond is sufficient to orient PLP for catalysis (Fig. [140]5c).

Fig. 5. Structural basis for YcfC C-S lyase.

   [141]Fig. 5
   [142]Open in a new tab

   a Surface representation of the YcfC homodimer, creating an active site
   for accommodating PLP and substrates. b Interaction network between PLP
   and YcfC, depicting π-π/anion-π stacking (yellow dashed lines) and
   hydrogen bonds (blue dashed lines). c Impact of mutations in putative
   PLP-binding residues on thioamide formation. The WT YcfC is regarded as
   possessing 100% enzyme activity. The mean is displayed ±SD for n = 3
   biological replicates. d Docking model of the YcfC with 6-Cys-GTP
   S-adduct, showing putative binding residues. e Activity assay of
   YcfC^Y19A and YcfC^E179A for thioamide formation. The WT YcfC is
   regarded as possessing 100% enzyme activity. The mean is displayed ±SD
   for n = 3 biological replicates. f Electrostatic potential map of YcfC,
   with 6-Cys-GTP S-adduct represented as sticks bound to the active site.
   g Proposed catalytic mechanism for YcfC-catalyzed C–S bond cleavage.
   The reaction begins with His147 deprotonating the substrate’s α-amine
   group, facilitating external aldimine formation by releasing PLP from
   Lys148. Freed Lys148 then mediates essential proton transfer reactions
   critical for C–S bond cleavage. Following reversible proton exchange
   with the solvent, His147 returns to its deprotonated state and
   subsequently acts again as a general base, deprotonating the
   substrate’s thiol (–SH) group to yield a thiolate anion. The protonated
   His147 can then transfer a proton to the N1 position of the resulting
   6-thioguanine product. Finally, Lys148 regenerates the internal Schiff
   base by nucleophilically attacking the C4’ atom of PLP, releasing the
   aminoacrylate product and resetting the enzyme for subsequent catalytic
   cycles. Source data are provided as a Source Data file.

   Although YcfC shares similar structural features with cysteine
   desulfurases^[143]25,[144]26, a DALI search identified
   C-DES^[145]27,[146]28, a C–S lyase, as its closest homolog
   (Supplementary Fig. [147]8a). Unlike cysteine desulfurases, YcfC lacks
   the conserved cysteine residue required for persulfide transfer
   (Supplementary Fig. [148]8b). Serine substitution of all cysteine
   residues in YcfC had no effect on product yield (Supplementary
   Fig. [149]9a), ruling out a desulfurase-like pathway. Furthermore, no
   direct interactions between YcfA and YcfC were detected (Supplementary
   Fig. [150]9b), unlike typical persulfide transfer systems. These
   findings confirm that YcfC functions as a C–S lyase, catalyzing the
   conversion of the 6-Cys-GTP S-adduct produced by YcfA into 6-thio-GTP.

   Molecular docking simulations provided insights into YcfC’s substrate
   recognition. Tyr19 and Glu179 were identified as key residues,
   interacting with the guanosine and ribose moieties, respectively. Tyr19
   forms a π–π stacking interaction with the guanine base, while Glu179
   forms hydrogen bonds with the ribose hydroxyl group (Fig. [151]5d).
   Substituting Tyr19 with alanine significantly impaired 6-thio-GTP
   formation, whereas mutating Glu179 had a minor effect (Fig. [152]5e),
   underscoring the critical role of Tyr19 in substrate positioning. A
   positively charged region comprising Arg17, Lys20, Lys77, Arg175,
   Lys247, and His256 at the entrance of the substrate-binding pocket,
   likely facilitates electrostatic interactions with phosphate groups of
   the substrate (Fig. [153]5f). However, mutations in these residues had
   minimal impact, indicating sufficient flexibility within this charged
   patch for substrate recruitment (Supplementary Fig. [154]9c).

   His147, positioned adjacent to the PLP–Schiff base, is essential for
   catalytic activity, as demonstrated by the complete loss of enzymatic
   function upon alanine substitution (Supplementary Fig. [155]9d, e).
   Aligning with the established mechanisms of PLP-dependent C–S bond
   cleavage enzymes^[156]29–[157]31, we propose a detailed catalytic
   mechanism for YcfC (Fig. [158]5g). The reaction cycle initiates as
   His147 deprotonates the α-amine group of the incoming 6-Cys-guanine
   nucleotide S-adduct, promoting substrate activation. This facilitates
   transaldimination with PLP, resulting in the displacement of the
   ε-amino group of Lys148 and formation of an external aldimine
   intermediate. Subsequently, the liberated Lys148 acts as a general base
   to abstract the α-proton from the substrate, triggering electron
   delocalization into the PLP conjugated system. This electron
   redistribution destabilizes the adjacent C–S bond, preparing it for
   cleavage–a critical step toward forming the PLP-aminoacrylate
   intermediate. Following reversible proton exchange with the solvent,
   His147 returns to its deprotonated state and subsequently acts again as
   a general base, deprotonating the substrate’s thiol (–SH) group to
   yield a thiolate anion. The protonated His147 can then transfer a
   proton to the N1 position of the resulting 6-thioguanine product.
   Finally, the ε-amino group of Lys148 executes a nucleophilic attack at
   the C4’ atom of the PLP–aminoacrylate intermediate, restoring the
   internal Schiff base, releasing the aminoacrylate product, and
   regenerating the active enzyme state. While we cannot definitively rule
   out a primary role of His147 in substrate positioning or stabilizing
   the catalytic site, we highlight its mechanistic resemblance to His114
   in the previously reported PLP-dependent L-cystine C–S lyase from
   Synechocystis^[159]29, which acts as a general base and stabilizes a
   labile persulfide intermediate. However, definitive experimental
   evidence confirming an analogous catalytic role for His147 in YcfC
   remains to be established.

