Structural characterization and bioactivity analysis of two-component lantibiotic flv system from a ruminant bacterium

Summary

The discovery of new ribosomally synthesized and post-translationally modified peptide natural products (RiPPs) has greatly benefitted from the influx of genomic information. The lanthipeptides are a subset of this class of compounds. Adopting the genome mining approach revealed a novel lanthipeptide gene cluster encoded in the genome of Ruminococcus flavefaciens FD-1, an anaerobic bacterium that is an important member of the rumen microbiota of livestock. The post-translationally modified peptides were produced via heterologous expression in Escherichia coli. Subsequent structural characterization and assessment of their bioactivity revealed features reminiscent of and distinct from previously reported lanthipeptides. The lanthipeptides of R. flavefaciens FD-1 represent a unique example within two-component lanthipeptides, consisting of a highly conserved α-peptide and a diverse set of eight β-peptides.

eTOC summary

An unusual gene cluster from Ruminococcus flavefaciens contains twelve substrate and two lanthipeptide synthetase genes. The post-translationally modified peptides were produced in E. coli and comprise four structurally-conserved lipid II binding peptides and eight structurally diverse β-peptides, some of which displayed synergistic antimicrobial activity.

Introduction

The human microbiome has received much attention in recent years for its connection to human health (Garrett, 2015; Wlodarska et al., 2015; Yurkovetskiy et al., 2015). Increasingly, the mutualistic relationship between the microbiome and host has been demonstrated (Hooper et al., 2012; Maynard et al., 2012; Nicholson et al., 2005; Tremaroli and Backhed, 2012). One of the beneficial roles of the gut microbiota of a healthy individual is believed to be the resistance that is provided against colonization by pathogens (Vogt et al., 2015). Similarly, the microbiota of livestock is critical for animal health, and a better understanding of the mechanisms that confer pathogen resistance in this setting is desired. Ruminant animals such as cattle and sheep have a symbiotic relationship with ruminal microorganisms that can degrade cellulose and/or hemicellulose (Hungate, 1966; White et al., 2014). Among the bacteria with cellulose activity in the rumen are the anaerobic ruminococci (Weimer, 2015). Some Ruminococcus strains have been reported to produce bacteriocins (Chen et al., 2004; Dabard et al., 2001; Gomez et al., 2002; Marcille et al., 2002; Pujol et al., 2011), a class of ribosomally produced antimicrobial compounds that may be important for maintaining a niche in the competitive microbial environment of the rumen as well as in the human gastrointestinal tract (Russell and Mantovani, 2002). Ruminococcus flavefaciens FD-1 has a particularly high cellulolytic activity (Shi and Weimer, 1996). It has not been reported to produce any antimicrobial peptides, but its genome was recently sequenced (Berg Miller et al., 2009), providing a view of the genetic capability to produce such compounds. Here we show that its genome encodes an unusual group of lanthipeptides that are composed of four highly conserved copies of a peptide that likely binds lipid II and a diverse set of eight additional peptides, some of which act synergistically with the lipid II-binding peptide. The possible functional implications of the system in the context of the rumen environment are discussed.

Lanthipeptides are a class of ribosomally synthesized and post-translationally modified peptides (RiPPs) that are characterized by the presence of thioether linkages. These crosslinks are formed by the Michael-type addition of cysteine thiols onto dehydroamino acids, and the resulting linkages are critical for their bioactivity. Lanthipeptides that display antibacterial activity are called lantibiotics, and interest in these compounds has stemmed from their potent antibacterial activity and low propensity for the development of resistance (Lubelski et al., 2008). Although the prospects of RiPP production in anaerobic organisms has been highlighted (Letzel et al., 2014), relatively few such compounds produced by the genus Ruminococcus have been reported thus far (Crost et al., 2011; Dabard et al., 2001; Kalmokoff and Teather, 1997). We designate the lanthipeptide biosynthetic gene cluster in R. flavefaciens FD-1 flv (Fig. 1A).

Figure 1.

Comparison of the flv system with known two-component lantibiotics. All alignments were performed using Clustal Omega. (A) The flv biosynthetic cluster encoded within the genome of R. flavefaciens FD-1. (B) Alignment of the full length FlvA peptides. The Glu at which endoproteinase Glu-C cleaves in each of the modified FlvA2.x peptides is shown in red. An arrow indicates the predicted endogenous FlvT cleavage site in the FlvA peptides. (C) Alignment of the FlvA1.a core peptide with previously reported α peptides of two-component lantibiotics. Dehydratable residues in the core peptides are highlighted in red and cysteine residues are highlighted in blue. The reported ring topology of lichenicidin is indicated above the sequence, and the proposed ring topology for the modified FlvA1 peptide is shown below the sequence. Hal denotes the locus for haloduracin, Ltn denotes the locus for lacticin 3147, and Lic the locus for lichenicidin. The conserved lipid II binding motif is indicated with a dotted line. (D) Alignment of the FlvA2.x peptides with reported β peptides of two-component lantibiotics. The fully conserved Pro is indicated with an asterisk.

