Ruminococcin C, an anti-clostridial sactipeptide produced by a prominent member of the human microbiota Ruminococcus gnavus

The human microbiota plays a central role in human physiology. This complex ecosystem is a promising but untapped source of bioactive compounds and antibiotics that are critical for its homeostasis. However, we still have a very limited knowledge of its metabolic and biosynthetic capabilities. Here we investigated an enigmatic biosynthetic gene cluster identified previously in the human gut symbiont Ruminococcus gnavus. This gene cluster which encodes notably for peptide precursors and putative radical SAM enzymes, has been proposed to be responsible for the biosynthesis of ruminococcin C (RumC), a ribosomally synthesized and posttranslationally modified peptide (RiPP) with potent activity against the human pathogen Clostridium perfringens. By combining in vivo and in vitro approaches, including recombinant expression and purification of the respective peptides and proteins, enzymatic assays, and LC-MS analyses, we determined that RumC is a sulfur-to–α-carbon thioether-containing peptide (sactipeptide) with an unusual architecture. Moreover, our results support that formation of the thioether bridges follows a processive order, providing mechanistic insights into how radical SAM (AdoMet) enzymes install posttranslational modifications in RiPPs. We also found that the presence of thioether bridges and removal of the leader peptide are required for RumC's antimicrobial activity. In summary, our findings provide evidence that production of the anti-Clostridium peptide RumC depends on an R. gnavus operon encoding five potential RumC precursor peptides and two radical SAM enzymes, uncover key RumC structural features, and delineate the sequence of posttranslational modifications leading to its formation and antimicrobial activity.

Despite its growing importance in biology, study of the human microbiome remains a challenging area of investigation. Recently, it has emerged that environmental rather than genetic factors play a major role in shaping this complex eco-system, one of the densest on earth (1). Among the molecular determinants underpinning the normal equilibrium within the microbiota (eubiosis), it has been predicted that antimicrobial substances should play a central role (2,3). However, to date, only few antimicrobial peptides from the human microbiome have been identified and characterized. Among these, ribosomally synthesized and posttranslationally modified peptides (RiPPs) 3 represent a growing family of natural products that has attracted considerable interest (4), propelled by the need for novel antibiotics and their involvement in homeostasis of the microbiota (5,6). Ruminococcus gnavus is an inhabitant of the human digestive tract, with some strains able to degrade mucins and host glycans (7), similar to the prominent gut symbiont Bacteroides thetaiotaomicron (8,9). However, despite its widespread distribution in humans and its potential role in human physiology, the metabolic properties of this Gram-positive bacterium are just starting to be unraveled. For instance, R. gnavus has been shown to produce several antimicrobial peptides, including RumA (10) and RumC (11,12), which are both active against Clostridium species. RumA has been shown to be a RiPP containing three lanthionine bridges (i.e. three ␤-thioether bonds) and, thus, to belong to the large class of lanthipeptides (13). Lanthipeptides are well-known to exhibit a wide range of biological activities spanning from antimicrobial activities to antiviral, antinociceptive, and antiallodynic functions (14). Some lanthipeptides, such as nisin, exert their antimicrobial activity by binding to lipid II (15). However, for the vast majority of them, we still have limited knowledge of their mode of action. Lanthionine bridges are installed by a two-step mechanism involving dehydration of a Ser or Thr residue followed by stereoselective intramolecular Michael addition of the thiol group of a remote Cys residue. Intriguingly, other thioether-containing peptides called sactipeptides (sulfurto-␣-carbon thioether-containing peptides) have been described recently (5,16). In contrast to lanthipeptides, formation of thioether bridges in sactipeptides involves a radical-based mechanism catalyzed by radical SAM enzymes (5,6,17) and leads to the formation of ␣-thioether bridges (18). By combining in vivo and in vitro approaches, we succeeded to unveil the structure of the elusive bacteriocin RumC. Our data show that RumC is a sactipeptide, the first one isolated from the human microbiota, and that it possesses a distinctive architecture. In addition, our study sheds new light on how radical SAM enzymes install posttranslational modifications in RiPPs.
Several recent reports have shown that RiPPs produced by Gram-positive bacteria can be efficiently modified by their cognate radical SAM enzymes when expressed in Escherichia coli (25)(26)(27). For heterologous expression of RumC, we selected c1 and c2, as these two genes have been shown to be highly induced in R. gnavus when this bacterium colonizes the digestive tract of rats (11,12). In addition, both genes are cotranscribed with the mc1 and mc2 genes, coding for the respective radical SAM enzymes MC1 and MC2 (Fig. 1A). We expressed C1 and C2 as His tag fusion peptides with a tobacco etch virus cleavage site to perform their expression in E. coli and purification by affinity chromatography.
As  Fig. 2A and Fig. S3 and Table S2). The fragmentation pattern of the C2 peptide was essentially identical to the one of the C1 peptide ( Fig. 2B and Fig. S3). In addition, the fragments y44 (Ϫ4 Da) and y18 allowed us to precisely pinpoint the mass loss on the region containing the four conserved cysteine residues. Finally, no b or y ions from the cysteine-rich domain were present in the MS spectrum, confirming the formation of internal disulfide bridges. Thus, when expressed in E. coli, both peptides were essentially purified under an oxidized form with two disulfide bridges.
The fact that only a low amount of the C2 peptide was modified when coexpressed with MC2 was intriguing ( Fig. 2A), considering that C1 was fully matured when coexpressed with MC1 ( Fig. 1D) and the high sequence identity between the two enzymes and peptides. To distinguish between a peptide and an enzyme issue, we coexpressed C1 with MC2 or C2 with MC1 in E. coli (Fig. 2D). As shown, MC1 was not able to modify C2 in vivo, whereas MC2 was active toward C1, albeit with a lower efficiency than MC1 (Fig. 2D). Interestingly, the C1 MC2 peptide had the same spectroscopic signature than the C1 MC1 peptide, supporting that both peptides contain the same posttranslational modifications. Altogether, these data support that both MC1 and MC2 install the same posttranslational modifications and that the low amount of modified C2 peptide produced (i.e. C2 MC2 and C2 MC1 ), is mainly due to its sequence.