Antibacterial activity and potential mechanism of 6-seleno-GTP

   Building on the antibacterial effect against S. aureus of
   6-thioguanine^[160]10, we investigated the antimicrobial potential of
   enzymatically synthesized 6-thio-GTP and 6-seleno-GTP against ESKAPE
   pathogens^[161]32. While 6-thio-GTP exhibited negligible activity,
   6-seleno-GTP demonstrated potent antibacterial effects specifically
   against S. aureus (Fig. [162]6a), indicating a selenoamide-dependent
   mechanism. To elucidate this mechanism, we performed metabolomic,
   transcriptomic, and proteomic analyses on the S. aureus Newman strain
   treated with 6-seleno-GTP. Metabolomics identified differentially
   expressed metabolites, primarily affecting amino acids, nucleotide, and
   purine metabolism (Supplementary Fig. [163]10a, d; Supplementary
   Data [164]1). Transcriptomic analysis revealed differentially expressed
   genes (Supplementary Fig. [165]10b, e; Supplementary Data [166]1), with
   upregulation in pathways related to purine, amino acids, and vitamin B₆
   metabolism, and downregulated in glycolysis/gluconeogenesis,
   two-component signaling, and quorum sensing pathways (Fig. [167]6b).
   Proteomics further identified differentially expressed proteins,
   including upregulated transporters and transcriptional regulators,
   alongside downregulated permeases and antiporters (Supplementary
   Fig. [168]10c, f; Supplementary Data [169]1).

Fig. 6. Antibacterial activity and potential mechanism of 6-seleno-GTP.

   [170]Fig. 6
   [171]Open in a new tab

   a MICs of 6-Seleno-GTP and 6-Thio-GTP against ESKAPE pathogens are
   shown. Blue arrows indicate MICs exceeding the highest tested
   concentration. b KEGG pathway enrichment analysis of differentially
   expressed genes is presented. Red represents upregulation; blue
   represents downregulation. c, d Venn diagrams show the overlap between
   differentially expressed genes (transcriptomics) and proteins
   (proteomics). c Nine upregulated genes were identified in both
   datasets. d One downregulated gene was identified in both datasets.
   Source data are provided as a Source Data file.

   Integrative analysis highlighted significant upregulation of purine
   biosynthesis genes (purQ, purF, purN, purH, purD) and arginine
   metabolism genes (arcA, arcB) (Fig. [172]6c), while the cold shock
   protein cspC was markedly downregulated (Fig. [173]6d). These findings
   suggest that 6-seleno-GTP disrupts purine and arginine metabolism while
   impairing stress response pathways in S. aureus. Given the efficacy of
   this disruption in suppressing the S. aureus Newman strain, these
   pathways represent promising targets for developing antimicrobial
   therapies against drug-resistant S. aureus.

Discussion

   Our findings, alongside recent studies, redefine the paradigm of
   thioamide biosynthesis by revealing a mechanism that diverges from the
   classical persulfide transfer model typically associated with cysteine
   desulfurases^[174]33. In E. amylovora, the YcfA–YcfC system employs a
   distinctive catalytic strategy: YcfA first adenylates guanine
   nucleotides and subsequently facilitates the direct addition of L-Cys
   to produce a 6-Cys-guanine nucleotide S-adduct (Fig. [175]2k). YcfC
   then acts as a C–S lyase, cleaving this intermediate to generate the
   thioamide product (Fig. [176]5g). Intriguingly, this system also
   synthesizes selenoamides, offering an alternative route for producing
   rare selenium-containing compounds^[177]34 with potential application
   in antimicrobial therapy (Fig. [178]1b; Supplementary Fig. [179]1d).