The LanA substrate peptides that are converted into lanthipeptides are divided into an N-terminal leader and C-terminal core peptide (Knerr and van der Donk, 2012). The leader peptide does not undergo post-translational modification during lanthipeptide maturation, whereas select Ser, Thr, and Cys residues within the core peptide are enzymatically cyclized to generate thioether linkages. Crosslink formation involves the dehydration of Ser and Thr residues, resulting in dehydroalanine (Dha) and dehydrobutyrine (Dhb), respectively. Subsequent attack of Cys thiols onto a Dha results in a lanthionine (Lan), whereas cyclization with a Dhb residue yields a methyllanthionine (MeLan). Both dehydration and cyclization reactions are catalyzed by lanthipeptide synthetases whose functional domains form the basis for a classification scheme for lanthipeptide systems. In the case of class II lanthipeptides, a single bifunctional enzyme termed LanM carries out the dehydration and cyclization reactions within the core peptide of the LanA substrate (Siezen et al., 1996; Xie et al., 2004). Compared to known lanthipeptide biosynthetic gene clusters, the flv cluster possesses unusual characteristics that drew our attention. The cluster encodes twelve putative LanA genes and two putative LanM genes (Fig. 1A). The large number of substrates is reminiscent of the prochlorosin-like systems (Li et al., 2010; Zhang et al., 2014b) and the presence of two LanM synthetases resembles two-component lantibiotic systems such as lacticin 3147 and haloduracin (McAuliffe et al., 2000; McClerren et al., 2006). Most of the two-component lanthipeptides investigated to date consist of a lipid II-binding α-peptide and a β-peptide that acts synergistically to affect pore formation via a poorly defined mechanism (Begley et al., 2009; McAuliffe et al., 1998; Oman et al., 2011b; Oman and van der Donk, 2009; Wiedemann et al., 2006). The gene clusters for two-component lantibiotics usually encode a single copy of an α and a β-peptide, and hence the twelve FlvA peptides encoded within the flv gene cluster are unprecedented. In light of these unusual characteristics, we set out to obtain the modified FlvA peptides. Systematic production of these peptides facilitated their structural characterization and bioactivity analysis.

Results

Bioinformatic analysis of the FlvA substrates

Inspection of the FlvA amino acid sequences revealed that four out of the twelve FlvA peptides possessed nearly identical amino acid sequences, with a single and two amino acid differences in the leader and core peptide, respectively (Fig. 1B). Querying the non-redundant protein database from the National Center for Biotechnology Information using the Basic Local Alignment Search Tool (Altschul et al., 1990) with the core peptides from this nearly identical set of FlvA peptides returned the LicA1 peptide sequence (Fig. 1C) (Begley et al., 2009; Dischinger et al., 2009; Shenkarev et al., 2010). The modified LicA1 core peptide forms the α-peptide of the two-component lantibiotic lichenicidin. An alignment of the lichenicidin peptides and all of the FlvA peptides using the alignment program Clustal Omega (Sievers et al., 2011) indicated that the latter could be roughly divided into a set of putative α and β precursor peptides (Fig. 1C,D). The four nearly identical FlvA peptides (Fig. 1B) make up the putative α-peptides and were designated FlvA1.x (where x = a–d), whereas the remaining eight FlvA peptides were hypothesized to be the corresponding β-peptides and were designated FlvA2.x (where x = a–h). Whereas the FlvA1.x core peptide amino acid sequences resemble that of lichenicidin α, the core peptides of FlvA2.x exhibited more limited sequence similarity with lichenicidin β, although some of the Ser, Thr, and Cys residues in the C-termini were similarly positioned (Fig. 1D). In general, the core peptides of the FlvA peptides are longer than those of the previously characterized α and β-peptides.

Further comparison of the FlvA peptides with other two-component lantibiotics revealed additional shared and unusual features. The FlvA1.a peptide has Ser20 and Cys30 positioned to form a putative lipid II binding motif (Fig. 1C, underlined) found in the lantibiotics mersacidin, lacticin 481, and nukacin ISK-1 (Dufour et al., 2007; Hsu et al., 2003; Islam et al., 2012; Szekat et al., 2003) and in the α-peptide of two-component lantibiotics (Begley et al., 2009; Dischinger et al., 2009; Martin et al., 2004; Oman and van der Donk, 2009). However, compared to the other α-peptides, the FlvA1.x peptides contain more Cys residues in the core peptide (Fig. 1C). The involvement of these Cys residues in thioether rings or disulfide bridges could endow the FlvA1.a peptide with increased stability, which may be important in the ruminal environment. Similar to the FlvA1.x peptides, many of the FlvA2.x core peptides (c-e, g, and h) also have additional Ser and Thr residues compared to other two-component β-peptides (Fig. 1D). Notably, some of these residues form a putative Dhx-Dhx-Xxx-Xxx-Cys motif (where Dhx = Dha or Dhb and Xxx = any amino acid), which has been shown in some lanthipeptides to result in MeLan rings with LL stereochemistry (Lohans et al., 2014; Tang et al., 2015; Tang and van der Donk, 2013). Corollary to the structural diversity of the FlvA2.x peptides would be a high substrate tolerance of the FlvM enzyme(s) responsible for installing the (Me)Lan rings.