Structure of RumC2
To determine the connectivity between the ␣-positions of the four target residues (i.e. Ala 31 , Asn 35 , Arg 53 , and Lys 61 in C1 or Glu 31 , Asn 35 , Arg 53 , and Arg 61 in C2) and the four cysteine residues (i.e. Cys 22 , Cys 24 , Cys 41 , and Cys 45 ), we devised an in vitro strategy. As shown above, C2 was efficiently produced without modification in E. coli and was thus a suitable source of Ruminococcin C, a new sactipeptide from the human microbiota Green, c1 to c5, genes predicted to encode RumC peptides; red, mc1 and mc2, genes predicted to encode tailoring radical SAM enzymes; orange, genes predicted to be involved in immunity; blue and gray, putative exporters. B, sequences of the C1 and C2 peptides. Amino acid residues from the predicted leader sequence are in gray. Conserved amino acid residues from the core sequence are in light blue, and nonconserved amino acid residues are in black. The four conserved cysteine residues are in red. Numbers indicate relative position to the sequence. substrate for in vitro experiments. In addition, we generated the three cysteine variants C2 A22A24 , C2 A41A45 , and C2 A24 , in which the corresponding cysteine residues Cys 22 and Cys 24 , Cys 41 and Cys 45 , and Cys 24 were replaced with alanine residues, respectively (Fig. 3). As shown, after purification, the C2 A22A24 variant had a mass of [Mϩ6H] 6ϩ obs 1061.20, indicating the formation of one disulfide bridge ( Fig. 3A and Fig. S4). After incubation with MC2, the peptide mass shifted to [Mϩ6H] 6ϩ obs 1060.86 (⌬ m ϭ Ϫ4.02 Da), consistent with the formation of two thioether bridges. LC-MS/MS analysis showed that only the residues Arg 53 and Arg 61 were modified (Fig. 3A, (Table S3). Thus, in contrast to all known sactipeptides such as subtilosin A (18,29,30), thuricin CD (31), and thurincin H (28), RumC2 contains two hairpin-like domains: one domain with Cys 22 and Cys 24 connected to Glu 31 and Asn 35 and a second domain with Cys 41 and Cys 45 connected to Arg 53 and Arg 61 .
When we assayed the C2 A24 variant, its mass shifted from [Mϩ6H] 6ϩ obs 1066.53 (oxidized form) to 1065.86, consistent with the formation of three thioether bridges. LC-MS/MS analysis showed that these three thioether bridges involved Arg 53 , Arg 61 , and likely Glu 31 but not Asn 35 ( Fig. 3C and Table S4), supporting that Glu 31 was connected to Cys 22 . We failed to express a fourth variant in which Cys 22 was replaced with an alanine residue (i.e. the C2 A22 peptide). However, we succeeded to obtain small amounts of the corresponding C1 mutant (i.e. the C1 A22 peptide) coexpressed with MC1. Surprisingly, with this variant, LC-MS/MS analysis was consistent with the formation of a thioether bridge between Cys 24 and Ala 31 ( Fig. S5 and Table S4). This last result supports that, in the absence of the target cysteine residue, thioether bridges might form with a nearby cysteine residue.
To accurately determine the connectivity of the thioether bridges, we designed, based on the C2 sequence, shorter peptide substrates containing either the first (i.e. residues 1-40, C2  or the second hairpin domain (i.e. residues 28 -63, C2 28 -63 ) and assayed them with MC2 in vitro. The C2 1-40 peptide proved to be an extremely poor substrate and was not amenable to LC-MS/MS analysis. However, the C2 28 Table S1). LC-MS/MS analysis confirmed the formation of two thioether bridges located on Arg 53 and Arg 61 (numbered as in the C2 sequence), a result in agreement with the ones obtained with the full-length-peptide ( Fig. 2 and Table S5). When we replaced Cys 41 with an alanine residue, incubation of the C2 28 -63 A 41 variant with MC2 led to a product containing a single thioether bridge between Cys 45 and Arg 53 (Fig. 4B). Intriguingly, the C2 28 -63 A 45 variant did not lead to formation of any product when incubated with MC2 (Fig. 4C). Similarly, no product was obtained with the double mutant C2 28 -63 A 41 A 45 . These results support a sequential order for formation of the thioether bridges, with the Cys 45 -Arg 53 bridge being required first before formation of the Cys 41 -Arg 61 bridge could take place.

