A ribosomally synthesized and posttranslationally modified peptide with ADP-ribosylation

Significance

Adenosine diphosphate ADP-ribosylation has yet to be extensively investigated in natural product biosynthesis. Here, we report minviopeptin, a triceptide modified by the ADP-ribosyltransferase MicD, marking a significant breakthrough as an instance where ADP-ribosylation serves as an intrinsic biosynthetic step in a ribosomally synthesized and posttranslationally modified peptide (RiPP) pathway. Additionally, we introduce an αKG-dependent nonheme iron (αKG-NHFe) enzyme, MicC, which effectively catalyzes oxidative N–Cα bond cleavage, leading to C-terminal amidation in a RiPP maturation. Our study reveals a posttranslational modification in RiPPs and expands the enzymatic chemistry known in natural product biosynthesis.

Abstract

Ribosomally synthesized and posttranslationally modified peptides (RiPPs) are a fertile ground for uncovering new enzymatic chemistry and structural complexity. Here, we describe minviopeptin, an unusual ADP-ribosylated triceptide accessed through heterologous expression of a cryptic biosynthetic gene cluster. Structural and functional analyses reveal a combination of crosslinking, ADP-ribosylation, and oxidative peptide cleavage, underscoring the capacity of RiPP pathways to generate densely functionalized molecular scaffolds. By revealing ADP-ribosylation as a previously unrecognized RiPP modification and exposing reactivity within radical SAM and nonheme iron enzymes, this work broadens the landscape of RiPP biosynthetic chemistries and offers opportunities for natural product diversification and peptide engineering.

Ribosomally synthesized and posttranslationally modified peptides (RiPPs) constitute a rapidly expanding superfamily of natural products present in all domains of life (1, 2). RiPP biosynthesis begins with genetically encoded precursor peptides that typically contain an N-terminal leader and a C-terminal core peptide. The leader sequence generally serves as a recognition signal that guides biosynthetic enzymes to modify the core peptide and is ultimately cleaved by dedicated peptidases to release the mature RiPP product (1, 2). Due to the high variability of core peptide sequence and the diverse modification enzymes encoded within the biosynthetic gene clusters (BGCs), RiPPs exhibit vast structural diversity and a broad range of bioactivities spanning antimicrobial, anticancer, and antiviral properties (3–9). Moreover, RiPP biosynthesis represents a privileged system for the discovery of novel enzymology, continuously expanding the known repertoire of enzymatic transformations in natural product assembly (3–9). The rich abundance and genomic accessibility of RiPP BGCs across highly diverse species underscore their potential as a vast, yet largely untapped reservoir of diverse chemical structures and biosynthetic chemistries.

ADP-ribosylation involves the covalent transfer of an ADP-ribose (ADPr) moiety from the cosubstrate β-nicotinamide adenine dinucleotide (NAD⁺) onto the target substrates (10–13). This reaction is catalyzed by enzymes of the ADP-ribosyltransferase (ART) superfamily, which is phylogenetically ancient and widespread across all domains of life (12, 14). ADP-ribosylation plays crucial roles in diverse cellular processes ranging from DNA repair and signal transduction to bacterial virulence (15–20). Bacterial ARTs were originally discovered as toxins, with diphtheria toxin (DT) secreted by Corynebacterium diphtheriae serving as the earliest identified prototype (21). These toxins inject directly or trigger an endocytosis mechanism into the host cell where they catalyze covalent ADP-ribosylation of specific host proteins, altering their activities and thereby perturbing essential cellular pathways (19, 22). Despite its fundamental biological significance and expanding substrate scope, ADP-ribosylation remains largely underexplored in natural product biosynthesis. To date, the only documented natural products known to undergo ADP-ribosylation are rifamycin antibiotics, in which the rifamycin ADP-ribosyltransferase (Arr) acts as a dedicated resistance enzyme by regioselectively transferring ADPr to the C23–OH of rifamycin antibiotics (e.g., rifampin) (Fig. 1A), thereby directly inactivating the antibiotic (23, 24). Arr homologs are frequently plasmid-encoded and disseminated among pathogens, indicating that this modification functions primarily as an evolved resistance mechanism rather than as an integral component of natural product construction (25). Thus far, ADP-ribosylation has not been identified in natural product biosynthetic pathway besides rifamycin resistance.

Fig. 1.

ADP-ribosylation in natural products. (A) Arr-mediated ADP-ribosylation as a resistance mechanism, which inactivates the rifamycin antibiotics (e.g., rifampin) through ADP-ribosylation at the C23–OH position. (B) The mic BGC from M. inviolabilis, which encodes a protein of unknown function MicC and a predicted ADP-ribosyltransferase (ART) MicD, in addition to the precursor peptide MicA and the PasB-like radical SAM enzyme MicB.

Herein, we report the identification and biosynthetic characterization of minviopeptin, an ADP-ribosylated RiPP natural product from Mucilaginibacter inviolabilis. Besides the precursor peptide MicA and a radical SAM enzyme MicB, the BGC (mic) also encodes two unusual auxiliary enzymes, a putative ART MicD and a hypothetical protein MicC. Heterologous expression of the mic cluster in Escherichia coli yielded a product featuring an ADP-ribosylation modification, which is catalyzed by MicD. Furthermore, we demonstrate that the hypothetical protein MicC functions as an α-ketoglutarate-dependent nonheme iron (αKG-NHFe) enzyme, catalyzing the cleavage of the C-terminal follower region to result in the mature peptide with an amide terminus. This work provides an example of a RiPP-associated ADP-ribosylation, representing an intrinsic biosynthetic modification and expanding the landscape of posttranslational diversification in ribosomal peptides.