Attempted detection of modified FlvA production in R. flavefaciens FD-1

Liquid cultures of anaerobically grown R. flavefaciens FD-1 were desalted and analyzed for the presence of FlvA peptides. MALDI-TOF-MS analysis did not provide clear indication of masses corresponding to putative peptide products of the flv cluster. In alternative attempts to elicit peptide production, R. flavefaciens FD-1 and Ruminococcus albus 7 were co-cultured as well as grown in proximity to one another on solid media, an approach that has been successful for other genera (Traxler et al., 2013; Yang et al., 2011). However, no new ions with m/z > 2000 could be observed by MALDI-TOF MS. These observations are representative of a commonly encountered hurdle in natural product research: the environmental triggers for expression of biosynthetic genes are generally poorly understood and often cannot be replicated in the laboratory (Bode and Müller, 2005; Winter et al., 2011). The inability to detect potential products of the flv cluster prompted consideration of alternate strategies to produce and assess the products generated from the flv genes.

Heterologous production of modified FlvA peptides

To access the products of the Flv system, a validated heterologous production strategy using Escherichia coli was employed. This approach has been successfully applied to the production of several lantibiotics and involves the co-expression of a LanA peptide and cognate LanM enzyme in E. coli, which typically results in full posttranslational modification of the peptide. The modified full length LanA peptide is then purified from E. coli and the leader peptide is removed in vitro using a variety of proteolytic strategies (Basi-Chipalu et al., 2015; Caetano et al., 2011; Garg et al., 2012; Lin et al., 2011; Nagao et al., 2005; Okesli et al., 2011; Shi et al., 2011; Tang and van der Donk, 2013; Wang et al., 2014b). The sequence similarity of FlvM1 with characterized two-component LanM1 enzymes is greater than the similarity of the latter enzymes with FlvM2. Thus, we hypothesized that FlvM1 was responsible for the modification of the putative FlvA1.x peptides and FlvM2 for the FlvA2.x peptides. In order to probe this hypothesis, pilot co-expression studies of an arbitrarily chosen peptide FlvA2.g with either FlvM1 or FlvM2 were undertaken. Co-expression of FlvA2.g with FlvM2 resulted in dehydrated peptides whereas co-expression of FlvA2.g with FlvM1 yielded no dehydration (Fig. S1A). Further experiments with all of the FlvA2.x peptides showed they were also dehydrated upon co-expression with FlvM2 in E. coli. Similarly, FlvA1.a was dehydrated upon co-expression with FlvM1 (Fig. S1B). Identical results were obtained in vitro with enzymes and substrates that were expressed as His₆-tagged fusion proteins and purified by immobilized affinity chromatography (e.g. Fig. 2A and S2). In all cases the number of dehydrations was observed to be less than the total number of Ser and Thr residues within the core peptide (Table 1, Fig. 2B, Fig. S3), but Ser and Thr residues that escape dehydration are not unusual in lanthipeptides (Rink et al., 2005) including the lichenicidins (Caetano et al., 2011; Shenkarev et al., 2010).

Figure 2.

Modification of the FlvA peptides. Modification of the FlvA peptides was observed as a loss of an integer number of 18 Da from the calculated, unmodified peptide mass. (A) In vitro modification of FlvA1.a by FlvM1. The red spectrum was observed upon exposure of FlvA1.a to FlvM1, whereas the black spectrum was observed in the absence of FlvM1. FlvA1.a six-fold dehydrated, calculated average [M+H]⁺ mass: 9832.74 m/z, observed average mass: 9833.65 m/z. FlvA1.a unmodified, calculated average [M+H]⁺ mass: 9941.07 m/z, observed average mass: 9937.60 m/z. For FlvA1.a and FlvM1 heterologous co-expression, see Fig. S1B. (B) In vivo modification of heterologously expressed FlvA2.b peptide (for the other FlvA2.x peptides, see Fig. S3). FlvA2.b six-fold dehydrated, calculated average [M+H]⁺ mass: 8463.51 m/z, observed mass: 8463.42 m/z. * Denotes ions with masses that correspond to gluconoylated starting material or product (Geoghegan et al., 1999).

Table 1.

Extent of dehydrations observed after co-expression of FlvA2.x with FlvM2 and in vitro treatment of FlvA1.a with FlvM1. The yields obtained for the core portion of the modified FlvA peptides range from 100–800 μg of purified peptide from 2 L of E. coli culture, depending on the method of removal of the leader peptide.

