Post-translational modification of ribosomally synthesized peptides by a radical SAM epimerase in Bacillus subtilis

Abstract

Ribosomally-synthesized peptides are built out of l-amino acids while d-amino acids are generally the hallmark of nonribosomal synthetic processes. Here we show that the model bacterium Bacillus subtilis is able to produce a novel type of ribosomally synthesized and post-translationally modified peptide containing d-amino acids that we propose to call epipeptides. We demonstrate that a two [4Fe-4S] clusters radical SAM enzyme converts l-amino acids into their d-counterparts by catalyzing C_α H-atom abstraction and using a critical cysteine residue as H-atom donor. Unexpectedly, these d-amino acid residues proved to be essential for the bioactivity of a peptide inducing the expression of LiaRS, a major component of the bacterial cell envelope stress response system. Present in B. subtilis and in several members of the human microbiome, these epipeptides and radical SAM epimerases broaden the landscape of peptidyl structures accessible to living organisms.

Bacterial cell envelop integrity is critical for bacteria survival and adaptation to the environment. Consequently, bacteria have developed several systems to sense membrane alterations. In Bacillus subtilis, a major cell envelope stress module is the Lia system composed of a two-component system (LiaRS) and an inhibitory protein (LiaF)^1,2. This genetic system is highly conserved among Firmicutes and part of the complex regulatory network orchestrating the cell wall stress response^1,2. Although its regulation has been described in great details, its precise physiological role in B. subtilis is not fully understood. LiaRS is specifically and strongly induced by antibiotics targeting the cell wall such as nisin, vancomycin or bacitracin³ and has been developed as a biosensor and high-throughput screen⁴. Upon antibiotic sensing, LiaRS transduces cell envelop stress signals activating gene expression presumably to maintain cell wall integrity. However, LiaRS does not appear to confer antibiotic resistance.

In an attempt to identify genes involved in LiaRS regulation, a mutagenesis study was undertaken in B. subtilis leading to the discovery of the yydFGHIJ operon⁵. This operon shows positive regulation on LiaRS and possesses all the characteristic features of a genetic system encoding a putative peptide (YydF) modified by a radical SAM enzyme (YydG) and a protease (YydH). The modified peptide is then predicted to be exported in the extracellular medium by an ABC-type transporter (YydIJ) even though none of these components has been isolated or investigated so far (Fig. 1a). Among these five genes, the function of the putative radical SAM enzyme (YydG) was unclear and predicted to be a peptide modifying enzyme⁵. However, it does not possess significant homology with any known radical SAM enzyme.

Figure 1. YydG is a radical SAM enzyme catalyzing the modification of the YydF peptide.

(a) Structure of the yydFGHIJ operon. yydF: putative peptide, yydG: radical SAM enzyme, yydH: protease, yydIJ ABC-type transporter. (b) Gel electrophoresis analysis of purified YydG expressed in E. coli. (c) UV-visible spectrum of anaerobically reconstituted YydG (blue line) and reduced with sodium dithionite (red line). The symbol * indicates absorbance due to reduced sodium dithionite. (d) Sequence of the YydF peptide from B. subtilis. In grey, region with a low conservation; in blue, strictly conserved amino acid residues (see Supplementary Fig. S2). (e) HPLC analysis of YydF_18-49 incubated with YydG after 90 min under anaerobic conditions in the absence (upper trace) or the presence (lower trace) of sodium dithionite as one-electron donor. In the absence of sodium dithionite (upper trace), only the YydF_18-49 peptide substrate was monitored. The m/z of each peptide is indicated above the corresponding peaks. See Supplementary Information for experimental conditions. (f) HPLC analysis of SAM incubated with YydG and YydF_18-49 after 90 min under anaerobic conditions in the absence (upper trace) or in the presence (lower trace) of sodium dithionite. See Supplementary Information for experimental conditions.

Radical SAM enzymes are emerging as a novel superfamily of enzymes catalyzing a broad range of radical-based reactions, many of them having no precedent in biology^6–8. They are characterized by the presence of at least one [4Fe-4S] center which binds S-adenosyl-l-methionine (SAM) in a bidendate fashion via its carboxylate and amino moities^9,10. Recent studies have highlighted the structural diversity of these enzymes which often contain additional [4Fe-4S] centers and/or cofactors such as PLP¹¹ or cobalamin¹². Upon reduction, the radical SAM [4Fe-4S]¹⁺ center transfers one electron to SAM resulting in its reductive fragmentation¹³ and usually, to the generation of the 5'-deoxyadenosyl radical (5'-dA^•), although this species has never been characterized¹⁴. The 5'-dA^• then abstracts a substrate H-atom and, in most of the cases, leads to the formation of one equivalent of 5'-deoxyadenosine (5'-dA), although few enzymes, including lysine 2,3-aminomutase¹⁵ and spore photoproduct lyase^16–18, have been reported to recycle SAM.