   Unlike canonical AANH-like enzymes that utilize reactive persulfides
   derived from L-Cys via desulfurases as sulfur donors^[180]16, YcfA
   directly employs L-Cys as the nucleophile to attack the adenylated
   guanine nucleotide substrates (Fig. [181]1c–e). Since L-Cys is
   inherently more stable but less reactive than persulfides, YcfA must
   overcome a significantly higher activation energy barrier. This
   challenge is mitigated by Gln129 in YcfA, which facilitates a proton
   bridge between L-Cys and the adenylated guanine, enabling L-Cys
   addition and subsequent AMP release (Fig. [182]2i, k). This dual
   function distinguishes YcfA from other AANH-like enzymes.

   Phylogenetic analyses (Supplementary Fig. [183]11) suggest that YcfC
   and canonical cysteine desulfurases likely diverged from ancestral
   protein but evolved under distinct selective pressures. Structurally,
   YcfC shares similarities with cysteine desulfurases^[184]25,[185]26 and
   C–S/C–Se lyases, such as CsdA, CsdB, and IscS (Supplementary
   Fig. [186]8a). However, unlike desulfurases, YcfC lacks the conserved
   cysteine residue critical for enzyme–persulfide intermediate formation
   (Supplementary Fig. [187]8b). Instead, YcfC operates as a C–S lyase,
   directly cleaving the C–S bond in the 6-Cys-guanine S-adduct
   (Fig. [188]5g), and a reaction that occurs with a lower activation
   energy compared to the desulfurization of L-Cys by cysteine
   desulfurases. This suggests that the YcfA–YcfC system has evolved to
   bypass the need for reactive sulfur carriers, reflecting an adaptation
   to improve catalytic efficiency.

   Our study also elucidates the dynamic oligomerization of YcfA, which
   assembles into a two-layered heptameric structure with full enzymatic
   activity (Fig. [189]3; Supplementary Fig. [190]5). Modifications to
   surface residues near the interface significantly influence this
   assembly. For instance, the L173D mutation stabilizes a dimeric form,
   while the D166R/Y170N double mutation enhances interfacial
   interactions, leading to a tetradecameric assembly with improved
   catalytic efficiency (Fig. [191]3c, d). Two key residues, Lys169 and
   Lys180, play pivotal roles in regulating oligomerization (Supplementary
   Fig. [192]6d). We identified four conditions that stabilize the
   tetradecameric structure: nucleotide substrates binding, PLP attachment
   to Lys169, mutation at Lys180, and the D166R/Y170N double mutation
   (Figs. [193]3c, e, f and [194]4a, b; Supplementary Fig. [195]5e). All
   these factors converge at the catalytic interface, facilitating the
   formation of a complete catalytic cavity.

   The highly ordered tetradecameric architecture of YcfA is critical for
   constructing functional catalytic cavities. This organization not only
   increases the local concentration of catalytic units and substrates but
   also safeguards reactive intermediates, particularly adenylated guanine
   nucleotides. The dynamic nature of this assembly acts as a regulatory
   mechanism, fine-tuning 6-TG production. Nucleotide substrates and the
   cofactor PLP exert opposing influences on the active assembly of YcfA:
   while guanine substrates promote active oligomerization, PLP inhibits
   the enzymatic activity despite promoting assembly (Figs. [196]3e, f and
   [197]4c, d). This competitive interplay ensures that the catalytic
   process remains efficient and synchronized with substrate availability,
   preventing the unnecessary expenditure of ATP and protecting transient
   intermediates.

   We hypothesize that this regulatory mechanism between YcfA and YcfC may
   allow E. amylovora to dynamically balance 6-TG production under varying
   metabolic conditions, potentially linking thioamide biosynthesis to
   bacterial virulence modulation during infection. In the early stages,
   when bacterial populations are low, the inhibitory effect of PLP could
   limit 6-TG production, delaying the expression of virulence factors and
   preventing premature activation of host defenses. As the infection
   progresses and guanine substrates accumulate, 6-TG production could be
   amplified, enabling the bacterial population to overcome host defenses.
   While these findings establish a foundation for understanding PLP’s
   role in this pathway, additional in vivo studies are required to
   validate its physiological relevance and broader implications for
   metabolic adaptation in pathogenic bacteria.

   Our discovery of 6-Se-GTP’s potent anti-methicillin-resistant S. aureus
   (MRSA) activity highlights its potential as a precision antimicrobial
   agent. Multi-omics analyses uncovered significant disruptions in key
   metabolic pathways of S. aureus, including purine and arginine
   metabolism, alongside impaired stress response pathways. While the
   exact selectivity of 6-Se-GTP remains to be fully characterized, its
   ability to target bacterial purine metabolism suggests that it could
   serve as a promising therapeutic candidate for combating
   antibiotic-resistant pathogens. These findings align with ongoing
   efforts to develop precision-based antimicrobial agents that target
   specific metabolic vulnerabilities to pathogens, offering a potential
   pathway for the treatment of resistant bacterial infections.