Substrate	Major product	Number of Ser and Thr in the core peptide	Number of Cys in the core peptide	Lan	MeLan
FlvA2.a	−7 H2O	8	1	DL	–
FlvA2.b	−6 H2O	9	3	–	DL
FlvA2.c	−5 H2O	10	4	LL	DL
FlvA2.d	−7 H2O	9	5	DL, LL	DL
FlvA2.e	−7 H2O	10	4	LL	DL
FlvA2.f	−4 H2O	5	2	–	DL
FlvA2.g	−7 H2O	9	4	LL	DL
FlvA2.h	−7 H2O	10	4	DL, LL	DL
FlvA1.a	−6 H2O	8	6	DL	DL

FlvM1 is selective for FlvA1 and FlvM2 is selective for the FlvA2 peptides

In order to assess the substrate tolerance of FlvM1, FlvA2.a and FlvA2.g were also incubated in vitro with FlvM1 under identical conditions. Analysis of the assay revealed little to no modification of either peptide by FlvM1 (Fig. S2). Similarly, FlvM2 was unable to modify FlvA1.a to any appreciable extent when exposed to the peptide in vitro under the standard reaction conditions (Fig. S2). The inability of FlvM1 to modify FlvA2.a and 2.g and FlvM2 to modify FlvA1.a suggests a high degree of substrate selectivity that is probably governed by recognition of the different leader peptides (Fig. 1B) (Thibodeaux et al., 2015; Yang and van der Donk, 2013).

Partial FlvA2.x leader peptide removal and structural analysis

The alignment of the amino acid sequences of the FlvA peptides shows a conserved GA/GG motif located in the middle of the peptides (Fig. 1B). Based on previous studies of other class II lanthipeptides, this motif likely marks the boundary between the FlvA leader and core peptides (Uguen et al., 2005). Usually, leader peptides ending in the GA/GG motif are removed by the Cys protease domain of LanT transporters (Håvarstein et al., 1995), and indeed such a protein is present in the gene cluster (FlvT, Fig. 1A). Evidence that the GA/GG sequence is indeed the cleavage motif for the FlvA peptides was provided by co-expression of FlvT with FlvM2 and FlvA2.g in E. coli. The supernatant of the co-expression was analyzed by MALDI-TOF MAS, which resulted in detection of masses corresponding to cleavage of modified FlvA2.g after the GA sequence (Fig. S6H). These masses were not observed in the culture co-expressing only FlvM2 and FlvA2.g. This observation is consistent with the previously reported removal of the leader peptide of modified LicA2 upon expression of the lic gene cluster, which contains licT, in E. coli (Caetano et al., 2011). Although the use of FlvT confirmed the predicted leader peptide cleavage site, this strategy unfortunately did not result in producing sufficient quantities of the Flvα and β-peptides for structure determination, in part because of the difficulty associated with purifying the desired, modified peptide from partially dehydrated peptides that were also secreted.

We therefore turned towards in vitro removal of the putative leader peptides from the His₆-tagged FlvA2.x peptides after co-expression with FlvM2 and purification by immobilized nickel affinity chromatography. The purified peptides were first subjected to proteolysis by endoproteinase Glu-C. Although FlvA2.f-2.h have Glu residues within their core peptides, appreciable proteolysis at these sites was not observed, probably because they are protected by the post-translational modifications. Instead, treatment with Glu-C resulted primarily in fragments arising from proteolysis within the leader peptide sequence, specifically at Glu−5 for FlvA2.a, Glu−6 for FlvA2.b through FlvA2.e and FlvA2.g, Glu−7 for FlvA2.h, and Glu−10 for FlvA2.f (for sequences, see Fig. 1B). The Glu-C-generated C-terminal fragments of the modified FlvA2.x peptides were purified by reversed-phase high performance liquid chromatography (RP-HPLC) and analyzed by electrospray ionization tandem mass spectrometry (ESI-MS/MS) (e.g. Fig. 3A).

Figure 3.

Structural analysis of the FlvM2-modified FlvA2 peptides. (A) Tandem mass spectrum of the modified FlvA2.b peptide after treatment with the endoprotease Glu-C. [M+H]⁺ denotes the peak corresponding to the protonated, monoisotopic parent mass. Fragment ions were not observed at the C-terminus of the peptide. For tandem mass spectra of the remaining FlvA peptides, see Fig. S4. (B) GC/MS, selected ion monitoring at 379 Da, indicating the presence of MeLan. The identity of the eluting material was confirmed by co-elution with an authentic MeLan standard (Fig. S5B).

A lack of fragmentation was observed in the C-termini of all of the modified FlvA2.x peptides analyzed, suggesting the presence of thioether rings (Fig. 3A and S4). ESI-MS/MS analysis of modified FlvA1.a also indicated the absence of fragmentation in the C-terminus of the peptide, consistent with cyclization in this region of the peptide (Fig. S4). To confirm the presence of (Me)Lan residues within the FlvM-modified FlvA1.a and FlvA2.x peptides, the peptides were hydrolyzed and the resulting residues were derivatized to the corresponding pentafluoropropionamide methyl esters. These volatile derivatives were then separated by chiral gas chromatography monitored by mass spectrometry (GC/MS), which confirmed that the modified FlvA1.a and FlvA2.x peptides indeed contained (Me)Lan residues (Fig. 3B and Fig. S5). The stereochemistry of the observed (Me)Lan residues was determined by using synthetic standards with known stereochemistries (Ross et al., 2010), and is listed in Table 1. As anticipated based on previous predictions (Tang and van der Donk, 2013), peptides containing a Dhx-Dhx-Xxx-Xxx-Cys motif resulted in LL-Lan whereas all other Lan and MeLan residues had the DL stereochemistry.