Results

Identification of mic as a Triceptide BGC.

Previous studies revealed that gene clusters homologous to the darobactin BGC are widely distributed in nature, producing darobactin-like peptides (daropeptide) with vast structural diversity (26, 27). Using the monocyclic daropeptide photorhaptin synthetase PasB (27) as a query, we conducted a comprehensive genome mining analysis to explore potential daropeptide BGCs. Detailed sequence similarity network (SSN) and genomic neighborhood analysis (28, 29) revealed a distinct BGC from M. inviolabilis (designated mic), which, besides the precursor peptide MicA and PasB-like radical SAM enzyme MicB, encoded a protein of unknown function MicC, and a predicted ART MicD (Fig. 1B and SI Appendix, Fig. S1).

The putative core peptide sequence of MicA lacked the canonical Ω1-X2-Ω3 motif characteristic of daropeptides, but instead contained three Ω1-X2-X3 motifs (i.e., YSK, HAS, and FTK) (Fig. 1B), suggesting it might be a triceptide rather than a daropeptide (26, 30, 31). To test this hypothesis, we coexpressed MicA with the radical SAM enzyme MicB in E. coli. High-resolution mass spectrometry (HR-MS) analysis of the resulting product, MicA^B (the superscript B denotes MicB-modified MicA), revealed a mass shift of −6 Da compared to unmodified MicA (SI Appendix, Fig. S2), consistent with the formation of three C–C crosslinks. Further LC-MS/MS analysis of a GluC-digested 28-mer C-terminal fragment localized −2 Da mass shifts to each of the YSK, HAS, and FTK motifs, suggesting C–C bond formation at the three Ω1-X2-X3 sites (SI Appendix, Fig. S3).

Of note, GluC digestion of the native MicA precursor yielded a 28-mer peptide fragment, and such high molecular weight significantly hindered efficient MS/MS fragmentation and unambiguous spectral assignment. We observed that MicA contains a Pro residue (Pro45) immediately preceding the YSK motif where the first cyclophane forms. In RiPP biosynthesis, Pro residues frequently serve as specific cleavage sites for proteases responsible for releasing mature natural products, as exemplified by prolyl oligopeptidases (32). Moreover, all currently characterized triceptides thus far undergo proteolytic maturation at a site preceding the first cyclophane-forming motif (30, 31). To facilitate structural characterization and enable direct access to the mature natural product, we engineered a MicA P45E mutant (MicA*) by replacing Pro45 with Glu to create a GluC cleavage site. This strategy, previously validated in other RiPP expression systems (33–37), allowed generation of a 14-mer core peptide (YSKIHASGGFTKDL, 1) by GluC digestion. Upon GluC treatment of MicA*^B obtained from coexpression of MicA* with MicB, HR-MS analysis of the resulting product 2 revealed a mass shift of −6 Da compared to the unmodified peptide (Fig. 2A, traces i and ii), consistent with triple cyclophane formation. Subsequent HR-MS/MS analysis unambiguously localized the −2 Da mass shifts to each of the YSK, HAS, and FTK motifs (SI Appendix, Fig. S4), confirming the formation of three characteristic C–C crosslinks. These results not only demonstrate that the mic BGC encodes a triceptide but also support our earlier observation that three-residue cyclophane forming enzymes (3-CyFEs) and daropeptide maturases share a common evolutionary origin (27).

Fig. 2.

In vivo investigation of MicB, MicC, and MicD in E. coli. (A) Functional characterization of MicD, showing HR-MS spectra of the full-length precursor peptide MicA* (i) expressed alone, or coexpressed with (ii) MicB, (iii) MicD, and (iv) MicB and MicD. (B) Functional characterization of MicC, showing HR-MS spectra of purified full-length precursor peptide MicA* (i) expressed alone, or coexpressed with (ii) MicC, (iii) MicB and MicC, and (iv) MicB, MicC, and MicD. (C) Proposed biosynthetic pathway for mic. The peptide fragments resulting from GluC digestion are the predicted core sequence numbered starting from Tyr1.

MicD is an ART in RiPP Biosynthesis.

Initial attempts to characterize MicD function by coexpressing MicA* with MicD alone in E. coli failed to yield any modified product (Fig. 2A, trace iii and SI Appendix, Fig. S5). However, when MicA* was coexpressed with both MicB and MicD, the resulting product MicA*^BD exhibited a net mass increase of 535 Da relative to unmodified MicA* (Fig. 2A, trace iv). This observation, combined with the predicted ART function of MicD, led us to propose a two-step modification process, in which MicB first mediates triple cyclophane formation (−6 Da total), followed by the MicD-catalyzed ADP-ribosylation (+541 Da) (Fig. 2C). HR-MS/MS analysis of the GluC-digested product of MicA*^BD (product 3) confirmed ADP-ribosylation in the putative core peptide (i.e., YSKIHASGGFTKDL) (SI Appendix, Fig. S6), though the precise modification site remained unresolved due to CID-induced lability of the ADPr linkage (38).