   Beyond structural biology, our work bridges fundamental enzymology with
   translational applications. The ability of the YcfA-YcfC system to
   bypass persulfide intermediates offers a biotechnological advantage for
   scalable thioamide production. Variants such as the D166R/Y170N mutant,
   which demonstrate enhanced catalytic efficiency, hold potential for the
   industrial-scale synthesis of 6-TG, reducing reliance on conventional
   chemical approaches that generate toxic byproducts. Furthermore, the
   modularity of this system, which is capable of incorporating both
   sulfur and selenium, opens opportunities for generating diverse
   thioamide and selenoamide libraries for drug discovery.

   In conclusion, our study provides a comprehensive view of thioamide
   biosynthesis via the YcfA–YcfC system, emphasizing its catalytic
   efficiency and regulatory mechanisms. We unveil a strategy involving
   YcfA oligomerization, wherein nucleotide substrates and PLP exert
   antagonistic effects to control catalytic activity (Fig. [198]7). These
   mechanisms may be exploited by E. amylovora to regulate 6-TG production
   and modulate virulence during host infection. Exploring this efficient
   thioamide biosynthesis pathway and its regulation could inform the
   development of antimicrobial agents, like rare selenium-containing
   compounds, and strategies for enhancing biosynthetic efficiency.

Fig. 7. A proposed working model of the YcfA-YcfC system.

   [199]Fig. 7
   [200]Open in a new tab

   This illustration depicts the dynamic interplay within the YcfA-YcfC
   system, emphasizing its dual catalytic roles and regulation by GTP and
   PLP. YcfA alternates between two conformational states: an active
   GTP-bound form and an inactive PLP-bound form. Binding of PLP induces
   oligomerization while simultaneously blocking the GTP binding site,
   acting as a regulatory switch. In its active state, YcfA facilitates a
   two-step catalytic process involving guanine nucleotides. First, ATP is
   used to adenylate the guanine nucleotide, releasing pyrophosphate
   (PPi). This modification enables the subsequent recruitment of L-Cys or
   L-SeCys, forming a 6-Cys/SeCys-guanine nucleotide S-adduct. YcfC,
   functioning as a C–S/C-Se lyase, cleaves the S-adduct to yield the
   final thioamide product. While in vitro data establishes
   GTP/PLP-mediated regulation of this system, the physiological relevance
   of PLP-induced oligomerization and metabolite homeostasis maintenance
   remains to be verified in vivo. Dashed arrows indicate hypothesized
   connections requiring experimental validation in cellular contexts.

Methods

Gene cloning and protein expression

   The ycfA and ycfC genes from E. amylovora ATCC 49946 were
   codon-optimized and synthesized by GENEWIZ (Hangzhou, China). YcfA,
   YcfC, and their mutants were overexpressed using the pET-21b vector.
   The interaction between YcfA and YcfC was investigated by co-expressing
   His-tagged YcfA and untagged YcfC in the pET-Duet-1 vector, which
   enabled subsequent affinity purification. Site-directed mutagenesis was
   introduced via quick-change PCR, and all the constructs were verified
   by Sanger sequencing (the plasmids and primers used are listed in the
   Supplementary Tables [201]4 and [202]5, respectively). The recombinant
   plasmids were transformed into E. coli BL21 (DE3) cells, which were
   subsequently cultured at 37 °C and 220 rpm to an OD[600] of 0.6.
   Protein expression was induced with 200 μM
   isopropyl-β-D-thiogalactopyranoside (IPTG) at 16 °C for 16 h.

Purification

   Harvested E. coli BL21(DE3) cell pellets were resuspended in buffer A
   (200 mM NaCl, 25 mM HEPES pH 8.0, and 10% glycerol) and lysed using a
   French press (AH-1500, ATS, China) at 4 °C. Following centrifugation
   (285,000 × g, 30 min), the supernatant was purified using
   nickel-affinity column chromatography (Ni-NTA; GE Healthcare, Little
   Chalfont, UK) with an imidazole gradient (10 and 20 mM), followed by
   elution with buffer B (200 mM NaCl, 25 mM HEPES pH 8.0, 300 mM
   imidazole, 5% glycerol). Further purification involved anion exchange
   chromatography (Source Q; GE Healthcare, Sweden; Q buffer A: 25 mM
   HEPES pH 8.0, 5% glycerol; Q buffer B: 1 M NaCl, 25 mM HEPES pH 8.0, 5%
   glycerol) and size-exclusion chromatography (Superdex 200 Increase
   10/30 GL; GE Healthcare, Sweden; gel filtration buffer: 15 mM HEPES pH
   8.0, 100 mM NaCl, 5% glycerol). Protein purity was assessed via sodium
   dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Crystallization

   Crystallization trials were conducted using purified YcfA (10 mg/mL)
   and YcfC (10 mg/mL). To obtain substrate-bound structures, YcfA and
   YcfA^D19A were pre-incubated with 5 mM substrates for 2 h at 4 °C.
   Crystals were grown by the hanging-drop vapor diffusion method at
   16 °C. Apo-YcfA crystallized in 3.5% Tacsimate and 25% PEG3350. The
   YcfA complexes (YcfA-ATP, YcfA-GTP, and YcfA^D19A-ATP-GTP) crystallized
   in 0.2 M sodium malonate pH 7.0 and 20% PEG3350. YcfC crystallized in
   50 mM sodium formate. Crystals were cryoprotected with 20% glycerol and
   flash-frozen in liquid nitrogen for data collection.