The possibility of incomplete cyclization was assessed by iodoacetamide (IAA) assays; iodoacetamide selectively alkylates free Cys residues and not thioethers. For all FlvM-modified peptides except FlvA1.a, treatment with IAA did not result in appreciable formation of adducts, which is indicative of complete cyclization; di-alkylation was observed with FlvM1-modified FlvA1.a (Fig. S6C). Based on the fragmentation pattern (Fig. S6D) and the alignment with other α-peptides (Fig. 1C), the di-alkylation product most likely involves alkylation at Cys8 and Cys29. Two non-cyclized Cys are also present in haloduracin α, but the Cys residues involved do not align with those of Flvα.a. The two free Cys residues in modified FlvA1.a presented the possibility of a disulfide bond in the structure of mature FlvAα.a. The antimicrobial activity of some lantibiotics has been reported to depend on the presence of such disulfide bonds (Kabuki et al., 2009; Lin et al., 2011; Wang et al., 2014b; Zhang et al., 2014a). Modified FlvA1.a was therefore incubated with oxidized and reduced glutathione in an attempt to oxidatively fold the peptide (Oman et al., 2011a). However, the peptide remained reduced under these conditions, as observed by ESI-MS, suggesting that unlike lantibiotics that contain disulfide bonds, the two Cys in modified FlvA1.a are not in a conformation in which they readily form a disulfide bond (Basi-Chipalu et al., 2015; Lin et al., 2011).

Proteolytic removal of the remaining portion of the FlvA2.x leader peptides

Complete removal of the leader peptide was challenging as introduction of artificial protease cleavage sites resulted in either incomplete modification by the FlvM enzymes, incomplete proteolysis because of the post-translational modifications near the cleavage site, or cleavage in the core peptides, which are problems that have been previously observed for other lanthipeptides (Garg et al., 2012; Plat et al., 2011; Tang and van der Donk, 2012). We therefore resorted to treating the Glu-C-generated FlvA2.x peptide fragments with aminopeptidase (Bindman and van der Donk, 2013; Majchrzykiewicz et al., 2010; Shi et al., 2012). It was anticipated that the proteolytic activity of aminopeptidase would terminate upon encountering the dehydrated residues featured at or near the predicted N-termini of most of the FlvA2.x peptides (Fig. 5). MALDI-TOF-MS analysis of aminopeptidase reactions indicated that for FlvA2.b, the predicted native product was formed. For FlvA2.c, FlvA2.e, and FlvA2.g, aminopeptidase removed all remaining residues of the leader peptides as well as the predicted first residue of the core peptides, resulting in N-terminal lanthionines. We term these products Δ1-Flvβ.c, Δ1-Flvβ.e, and Δ1-Flvβ.g. For FlvA2.d and FlvA2.h, the aminopeptidase treatment left two amino acids of the leader peptide on the final products, likely because of the high Gly content at the junction between the core and leader peptides, making these peptides less efficient substrates for aminopeptidase (Velásquez et al., 2011). We term these products +2Flvβ.d and +2Flvβ.h. For FlvA2.a, three residues remained after aminopeptidase treatment, resulting in +3Flvβ.a. Finally, in the case of FlvA2.f, seven residues from the leader peptide remained after aminopeptidase treatment (Fig. S6G; +7Flvβ.f).

Figure 5.

Tandem MS fragmentation patterns and proposed structures of the modified FlvA core peptides. The FlvA peptides are divided into the categories that are discussed in the text, and the alignments highlight the similar ring topology of the peptides in each category. X = Dha or Ala, when engaged in a Lan, and Z = Dhb or Abu, when engaged in a MeLan.

This sequential digestion approach works reasonably well when the N-terminus of the lanthipeptide is blocked by post-translational modifications, but if the N-terminus is a linear sequence as predicted for the product of FlvA1.a, the method does not work. Instead, removal of the FlvA1.a leader peptide was achieved via a single step proteolysis using chymotrypsin (Fig. S6B). Assuming that the GA/GG sequence marks the end of the leader peptide, chymotrypsin cleaves after Trp2 in the core peptide of FlvM1-modified FlvA1.a, and therefore the resulting peptide was designated Δ2-Flvα.a. In an effort to avoid removal of the two N-terminal amino acid residues of the FlvA1.a core peptide, a mutant FlvA1.a peptide with Glu inserted before the first residue of the core peptide was generated (termed FlvA1.a(A−1insE)), and co-expressed with FlvM1. Doing so resulted in six-fold dehydrated FlvA1.a(A−1insE), and treatment of this mutant peptide with Glu-C yielded the desired Flvα.a (Fig. S6E).