Sequential formation of the thioether bridges in RumC2
To get better knowledge regarding the sequential formation of the thioether bridges and the structure of RumC2, we performed a kinetics analysis by LC-MS/MS. Using an optimized gradient and short reaction times, we were able to identify and characterize several C2 reaction intermediates (Fig. 5A (Fig. 5B) and Glu 31 and Asn 35 (Fig. 5C), respectively. After 60 min, species A and AЈ had almost totally disappeared, whereas the main species, E ([Mϩ7H] 7ϩ obs 918.01), was identified as the mature C2 MC2 with four thioether bridges (Fig. 5D). The transient species D could not be unambiguously assigned, but LC-MS/MS analysis was consistent with a reaction intermediate containing three thioether bridges en route to conversion into species E.
Collectively, these experiments allowed us to propose a model for formation of the thioether bridges in RumC based on the following evidence. First, thioether bridges in the N-and C-terminal domains are installed independently, as shown by the experiments performed with the C2 A22A24 and C2 A41A45 peptides (Fig. 3) and with the truncated C2 28 -63 peptide (  (Fig. 5). Second, formation of the thioether bridges follows an N-to-C direction. Indeed, our data support that the Cys 45 -Arg 53 bridge is formed before the Cys 41 -Arg 61 bridge in the C-terminal domain (Fig.  4). Similarly, for the N-terminal domain, we identified several

Ruminococcin C, a new sactipeptide from the human microbiota
reaction intermediates having the Cys 24 -Glu 31 bridge (e.g. intermediates B, C, and E) but none with only the Cys 22 -Asn 35 bridge.