X-ray diffraction and structure determination

   X-ray diffraction data were collected at beamlines 18U1 and 19U1 of the
   Shanghai Synchrotron Radiation Facility at a wavelength of 0.9791 Å and
   processed using DIALS^[203]35 or XDS^[204]36. Structures were
   determined by molecular replacement using models generated by
   AlphaFold2^[205]37 and refined using COOT^[206]38 and Phenix^[207]39.
   The representative electron density maps for each crystal structure are
   illustrated in Supplementary Fig. [208]12 and omit maps for all
   ligand-bound crystal structures are illustrated in Supplementary
   Fig. [209]13.

Cryo-EM sample preparation

   For cryo-EM analysis, the YcfA protein sample (0.1 mg/mL) was applied
   to glow-discharged Quantifoil Au grids (40 s, 200-mesh, R2/1). The
   grids were blotted for 2 s at 100% humidity and plunge-frozen in liquid
   ethane using a Vitrobot Mark IV (Thermo Fisher).

Cryo-EM single particle data acquisition and data processing

   Data were acquired (Titan Krios microscope, 300 kV, Thermo Fisher)
   using EPU software (K2 Bioquantum director, 20 eV energy filter, 5
   frames/s, total exposure for 6 s, 49 e−/Å^2) to yield movie stacks
   (−1.2 to −1.7 μm defocus). Motion correction (Motioncor2^[210]40), CTF
   correction (CTFFIND4^[211]41), and particle extraction
   (EMAN2.31^[212]42) resulted in 85,057 particles. After 2D
   classification (Relion3.1^[213]43), the best classes were selected by
   visual examination. Then, a total of 67,556 particles were subjected to
   cryoSPARC4.4.1 to build the initial model and then subjected to
   heterogeneous refinement to yield a major class including 16,708
   particles. The major class including 16,708 particles were subjected to
   cryoSPARC4.4.1 homogeneous refinement. The particles were subjected to
   symmetry alignment and homogeneous refinement with C7 symmetry string
   to yield the final map. A sharpening B-factor of −102.6 Å^2 was applied
   to the resulting cryo-EM map to yield the final sharpened map at 3.44 Å
   resolution estimated by the 0.143 criterion of FSC curve. Data
   collection and processing statistics were included in Supplementary
   Table [214]2. Representative model building with cryo-EM densities of
   the YcfA are illustrated in Supplementary Fig. [215]14.

UV-Vis spectrophotometric activity assays

   YcfA and YcfC activity assays were performed using UV-Vis
   spectrophotometry. For endpoint assays, purified YcfA, YcfC, or their
   mutants (10 μM or 5 μM, respectively) were incubated in a reaction
   buffer (10 mM HEPES pH 8.0, 100 mM NaCl, and 5 mM MgCl[2]) with 100 μM
   substrates (ATP, GTP/GDP/GMP, and L-Cys/L-SeCys/other amino acids) at
   30 °C for 30 min. The reactions were quenched with methanol (80 μL for
   each 20 μL of reaction mixture) and centrifuged, and the supernatant
   was analyzed by a Nanodrop One spectrophotometer (Thermo Scientific,
   MA). For kinetic assays, reactions (1 mL) containing 20 μM YcfA, 500 μM
   ATP, 500 μM GTP, and 100 μM L-Cys in reaction buffer (100 mM NaCl,
   15 mM HEPES pH 8.0, 5% glycerol, and 5 mM MgCl[2]) were monitored at
   5-second intervals in a quartz cuvette using an Ultrospec 2100 pro
   UV-Vis spectrometer (Biochrom, UK). After the first 40 readings, 30 nM
   YcfC was added, and the reaction was monitored for an additional 60
   readings. In separate experiments, 42 nM YcfC, 500 μM ATP, 500 μM GTP,
   and 100 μM L-Cys were mixed, recorded at 5-s intervals for 40 reads,
   followed by the addition of 20 μM YcfA for another 60 readings. To
   analyze the effect of PLP on the enzyme activity of YcfA, we incubated
   YcfA with various concentrations of PLP (0 μM, 40 μM, 80 μM and 160 μM)
   for 10 min, followed by kinetic analysis. The PLP concentration is
   consistent with reported in vivo PLP levels^[216]44. The data were
   analyzed using GraphPad Prism 8.3.