Antimicrobial activity of the modified FlvA core peptides

The peptides produced by in vitro leader peptide removal were purified by RP-HPLC, and the modified FlvA peptides were first assayed for antimicrobial activity against the lantibiotic-sensitive, aerobic microorganism Micrococcus luteus DSM 1790. In an agar diffusion test, Δ1-Flvβ.c and Δ1-Flvβ.e exhibited antimicrobial activity that was not synergistically enhanced by Δ2−Flvα.a (Fig. 4). In contrast, combinations of Flvβ.b and Δ1-Flvβ.g with Δ2−Flvα.a displayed the synergistic antimicrobial activity that is characteristic of two-component lantibiotics. A comparable level of synergistic, antimicrobial activity was also observed for a combination of Flvα.a and Flvβ.b (Fig. S6F), suggesting that the removal of the additional two amino acids at the N-terminus of Δ2−Flvα.a is not detrimental for bioactivity. Since we could obtain considerably larger quantities of Δ2−Flvα.a, this peptide was used for all subsequent bioassays. Δ1-Flvβ.g also displayed activity by itself that was lower than that observed when spotted with Δ2−Flvα.a. The remainder of the Flvβ-peptides (+3Flvβ.a, +2Flvβ.d, +7Flvβ.f, and +2Flvβ.h) did not display antimicrobial activity, whether tested separately or in combination with Δ2−Flvα.a (Fig. 4). Enhanced activity beyond what was observed in pairwise combination was also not detected by combining all peptides. In light of their antimicrobial activity, Flvα.a, β.b, β.c, β.e, and β.g were designated flavecins.

Figure 4.

Bioassay of FlvM-modified FlvA core peptides against M. luteus DSM 1790. The modified Δ2-Flvα.a peptide was assayed at a concentration of 1 mM and the Flvβ.x peptides were assayed at a concentration of 0.5 mM. A volume of 1 μL was used for spots consisting of a single peptide while 0.5 μL of each solution was used for spots combining two peptides; nisin was assayed using 0.2 μL of a 100 μM solution.

M. luteus is often used as an indicator strain to test for antimicrobial activity of lipid II-targeting peptides such as two-component lantibiotics, but it is not a rumen bacterium. To investigate the activity of the peptides against bacteria that would be more relevant in the ruminal context, the flavecins were assayed against Ruminococcus albus 7 and R. flavefaciens C94 under anaerobic conditions. Interestingly, very similar patterns of activity were observed with Δ2−Flvα.a, Flvβ.b, and Δ1-Flvβ.e (Fig. S6A). As with M. luteus, the activity was weak, suggesting that these organisms are not the intended target of these compounds. Alternatively, the weak antimicrobial activity exhibited by the flavecins may suggest that these peptides mediate more subtle microbial interactions (D’Onofrio et al., 2010).

Discussion

The sequence similarity of LanM enzymes was used to survey genomes for unusual class II lanthipeptide biosynthetic systems, which resulted in the identification of the novel flv gene cluster in R. flavefaciens FD-1. Recently, Singh and Sareen also bioinformatically identified the flv cluster via the sequence similarity of FlvT to HalT (Singh and Sareen, 2014), but this previous study did not characterize the enzymes or the products of the cluster nor discussed its unusual characteristics. Some features of the flv cluster are reminiscent of two-component lantibiotics but the high number of substrate genes with diverse sequences is not consistent with a typical two-component system. Previous reports on antimicrobial peptides produced by species from the Ruminococcus genus (Chen et al., 2004; Odenyo et al., 1994; Russell and Mantovani, 2002) suggested that the mature FlvA peptides could represent antimicrobial defenses of R. flavefaciens FD-1. In addition, previous studies of the lantibiotics ruminococcin A and butyrivibriocin OR79A (Dabard et al., 2001; Kalmokoff et al., 1999; Marcille et al., 2002) suggested that anaerobic bacteria associated with the digestive system are a source of bioactive lanthipeptides. Thus, characterization of the flv gene cluster and the resulting modified FlvA peptides was undertaken.

The structure of Flvα.a suggested by sequence homology and tandem MS data is consistent with those of other α-peptides of two-component systems. Although a general lack of fragmentation from the middle of the peptide to the C-terminus makes definitive assignment of the ring structure in this region difficult (Fig. S4I), the ring pattern is very likely the same as those in Halα, Ltnα, and Licα (Fig. 1C,D) (Cooper et al., 2008; Martin et al., 2004; Shenkarev et al., 2010). The most N-terminal Cys residue does not appear to be cyclized since fragmentation is observed on either side of the residue, and IAA assays showed the presence of two free Cys (Fig. S4I and S6C). The possibility that these two Cys might be engaged in a disulfide bond was not supported by attempts to oxidatively cyclize them.

Based on the results obtained from tandem MS and GC/MS, the modified FlvA2 peptides can be divided into two groups (Fig. 5). The first group features an N-terminal ring, and the second category does not have the residues required to form this N-terminal ring. The assignment of Flvβ-peptides into the former group was based on the lack of fragmentation at the N-terminal extremity of these peptides as well as the LL stereochemistry of the Lan rings observed by GC/MS. Such a ring is also present in the β-peptides of haloduracin and lichenicidin (Cooper et al., 2008; Martin et al., 2004; Shenkarev et al., 2010). The central portion of all Flvβ-peptides contains a ring consisting of seven residues. The assignment of this ring is based on the tandem MS data that show a clear lack of fragmentation for all Flvβ-peptides in this region (Fig. S4). This ring also nearly aligns to the B ring of other two-component peptides, but the latter feature a smaller five-amino-acid ring in this region (Fig. 1D). The size of this central ring is not entirely clear for Flvβ.b, β.c, β.e, and β.g since a six-amino-acid ring is also possible. However, formation of a seven-amino-acid ring is more likely based on Flvβ.a, β.d, β.f and β.h, which unambiguously contain such a ring. Also, comparison of the fragmentation patterns of Flvβ.b, β.c, β.e and β.g imply that one Ser/Thr within this region consistently escapes modification (Fig. 5), suggesting the Ser/Thr that could result in a six-amino-acid ring is actually not dehydrated.