To biochemically validate the ART activity of MicD, the N-terminally His-tagged MicD was overexpressed in E. coli and purified to homogeneity (SI Appendix, Fig. S7). When incubated with NAD⁺ and MicD, the peptide substrate MicA*^B (obtained from MicA* coexpression with MicB) clearly exhibited a +541 Da modification consistent with ADP-ribosylation (Fig. 3A, trace iii). This modification did not occur in the control assay using the heat-inactivated enzyme or omitting NAD⁺ (Fig. 3A, traces i and ii, and SI Appendix, Fig. S8), or with the unmodified MicA* that lacks cyclophane (SI Appendix, Fig. S9), the latter is consistent with the coexpression study showing the MicB-catalyzed cyclophane formation is essential for subsequent ADP-ribosylation. Kinetic studies revealed that MicD modifies MicA*^B with moderate efficiency, exhibiting a K_m of 8.64 ± 1.55 μM and a k_cat of 3.04 ± 0.23 min⁻¹ (SI Appendix, Fig. S10). These results unequivocally establish MicD as a bona fide ART that specifically modifies the cyclophane-containing MicA precursor, representing an example of an ART enzyme participating in RiPP maturation.

Fig. 3.

In vitro reconstitution of MicC and MicD activity and structural elucidation of minviopeptin. (A) In vitro modification of MicA*^B by MicD. A +541 Da mass increase, consistent with ADP-ribosylation, was not observed in reactions with heat-inactivated MicD (i) or omitting NAD⁺ (ii). (B) In vitro modification of MicA*^BD by MicC, showing the HR-MS spectra of MicC in vitro reaction using MicA*^BD as substrate (i) with the heat-inactivated enzyme, (ii) in the absence of αKG, (iii) EDTA treatment to chelate Fe²⁺, (iv) the absence of ascorbate, (v) no extra Fe²⁺ added and (vi) the presence of αKG, Fe²⁺, and ascorbate. The −717.3 Da mass shift corresponds to the oxidative cleaved product with a C-terminal amide. (C) The planar structure of minviopeptin characterized based on key 2D NMR correlations. (D) The configurations of the YSK macrocycle based on key NOESY correlations. (E) The configurations of the HAS macrocycle based on key NOESY correlations. Protons positioned toward the same side of the plane are highlighted in blue and red, respectively. (F) The structure of minviopeptin.

MicC is an αKG-NHFe Enzyme Responsible for Peptide Oxidative Cleavage.

Initial attempts to characterize MicC activity through in vivo coexpression with either MicA* alone or with MicA* and MicB failed to yield detectable modifications (Fig. 2B, traces ii and iii, and SI Appendix, Figs. S11 and S12). When MicA* was coexpressed with the full biosynthetic machinery (MicBCD), the resulting peptide (MicA*^BCD) clearly exhibited a net mass loss of 182.3 Da compared to the unmodified MicA* in HR-MS analysis (Fig. 2B, trace iv), which is in striking contrast to the +541 Da shift observed for MicA*^BD. This significant discrepancy suggested a putative peptide cleavage alongside ADP-ribosylation. Detailed MS/MS analysis of the GluC-digested MicA*^BCD (product 4) mapped the mass changes to three modifications: MicB installed two cyclophanes (−4 Da) at YSK and HAS motifs, MicD appended ADPr (+541 Da), and MicC mediated excision of the C-terminal heptapeptide GGFTKDL (−719.3 Da net) (Fig. 2C and SI Appendix, Fig. S13). The −0.99 Da shift in y-series ions outside cyclophane regions suggest formation of an amide Ser7 in the C-terminus, implicating MicC in oxidative cleavage of MicA*^BD at Gly8.

MicC was annotated as a hypothetical protein with no discernible function from sequence analysis. Searching the DALI server using the AlphaFold3 structure of MicC, however, revealed the MicC homology to αKG-dependent nonheme iron (αKG-NHFe) enzymes such as KdoO (3.1 Å RMSD over 177 Cα, PDB 6A2E) (39, 40). Biochemical studies by incubation of MicA*^BD (obtained by coexpression of MicA* with MicB and MicD) with αKG, (NH₄)₂FeSO₄, and ascorbate indeed yielded the expected product MicA*^BCD with −717.3 Da modification (Fig. 3B, trace vi), which was confirmed by detailed HR-MS/MS analysis (SI Appendix, Figs. S14 and S15). The activity of MicC was strictly dependent on αKG (Fig. 3B, trace ii) and was completely abolished by EDTA addition (Fig. 3B, trace iii). Although enzyme activity was observed in the absence of Fe²⁺ (Fig. 3B, trace v) and ascorbate (Fig. 3B, trace iv), the addition of both compounds significantly enhanced enzymatic activity (SI Appendix, Fig. S15). Notably, MicC exhibited strict selectivity for the doubly modified substrate, as no reactivity occurred with MicA* or MicA*^B (SI Appendix, Fig. S16), suggesting that MicD catalyzed ADP-ribosylation is a prerequisite for MicC activity. To unambiguously define the requirement of ADP-ribosylation for MicC catalysis, we carried out sequential-addition in vitro assays. Consistent with our in vivo coexpression results, product 4 was observed only when MicD was preincubated with MicA*^B prior to the addition of MicC (SI Appendix, Fig. S17). Steady-state kinetic analysis using MicA^*BD as the substrate yielded a K_m of 84.83 ± 13.11 μM and a k_cat of 57.5 ± 5.4 min^-1 (SI Appendix, Fig. S18).