Stereochemistry of the ␣-carbon atoms in RumC1
The vast majority of known sactipeptides, such as subtilosin A (18,29) or thuricin CD (31), contain L-and D-configurated thioether bonds, the latter being formed by C ␣ atom configuration-inversion during catalysis. To establish the stereochemistry of the thioether bridges, we further purified the C1 MC1 peptide by HPLC to remove contaminating MC1 and trace amounts of unmodified C1. After purification, the C1 MC1 peptide was subjected to deuterated hydrochloric acid (DCl) hydrolysis and amino acid derivatization with N ␣ -(2,4-dinitro-5-fluorophenyl)-L-valinamide, and its amino acid content was analyzed by LC-MS as described previously (32). With this procedure, we could identify D-amino acid residues and determine whether they originated from the peptide backbone (unlabeled residues) or were produced during acid hydrolysis (incorporation of one deuterium atom). Indeed, it is well-known that, during peptide hydrolysis, free L-amino acids can spontaneously epimerize. Among the four residues involved in thioether bridges (i.e. Ala, Asn, Arg, and Lys), we only identified L-amino acid residues (Fig. S6), supporting that the four thioether bridges involve L-configurated amino acid residues.

Biological activity of RumC1
RumC was originally identified as a trypsin-dependent substance produced by R. gnavus and active against the Gram-positive bacteria Clostridium perfringens and Bacillus subtilis (33). This substance has been shown to be produced in vivo in the digestive tract of germ-free rats colonized by R. gnavus under the dependence of protease activity. To determine the active form of RumC, we took advantage of the C2 and C1 MC1 peptides, which we were able to recombinantly produce in significant amounts. Preliminary assay against C. perfringens ATCC 13124 (22) revealed that C1 MC1 was inactive. We thus treated C1 MC1 with trypsin to mimic the activation process reported in previous studies (12,33). Following trypsin hydrolysis, LC-MS analysis showed that C1 MC1 was truncated after Lys 19 , leading to formation of a peptide ([Mϩ4H] 4ϩ obs 1080.98) encompassing residues Trp 20 to Ala 63 and containing the four thioether bridges (Fig. 6A). This peptide, called RumC1, exerted antimicrobial activity toward C. perfringens, supporting this structure being the active form (data not shown).
To better characterize the antimicrobial properties of RumC1, we used B. subtilis, which is a nonpathogenic and aerobic bacterium sensitive to RumC (33). Similarly to C. perfringens, B. subtilis was sensitive to RumC1 but not to C1 MC1 (Fig.  6B). As an additional control, the C2 peptide, with or without trypsin treatment, was assayed against B. subtilis and proved to be inactive. This result supports that, in addition to removal of the leader peptide, the presence of the thioether bridges is mandatory for activity, although subtle differences between the sequences of the C1 and C2 peptides might also account for the lack of activity of the C2 peptide. The minimum inhibitory concentration of RumC1 against B. subtilis was determined to be 6 M (Fig. 6C), a value likely to be overestimated considering the contamination of the RumC1 peptide with the radical SAM enzyme MC1. Assayed against other Gram-positive bacteria, RumC1 (70 M) was moderately active against Enterococcus faecalis (Fig. 6D) but not against Staphylococcus aureus. Interestingly, assayed against E. coli at an identical concentration, RumC1 induced a lag phase, suggesting possible action toward Gram-negative bacteria (Fig. 6E). This last result was consistent with a recent study indicating possible activity of a peptide extract containing RumC against Salmonella enteridis (12).