Two combinations of peptides resulted in synergistic bioactivity. Adding Δ2-Flvα.a to either Flvβ.b or Δ1-Flvβ.g resulted in clearly enhanced zones of growth inhibition. A comparison of the structures determined for Flvβ.b and Δ1-Flvβ.g indicates that the similarities are all in the C-terminal region, suggesting that this segment of the peptide is responsible for the interactions of the Flvβ peptides with the Flvα peptide. This observation is in agreement with the extensive mutagenesis data on the β-peptide of lacticin 3147 (Cotter et al., 2006). Two possibilities were considered for the relatively low observed antibacterial activity. Perhaps the small deviations between the predicted leader peptide removal sites and the products obtained after in vitro leader peptide removal resulted in imperfect interactions of the peptides. We did not find this explanation very likely since the N-termini of the eight Flvβ-peptides (and other two-component lanthipeptides) are already highly heterogeneous by sequence (Figure 1D), and more importantly, the antimicrobial activity observed for the predicted Flvα.a and β.b core peptides was comparable to the combination of Δ2−Flvα.a and β.b (Fig. S6F). Alternatively, perhaps the flv cluster is a pseudo cluster. We find this possibility also highly unlikely. It seems no coincidence that the four copies of the α-peptide are highly conserved, consistent with the general model of two-component lantibiotics in which this peptide recognizes a structurally conserved target (lipid II). The observation that the cluster encodes four such copies and eight copies of the β-peptides in light of a 1:2:2 stoichiometry observed for lipid II:Halα:Halβ (Oman et al, 2011b) also seems non-coincidental, but attempts to obtain increased activity by various combinations of peptides have thus far not been successful.

The weak observed activity may instead reflect our inability to identify the physiological target organism(s). Ruminal two-component lantibiotics may be much more fine-tuned for specific target organisms than observed in other environments. It has been reported that bacterial diversity in vertebrate-associated communities is quite different from that in soil or aquatic environments, with 16S rRNA gene-based trees indicating extensive diversity at the species level rather than at higher levels as observed in the latter environments (Backhed et al., 2005; Dethlefsen et al., 2007; Kim et al., 2011; Ley et al., 2008). Perhaps such an environment has promoted the thus-far unique example of a multi-component lantibiotic system with eight diverse β-peptides. The corollary of this hypothesis is that it may be very difficult to identify the specific physiological target organism of these peptides, especially if the flavecins exhibit the narrow target specificity that is not unusual for RiPPs (Scholz et al., 2011; Wang et al., 2014a). In turn, this may explain the rather weak activity observed against the organisms tested thus far. At present, still very few lanthipeptide gene clusters are known from ruminant environments (Azevedo et al, 2015) and we have not observed any other two-component systems in the available genomes. As discussed below, this may be the result of the still limited available genome sequence information, and as more genomes are sequenced, additional such examples may be uncovered.

The current work on the genetic capacity of R. flavefaciens FD-1 to produce a diverse set of eight different β-peptides has similarities but also important differences compared to previous examples of combinatorial biosynthesis of lanthipeptides. In 2010, the first example was reported of a single enzyme making 30 different lanthipeptides in the cyanobacterium Prochloroccocus MIT 9313 (Li et al., 2010). At the time this was a curious finding isolated to one genome, but in the intervening time as more genomes have been sequenced, this combinatorial biosynthesis has been found in many different cyanobacteria genera (Zhang et al., 2014b). To date, none of the products has shown any antimicrobial activities and the products have all had very different structures. In contrast, the example described here is clearly different. Firstly, the four copies of the α-peptide are essentially identical and all contain the highly conserved lipid II binding motif (Fig. 1C), strongly suggesting that the products are antimicrobial peptides. Secondly, the system contains two enzymes that have clearly defined roles, unlike the one enzyme in the cyanobacterial systems that has evolved for promiscuity. Thirdly, unlike the one cyanobacterial enzyme, which in order to be able to accommodate a highly diverse set of substrates is a very slow enzyme (Thibodeaux et al., 2014), the FlvM2 enzyme is very efficient in processing the eight FlvA2.x substrates that are diverse but that have maintained a certain core set of rings at the C-terminus. All of these observations point at a two-component lantibiotic system where the diversification of the β-peptides has been beneficial in the specific growth environment of the producer organism.

In conclusion, we present the characterization of the ring topology and stereochemistry of nine different lanthipeptides encoded in the genome of R. flavefaciens FD-1. A subset of the modified Flvβ core peptides (Flvβ.b and Δ1-Flvβ.g) demonstrated enhanced antibacterial activity when used in combination with Flvα.a, demonstrating that at least a subset of these peptides constitute two-component lantibiotic systems. Although the FlvA2 peptides display appreciable variation in their amino acid sequence, their post-translational modification is orchestrated by only one FlvM2 enzyme that is responsible for the modification of eight FlvA2 peptides with quite diverse sequences and final ring topologies. Elucidation of the features of the Flv system sets the stage for more in depth investigation of the biological roles of this curious set of lanthipeptides.

Significance

A heterologous expression system was used to access post-translationally modified lanthipeptides from the anaerobic Ruminococcus flavefaciens FD-1, thus enabling characterization of their structure and activity. FlvM2 is a substrate-tolerant enzyme that dehydrates and cyclizes eight different FlvA2 peptides. A subset of the FlvA2.x modified peptides displays synergistic antimicrobial activity in the presence of the FlvM1-modified FlvA1.a peptide. Thus, the flv system is the first example of a two-component lantibiotic gene cluster with several unique β-peptides.

Experimental Procedures

For cloning and biochemical detailed procedures, see the Supplemental Experimental Procedures. Expression and purification of the FlvA peptides was carried out as previously described with modifications to the protocol (Li et al., 2009). The LanM in vitro, IAA assay, and oxidative folding reaction conditions were adapted from previous reports. MALDI-TOF-MS analysis of desalted samples was completed using a Bruker UltrafleXtreme MALDI-TOFTOF-MS maintained in the UIUC School of Chemical Sciences Mass Spectrometry Laboratory.

Proteolysis reactions ranged in volume from 20 μL to 20 mL and contained final concentrations of 250 mM Tris-HCl, pH 7.5, 2 mg/mL of peptide, and 0.2–1 μg/mL of protease. After approximately 1 h incubation at room temperature, the reactions were desalted by ZipTip, as per the manufacturer’s instruction (EMD Millipore). The content of the ZipTip was directly eluted onto a Bruker MTP 384 polished steel target plate with 1 μL of sinapic acid matrix. The sample was analyzed by MALDI-TOF-MS for the complete consumption of full length substrate peptide. The remainder of the peptide in the proteolysis reaction was desalted by SPE.

The pellet was removed from cultures obtained after the co-expression of FlvM2, FlvA2.g and FlvT by centrifugation at 11,900×g. The supernatant was then decanted and subjected to ZipTip and MALDI-TOF-MS analysis as stated above.

Analytical scale RP-HPLC was completed using an Agilient 1260 Infinity equipped with a Phenomenex Luna column (10 μm, C18(2), 100 Å, 250×4.6 mm) and managed using Agilent Instrument 1 (Online) software. Semi-preparative scale RP-HPLC was performed using a Shimadzu Prominence equipped with a Phenomenex Luna column (10 μm, C18(2), 100 Å, 250×10.0 mm), managed using the Shimadzu program LC Real Time Analysis. Peptide hydrolysis, Lan derivatization, and GC/MS was adapted from a previous report (Ross et al., 2010). A Waters Synapt ESI-QTOF coupled to an Acquity ultra high performance liquid chromatography (UPLC) system was used for tandem mass spectrometry analysis.

The liquid media used to culture R. flavefaciens FD-1 consisted of 4.0 g of cellobiose, 2.0 g of Bacto-tryptone, 50 mL mineral solution 1, 50 mL mineral solution 2, and 10 mL volatile fatty acid solution in 820 mL of distilled water (see Supplemental Experimental Procedures for solution compositions). After autoclaving but prior to inoculation, 50 mL of an 8% sodium carbonate solution and 20 mL of a 1.25% cysteine-sulfide solution was added.

Culturing procedures were performed using anaerobic technique and as previously described (Berg Miller et al., 2009). M. luteus was maintained as 40% glycerol stocks at −80 °C. The glycerol stocks were revived by first streaking onto solid media and incubating at 37 °C for 16–20 h. A single colony of the organism was transferred into 5 mL of liquid media and incubated at 37 °C, 200–220 rpm for 12–16 h. Agar diffusion assay plates were prepared by seeding 20 mL of molten solid media (~42 °C) with 50 μL of overnight liquid culture. Following thorough mixing, the seeded solution was poured into a petri dish and allowed to solidify at room temperature. Samples to be assayed were dissolved in DMSO and 1 μL aliquots were spotted directly onto the solidified media. When two peptide solutions were combined, only 0.5 μL aliquots of each solution were spotted in a single area. The Δ2−Flvα.a peptide was assayed at a concentration of 1 mM while the FlvA2.x peptides were assayed at a concentration of 0.5 mM. Nisin was assayed at a concentration of 100 μM and 0.2 μL of the solution was spotted on the plate. Once all spots were dry, the petri dish was incubated at 37 °C for 16–20 h prior to visualization using a Bio-Rad Molecular Imager XR+ Gel Doc Imaging System and the accompanying Quantity One 4.6.9 program.

Highlights.

• Production of nine different lanthipeptides from an anaerobic ruminant bacterium

• Novel example of substrate diversification in two-component lanthipeptides

• Flavecin synthetase FlvM2 converts eight diverse peptides into polycylic structures