Notably, we also observed a product 5 (m/z = 734.3347) corresponding to the cleaved heptapeptide (GGFTKDL −3.05 Da) (Fig. 2C and SI Appendix, Fig. S19A). HR-MS/MS analysis localized the mass loss to the cyclophane formed within the FTK motif (−2 Da) and the N-terminal oxidation (−1.03 Da) (SI Appendix, Fig. S19B), and the latter is further supported by observation of its hydrated gem-diol derivative (m/z = 752.3435) (SI Appendix, Fig. S19C). These observations suggest MicC cleaves the N–Cα bond at Gly8, releasing GFTKDL with an N-terminal oxalyl group while forming an amide at Ser7 in the C-terminus (Fig. 2C and SI Appendix, Fig. S20). This dual chemistry expands the known reactivity of αKG-NHFe enzymes, positioning MicC as a protease-like enzyme in this superfamily. Notably, the G8A variant of MicA* showed no detectable cleavage (SI Appendix, Fig. S21), establishing that MicC-catalyzed oxidative bond cleavage occurs strictly on a Gly residue.

Structural Characterization of Minviopeptin, an ADP-Ribosylated Triceptide.

Although MS/MS analysis supported the occurrence of ADP-ribosylation, the precise modification site had remained unclear. To definitively characterize the product of the mic BGC, MicBCD-modified MicA* was produced on a 25 L scale, digested with GluC, and purified by HPLC to afford ~5 mg compound 4 for NMR analysis. A comprehensive set of 1D and 2D spectra (¹H, ¹³C, ³¹P, DEPT-90, COSY, TOCSY, NOESY, HSQC, and HMBC) was acquired (SI Appendix, Figs. S22–S30 and Table S5).

For the YSK motif, DEPT-90 and HSQC spectra indicated the loss of a proton resonance at Tyr1-C3 (δC 124.5), and Lys3-Cβ (δC 50.1, δH 2.57) was converted from a methylene to a methine (Fig. 3C and SI Appendix, Figs. S24 and S29). Key HMBC correlations, including Tyr1-H2 (δH 6.61) and H6 (δH 6.88) to Tyr1-Cβ (δC 37.4), Lys3-Hβ (δH 2.57) to Tyr1-C4 (δC 156.1), and Tyr1-H2 (δH 6.61) to Lys3-Cβ (δC 50.1), established a C–C linkage between Tyr1-C3 and Lys3-Cβ (Fig. 3C and SI Appendix, Fig. S31A). Strong NOESY correlations between Tyr1-H2 and Lys3-Hβ further supported the C–C crosslinking (Fig. 3D). Within the HAS motif, His5 displayed a single protonated heteroaromatic carbon at C4 (δC 120.6, δH 7.36), assigned by DEPT-90 and HMBC correlations to adjacent quaternary carbons (δC 143.4, 125.7) (Fig. 3C and SI Appendix, Fig. S31 B, Left). HMBC from His5-Hβ (δH 3.22) to His5-C5 indicated substitution at C2 (SI Appendix, Fig. S31 B, Right), consistent with crosslink formation between His5-C2 and Ser7-Cβ. Supporting this, Ser7-Cβ (δC 67.5) appeared as a substituted methine (SI Appendix, Fig. S24), and a diagnostic HMBC correlation from Ser7-Hβ (δH 5.05) to His5-C2 (δC 143.4) confirmed the C-C crosslink (Fig. 3C and SI Appendix, Fig. S31 B, Right). Stereochemical assignments were established by combining NOESY correlations with Marfey’s analysis, which showed all α-carbons are L-configured (Fig. 3 D and E and SI Appendix, Table S6).

Product 4 (hereafter minviopeptin) exhibited a strong UV absorption at 258 nm characteristic of adenosine, which is absent in product 2 lacking ADPr (SI Appendix, Fig. S32). DEPT-90 and HSQC spectra indicated that ribose’-C1 (δC 89) was a methine rather than a methylene (SI Appendix, Fig. S24), and ³¹P NMR revealed the presence of phosphate groups (SI Appendix, Fig. S25). Critically, an HMBC correlation between ribose’-H1 (δH 6.5) and His5-C4 (δC 120.6) established a C–N bond between ADPr and His5-Nτ (Fig. 3C and SI Appendix, Fig. S31 B, Right). The key NOE correlation between His5-H4 (δH 7.36) and ribose’-H1 (δH 6.5) provides additional support for modification at the Nτ position of His5 (Fig. 3C and SI Appendix, Fig. S31C). Aqueous solutions of minviopeptin containing 0.1 M TFA, 0.1 M formic acid, or 0.1 M NaOH were incubated overnight at room temperature. LC-MS analysis revealed no detectable cleavage of the N-glycosidic linkage under the acidic conditions, and approximately 60% of the peptide remained intact following NaOH treatment (SI Appendix, Fig. S33). These results indicate that the N-ADP-ribosylated linkage in minviopeptin is relatively stable. Although spectral overlap precluded unambiguous assignment of the terminal amide carbon, MS/MS fragmentation strongly supported the presence of a C-terminal amidation. Collectively, these data define minviopeptin as a triceptide bearing three distinct modifications: i) Tyr1-Lys3 and His5-Ser7 cyclophanes formation, ii) His5-linked ADP-ribosylation, and iii) C-terminal amidation. This constitutes a structurally characterized ribosomal peptide featuring ADP-ribosylation.

Crystal Structure of MicC Revealed a Conserved Fold of αKG-NHFe Enzyme.