Discussion
Early studies have shown that R. gnavus, a commensal bacterium from the human microbiota (34), produces various anticlostridial substances. The first substance identified was called Ruminococcin A (RumA) and proved to be a lanthibiotic (10,35) characterized by the presence of three ␤-thioether bridges (13). Intriguingly this lanthipeptide was produced only when trypsin was added to the bacterial growth medium (35). Later, it was shown that R. gnavus produces an additional anti-clostridial substance when it colonizes the gastrointestinal tract of mono-associated rats (i.e. germ-free rats colonized with R. gnavus only) (12). Production of this elusive substance, called Ruminococcin C (RumC), was shown to be dependent on an operon notably encoding two putative radical SAM enzymes and five potential peptide precursors (C1-C5) (Fig. 1A) and to require trypsin for activity (11,33). Radical SAM enzymes have been shown to introduce a broad range of posttranslational modifications in RiPPs (5, 16), including methylation (36,37), epimerization (32,38,39), and carbon-carbon (40,41) and thioether bonds (18,25,30,42). Our results support that the C1 and C2 peptides from the RumC operon are modified by the radical SAM enzymes MC1 and MC2, respectively. These enzymes introduce posttranslational modifications on Ala 31 , Asn 35 , Arg 53 , and Lys 61 in the C1 peptide (Fig. 1E) and on Glu 31 , Asn 35 , Arg 53 , and Arg 61 in the C2 peptide (Fig. 2B). Thus, both enzymes introduce posttranslational modifications at the same locations despite involving different amino acid residues. In addition, our results demonstrate that both MC1 and MC2 can modify the same substrate (i.e. C1 peptide), leading to formation of an identical product (Fig. 2D). Collectively, these data support that the complex RumC biosynthetic operon might result from gene duplication and rearrangement events. The thuricin CD (31) and thurincin H (43) biosynthetic operons share similar features. However, although the thurincin H biosynthetic operon encodes for one radical SAM enzyme and three identical peptide precursors (31), the thuricin CD operon encodes for two radical SAM enzymes and leads to production of two sactipeptides with synergistic antimicrobial activities (43).  High-resolution LC-MS/MS analysis of C1 and C2 peptides after modification by MC1 or MC2 showed that both peptides contain ␣,␤-dehydro-amino acid residues, the hallmark of C␣-thioether bonds. Indeed, although radical SAM enzymes have been shown recently to be able to introduce ␣-, ␤-, and ␥-thioether bonds in RiPPs (18,26,27,30), because of their lower stability, only S-C␣ thioether bonds open during LC-MS/MS analysis, with concomitant formation of characteristic dehydro-amino acid residues (26).

Ruminococcin C, a new sactipeptide from the human microbiota
Our results unambiguously establish that Cys 41 and Cys 45 are connected to residues 53 and 61 (i.e. Arg 53 and Lys 61 in C1 and Arg 53 and Arg 61 in C2, respectively), defining a C-terminal hairpin domain (Fig. 3). Regarding the N-terminal domain, it was more challenging to determine the connectivity of Cys 22 and Cys 24 . Indeed, when we assayed peptides mutated for each residue (i.e. C2 A24 and C1 A22 ), a thioether bridge involving the residue in position 31 was always formed. This is likely due to the close proximity between Cys 22 and Cys 24 , which can react with the residue in position 31 after its radical activation. We currently favor for the structure of RumC, a model based on two symmetrical hairpin domains, as shown in Fig. 6A, although an alternate model with Cys 22 connected to Ala 31 and Cys 24 to Asn 35 cannot be ruled out completely.
Our data also support that formation of the thioether bridges in RumC follows a processive order with an N-to-C directionality and that the N-and C-terminal domains are processed independently. Indeed, using either full-length or truncated peptides (Figs. 3 and 4), we have shown that the formation of the thioether bridge involving Cys 41 is under the dependence of the Cys 45 -Arg 53 bridge, supporting that Arg 53 must be modified before Arg 61 . Of note, the efficient modification of a truncated peptide (i.e. the C2 28 -63 peptide) by the MC2 enzyme

Ruminococcin C, a new sactipeptide from the human microbiota
demonstrates that its activity is leader peptide-independent, a trend encountered in a growing number of radical SAM enzymes catalyzing RiPP posttranslational modifications (18,32,39,44).
In conclusion, we have deciphered the structure of RumC, a sactipeptide containing four C␣-thioether bridges in the L configuration. Although sactipeptides with up to four thioether bridges, including thurincin H (28) and huazacin (26), have been reported, the structure of RumC is unique, as it is based on two hairpin domains. Indeed, all sactipeptides described so far are folded as a single hairpin domain with thioether bridges linking cysteine residues from the N-terminal domain to residues from the C-terminal domain. Our data also support that, in RumC, installation of the thioether bridges follows a precisely defined order suggestive of a processive mode of action. Indeed, although a growing number of radical SAM enzymes have been shown to be involved in the biosynthesis of S-C␣linked (18,28,30,31,45,46), S-C␤-linked, and S-C␥-linked thioether-containing peptides (26), it is currently unknown whether the formation of these bridges is random. We have shown here that MC2, similarly to the radical SAM epimerase PoyD (32), introduces modifications in a sequential order. However, contrary to PoyD, MC2 has an N-to-C directionality. Finally, we have established that the presence of thioether bridges and removal of the leader peptide are both required for the antimicrobial activity of RumC. Although several hypotheses have been proposed for this dependence, our data support that the leader peptide impedes the biological activity of RumC. Further investigations regarding the other precursor peptides (i.e. C3 to C5) should also clarify whether these peptides, after posttranslational modification, have similar activity than RumC1 or exhibit synergistic properties, as reported for thuricin CD. RumC thus delineates a novel class of sactipeptides and is the first member of this natural product family to be isolated from the human microbiota. Intriguingly, by targeting Grampositive and Gram-negative bacteria, it could contributes to shape this unique and complex ecosystem (6).

Cloning, expression, and purification of C1 and C2 peptides
The c1 and c2 genes were synthesized by Life Technologies (Thermo Fisher GeneArt) and ligated in the pRSFDuet-1 plasmid with a His 6 tag fusion and transformed in E. coli BL21 (DE3) (Life Technologies). Peptide production was performed in Luria-Bertani, and cells were harvested after 20 h by centrifugation (5500 rpm, 10 min at 4°C). Cells were suspended in buffer A (50 mM Tris and 300 mM KCl (pH 7.5)) supplemented with 1% v/v Triton X-100. Cells were disrupted by sonication, followed by ultracentrifugation (45,000 rpm for 1.5 h at 4°C) to remove cell debris. The supernatant was loaded on Ni-NTA Fast Flow Gel (Qiagen) equilibrated previously with buffer A. The Ni-NTA gel was washed successively with buffer A containing 25 mM and 75 mM imidazole. The peptide was eluted with buffer A containing 500 mM imidazole. This fraction was loaded on an NAP10 column equilibrated previously with buffer A. The peptide was concentrated in an Amicon concentrator (molecular weight cutoff of 3 kDa, Millipore) and stored at Ϫ80°C. Peptide purity was assessed by SDS-PAGE (18% (w/v)), and the concentration was determined using a Nano-Drop spectrophotometer.

Cloning, expression, and purification of C1 and C2 coexpressed with MC1 and MC2
Plasmids pRSFDuet-His-C1 and pRSFDuet-His-C2 were used as a template. The plasmids were digested with NdeI/ XhoI, and the mc1 and mc2 genes (synthesized by Life Technologies) were ligated into the respective construct. Protein expression and purification were performed as described above. Purity was assessed by SDS-PAGE and LC-MS analysis. Peptide concentration was determined by NanoDrop spectrophotometer.

Cloning, expression, and purification of the radical SAM enzyme MC2
The RumMC2 gene was synthesized by Life Technologies and ligated in the pET-28a(ϩ) plasmid. After sequencing, the construction was transformed in E. coli BL21 (DE3) star cells (Life Technologies). Protein expression and purification were performed as described above. Proteins were concentrated in an Amicon concentrator (molecular weight cutoff of 10 kDa, Millipore) and stored at Ϫ80°C. 12% (w/v) SDS-PAGE was run to confirm the purity of the protein, and concentration was determined by NanoDrop spectrophotometer.

Ruminococcin C, a new sactipeptide from the human microbiota
ogies). Expression and purification of the peptides were performed as detailed above.