Radical SAM enzymes have been shown to be involved in a wide range of protein and peptide post-translational modifications including the formation of glycyl radicals¹⁹, oxidation^20–23, methylthio transfer^24,25 or thioether bond formation^26,27. It is hence of no surprise if recent studies have highlighted their broad distribution in the biosynthetic pathways of the so-called RiPPs (ribosomally synthesized and post-translationally modified peptides)²⁸, where they have been proposed to catalyze unprecedented reactions such as unusual methyl transfer^29,30 or epimerization reactions²⁹.

Results and Discussion

YydG is a two [4Fe-4S] cluster radical SAM enzyme catalyzing peptide post-translational modifications

To investigate the catalytic function of the putative radical SAM enzyme YydG and the biological role of the YydF peptide, we expressed the protein in E. coli and assayed its activity against YydF. The purified and anaerobically reconstituted protein had the distinctive spectroscopic properties of radical SAM enzymes with charge transfer absorption bands at 320 and 420 nm (Fig. 1b & c). Iron-sulfur determination indicated that, after anaerobic reconstitution, YydG contained 7.3± 0.1 Fe and 7.7± 0.5 sulfide per polypeptide chain, in agreement with the UV-visible spectrum of the protein (Fig. 1c & Supplementary Fig. S1). These results supported that YydG contains two [4Fe-4S]²⁺ centers. Anaerobic reduction of the reconstituted protein by sodium dithionite resulted in ~20% decreased intensity in the 350-700 nm region (Fig. 1c), similarly to what has been reported for the reduction of [4Fe-4S]²⁺ to [4Fe-4S]¹⁺ cluster in other radical SAM enzymes³¹. As expected, exposure to atmospheric oxygen resulted in cluster re-oxidation as shown by the increase of the absorbance in the 350-700 nm region (Supplementary Fig. S1).

Genome mining revealed that yydF and yydG genes are also present in Gram-positive pathogens including Enterococcus faecalis and several Staphylococcus and Streptococcus species (Supplementary Fig. S2). Sequence alignment of selected YydF homologs indicated a highly conserved region from the residues 23 to the C-terminus end of the peptide (Fig. 1d and Supplementary Fig. S2). To probe the activity of YydG, we assayed the reconstituted enzyme with the full-length YydF peptide or a truncated form, encompassing the conserved amino acid residues from position 18 to 49 (peptide YydF_18-49). As shown, under anaerobic and reducing conditions, YydG converted ~30% of the YydF peptide (Supplementary Fig. S3) and >80% of the YydF_18-49 peptide (Fig. 1e) into several peptide products: YydF_a, YydF_b and YydF_c eluting at 41.7, 42.2 and 42.8 min, respectively. Concomitantly, SAM was converted into 5'-deoxyadenosine (5'-dA) during the reaction (Fig. 1f). As expected, if we omitted the enzyme or sodium dithionite, no peptide modification was observed, demonstrating the radical nature of the reaction (Supplementary Fig. S4). Interestingly, we also evidenced that YydG co-purified with large amounts of SAM as shown by HPLC analysis and SAM cleavage assay (Supplementary Fig. S5). Altogether, these results established that YydG is a two [4Fe-4S] centers radical SAM enzyme catalyzing the modification of the YydF and YydF_18-49 peptides.

Because YydF_18-49 was a more soluble substrate, we used this truncated peptide to characterize the modification catalyzed by YydG. Mass spectrometry inspection of the three peptides formed (YydF_a, YydF_b and YydF_c) revealed no mass difference compared to the substrate (YydF_18-49 [M+3H]³⁺ = 1258.3) (Fig. 1e). This result was consistent with previous report on radical SAM enzymes catalyzing isomerization¹⁵, substrate rearrangement^17,18 or epimerization²⁹. Tryptic peptide mapping of the substrate (i.e. YydF_18-49) gave three major products (Peptide 1 [M+H]⁺ = 931.7, Peptide 2 [M +2H]²⁺ = 592.6 and Peptide 3 [M+H]⁺ = 768.6) eluting at 22, 27 and 19.2 min respectively (Fig. 2a). Comparison with the enzymatically modified peptide showed the formation of two additional peptides (named Peptide 2* and Peptide 3*) having the same molecular weight as Peptides 2 and 3 but eluting at 26.5 and 23.5 min, respectively (Fig. 2b). This result supported that YydG had introduced two modifications one located internally (in Peptide 2) and one in the C-terminus end of the peptide (in Peptide 3).

Figure 2. YydG catalyzes H-atom transfer to the peptide backbone.

Tryptic peptide mapping and LC-MS analysis of (a) YydF_18-49 or (b) YydF_18-49 after incubation with YydG. Numbers indicate the m/z value for each peptide. Sequences in green indicate the relevant peptide identified by LC-MS (i.e. Peptide 1: Ac-GLLDESQK, [M+H]⁺ = 931.7 ; Peptide 2: VNDLWYFVK [M+2H]²⁺ =592.6 and Peptide 3: WILGSGH-NH₂, [M+H]⁺ =768.6). Sequences in pink represent the peptides modified by YydG. (c) LC-MS analysis of the peptide YydF_18-49 after incubation with YydG in deuterated buffer. (d) Tryptic peptide mapping and LC-MS analysis of YydF_18-49 after incubation with YydG in deuterated buffer. (e) LC-MS/MS analysis of Peptide 2 (upper left panel), Peptide 2* (lower left panel), Peptide 3 (upper right panel) and Peptide 3* (lower right panel) obtained after tryptic hydrolysis of YydF_18-49 incubated with YydG in deuterated buffer (see full assignment in Supplementary Table S2-5).

To identify the nature and position of the modifications catalyzed by YydG, we repeated the reaction in >90% deuterated buffer since radical SAM enzymes are known to abstract but also, in some cases, to exchange H-atoms during catalysis. In deuterated buffer, YydG produced a similar pattern of products (i.e. YydF_a, YydF_b & YydF_c) with YydF_c being the most abundant and YydF_a the minor product (Fig. 2c). LC-MS analysis of the reaction showed that under these conditions, YydF_a and YydF_c had a molecular weight of [M+3H]³⁺ = 1258.6 and YydF_b a molecular weight of [M+3H]³⁺ = 1259.0 corresponding to one and two Dalton units more than the YydF_18-49 substrate ([M+3H]³⁺ = 1258.3), respectively. This result unambiguously demonstrated that one deuterium atom was incorporated into YydF_a and YydF_c while two deuterium atoms were added into YydF_b. Tryptic peptide mapping allowed to localize deuterium incorporation exclusively in Peptide 2* and Peptide 3* whose molecular masses shifted by one Dalton unit (i.e. [M+2H]²⁺ =593.0 and [M+H]⁺ = 769.6, respectively) (Fig. 2d). LC-MS/MS fragmentation of these two peptides showed that one deuterium atom was incorporated in Val₃₆ (as proved by the characteristic ions: y₁, y₂+1 and b₇, b₈+1) and one in Ile₄₄ (as shown by identification of the ions: y₅ and y₆+1) (Fig. 2e) (see Supplementary Fig. S6-7 and Supplementary Tables S1-5 for full assignment). Altogether these results demonstrated that YydG catalyzes the replacement of two H-atoms located in Val₃₆ and Ile₄₄ by two solvent exchangeable H-atoms during catalysis.

YydG is a novel type of radical SAM epimerase

To determine the nature of the modification, we performed acid hydrolysis of the enzymatically modified peptides and analyzed the amino acid content by LC-MS/MS after derivatization with N-α-(2,4-dinitro-5-fluorophenyl)-l-valinamide (l-FDVA) (see Supplementary methods). The YydF_18-49 peptide contains one Ile and five Leu residues (Fig. 1d) which not only have the same molecular weight than Ile but also eluted at similar retention times. As shown on Figure 3a & b, optimized LC-MS/MS conditions allowed the separation and characterization of l-Ile, l-Leu and l-Val but also of their d-configured counterparts (d-allo-Ile, d-Leu and d-Val) as diastereoisomers. Analysis of the enzymatically modified peptides showed that, in addition to l-Ile and l-Leu, another product eluting at 27.7 min was formed corresponding to d-allo-Ile as shown by its retention time, its molecular weight and the formation of a major ion fragment at m/z= 366. 2 in H₂O buffer and m/z= 367.2 in D₂O buffer (Fig. 3a & c). The d-allo-Ile/l-Ile ratio ranged from 20% in deuterated buffer (Fig. 3a) to 70% under standard assay conditions (Supplementary Fig. S8). Among the other amino acids, we only identified valine as a mixture of l- and d-amino acids (Fig. 3b). To definitely assert the nature of the modification, we synthesized a peptide containing one d-Val and one d-allo-Ile at positions 36 and 44, respectively. Tryptic peptide mapping and amino acid analysis of this synthetic peptide perfectly reproduced the ones obtained with the enzymatically modified peptide (Supplementary Fig. S9-11). We hence demonstrated that YydG is a radical SAM peptide epimerase, the first one shown to be active in vitro on a peptide backbone.

Figure 3. YydG catalyzes amino acid epimerization.

LC-MS/MS analysis of (a) L-/D-Leu and L-Ile/D-allo-Ile or (b) L-/D-Val. Upper traces were obtained with standard amino acids while lower traces correspond to amino acids obtained after incubation of YydF_18-49 with the radical SAM enzyme YydG in deuterated buffer. Amino acids were analyzed after hydrolysis and derivatization by N-α-(2,4-dinitro-5-fluorophenyl)-L-valinamide (L-FDVA). MS spectra of (c) L-Ile-FDVA, (d) L-Val-FDVA (upper traces) and their D-epimers (lower traces) obtained after incubation of YydF_18-49 with YydG in deuterated buffer. Mass spectra correspond to the dominant ions fragments: m/z= 366 and 352 for Ile- and Val-FDVA derivatives, respectively. The amino acids were derivatized by L-FDVA and detected by LC-MS after ion current extraction in MS/MS experiments using the transition 412>366 and 398>352 for Ile/Leu and Val-FDVA derivatives, respectively. (e) LC-MS/MS analysis of the deuterated peptide YydF_18-49-VD₈ incubated with YydG. After incubation with YydG, YydF_18-49-VD₈ ([M+2H]²⁺ = 1887.7) (trace 1) was converted in peptides containing one D-valine ([M+2H]²⁺ = 1887.2) (traces 2&3) or two D-Val [M+2H]²⁺ = 1886.7 (trace 4). In the sequences, the octadeuterated L-valine residues are labeled in light blue and the heptadeuterated D-valine in red. Masses indicate mono-isotopic ions. (f) LC-MS/MS analysis of 5'-dA produced by YydG incubated with YydF_18-49 (upper trace) or YydF_18-49-VD₈ (lower trace). (g) Time-course production of 5'-dA (blue symbol) and epimerized peptide (red symbol). See supplementary Information for experimental conditions.

YydG catalyzes C_α H-atom

Of major interest, when we analyzed by LC-MS/MS the d-amino acids produced in deuterated buffer, we monitored a mass increment of +1 Da only in d-allo-Ile and d-Val (Fig. 3c & d), supporting that the epimerization process resulted from the replacement of the C_α H-atom by a solvent exchangeable H-atom (Supplementary Fig. S10-11). Interestingly, the 5'-dA produced during the reaction in deuterated buffer contained no significant labeling, as shown by high-resolution LC-MS/MS (Supplementary Fig. S12). This result suggested that YydG abstracts a non-exchangeable H-atom and that a tight coupling occurred between SAM cleavage and peptide post-translational modification.

To further demonstrate that YydG catalyzes direct H-atom abstraction, we synthesized a peptide variant (YydF_18-49-VD₈) containing two perdeuterated valines in position 36 and 44. The replacement of l-Ile₄₄ by an l-Val residue mirrored what is found in several YydF variants (Supplementary Fig. S2) and simplified the labeling procedure. Incubation of YydG with this novel substrate ([M+2H]²⁺= 1887.7) led to the formation of 3 products with a mass loss of one ([M+2H]²⁺= 1887.2) or two ([M+2H]²⁺= 1886.7) Dalton units (Fig. 3e & Supplementary Fig. S13). This result definitively established that residues 36 and 44 are the targets of the enzyme. More importantly, the 5'-dA produced in the presence of YydF_18-49-VD₈ had a mass increment of +1 Da ([M+H]⁺ = 253.1) (Fig. 3f) demonstrating that YydG catalyzes direct C_α H-atom abstraction. Of note, the magnitude of the labeling (>85%) confirmed the tight coupling between SAM cleavage and H-atom abstraction. This result was in sharp contrast with other radical SAM enzymes catalyzing peptide post-translational modifications which exhibit strong uncoupling between 5'-dA production and peptide modification^22,27.

We further assayed the activity of YydG on synthetic peptides containing one or two epimerized amino acids. YydG was able to convert the mono-epimerized peptides but not the peptide containing two epimerized residues (Supplementary Fig. S14). This result proved that, contrary to the eukaryotic peptide epimerases³², the reaction catalyzed by YydG is irreversible. In addition, this showed that no other modification is catalyzed by YydG.

Based on these analyses, we were able to assign YydF_a as a peptide containing a d-Val in position 36, YydF_c as peptide containing a d-allo-Ile in position 44 and YydF_b as peptide containing a d-Val and a d-allo-Ile in positions 36 and 44, respectively. Hence, YydG produces a mixture of peptides containing either one or two modified amino acids, with Ile₄₄ being the favored target in vitro (Fig. 1e). Kinetic analysis of the reaction showed that YydG had an activity of 2.7 nmol.min^-1.mg^-1 and of 1.7 nmol.min^-1.mg^-1 for the production of 5'-dA and epimerized peptides, respectively, in line with its propensity to add 1 to 2 epimerizations during in vitro catalysis (Fig. 3g). A rate constant of 0.15 min^-1 was determined, similar to what has been measured for other radical SAM enzymes catalyzing peptide modifications⁸. Altogether, these results supported that YydG uses one molecule of SAM to epimerize one amino acid residue through a radical-based mechanism.

YydG defines a novel type of radical SAM epimerases present within the human microbiome

Recently, it has been shown that during the biosynthesis of the so-called proteusins, putative radical SAM enzymes catalyze the in vivo epimerization of amino acid residues^29,33. Beside the three cysteine residues from the radical SAM cluster, multiple sequence alignment failed to reveal any sequence homologies between YydG and the proteusin epimerases (Supplementary Fig. S15). Further analysis using the HHPred server showed that, in sharp contrast with known radical SAM enzymes catalyzing RiPP post-translational modifications, YydG is deprived of the recently identified RiPP precursor peptide recognition element (i.e. RRE or PqqD-like domain)³⁴ (Supplementary Fig. S16). The RRE domain has been shown to be a critical component for interaction between a vast array of enzymes and the leader peptides commonly encountered in RiPPs. Its absence in YydG is consistent with the unique ability of this enzyme to modify full length and truncated peptide substrates.

Another major difference between YydG and the proteusin epimerases is the presence in YydG of a putative SPASM-domain^35,36. This domain is likely involved in the coordination of the second [4Fe-4S] center as predicted by structural modeling (Supplementary Fig. S17). Multiple sequence alignment failed to evidence such additional cluster in proteusin epimerases (Supplementary Fig. S15). Interestingly, a similar SPASM-domain has been identified in the carbohydrate epimerase NeoN catalyzing the last step in the biosynthesis of the aminoglycoside neomycin B³⁷. To further probe the sequence and structural relationships between these enzymes, we performed a phylogenetic analysis^38–40 including all known radical SAM epimerases and key representative members from the radical SAM superfamily of enzymes (Fig. 4a). Interestingly, YydG did not cluster with other epimerases (i.e. carbohydrate or proteusin epimerases) but with a sub-group of radical SAM enzymes recently shown to catalyze thioether bond formation (i.e. AlbA, SkfB & ThnB). This result appears meaningful considering that all these enzymes are predicted to abstract a C_α H-atom during catalysis^26,27. Finally, Proteusins are large peptides (>150 amino acids) characterized by the presence of a nitrile hydratase-type leader peptide^29,33 while YydF is a typical bacterial peptide.

Figure 4. Activity of YydG mutants.

(a) Molecular phylogenetic analysis of radical SAM epimerases and representative members of the radical SAM enzyme superfamily: The evolutionary history was inferred by using the Maximum Likelihood method based on the JTT matrix-based mode³⁹. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test⁴⁰ (1000 replicates) are shown next to the branches. Initial tree(s) for the heuristic search were obtained by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value (see Supplementary Information for full methodology. (b) Gel electrophoresis analysis of the purified YydG mutants. Mutants AxxxAxxA (A3), C22A, C222A and C223A. (c) UV-visible spectra of A3 (blue trace), C22A (red trace), C222A (green trace) and C223A (purple trace) mutants after anaerobic reconstitution. (d) HPLC analysis of the reaction after incubation of YydF_18-49 and each mutant protein. (e) LC-MS analysis of the reaction catalyzed by the wild-type enzyme or the C223A mutant in the absence of DTT. YydG produced peptides at [M+3H]³⁺= 1258.3 while the C223A mutants produced additional peptides at [M+Na]⁺ = 603.2861, [M+2H]²⁺= 1646.3838 and [M+2H]²⁺= 1125.7 identified as I₄₄^Δ1-H₄₉, G₁₈-I₄₄^Δ30 and G₁₈-V₃₆^Δ30, respectively (see Supplementary Fig. S22-25 and Supplementary Table S7-12 for full analysis and assignment). (f) Radical fragmentation mechanism of the YydF_18-49 peptide leading to the production of the I₄₄^Δ1-H₄₉ and the G₁₈-I₄₄^Δ30. The theoretical masses for the hydrolytic products (I₄₄-H₄₉ & G₁₈-I₄₄) and the experimental masses measured (I₄₄^Δ1-H₄₉ & G₁₈-I₄₄^Δ30) are indicated (See Supplementary Table S7 for comparison with theoretical products).

Collectively, these data demonstrate that YydF is the member of a distinct family of RiPPs²⁸ that we propose to call Epipeptides. Blast search retrieved 216 sequences homologous to YydG (E-value >60) in Gram-positive bacteria mainly related to the human microbiota (Supplementary Table S6). In support with this observation, search in the Human Microbiome Project database allowed retrieving peptides sharing high sequence identity with YydF. We further built an YydG sequence similarity network and a genome neighborhood network using the EFI-EST server⁴¹. The genome neighborhood network showed a strong association between YydG homologs and a putative peptidase and the two sub-units of an ABC transporter (Supplementary Fig. S18). This association was further confirmed by KEGG gene cluster prediction which revealed perfect co-occurrence between the 5 genes of the yydFGHIJ operon across Bacillus, Staphylococcus, Corynebacterium and Streptococcus species.

YydG contains a critical H-atom donor

The last question which remained to be solved was the origin of the exchangeable H-atom introduced during catalysis. Indeed, the carbon-centered radical produced by YydG, after C_α H-atom abstraction, was unlikely to interact with a buffer component. This highly reactive species is presumed to be kept sealed within the enzyme active-site. Close inspection of the YydG sequence pointed out that, in addition to the three cysteine residues from the radical SAM motif, only six cysteines were present in the protein sequence (Supplementary Fig. S19). One cysteine residue was adjacent to the radical SAM motif (i.e. Cys₂₂), two cysteine residues were side-by-side in the sequence (Cys₂₂₂ and Cys₂₂₃) while the three other cysteine residues were reminiscent of motifs involved in the coordination of [4Fe-4S] centers³⁵. Homology modeling showed that the 5 cysteine residues, located in the C-terminal part of the protein, had a correct position for the coordination of a SPASM [4Fe-4S] center (Supplementary Fig. S17). To probe their function, we substituted Cys₂₂, Cys₂₂₂, Cys₂₂₃ and the three cysteine residues of the radical SAM motif (Cys₁₄, Cys₁₈ and Cys₂₁) by alanine residues. The four designed mutants (i.e.C22A, C222A, C223A and A3) were successfully purified although the C222A mutant proved to be recalcitrant to purification and was produced partly as a truncated form (Fig. 4b). Spectroscopic analysis and iron assay showed that, in agreement with our structural model, the A3 mutant contained after anaerobic reconstitution 4.6± 0.1 Fe and 5.7± 0.2 sulfide per polypeptide chain indicating the presence of one [4Fe-4S] center (Fig. 4c and Supplementary Fig. S17 & S20).

The UV-visible spectra of C22A and C223A mutants perfectly superimpose with the wild-type enzyme. The C22A mutant contained 8.6± 0.3 Fe and 7.9± 0.5 sulfide and the C223A mutant 8.2± 0.3 Fe and 7.1± 0.2 sulfide per polypeptide. In contrast, the C222A mutant did not contain detectable amounts of iron-sulfur center, even after anaerobic reconstitution (Fig. 4c & Supplementary Fig. S20). Furthermore, its absorption maximum was shifted toward 250 nm indicating that the protein was likely misfolded as reported for many radical SAM enzymes^35,37, when cysteine residues involved in [4Fe-4S] coordination are mutated.

The activity of all the mutants was assayed under standard assay conditions with the YydF_18-49 substrate (Fig. 4d). As expected, the A3 mutant was unable to cleave SAM and to convert the peptide substrate. Similarly, the activity of the C222A mutant was impaired consistent with the absence of [4Fe-4S] center in this enzyme. In contrast, the C22A mutation did not affect the activity of the enzyme as shown by the formation of the three epimerized peptides (i.e. YydF_a, YydF_b and YydF_c) (Fig. 4d). The C223A mutant exhibited also enzyme activity. However, this enzyme variant showed a different pattern of products, with only the YydF_c peptide being efficiently produced (Fig. 4d). If we substituted this residue by an isosteric serine residue (C223S mutant), the activity was further inhibited (Supplementary Fig. S21) suggesting a critical role for this residue.

These results were reminiscent of the spore photoproduct lyase (SP lyase), a radical SAM enzyme catalyzing DNA repair¹⁷. We and others have shown that SP lyase uses a cysteine residue to quench the radical substrate intermediate and terminate the reaction^16–18,42. When mutated, SP lyase produced, along its expected product, several DNA adducts⁴³. The structural study of SP lyase confirmed that a conserved cysteine residue was perfectly positioned to act as an H-atom donor^18,42. Additional studies have shown that the activity of a mutated SP lyase could be rescued by precisely positioning a cysteine residue within the enzyme active site⁴².

We thus compared the activity of YydG (wild-type and C223A mutant) in the presence or the absence of DTT, a potential source of H-atoms. While the wild-type enzyme still produced the epimerized peptides, the C223A mutant produced other peptide derivatives in the absence of DTT (Fig. 4e). LC-MS analysis showed that these peptides derived from YydF_18-49 and contained a modified Val₃₆ or Ile₄₄ at their N-terminus or C-terminus ends with a mass shift of -30.005 Da and -1.032 Da (Fig. 4e & f, Supplementary Fig. S22-25).

We identified the modification as the loss of a carboxyl or amino group, resulting from the rupture of either the C_α-N or the C_α-CO bonds, and the addition of a double-bonded oxygen-atom on the C_α-atom (Fig. 4f). No other product, resulting from peptide cleavage, hydrolysis or rearrangement could match our MS analysis (Supplementary Table S7). The peptides were determined as being: I^Δ1LGSGH-NH₂ (peptide I₄₄^Δ1-H₄₉, [M+Na]⁺= 603.2861), Ac-GLLDESQKLAKVNDLWYFVKSKENRWI^Δ30 (peptide G₁₈-I₄₄^Δ30, [M+2H]²⁺= 1646.3838) and Ac-GLLDESQKLAKVNDLWYFV^Δ30 (peptide G₁₈-V₃₆^Δ30, [M+2H]²⁺= 1125.7) (See Supplementary Fig. S22-25 & Supplementary Table S8-12 for full assignment and analysis).

These results were reminiscent of the oxygen-induced cleavage at the sites of the glycyl radical of the pyruvate formate lyase⁴⁴ and the ribonucleotide reductase⁴⁵. However, to date, the structure of these peptides containing an oxygen atom double-bonded to the C_α-atom is unique among radical SAM enzymes. Altogether, these results are in perfect agreement with our hypothesis that YydG generates a carbon-centered radical on the C_α-atom of Val₃₆ and Ile₄₄ and that Cys₂₂₃ plays a critical role for the termination of the reaction. In light of previous work on SP lyase^{16–18,42,43}, we interpret the role of Cys₂₂₃ as the critical H-atom donor required for the termination of the reaction.

We propose an unprecedented mechanism on a peptide backbone leading to peptide epimerization (Fig. 5). After reductive cleavage of SAM, YydG generates a 5′-dA radical which abstracts the C_α H-atom of l-Val or l-Ile. This process is thermodynamically favourable considering the bond dissociation energies (BDE) of 354.6 and 351.3 kJ.mol^-1 for the H-C_α bonds of L-Val and L-Ile, respectively⁴⁶. 5′-dA and a carbon-centered radical are then formed, leading to the loss of the C_α atom stereochemistry. This radical intermediate is then quenched by a thiol H-atom provided by a protein cysteine residue (i.e. Cys₂₂₃). Although the BDE of the thiol H-atom is slightly higher (368.8 kJ.mol^-1), such radical transfers have been extensively reported notably in the case of the ribonucleotide reductase⁴⁷ and is supported here by the stable incorporation of a solvent derived H-atom into the peptide backbone. In addition, we showed that in the absence of this critical cysteine residue (Cys₂₂₃), polypeptide fragmentation occurs at the two radical sites leading to the formation of unusual peptide fragments. For the next catalytic cycle, the thiyl radical intermediate located on Cys₂₂₃ has to be reduced. In the case of the carbohydrate epimerase NeoN³⁷, it has been proposed that the reduction of the thiyl radical involves a remote cysteine and the formation of a disulfide anion radical, similarly to ribonucleotide reductase⁴⁸. However, assuming the SPASM [4Fe-4S] center has full coordination, like in all characterized SPASM-domain radical SAM enzymes³⁶, this would require bond breakage between a cysteine residue and the [4Fe-4S] center followed by further rearrangements. The proximity of a [4Fe-4S] center with a thiyl radical is reminiscent of ferredoxin:thioredoxin reductase⁴⁹ and heterodisulfide reductases^50,51 for which spectroscopic and structural⁵² studies have pointed that the function of a [4Fe-4S]²⁺ center is to stabilize a thiyl radical leading to the formation of a pentacoordinated iron with unique properties^50,53. A similar mechanism might be involved to reduce the thiyl radical generated on Cys₂₂₃ during catalysis by YydG. In support of this hypothesis, we showed that the A3 mutant, containing only the SPASM [4Fe-4S] center, could not be reduced by sodium dithionite (Supplementary Fig. S26), a typical feature of ferredoxin:thioredoxin reductase⁴⁹ and heterodisulfide reductases⁵³. Hence, we proposed that during catalysis, Cys₂₂₃ might coordinate one Fe atom of the SPASM [4Fe-4S] center in order to regenerate the thiolate group for the next catalytic cycle.

Figure 5. Proposed mechanism of the radical SAM peptide epimerase YydG.

YydG generates a 5′-dA• radical which abstracts an amino acid C_α H-atom. A carbon-centred radical is generated and quenched by the thiolate H-atom of Cys₂₂₃ leading to the formation of a D-amino acid residue. The SPASM [4Fe-4S] center likely assists the radical quenching by reducing the thiyl radical. Further reduction of Cys₂₂₃ would be required to allow another catalytic cycle.

Biological activity of the YydF epipeptide

Finally, since the yydFGHIJ operon (Fig. 1a) was shown to activate the B. subtilis Lia system, possibly by perturbing the bacterial cell wall⁵, we assayed the activity of the enzymatically modified YydF_18-49 peptide on B. subtilis. LB medium was inoculated with B. subtilis in the presence of peptide containing one, two or no epimerized residue. As shown, the unmodified YydF_18-49 peptide or the peptides containing either a d-Val₃₆ or a d-allo-Ile₄₄ (YydF_a and YydF_c, respectively) did not affect the growth of B. subtilis (Fig. 6a). In sharp contrast, the peptide containing both a d-Val₃₆ and a d-allo-Ile₄₄ residue (YydF_b) strongly inhibited bacterial growth, supporting a key biological function for these 2 epimerized residues. We reproduced the experiment by adding YydF_b during mid-exponential phase, when bacterial cells have reached a high density. As shown, addition of the two epimerized peptide led to a surprising fast decay of the bacterial population (Fig. 6b).

Figure 6. Activity of the epipeptides on Bacillus subtilis.

(a) Growth of B. subtilis in liquid LB medium in the presence of YydF_18-49 or peptides containing one or two epimerized residues in position 36 and 44. B. subtilis was grown in LB medium alone (black squares), in the presence of YydF_18-49 (white squares), YydF_a (purple squares), YydF_c (Green squares) or YydF_b (red squares). Each measurement is the mean of three growth experiments with the SD indicated. (b) Growth of B. subtilis in the presence of YydF_18-49 at initial time or mid-exponential phase. B. subtilis was grown in LB medium alone (black squares), in the presence of YydF_18-49 (white squares), YydF_b (red squares) or after addition of YydF_b at mid-exponential phase (indicated by the red arrow) (light blue squares). Each measurement is the mean of three growth experiments with the SD indicated. (c) LC-MS/MS analysis of the purified peptide from B. subtilis. (see Supplementary methods for peptide purification). (d) LC-MS analysis of YydF_33-49, YydF_33-49DD synthetic peptides (traces 1&2) and the peptide isolated from B. subtilis (trace 3). (See Supplementary Fig. S28 & S29 for complete assignment). (e) Growth of B. subtilis in liquid LB medium in the presence of YydF_33-49DD. B. subtilis was grown in LB medium alone (black squares), in the presence of YydF_33-49 (white squares) or YydF_33-49DD (red squares). Each measurement is the mean of three growth experiments with the SD indicated. Growth ratio of B. subtilis in the presence of (f) YydF_18-49 (100 µM) or YydF_b and (g) YydF_33-49 or YydF_33-49DD. Ratios were determined by comparison with growth in the absence of peptide.

In order to demonstrate if B. subtilis actually produces epimerized peptides and to determine the relevant physiological form of the YydF peptide, we tried to isolate the naturally secreted peptide. Large scale transcriptomic analysis designed to cover various adaptative features of B. subtilis revealed that yydF is strongly expressed under nutrient depletion conditions⁵⁴ (Supplementary Fig. S27). We grew B. subtilis in synthetic medium under conditions favoring optimal induction of the yydF, yydG, yydH, yydI and yydJ genes⁵⁴ (see Supplementary information). The supernatant was purified by solid phase extraction and peptides were searched by mass spectrometry analysis using the X!tandem suite⁵⁵ and manual inspection. Only one relevant peptide ([M+3H]³⁺: 703.1; Fig. 6c) was identified. The sequence of this 17mer peptide was determined to encompass residues 33 to 49 including the highly conserved domain found in YydF and the two amino acid residues, target of YydG (Fig. 6c and Supplementary Fig. S28). In order to determine if the purified peptide was actually epimerized, we synthesized two peptides containing either only l-amino acid residues (YydF_33-49) or a d-Val and a d-allo-Ile (YydF_33-49DD) in position 36 and 44 (respective positions to YydF). Interestingly, the peptide isolated from B. subtilis exhibited the same fragmentation pattern than the synthetic peptide YydF_33-49DD containing two d-amino acid residues (Supplementary Fig. S28). In addition, the retention time of the peptide isolated from B. subtilis (20.5 min) perfectly matched with the retention time of YydF_33-49DD but not with the one of YydF_33-49 (18.7 min) (Fig. 6d). Finally, a tryptic peptide obtained after digestion of the peptide produced by B. subtilis eluted at the same time than Peptide 3*, containing a d-allo-Ile (Supplementary Fig. S29). Altogether, these results demonstrate that B. subtilis produces peptides containing d-amino acids.

Finally, since we identified the mature form of YydF, we aimed at comparing its activity with the previously synthesized epimerized peptide (YydF_b). Surprisingly, YydF_33-49DD proved to be a stronger inhibitor leading to the total inhibition of B. subtilis growth (Fig. 6e). Consistent with our previous experiments, a peptide with identical sequence but devoid of d-amino acid residues did not exhibited activity toward B. subtilis (Fig. 6e). In order to determine the minimum inhibitory concentrations (MIC) of YydF_b and YydF_33-49DD, we performed serial dilution assay. YydF_33-49DD proved to be at least 100 times more potent than YydF_b with a MIC < 2µg.mL^-1 (Fig. 6f & g). This value is 120 times smaller than the MIC determined for bacitracin⁵⁶, a well-known antibiotic peptide inducing LiaRS⁵⁷ and inhibiting B. subtilis growth. To the best of our knowledge, it is the first time that a naturally short epimerized peptide, proved to be the active form of a bacterial peptide with regulatory and/or antimicrobial activity. Interestingly, the activity of YydF_33-49DD is in the range of the most potent inhibitory peptides⁵⁸.

Although the physiological role of YydF has yet to be deciphered, its expression is perfectly correlated with sdpC, a gene encoding a B. subtilis killing factor⁵⁹. Among the 4175 genes of B. subtilis no other gene exhibits a better correlation (Supplementary Fig. S27). Interestingly, the expression of both genes has been reported to be under the negative regulation of AbrB, a regulator of the central metabolism⁵⁹. It is tempting to speculate that YydF is a novel regulator of B. subtilis complex metabolic plasticity. Our data support that the YydF peptide is first epimerized by the radical SAM enzyme YydG, processed by the YydH membrane protease and then exported by the YydIJ ABC-transporter. In support of this conclusion, assays performed between the YydG epimerase and the short YydF_33-49 peptide, revealed only low level of activity suggesting that it is not a relevant enzyme substrate.

Conclusion

This study demonstrates that peptides containing d-amino acids, are much more widespread in living organisms than previously anticipated. Indeed, recent reports suggested that such type of modification was restricted to bacteria with exceptional metabolic capacities such Entotheonella and related organisms producing proteusins⁶⁰. We show here that epimerized peptides are also produced by the model bacterium B. subtilis and likely by several pathogenic species and members of the human microbiome including Enterococcus faecalis, Streptococcus agalactiae and Staphylococcus epidermidis.

Among known bio-active peptides and bacteriocins⁶¹, YydF represents the first member of a novel class of RiPPs, that we propose to call epipeptides. As we demonstrated here, the activity of epipeptides is dependent only of discrete epimerizations and requires no other post-translational modification. Peptides containing d-amino acids have thus to be considered when trying to identify the active forms of bacterial peptides with unknown functions notably within the human microbiome. Finally, radical SAM enzymes appear to be widespread catalysts employed by bacteria to increase peptide diversity through the control of the chirality of life.

Methods

Protein cloning & purifications, enzymatic assays, analytical techniques & peptide synthesis are described in Supplementary Information.

Supplementary Material

Supplementary information is available in the online version of the paper.