To gain mechanistic insight into the MicC-catalyzed reaction, we determined its crystal structure at 2.20 Å resolution (PDB ID: 9VKM). The structure belongs to the P22₁2₁ space group and contains two monomers in the asymmetric unit (SI Appendix, Table S10). MicC adopts a canonical double-stranded β-helix (DSBH) fold, with a conserved 2-His-1-Asp facial triad (H160-D162-H253) positioned at the open end of the DSBH core (Fig. 4A), a characteristic commonly found in αKG-NHFe enzymes (41–44). Structure-based similarity searches with the Dali server identified kanamycin B dioxygenase KanJ (PDB: 7CL4) (45) as the closest structural homolog, with an RMSD of 3.6 Å across 170 Cα atoms despite only 13.9% sequence identity. Phylogenetic analysis revealed that MicC does not group with previously characterized RiPP-associated αKG-NHFe enzymes, but instead clusters with MovX (Fig. 4B), a multinuclear non-heme iron dependent oxidative enzyme (MNIO) that catalyzes N–Cα bond cleavage to yield a C-terminal amide (46), suggesting possible convergent evolution between αKG-NHFes and MNIOs toward similar amidation chemistry.

Fig. 4.

Structural and phylogenetic analysis of MicC and substrate specificity analysis of MicC and MicD. (A) Overview of the MicC structure. The experimental electron density (2F_o–F_c) contoured to 3.0 σ around the citrate is represented as gray mesh. The terminal carboxylate of citrate is positioned within 5 Å of the 2-His-1-Asp facial triad residues. (B) Phylogenetic analysis of MicC with representative αKG-NHFe enzymes. Two α-hydroxylating proteins, MovX (a multinuclear nonheme iron-dependent oxidase) and CrPHM (the peptidylglycine α-hydroxylating monooxygenase domain of the peptidylglycine α-amidating monooxygenase from Chlamydomonas reinhardtii), were included in the analysis. The sequences used in this analysis are listed in SI Appendix, Table S7. (C) In vitro analysis of MicD with different cosubstrates. Conversion rates are normalized to NAD⁺ as the cosubstrate. (D) In vitro modification of peptide 2 by MicD. (E) In vitro modification of peptide 3 by MicC.

Notably, a citrate molecule was observed in the active-site cavity, positioned adjacent to the putative Fe²⁺-binding 2-His-1-Asp facial triad (Fig. 4A). This citrate likely acts as a structural surrogate for αKG, occupying its canonical binding pocket. To further delineate the residues critical for αKG recognition, we removed citrate from the structure and performed molecular docking with αKG. The resulting model showed αKG is situated near the inner end of a largely hydrophobic pocket. Its C5 carboxylate, oriented distal to the catalytic triad, is anchored by the side chains of Arg188, Phe190, Tyr244, and Lys266, among which Arg188 forms a bidentate ionic contact with both C5-carboxylate oxygens (SI Appendix, Fig. S34). Mutagenesis studies were performed by replacing R188, F190, Y244, and K266 to Ala, respectively (SI Appendix, Fig. S35). The results showed that R188A and K266A variants were catalytically inactive, indicating the critical roles of these basic residues in mediating electrostatic interactions with αKG. Such salt-bridge contact has also been observed in other αKG-NHFe enzymes (47). In contrast, Y244A had little effect on enzyme activity. The F190 A variant resulted in insoluble protein, implying a possible role for proper protein folding. Notably, alanine substitution of each residue in the 2-His-1-Asp facial triad (H160A, D162A, and H253A) led to complete loss of activity (SI Appendix, Fig. S36), demonstrating its strictly essential role in enzyme catalysis.

Cosubstrate Binding and Substrate Specificity of MicD.

Efforts to solve the crystal structure of MicD were unsuccessful despite extensive trials. We hence employed AlphaFold3 to generate two structural models: (i) MicD bound to NAD⁺ (SI Appendix, Fig. S37), and (ii) MicD in complex with both NAD⁺ and the MicA (SI Appendix, Fig. S38). These analyses showed the MicD adopts a typical ART fold characterized by a split β-sheet with each half containing three strands (48, 49), and the NAD⁺ cosubstrate was likely bound to four residues (His7, Asp11, Glu101, and Phe41). Moreover, the Phe₃₀-X₁₀-Tyr₄₁ arrangement in MicD is comparable to an aromatic stacking motif reported in other ARTs, where two aromatic residues (Tyr-X₁₀-Tyr) interact with nicotinamide via π–π stacking (50). To evaluate the functional importance of these residues, we generated individual alanine substitutions and analyzed their biochemical properties (SI Appendix, Fig. S39). The F41A and Y30A variants showed markedly reduced soluble expression, suggesting their crucial role in maintaining protein stability. While the D11A variant maintained wild-type (WT) catalytic activity, both H7A and E101A substitutions completely abolished ADP-ribosylation activity (SI Appendix, Fig. S40), demonstrating their essential catalytic roles in MicD activity.

To investigate the cosubstrate specificity of MicD, we performed in vitro enzymatic assays using NAD⁺ analogs (NADH, NADPH, and NADP⁺). The result showed that only NADH supported the MicD-catalyzed ADP-ribosylation, showing an approximately 30% efficiency relative to NAD⁺ (Fig. 4C and SI Appendix, Fig. S41). This finding reveals an unexpected flexibility in MicD cosubstrate recognition, as canonical ARTs (e.g., Arr) exhibit strict selectivity for NAD⁺ (24). The ability of MicD to utilize NADH, though with reduced efficiency, demonstrates a unique promiscuity among characterized ARTs.