Iron-sulfur cluster reconstitution
Protein reconstitution was performed under anaerobic conditions at 4°C using a 12 molar excess of NH 4 ) 2 Fe(SO 4 ) 2 (Sigma-Aldrich) and Na 2 S (Sigma-Aldrich). The excess of unbound iron and sulfur was removed by Sephadex G25 column (GE Healthcare) equilibrated with buffer A. Proteins were concentrated using an Amicon concentrator (molecular weight cutoff of 10 kDa, Millipore).

Enzymatic assays
All assays were performed in an anaerobic chamber. Freshly reconstituted protein was used to perform the assays. Protein was concentrated at 200 M and incubated with 1 mM peptide. Reactions were quenched by adding 0.1% (v/v) of formic acid for LC-MS analysis.

HPLC analysis and purification
Peptides were analyzed and purified using a Zorbax Eclipse Plus C18 Rapid Resolution HT (2 ϫ 50 mm, 1.8 m, 100 Å, Agilent) by loading 10 -20 l of each sample diluted 10 times in 0.1% (v/v) formic acid. Elution was performed at a flow rate of 0.3 ml/min using an acetonitrile gradient between 10% to 30% (v/v) of acetonitrile 80% (v/v), formic acid 0.1% (v/v). Peptide UV detection was performed at 215 nm.

LC-MS analysis
Each peptide was analyzed by LC-MS using a Q Exactive Focus mass spectrometer (Thermo Fisher Scientific). Peptide separation was performed on Ultimate 3000 nanoHPLC and Vanquish ultra-high-performance liquid chromatography systems. The C1, C1 MC1 , and C2 MC2 peptides were analyzed using a Proswift RP4H (Thermo Fisher) monolithic nanocolumn (0.1 ϫ 250 mm), and the C2 MC1 and C1 MC2 peptides were analyzed using a Zorbax Eclipse Plus C18 Rapid Resolution HT column (2 ϫ 50 mm, 1.8 m, 95 Å). Acetonitrile gradients between 10%-30% and 15%-25% in formic acid (0.1%) were used. Mass analysis was performed at a resolution of 35,000 (m/z, 200) with a MS range of 500 -1300 and MS/MS analysis. The collision energy was optimized for each peptide (between 22% and 25%) to reduce formation of internal fragments. The lock mass option was activated to enhance mass accuracy. For each peptide, several scans (between 5 and 10) were merged to enhance the quality of the data. Data were deconvoluted using Xtract tools included in the Freestyle software suite, version 1.3 (Thermo Electron). All daughter ions observed were verified and annotated manually.

Determination of thioether bridge configuration
The C1 MC1 peptide was purified by HPLC and dried using a centrifugal vacuum concentrator. Hydrolysis was performed in DCl (6 N) under vacuum conditions at 110°C for 12 h. Reaction mixtures were incubated for 1 h at 42°C after addition of 4 l of NaHCO 3 1 M, 3 l of N-␣-(2,4-dinitro-5-fluorophenyl)-L-valinamide at 10 mg/ml, and 13 l of H 2 O. Final mixtures were diluted in 0.1% (v/v) formic acid before LC-MS/MS analysis as described previously (32).

Antimicrobial assay
Peptide purity was assessed by LC-MS/MS and HPLC analysis. Each peptide was analyzed by data-dependent top-down analysis MS/MS fragmentation analysis on a Q-exactive Focus mass spectrometer. Peptide digestion was performed using trypsin (1% (v/v)) at 37°C for 12 h. Growth experiments were performed using M17 (Glucose 1% (v/v)) medium for E. faecalis and Mueller-Hinton broth for E. coli (K12), B. subtilis 168, and S. aureus, respectively. An overnight culture was used to inoculated the growth medium at a final A 600 ϭ 0.1. After 5 h of growth, a fresh culture was prepared (A 600 ϭ 0.05) for the antimicrobial assay. 10 l of sample was added to 90 l of inoculated medium. Growth analysis was performed using a Bioscreen apparatus for 24 h at 37°C.