High-performance liquid chromatography (HPLC) analysis of the reaction
products

   The reaction products of the YcfA-YcfC system were analyzed by HPLC
   following methanol treatment and centrifugation to remove precipitates.
   A Shimadzu LC-2030 system equipped with a Prodigy ODS-3 analytical
   column (4.6 × 250 mm, 5 μm) was used for analysis. The mobile phase
   consisted of solvent A (30 mM KH[2]PO[4], 10 mM tetrabutylammonium
   hydrogen sulfate in water, pH 7.5) and solvent B (methanol). Gradient
   elution was performed as follows: 0.5% solvent B (0-10 min), a linear
   gradient to 70% solvent B (10–23 min), and a final increase to 99%
   (23–28 min). Detection was performed at 340/360/315 nm, with a flow
   rate of 1 mL/min at room temperature.

Product identification by mass spectrometry

   The reaction products, prepared as described above, were dissolved in a
   1:1 (v/v) methanol: water mixture. An Ultimate 3000 system coupled with
   a Q Exactive Plus Q-Orbitrap high-resolution mass spectrometer (Thermo
   Fisher Scientific, Waltham, MA, USA) was used for analysis. After
   injection of 10 μL of sample, isocratic elution was performed using
   solvent A (water containing 10 mM NH[4]Ac) and solvent B (acetonitrile:
   water = 90:10 (v/v) with 10 mM NH[4]Ac). A heated electrospray
   ionization (HESI) source was used with the following parameters: sheath
   gas, 35 arb; auxiliary gas, 10 arb; spray voltage, 3.2 kV
   (positive/negative mode); capillary temperature 320 °C; and auxiliary
   gas heater temperature, 350 °C. Data were collected within a mass range
   of 50–750 m/z with a mass resolution of 70,000.

Identification of the 6-Cys-GTP S-adduct intermediate

   The reaction mixture (50 μL), containing reaction buffer (100 mM NaCl,
   15 mM HEPES, pH 8.0, and 5 mM MgCl[2]), 100 μM YcfA, 1 mM ATP, 1 mM
   GTP, and 1 mM L-Cys, was incubated at room temperature for 30 min.
   Then, the mixture was quenched with methanol (100 μL) and centrifuged
   (160,000 × g, 20 min). For MS analysis, 5 µL samples were separated on
   a BEH amide column (2.1 × 100 mm, 1.7 µm, Waters, USA) at 40 °C at a
   flow rate of 0.3 mL/min). The mobile phase comprised solvent A (H[2]O)
   and solvent B (90:10 acetonitrile: H[2]O), both of which contained
   10 mM ammonium formate and 0.15% formic acid. The gradient elution
   profile was as follows: 100% B (0–2 min), 100% B to 85% B (2–9 min),
   85% B to 50% B (9–14 min), isocratic elution with 50% B (14–19 min),
   and 100% B (19–27 min). A heated electrospray ionization (HESI) source
   was used with the following parameters: sheath gas, 35 arb; auxiliary
   gas, 10 arb; spray voltage, 3.2 kV (positive/negative mode); capillary
   temperature 320 °C; and auxiliary gas heater temperature, 350 °C. Data
   were collected within a mass range of 50–750 m/z with a mass resolution
   of 70,000. The dehydride ion at m/z 624.9925 was fragmented using DDA
   mode with specific MS/MS parameters.

Limited proteolysis assay of YcfA

   The YcfA used in this assay was adjusted to a molar concentration of
   about 20 µM. YcfA with or without ATP and GTP was subjected to trypsin
   digestion. After 30 min of incubation on ice, the resulting proteolyzed
   products were analyzed via SDS-PAGE.

MS analysis of YcfA-PLP covalent binding

   Covalent binding of PLP to YcfA was analyzed using a previously
   described method^[217]24. Briefly, 100 µg of YcfA was incubated with
   1 mM PLP for 30 min, followed by precipitation with 1 mL of acetone,
   and storage overnight at −80 °C. After centrifugation (13,000 × g,
   15 min, 7 °C), the protein pellet was solubilized in 100 mM HEPES (pH
   7.8), reduced with 10 μL of 1 M NaBH[4] (37 °C, 1 h), and then
   precipitated again. The resulting sample was dissolved in S-Trap lysis
   buffer, reduced with 4.5 mM DTT, alkylated with 10 mM iodoacetamide
   (quenched with DTT), and purified using an S-Trap column. Trypsin
   digestion was performed overnight at 37 °C. The resulting peptides were
   separated using a Thermo Scientific EASY nLC-1200 system and analyzed
   by a Q Exactive Hybrid mass spectrometer. The data was processed using
   Thermo Xcalibur and search parameters included carbamidomethyl
   (+57.021 Da) as a static modification and PLP-binding (+231.0297 Da) as
   a dynamic modification; all other parameters were set to the default
   values.

Molecular docking simulations

   Molecular dynamics simulations were performed to obtain the
   YcfA-adenylated-GTP complex. YcfA with a 2.34 Å crystal structure
   resolution was used as a template. The protonation states of the
   residues were determined using H^++ server, in accordance with the
   hydrogen bonding network, and the force fields for amino acid residues
   and solvent water molecules were assigned using GROMACS 2022.5^[218]45.
   The system was placed in a cubic box with a minimum distance of 10 Å
   between the protein surface and the box edge, then solvated with TIP3P
   water molecules and neutralized with sodium and chloride ions. Energy
   minimization was conducted using a 5000-step steepest descent method to
   remove unreasonable contacts, followed by 5000 steps of gradient
   conjugation. Subsequently, the system was heated from 0 K to 310 K over
   100 ps using the NVT ensemble. This was followed by 100 ps of NPT
   ensemble equilibration at 1 bar pressure, 310 K. Hydrogens bonds were
   constrained using the LINCS algorithm and the time step for all
   simulations was 2 fs. A 100 ns production MD simulation was then
   performed under NPT ensemble conditions without any restraints.
   Trajectory analysis was performed with the VMD 1.9.3 software. The
   average RMSD of YcfA, adenylated-GTP, and YcfA-adenylated-GTP complex
   is provided in Supplementary Fig. [219]15. Molecular docking
   simulations were performed using AutoDock Vina 4.2.0^[220]46. Ligands
   were prepared using IQmol Molecular Viewer. For the YcfA-adenylated-GTP
   docking with L-Cys, the binding site was defined with the following
   coordinates: center_x: 46.396, center_y: 38.633, and center_z: 6.731.
   For YcfC docking with the 6-Cys-GTP S-adduct, the binding site
   coordinates were as follows: center_x: 12.399, center_y: 47.636, and
   center_z: -5.996. All other parameters were set the default values.
   PyMOL 2.5^[221]47 was used for visualization and image processing. All
   the coordinate files of the final docked models and MD data are
   provided in Supplementary Data [222]2.

Size-exclusion chromatography (SEC) analysis

   Size-exclusion chromatography was performed using a Superdex 200
   Increase 10/30 GL column (GE Healthcare) equilibrated with gel
   filtration buffer containing 200 μM of the appropriate substrates.
   Purified YcfA and its variants were preincubated with the corresponding
   substrates (200 μM) before injection. As for the interaction between
   PLP and YcfA, YcfA (1 mg/mL) and PLP (200 μM) were used for SEC
   analysis.

Isothermal titration calorimetry (ITC)

   ITC experiments were conducted using a Nano ITC (TA instruments, USA)
   at 16 °C. To analyze the binding of GTP to YcfA (or PLP-bound YcfA),
   the concentration of YcfA was used at 260 μM, and the GTP solution was
   diluted to 4 mM. GTP (2.5 µL) was injected into 300 μL of YcfA (or
   PLP-bound YcfA) twenty times every 120 seconds with a stirring speed of
   200 rpm. The data were analyzed using the one-site binding mode in
   NanoAnalyze Data Analysis software (version 3.8.0).

Analytical ultracentrifugation (AUC)

   The SV-AUC experiments were conducted using a 12 mm charcoal-filled
   Epon centerpiece and an eight-hole An50 Ti rotor at 56,500 × g (Optima
   AUC analytical ultracentrifuge, Beckman Coulter). The absorbance was
   monitored at 280 nm. The data were analyzed using SEDFlIT software to
   determine the sedimentation coefficient distribution C(S).

Phylogenetic analysis of YcfC

   All protein sequences used for analysis were obtained from the NCBI
   server ([223]https://www.ncbi.nlm.nih.gov/). The protein data used for
   phylogenetic analysis is listed in Supplementary Table [224]6. The
   amino acid sequences of these proteins were aligned using the MAFFT
   ver. 7 server ([225]https://mafft.cbrc.jp/alignment/server/) with the
   default settings. The unrooted fast maximum-likelihood-based tree was
   generated using IQ-TREE^[226]47 with automatic model selection mode
   (ModelFinder)^[227]48, where LG + R4 was selected. Phylogenetic
   bootstrap analysis was performed by ultrafast approximate bootstrap
   with 1000 bootstrap replicates. The tree was displayed using
   MEGA7^[228]49 software and then beautified with iTOL.

Minimum inhibitory concentration (MIC) determination

   MIC values were determined via a standard broth microdilution method.
   The bacterial cultures were adjusted to ~5 × 10^5 CFU/mL, and 2-fold
   serial dilutions of each antibiotic were dispensed into 96-well plates.
   After a 24 h of incubation at 37 °C, the MIC was defined as the lowest
   concentration of antibiotic with no visible bacterial growth. The
   experiments were performed in triplicate.

Omics sample preparation

   S. aureus Newman was grown overnight then diluted to an OD[600] of 0.1
   in TSB medium. For the drug treatment groups, 6-seleno-GTP was added at
   a final concentration of 4 μg/mL, and the cultures grew to the
   mid-exponential phase. Ten duplicate samples were prepared for
   metabolomics analysis, whereas three duplicate samples were prepared
   for transcriptomics and proteomics analyses.

Metabolite extraction

   Treated bacterial suspensions were centrifuged and washed twice with
   PBS. Then, 100 μL of prechilled methanol: water (8:2) mixture and 10 μL
   of internal standard were added. The mixture was subjected to three
   freeze-thaw cycles (liquid nitrogen and ice), followed by 15 min of
   sonication in an ice bath. An additional 400 μL of prechilled methanol:
   water (8:2) was added, and the mixture was vortexed for 3 min at 4 °C
   and 2500 g. After quenching at −80 °C for 20 min, the sample was
   sonicated for 5 min at 4 °C and centrifuged at 4 °C and 189,000 × g for
   20 min. The resulting supernatant (200 μL) was collected, vacuum-dried,
   and prepared for GC-MS analysis.

Metabolomics analysis

   The samples were analyzed via GC–TOF–MS using an Agilent 8890/LECO
   Pegasus BT instrument. One microliter of each sample was separated on
   an Rxi-5ms column (30 m × 0.25 mm × 0.25 μm) with helium as the carrier
   gas (1 mL/min). The inlet was operated in splitless mode at 270 °C. The
   GC oven temperature program was as follows: 50 °C for 0.5 min, followed
   by an increase of 5 °C/min to 300 °C, where it was held for 6 min. The
   transfer line and ion source temperatures were 280 °C and 250 °C,
   respectively. MS analysis was conducted in the m/z range of 60–700 with
   an ionization energy of 70 eV. The data were compared to an in-house
   data library and analyzed using tidyverse^[229]50. Metabolites with
   |log[2]FoldChange|> 0.5 and P-value < 0.05 were considered
   significantly differentially expressed.

Transcriptomics analysis

   RNA extraction and transcriptome sequencing were performed by Shanghai
   Personal Biotechnology Co., Ltd. (Shanghai, China). Total RNA was
   isolated using TRIzol Reagent, and rRNA was removed using the Zymo-Seq
   RiboFree Kit. cDNA was synthesized using SuperScript III and DNA
   Polymerase I. cDNA fragments were purified using the AMPure XP system
   (Beckman Coulter, Beverly, CA, USA), and sequencing libraries were
   prepared using the Illumina PCR Primer Cocktail and sequenced on the
   NovaSeq 6000 platform (Illumina). The transcriptomic datas were aligned
   to the S. aureus Newman genome using Bowtie2. Differentially expressed
   mRNAs were identified using DESeq2^[230]51, with significance defined
   as |log[2]FoldChange|>0.5 and P-value < 0.05. GO enrichment analysis
   was performed using topGO^[231]52, and KEGG pathway enrichment analysis
   was performed using clusterProfiler 4.0^[232]53, with significance
   defined as a P-value < 0.05.

Proteomics analysis

   Protein extraction and proteomics analysis were performed by Shanghai
   Personal Biotech Co., Ltd. (Shanghai, China). Samples were extracted
   via homogenization and SDT lysis, and the protein was quantified via
   the BCA method, and separated by SDS-PAGE. Use FASP to digest the
   protein. After drying the digested peptides, dissolve them with 0.1%
   FA, and determine the peptide concentration for LC-MS analysis. The
   Vanquish Neo UHPLC (Thermo Scientific) system and an Orbitrap Astral
   high-resolution mass spectrometer were utilized for the purpose of
   proteomic analysis. The sample is injected into the Trap Column (PepMap
   Neo 5 µm C18 300 µm × 5 mm, Thermo Scientific) and then subjected to
   gradient separation using the analytical column (μPAC Neo High
   Throughput column, Thermo Scientific). The mass spectrometry was
   performed in Data Independent Acquisition (DIA) mode with parameters
   including a scan range of m/z 380–980, a resolution of 240,000 for the
   first stage, a resolution of 80,000 for the second stage, an isolation
   window of 2 Th, and an HCD collision energy of 25%. DIA raw files were
   analyzed via DIA-NN^[233]54. NCBI S. aureus Newman (access number
   426430) as reference database. The analysis yielded 2246 protein groups
   and 32,906 PSMs. Proteins with |log[2]FoldChange|>0.5 and
   P-value < 0.05 were considered significantly differentially expressed.
   All the gene models were annotated and classified using our draft
   genome database draft for COG and KEGG annotation.

Reporting summary

   Further information on research design is available in the [234]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [235]Supplementary Information^ (21.4MB, pdf)
   [236]41467_2025_63937_MOESM2_ESM.pdf^ (88.8KB, pdf)

   Description of Additional Supplementary Files
   [237]Supplementary Data 1^ (47.7KB, xlsx)
   [238]Supplementary Data 2^ (1.5MB, zip)
   [239]Reporting Summary^ (75.1KB, pdf)
   [240]Transparent Peer Review file^ (677.3KB, pdf)

Source data

   [241]Source data^ (29.8MB, xlsx)

Acknowledgements