In the AlphaFold3 model of the MicD in complex with NAD⁺ and MicA, the MicA leader peptide shows only peripheral association with MicD (SI Appendix, Fig. S38). This observation, combined with the absence of discernible RiPP recognition element (RRE) motif in MicD, suggested the leader peptide might be nonessential for catalysis. To test this hypothesis, we performed in vitro assays using 2 (the GluC-digested core fragment of MicA*^B). Remarkably, incubation of 2 with MicD and NAD⁺ resulted in near-quantitative ADP-ribosylation (Fig. 4D and SI Appendix, Fig. S42), demonstrating that the leader peptide is dispensable for MicD activity. Since MicC-catalyzed peptide cleavage occurs downstream of MicD-mediated ADP-ribosylation, we hypothesized that MicC activity, like MicD, would be independent of the leader peptide. Consistent with this, in vitro assays using 3 (the GluC-digested core fragment of MicA*^BD) demonstrated complete processing by MicC (Fig. 4E and SI Appendix, Fig. S43). These results collectively establish that neither MicD nor MicC requires the leader peptide for catalytic activity.

As discussed above, MicB-catalyzed cyclophane formation is essential for subsequent MicD activity. To identify which specific cyclophane structure(s) are required for MicD recognition, we generated three core peptide mutants (S2A, S7A, F10A) designed to disrupt individual cyclophane rings. Characterization of these variants revealed distinct cyclization patterns: The S2A (SI Appendix, Fig. S44) and F10A (SI Appendix, Fig. S45) mutants each retained only one cyclophane (at the F₁₀T₁₁K₁₂ and Y₁S₂K₃ motifs, respectively), while the S7A mutant maintained two cyclophanes (at Y₁S₂K₃ and F₁₀T₁₁K₁₂) (SI Appendix, Fig. S46). This hierarchical cyclization pattern resembles the regioselectivity of ChlB (51, 52) but differs from other 3-CyFEs like PauB (53) and WprB (54), highlighting mechanistic diversity in triceptide maturation. Most significantly, none of these partially cyclized intermediates were ADP-ribosylated by MicD, demonstrating that full installation of all three crosslinks by MicB is required to generate a competent substrate for MicD.

Discussion

In this study, we uncover an RiPP biosynthetic pathway that introduces two unique enzymatic transformations. The mic cluster encodes a sequential modification cascade initiated by the radical SAM enzyme MicB, which installs three cyclophanes within the core peptide. This rigidified scaffold then serves as the substrate for MicD, an ART that modifies a histidine residue within one of the macrocycles. Functioning as an integral middle step in minviopeptin biosynthesis, MicD-catalyzed ADP-ribosylation represents a biosynthetic modification, distinct from the Arr-mediated resistance mechanism in rifamycin where ADP-ribosylation evolved as an exogenous detoxification strategy. After ADP-ribosylation, the minviopeptin biosynthetic pathway proceeds through MicC, an αKG-NHFe enzyme that catalyzes oxidative cleavage of the glycine N–Cα bond, ultimately generating a C-terminally amidated peptide product (SI Appendix, Fig. S47).

In attempts to obtain the mature product of the mic pathway, we incorporated an ABC transporter MicT encoded within the BGC into the MicABCD coexpression system, as sequence analysis suggested that this protein contains a protease domain. However, LC-MS and LC-MS/MS analysis of both culture supernatants and cell pellets, together with molecular networking analysis using the Global Natural Products Social Molecular Networking (GNPS) platform (55), failed to detect any fragments that match to the MicA-derived peptide. Moreover, Ni-affinity purification from the MicAT, MicABT, MicABDT, and MicABCDT expression combinations consistently yielded only the full-length precursor peptides, without detectable truncated forms (SI Appendix, Fig. S48). These observations suggest that MicT is either nonfunctional in our E. coli heterologous expression system or does not serve as the maturation enzyme in the mic pathway.

Despite extensive biological profiling, no discernible inhibitory activity was detected for minviopeptin against a panel of test organisms (six Gram-positive and five Gram-negative bacterial strains, along with one fungal species) or in cytotoxicity assays against five human cancer cell lines (SI Appendix, Fig. S49 and Table S8). Moreover, no anti-inflammatory activity was observed based on nitric oxide (NO) analysis in compound-treated RAW264.7 macrophages (SI Appendix, Table S9). This observed inactivity may stem from the limitations of our heterologous production system, as native host peptidases could process the metabolite into alternative bioactive forms. Interestingly, comparison of the antibacterial activities of compound 2 and minviopeptin showed that only the cyclized but non-ADP-ribosylated 2 displayed weak activity against Acinetobacter baumannii, Escherichia coli, and Micrococcus luteus (SI Appendix, Table S12). These observations suggest that ADP-ribosylation may attenuate antibacterial activity, raising the possibility that this modification could contribute to a resistance-related or self-protection mechanism. Alternatively, ADP-ribosylation may itself confer specific ecological functions, such as modulating interaction with specific bacterial receptors or enhancing environmental persistence. Distinguishing between these possibilities warrants future investigation in a native or ecologically relevant context.

MicD catalyzes a rare histidine-targeted ADP-ribosylation. Although ADP-ribosylation has been reported on nearly all nucleophilic amino acids (e.g. Cys, Asp, Glu, Lys, Arg, Ser, Thr, and Tyr) (49, 56), histidine remains an uncommon target. The known example involves bacterial toxins, including DT and exotoxin A, which transfer the ADPr moiety to the imidazole nitrogen of diphthamide (a modified histidine) located at translation elongation factor 2 (eEF2) (57, 58). Notably, in contrast to DT ART and other known ARTs, which are typically larger than 40 kDa, MicD is unusually small, with a molecular weight of only ~23 kDa.

MicC catalyzes the cleavage of the glycine N–Cα bond to liberate the C-terminally amidated peptide, a transformation that is reminiscent of the reactions catalyzed by peptidylglycine α-amidating monooxygenase (PAM) (59–61) and multinuclear nonheme iron-dependent oxidative enzyme (MNIO) MovX (46). PAM uses distinct catalytic domains to catalyze the amidation reaction in two steps. The first involves peptidylglycine α-hydroxylating monooxygenase (PHM) domain which relies on copper and ascorbate to hydroxylate the α-carbon of C-terminal glycine. The second step is catalyzed by the peptidyl-α-hydroxyglycine α-amidating lyase (PAL) domain, a zinc-dependent lyase, which mediates an N-dealkylation reaction to yield an amidated peptide and glyoxylate (59, 62, 63) (SI Appendix, Fig. S50B). In contrast, MovX contains a single catalytic domain coupled with a C-terminal extension that may facilitate substrate recruitment (46). This enzyme was shown to act on asparagine by cleaving the N–Cα bond, likely involving a key Fe(III)-superoxo intermediate (46) (SI Appendix, Fig. S50A). Moreover, several nonribosomal peptides, such as skyllamycin (64), nyuzenamide (65), melithiazol (66), and dolyemycin (67) utilize flavin- or F₄₂₀-dependent NRPS domains or post-NRPS enzymes to achieve similar reactions. A mechanism was proposed for MicC (SI Appendix, Fig. S51), which involves hydroxylation of the glycine Cα via hydroxyl radical rebound, a process common in αKG-NHFe enzymes (41, 68).

Bioinformatic analysis using MicC as a query revealed its frequent genomic association with radical SAM enzymes in Actinomycetota and Bacteroidota BGCs (SI Appendix, Fig. S52). These clusters often encode diverse auxiliary enzymes, including transferases, αKG-HExxH proteins (51), ATP-grasp enzymes, and lanthipeptide dehydratases, suggesting elaborate PTM cascades. Notably, some BGCs contain one or multiple MicC homologs alongside radical SAM enzymes, hinting at functional diversification beyond the oxidative peptide fragmentation as observed for MicC (SI Appendix, Fig. S53). This finding holds particular significance given the established role of C-terminal amidation in modulating peptide properties, from membrane interaction to structural stability and bioactivity. Our finding of the minviopeptin pathway, with its unique ADP-ribosylation and oxidative cleavage steps, establishes a paradigm for exploring unconventional RiPP modifications and expands the toolbox for peptide engineering.

Materials and Methods

Detailed information on instrumental settings, culture conditions, gene cloning, mutant construction, protein expression, and purification is provided in SI Appendix.

Bioinformatics.

We employed the previously characterized radical SAM enzyme PasB as a query sequence for SSN analysis. The analysis was conducted using the Enzyme Function Initiative-Enzyme Similarity Tool (EFI-EST) with a minimum alignment score of 110 and an E-value cutoff of –5 (28, 29). This analysis generated a network comprising 3,110 nodes, which was then subjected to genome neighborhood analysis using EFI-GNT to examine the ± 20 open reading frames (ORFs) flanking each target gene. A distinct cluster of 3-CyFEs was observed, within which a gene cluster from M. inviolabilis was found to encode a radical SAM enzyme (MicB) colocalized with an ART (MicD) and an uncharacterized protein (MicC).

To uncover additional 3-CyFE-RiPP BGCs encoding MicD-like proteins, position-specific iterative BLAST (PSI-BLAST) was performed for MicD against the NCBI RefSeq protein database over three iterations to retrieve homologous sequences. The resulting hits were analyzed using the RODEO web server to obtain the genome neighborhood (69). Manual inspection of the identified gene clusters was performed to single out entries including small open reading frames (sORFs) with Ω1-X2-X3 (Ω1 = Phe, Trp, His, and Tyr) motif and proximal radical SAM enzymes. This analysis resulted in four orthologous clusters and corresponding precursor peptides were extracted from the RODEO output files.

To expand the repertoire of radical SAM enzyme–αKG-NHFe enzyme homolog pairs in other BGCs, PSI-BLAST searches were performed using MicC as the query sequence. Six iterative rounds of PSI-BLAST were executed with a threshold cutoff of 0.005, retrieving up to 2,000 sequences per iteration. The resulting hits were subsequently analyzed through the RODEO web platform. For each candidate locus, the ORFs flanking the target gene were manually inspected, and only BGCs that simultaneously encoded a small ORF, a radical SAM enzyme, and a MicC homolog were retained. The resulting MicC homolog sequences were subjected to SSN analysis with a minimum alignment score of 56 and an E-value cutoff of –5. The network visualization and annotation were performed in Cytoscape (70), with the nodes colored according to the genus.

Production of Minviopeptin (compound 4).

Constructed plasmids pRSF_His₆-micA^*C and pCDF_micBD were cotransformed into E. coli BL21 Star (DE3) cells to express the peptide product. The expression and purification of the modified peptide were performed following the procedure described in SI Appendix, with the culture volume scaled up to 25 L. The peptide MicA*^BCD, obtained in a yield of over 100 mg, was subsequently digested on a large-scale using GluC. After digestion, the solutions were pooled and lyophilized to dryness. The concentrated digests were resuspended in Milli-Q water and injected into a reverse-phase Elite SinoChrom ODS-AP C18 HPLC column (10 μm, 9.4 × 250 mm) embedded in an Agilent 1260 Infinity II LC system. A gradient elution operating at a flow rate of 2.4 mL/min was used under the following condition: t = 0 min, 2% B; t = 4 min, 2% B; t = 30 min, 12.6% B; t = 30.1 min, 98% B; t = 36 min, 98% B; t = 36.1 min, 2% B; t = 40 min, 2% B (A = distilled water with 0.1% TFA, B = acetonitrile). The UV absorbance was monitored at 220 nm and 254 nm. Fractions containing minviopeptin were pooled, then concentrated by rotary evaporation and lyophilized to dryness. The obtained white powder was used for further NMR analysis or stored at −80 °C for further use.

In Vitro Assay for MicC and its Mutants.

For the in vitro reconstitution of MicC and its mutant activity, peptide substrates used for enzymatic reactions, including MicA*, MicA*^B, and MicA*^BD, were prepared as described in SI Appendix. All reactions in this study were performed as follows, unless noted otherwise. Reactions were carried out in a 100 μL scale with 50 mM Tris (pH 8.0), 50 μM peptide substrate, 10 μM WT or mutant MicC, 0.2 mM (NH₄)₂Fe(SO₄)₂·6H₂O, 1 mM αKG, and 2 mM ascorbate. To assess the contribution of each component, control reactions were conducted by systematically excluding one component from the full reaction setup, except the buffer. Enzyme reactions were incubated overnight at 25 °C with shaking, followed by heat quenching at 85 °C for 3 min. After centrifugation to remove any precipitate, the reaction mixtures were either directly analyzed by LC-MS or subjected to proteolytic digestion before MS/MS analysis.

In Vitro Assay for MicD and its Mutants.

For in vitro reconstitution of MicD activity, peptide substrates used for enzymatic reactions, including MicA* and MicA*^B, were prepared as described in SI Appendix. All reactions in this study were performed as follows, unless noted otherwise. Reactions were carried out in a 100 μL scale with 50 mM HEPES (pH 7.5), 20 μM peptide substrate, 5 μM MicD, and 2.5 mM NAD⁺. Control reactions were conducted under the same conditions, using either boiled MicD or in the absence of NAD⁺. After overnight incubation at room temperature, reactions were heated at 85 °C for 3 min. Precipitates were removed by centrifugation. The reaction mixtures were then either directly analyzed by LC-MS or subjected to proteolytic digestion before MS/MS analysis.

For MicD mutant activity assays, the reactions were performed under the same conditions as described for the WT protein.

Marfey’s Analysis.

The absolute configuration for virtually all constituent amino acids in minviopeptin was determined by advanced Marfey’s analysis (71). Dried 0.1 mg of compound 4 was hydrolyzed in 6 N HCl (1 mL), additionally containing 0.4% (v/v) β-mercaptoethanol in a glass vial and heated at 110 °C in an oil-bath for 16 h. After concentration by rotary evaporation, the crude hydrolysate was dissolved in 100 μL of water and subsequently divided into two equal portions.

For each one, 1 M NaHCO₃ (20 μL) and 1% w/v of L-FDAA (Marfey’s reagent, Nα-(2,4-dinitro-5-fluorophenyl)-alaninamide) dissolved in acetone (100 μL) was added and incubated at 40 °C for 1 h. The reaction mixtures were then quenched by the addition of 1 M HCl (20 μL) and diluted twofold with acetonitrile. After centrifugation to remove any precipitate, the derivatives were analyzed by LC-MS. Separations were carried out on an Agilent Eclipse XDS-C18 column (5 µm, 150 × 4.6 mm) attached to an Agilent 6546 LC/Q-TOF system. For Ile analysis, a gradient elution operating at a flow rate of 0.4 mL/min was used under the following condition: t = 0 min, 20% B; t = 3 min, 20% B; t = 30 min, 65% B; t = 30.1 min, 95% B; t = 35 min, 95% B; t = 35.1 min, 20% B; t = 38 min, 20% B (A = distilled water with 0.1% TFA, B = acetonitrile). For Ser analysis, elution was performed with an isocratic step of 10% acetonitrile in water (with 0.1% TFA) for 3 min, followed by a gradient step of 10 to 47% acetonitrile in water (with 0.1% TFA) over 27 min. For His, Tyr, Ala, and Lys analysis, elution was performed with an isocratic step of 5% acetonitrile in water (with 0.1% TFA) for 3 min, followed by a gradient step of 5 to 60% acetonitrile in water (with 0.1% TFA) over 27 min. The UV absorbance was monitored at 340 nm. The standards of L-amino acid or D-amino acid were subjected to analogous derivatization and analysis conditions.

NMR Analysis of Minviopeptin.

5 mg sample was dissolved in 600 μL of DMSO-d₆ containing 0.03% v/v TMS (99.9 atom %, Cambridge Isotope Laboratories), and subsequently transferred into an NMR tube. One and two-dimensional NMR data including ¹H, ¹³C, ¹H-¹H COSY, ¹H-¹H TOCSY, ¹H-¹H NOESY, ¹H-¹³C HSQC, and ¹H-¹³C HMBC were acquired on a Bruker Avance NEO 600 MHz spectrometer using a 5 mm Prodigy BBO cryoprobe at 298 K. Additionally, ³¹P NMR spectra were obtained on a Bruker AVANCE III HD 400 MHz using a Prodigy TCI™ cryoprobe under the same temperature conditions. All spectral data were processed using MestReNova, and the chemical shift assignments derived from the full set of 1D and 2D experiments are summarized in SI Appendix, Table S5.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix.