Genetic Determinants of Reutericyclin Biosynthesis in Lactobacillus reuteri

Abstract

Reutericyclin is a unique antimicrobial tetramic acid produced by some strains of Lactobacillus reuteri. This study aimed to identify the genetic determinants of reutericyclin biosynthesis. Comparisons of the genomes of reutericyclin-producing L. reuteri strains with those of non-reutericyclin-producing strains identified a genomic island of 14 open reading frames (ORFs) including genes coding for a nonribosomal peptide synthetase (NRPS), a polyketide synthase (PKS), homologues of PhlA, PhlB, and PhlC, and putative transport and regulatory proteins. The protein encoded by rtcN is composed of a condensation domain, an adenylation domain likely specific for d-leucine, and a thiolation domain. rtcK codes for a PKS that is composed of a ketosynthase domain, an acyl-carrier protein domain, and a thioesterase domain. The products of rtcA, rtcB, and rtcC are homologous to the diacetylphloroglucinol-biosynthetic proteins PhlABC and may acetylate the tetramic acid moiety produced by RtcN and RtcK, forming reutericyclin. Deletion of rtcN or rtcABC in L. reuteri TMW1.656 abrogated reutericyclin production but did not affect resistance to reutericyclin. Genes coding for transport and regulatory proteins could be deleted only in the reutericyclin-negative L. reuteri strain TMW1.656ΔrtcN, and these deletions eliminated reutericyclin resistance. The genomic analyses suggest that the reutericyclin genomic island was horizontally acquired from an unknown source during a unique event. The combination of PhlABC homologues with both an NRPS and a PKS has also been identified in the lactic acid bacteria Streptococcus mutans and Lactobacillus plantarum, suggesting that the genes in these organisms and those in L. reuteri share an evolutionary origin.

INTRODUCTION

Reutericyclin, produced by Lactobacillus reuteri, is the only chemically characterized low-molecular-weight antibiotic produced within the genus Lactobacillus (1). It is an N-acylated tetramic acid (2) with bacteriostatic and bactericidal activity against Gram-positive bacteria, including Staphylococcus aureus, Listeria innocua, Enterococcus faecium, Clostridium difficile, and bacilli that cause ropy spoilage of bread (3, 4). Tetramic acids are produced mainly by fungi (5); known bacterial producers of tetramic acids include Streptomyces spp., Alteromonas spp., Stenotrophomonas spp., and Lysobacter enzymogenes, all of which are phylogenetically unrelated to lactobacilli (5, 6, 7). Tetramic acids are produced by polyketide synthases (PKS) and nonribosomal peptide synthetases (NRPS). PKS and NRPS are enzymes responsible for the biosynthesis of many secondary metabolites with many functions, e.g., as pigments, virulence factors, or infochemicals, or for defense (5, 8). However, genes coding for these enzymes are rarely found in genomes smaller than 3 Mbp (9). Lactobacillus reuteri and other Lactobacillus spp. have evolved by reduction of genome size to succeed in very narrow ecological niches (10, 11), and functional PKS/NRPS systems have not been reported for lactobacilli. Moreover, despite the extensive body of literature related to the antimicrobial activity of food-fermenting lactic acid bacteria, only a few strains of L. reuteri have been shown to produce reutericyclin, indicating that reutericyclin production is a rare trait within this bacterial group (1, 12, 13, 14). Reutericyclin production has been described for only four strains within the species L. reuteri: strains LTH2584, TMW1.106, TMW1.112, and TMW1.656 (15). These strains were isolated in 1988, 1994, and 1998 from the same industrial sourdough (SER), which has been maintained by continuous propagation (15). Another strain, L. reuteri LTH5448, which was isolated from a different sourdough produced in the same facility (16), does not produce reutericyclin, but its resistance to reutericyclin is comparable to the resistance of producing strains (17).

Although the chemical structure of reutericyclin had been elucidated 15 years ago (1, 2), the genes involved in reutericyclin biosynthesis were unknown. The aim of this study was therefore to determine the genetic basis of reutericyclin biosynthesis and immunity by combining a comparative genomic gene-finding approach with functional characterization of null mutants.

MATERIALS AND METHODS

Strains and culture conditions.

L. reuteri LTH2584, TMW1.106, TMW1.112, and LTH5448 were grown anaerobically in modified MRS (mMRS) broth at 37°C overnight. Each liter of mMRS broth contained 10 g tryptone, 10 g maltose, 5 g glucose, 5 g fructose, 5 g beef extract, 5 g yeast extract, 4.0 g potassium phosphate dibasic, 2.6 g potassium phosphate monobasic, 2 g tri-ammonium citrate, 0.5 g l-cysteine, 0.2 g magnesium sulfate, 0.05 g manganese sulfate, 1 g Tween 80, and 1 ml of a vitamin mixture containing 0.5 μg each of vitamins B₁, B₂, B₆, and B₁₂, folic acid, and pantothenic acid.

DNA extraction, genomic sequencing, and assembly.

Cells harvested by centrifugation from a 2-liter overnight culture were flash frozen at −80°C and were shipped to the SMRT Sequencing Laboratory, Institute for Computational Biomedicine, Weill Medical College of Cornell University (Ithaca, NY, USA), for genomic sequencing. Assemblies were produced by scaffolding PacBio long reads and medium-insert (∼550-bp) Illumina paired-end reads with the AHA scaffolding pipeline in SMRT Analysis software, version 2.0. Consensus sequences for the final drafts were produced with the Quiver algorithm. Final assemblies resulted in 48, 37, 25, and 24 scaffolds for L. reuteri TMW1.112, LTH5448, LTH2584, and TMW1.656, respectively, and the respective genome sizes were 2,033,533, 1,980,298, 1,944,170, and 2,066,054 bp. Genomes were annotated using the JGI annotation pipeline, and the genome sequences were deposited in GenBank.

Identification of genes involved in reutericyclin production.

Genomic analysis was performed using the Integrated Microbial Genomes (IMG/ER) system of the Joint Genome Institute (18). The genome sequences of the reutericyclin producers L. reuteri TMW1.656, TMW1.112, and LTH2584 were compared with those of 11 other L. reuteri strains, i.e., strains ATCC 53608, ATCC 55730, mlc3, I5007, 100-23, JCM1112/F275, TD1, Ipuph, CF48-3A, MM2-3, and MM4-1A. None of those 11 strains have been reported to produce reutericyclin. The loci identified were annotated using Geneious, version 6.1.6 (Biomatters Limited). Database searches were performed using the BLASTp and CD-search programs. The NRPS-PKS program (http://www.nii.res.in/nrps-pks.html) (19) was used to predict the modules and domains of NRPS and PKS and to predict the substrate specificity of the NRPS adenylation domains (A domains). The sequences of the gene clusters containing genes involved in reutericyclin biosynthesis were manually annotated and were deposited in GenBank.

Generation of mutants of L. reuteri TMW1.656.

The rtcN and rtcA genes in L. reuteri TMW1.656 were truncated by using the double-crossover method described previously (20). The rtcT and rtcRS genes were deleted in L. reuteri TMW1.656ΔrtcN by using the same method. The plasmids and primers used are listed in Tables 1 and 2. Gene deletions in all mutant strains were verified by Sanger sequencing.

TABLE 1.

Primers used to generate derivatives of L. reuteri TMW1.656 by double-crossover mutagenesis

Primer name	Primer sequence (5′–3′)
N1	AAGGATCCGGTTGTTGGAAGAGGATTTGAA
N2	TATAGAGTCGACTTAACTAGTAAGATTCATTGATAACATCCTT
N3	ATAACAGTCGACAAGGAGCATTAATTTCCATGAAA
N4	AACTGCAGTGTTCAGGAATGATTTTTGAGG
P1	GCAGGATCCCGACTTGGAGGAATATGGAA
P2	AACGGCGTCGACGTTAGACATTATCCCAATATCAGTCATTAT
P3	GCATTAGTCGACTTTATCTAAAGGAAGTTATAGTTATGACTGAAAA
P4	TGCTGCAGCACGTGCAATATTTCCACCA
T1	TTAGGATCCGGAAGAATGGCAGATTCCAA
T2	ATTCGAGTCGACTTATCCTTTCATCGGTACCTTTAATC
T3	ACGGATGTCGACTCTAAATCGTAATATTTATGCGTAATTACA
T4	GCCTGCAGGGGTTCAAATCCCCTTGTCT
R1	ATTAAGCTTGGTCCAATAATTGCAAAACCA
R2	ATTGGATCCTAACATTCTATTGGTCATTTATTGATTCTC
R3	GCAGGATCCAATGACTGATATTGGGATAATGTCTTA
R4	ACTTGAGTCGACATTGCTGTTGACCCAGAACC
A1	TGAGGATCCTGGACCATTTTCTGCAGTTG
A2	TTCAATGTCGACCTAATTCATTGGCTCATCAAG
A3	TCCGCTGTCGACGTAATTTAGACTAATAACGGGCTTGATG
A4	TATAAGCTTCTTTGCGCATAAGGATGACA

TABLE 2.

Bacterial strains and plasmids used to generate derivatives of L. reuteri TMW1.656 by double-crossover mutagenesis

Quantification of gene expression.

The expression of genes in L. reuteri TMW1.656 and its mutants was quantified by reverse transcription-quantitative PCR. Bacteria were grown in mMRS medium at 37°C to an optical density (OD) of 0.5. RNA was extracted using RNAprotect Bacteria Reagent and an RNeasy minikit (Qiagen, Mississauga, Canada) and was reverse transcribed to cDNA by using a QuantiTect reverse transcription kit (Qiagen) according to the manufacturer's instructions. Quantitative PCR was performed on a 7500 Fast real-time PCR system (Applied Biosystems) using SYBR green reagents (Qiagen). DNase-treated RNA samples served as negative controls. Relative gene expression was calculated as (E_target)^{ΔCP(control − sample)}/(E_reference)^{ΔCP(control − sample)}, where E_target is the PCR efficiency for the target gene, E_reference is the PCR efficiency for the reference gene, and ΔCP is the difference in the crossing points for control (wild type) and sample (mutant strains) reactions (21). Wild-type L. reuteri TMW1.656 was used as the control, and d-alanyl-alanine synthetase (ddl) gene was used as the reference gene. The primers used for the quantification of gene expression are shown in Table 3. Results are reported as means ± standard deviations for three biological replicates.

TABLE 3.

Primers used for quantification of gene expression

Gene	Directiona	Oligonucleotide sequence (5′–3′)
rtcP	F	GGGGGTGATAATCCAATTTTT
	R	TGTTCCATTTTGAACGAAGG
rtcK	F	CGCAACAACGCATTTTCTTA
	R	AGCGGAAGAAGCAAAGACAC
rtcC	F	GGACTTTTCAACGGGTTTCC
	R	TGCAGTTCCTTCAGGACAAA
rtcT	F	TCCTCCAATAGCTGGTCCAC
	R	CGTTTCTTTCAAGGGATTGG
rtcR	F	TGACCAATAGAATGTCAATACAAAA
	R	TTTCGAGAAATACCCGCAGT
rtcS	F	GGAAGAATGGCAGATTCCAA
	R	TCTCGCCTTCTTTTTGCATT

^aF, forward; R, reverse.

Determination of reutericyclin production and resistance to reutericyclin.

Reutericyclin production by wild-type and mutant strains of L. reuteri was determined by quantification of the antimicrobial activity of cell-free culture supernatants by a critical dilution assay as described previously (4). To obtain an assay that was specific for reutericyclin, Lactobacillus sanfranciscensis DSM 20451 and L. reuteri LTH5448 were used as reutericyclin-sensitive and reutericyclin-resistant indicator strains, respectively. The growth of indicator strains was judged by measuring the optical density at 595 nm. To determine the resistance of wild-type and mutant strains to reutericyclin, the compound was isolated from culture supernatants of L. reuteri TMW1.656 as described previously (4). In brief, overnight L. reuteri cultures were subcultured with a 10% inoculum and were incubated overnight at 37°C. Cells were collected by centrifugation, washed with phosphate buffer, and extracted with 30% (vol/vol) isopropanol. NaCl was added to the cell extract to saturation, and the organic phase was collected. The antimicrobial activity of reutericyclin was determined by using a critical dilution assay on microtiter plates (4). Twofold serial dilutions of the reutericyclin stock solution were prepared with mMRS medium, and solvents were evaporated under a flow of sterile air for 2 h. The microtiter plates were then inoculated with the strains to a cell count of about 10⁷ CFU/ml and were incubated overnight at 37°C. Lactobacillus sanfranciscensis DSM 20451 and L. reuteri LTH5448 were used as reutericyclin-sensitive and reutericyclin-resistant indicator strains, respectively. The growth of indicator strains was judged by measuring the optical density at 595 nm.

Bioinformatic and phylogenetic analysis of genes coding for NRPS-PKS in the order Lactobacillales and related organisms.

The sequences of all putative proteins encoded by the reutericyclin gene cluster were searched against the NCBI nonredundant (nr) database using BLASTP. NRPS and PKS sequences from strains that possess both proteins, with the same organization as that in L. reuteri, were collected. Likewise, sequences homologous to RtcA, RtcB, or RtcC of L. reuteri were collected from bacterial strains that have all three proteins. Protein sequences were aligned by Muscle (22). The phylogenetic trees were constructed using the best model estimated by MEGA6 (23).

Nucleotide sequence accession numbers.

The genome sequences for L. reuteri TMW1.112, LTH5448, LTH2584, and TMW1.656 have been deposited in GenBank under BioProject accession number PRJNA248653. The sequences of the gene clusters containing genes involved in reutericyclin biosynthesis have been deposited in GenBank under accession numbers KJ659887 (identical for L. reuteri TMW1.656, TMW1.112, and LTH2584) and KJ659888 (for L. reuteri LTH5448). The sequences of the reutericyclin biosynthesis gene clusters of the mutant strains derived from L. reuteri TMW1.656 have been deposited in GenBank as notes to the sequence for the wild-type strain under accession number KJ659887.

RESULTS

Identification of candidate genes involved in reutericyclin biosynthesis.

We used a comparative genomic approach to identify the genes responsible for reutericyclin biosynthesis. We identified genes that are present in the three reutericyclin-producing strains L. reuteri LTH2584, TMW1.656, and TMW1.112 but are absent in other L. reuteri strains. L. reuteri LTH5448 was initially excluded from this analysis. The analysis revealed 15 open reading frames (ORFs) unique to reutericyclin-producing strains, 12 of which were organized as a cluster (see Table S1 in the supplemental material) in a putative genomic island absent in reutericyclin-sensitive L. reuteri strains. The cluster spans about 14 kb and contains two putative transcriptional terminators located 8 bp downstream of rtcT and 950 bp upstream of rtcP. The sequences of the genomic islands in all reutericyclin-producing strains are identical, while that in the non-reutericyclin-producing strain L. reuteri LTH5448 differs at two sites (Fig. 1). A single base deletion at bp 1812 in L. reuteri LTH5448 rtcN introduces a stop codon, while rtcA is interrupted by a transposon at bp 850. Either of these two mutations may account for the loss of reutericyclin production in L. reuteri LTH5448 (17) (Fig. 1).

FIG 1.

Comparison of the reutericyclin gene clusters of the reutericyclin-producing L. reuteri strains with that of the reutericyclin-resistant, non-reutericyclin-producing L. reuteri strain LTH5448. Two mutations led to the loss of reutericyclin production in LTH5448: a single base pair deletion introducing a stop codon (indicated by an asterisk) and resulting in a truncated rtcN gene, and a transposon insertion in rtcA.

The cluster contains nine genes whose annotated functions relate to the synthesis of reutericyclin, its chemical structure, or resistance against its action (Fig. 2). rtcP encodes a 24.9-kDa protein predicted to be 4′-phosphopantetheinyl transferase, which activates polyketide synthases (PKS) and nonribosomal peptide synthases (NRPS) by transferring the 4′-phosphopantetheinyl arm to a conserved serine residue (24). RtcP is 27% identical to the 4′-phosphopantetheinyl transferase responsible for iturin A and surfactin biosynthesis in Bacillus subtilis (GenBank accession number P39144.1) (Table 4) (25).

FIG 2.

Graphic illustration of the rtc locus in reutericyclin-producing strains of L. reuteri. ORFs (indicated by arrows) with deduced functions in the synthesis and secretion of reutericyclin are designated with the prefix rtc. Sections of the locus assumed to be dedicated to transport/immunity, regulation, and synthesis are indicated. The PKS and NRPS domains involved in reutericyclin synthesis are indicated below the map. Letters denote domain functions: AT, acyltransferase; C, condensation; A, adenylation; T, thiolation domain (also known as PCP [peptidyl carrier protein] domain); KS, ketosynthase; ACP, acyl-carrier protein; TE, thioesterase; PP, phosphopantetheinyl transferase.

TABLE 4.

Proteins which are homologous to proteins encoded by the reutericyclin genetic island, and for which functional characterization has been published

L. reuteri protein	Similar protein in NCBI databasea	Organism	Identity (%)	GenBank accession no.
RtcP	4′-Phosphopantetheinyl transferase	Bacillus subtilis	27	P39144.1
RtcK	Mycosubtilin synthase subunit A	B. subtilis	32	Q9R9J1.1
RtcN	Plipastatin synthase subunit B	B. subtilis	28	P39846.1
RtcA	PhlA	P. fluorescens	33	BAD00178.1
RtcC	PhlC	P. fluorescens	32	AAY86549.1
RtcB	PhlB	P. fluorescens	48	AAM27407.1
RtcT	MFS-type transporter YusP	B. subtilis	32	O32182.1
RtcR	HTH-type transcriptional repressor BscR	B. subtilis	41	O08335.1
RtcS	HTH-type transcriptional regulator YerO	B. subtilis	35	O31500.1

^aRetrieved by BLASTP.

rtcK encodes a protein of 99.5 kDa. RtcK is composed of a ketosynthase (KS) domain, an acyl-carrier protein (ACP) domain, and a thioesterase (TE) domain. It is 32% identical to B. subtilis mycosubtilin synthase (GenBank accession number Q9R9J1.1) (Table 4), a hybrid enzyme that combines functional domains of peptide synthetase, aminotransferase, and fatty acid synthase.

rtcN encodes a protein of 115.8 kDa. The protein is composed of a condensation (C) domain, an adenylation (A) domain, and a thiolation (T) domain. The protein is 28% identical to B. subtilis plipastatin synthase subunit B (Table 4); the active sites are 33 to 38% identical to NRPS from Bacillus spp., Streptomyces spp., and Pseudomonas syringae (see Table S2 in the supplemental material). Analysis of the adenylation domain active-site residues with the NRPS-PKS program predicted specificity for leucine.

rtcA, rtcB, and rtcC encode proteins with estimated molecular masses of 38.2, 16.7, and 43.5 kDa, respectively. They share similarities with the PhlA (33% identical to RtcA) (GenBank accession number BAD00178.1), PhlC (32% identical to RtcC) (GenBank accession number AAY86549.1), and PhlB (48% identical to RtcB) (GenBank accession number AAM27407.1) (Table 4) proteins, which are involved in 2,4-diacetylphloroglucinol (2,4-DAPG) biosynthesis in Pseudomonas fluorescens (26).

rtcR and rtcS encode two distinct TetR family transcriptional regulators located adjacent to each other with molecular masses of 24.9 kDa and 21.8 kDa, respectively. RtcR is 41% identical to helix-turn-helix (HTH)-type transcriptional repressor BscR in Bacillus subtilis (GenBank accession number O08335.1), and RtcS is 35% identical to HTH-type transcriptional repressor YerO in Bacillus subtilis (GenBank accession number O31500.1) (Table 4).

rtcT encodes a 63.9-kDa protein belonging to the major facilitator superfamily (MFS). MFS proteins are often involved in the transport of synthesized compounds and confer immunity to various antibiotics (27). RtcT is 32% identical to Bacillus subtilis YusP (GenBank accession number O32182.1) (Table 4).

Three other genes were identified in the four reutericyclin-resistant strains, L. reuteri LTH2584, TMW1.656, TMW1.112, and LTH5448, but were absent in other strains of L. reuteri. These genes code for components of ABC transporters (see Table S1 in the supplemental material).

Gene function analysis.

To elucidate the functions of the genes on the genomic island, we attempted to delete rtcN, rtcA, rtcT, and rtcRS in L. reuteri TMW1.656 by double-crossover mutagenesis. Mutants with a deletion in rtcN or rtcA could be generated. The L. reuteri mutant strains TMW1.656ΔrtcN and TMW1.656ΔrtcA lost the ability to synthesize reutericyclin but were still reutericyclin resistant, providing direct evidence for the role of these genes in reutericyclin synthesis (Table 5). Attempts to delete rtcT or rtcRS in L. reuteri TWM1.656 were not successful. However, the same genes could be deleted in the reutericyclin-negative L. reuteri strain TMW1.656ΔrtcN (Table 5). The disruption of rtcT or rtcRS in the reutericyclin-producing wild-type L. reuteri strain TMW1.656 thus likely generated the lethal phenotype of reutericyclin sensitivity coupled to reutericyclin production. Disruption of rtcT eliminated reutericyclin resistance; the sensitivity of L. reuteri TMW1.656ΔrtcNΔrtcT was comparable to that of the reutericyclin-susceptible species L. sanfranciscensis. This finding suggests that RtcT is responsible for resistance against reutericyclin, probably through export of the substance. L. reuteri TMW1.656ΔrtcNΔrtcRS was also reutericyclin sensitive, suggesting that one or both of these regulators regulate the expression of reutericyclin resistance.

TABLE 5.

Mutants of L. reuteri TMW 1.656 created to identify the genes involved in reutericyclin biosynthesis and immunity

L. reuteri TMW1.656 mutant genotype	Presence or absencea of:
	Reutericyclin synthesis	Resistance to reutericyclin
ΔrtcN	−	+
ΔrtcA	−	+
ΔrtcT	NA	NA
ΔrtcRS	NA	NA
ΔrtcN ΔrtcT	−	−
ΔrtcN ΔrtcRS	−	−
ΔABC	+	+
ΔrtcN ΔABC	−	+

^aPlus and minus signs indicate presence and absence, respectively. NA, not applicable; these mutants could not be obtained.

The maintenance of reutericyclin resistance through active transport of the compound implies the expenditure of metabolic energy at the expense of growth. To determine the energetic consequences of reutericyclin production and resistance, the growth of L. reuteri TMW1.656 was compared to that of isogenic non-reutericyclin-producing mutants. The growth rates of L. reuteri TMW1.656ΔrtcN, TMW1.656ΔrtcNΔrtcT, and TMW1.656ΔrtcNΔrtcRS were 20% higher than that of L. reuteri TMW1.656, while the duration of the lag phase in the same strains was reduced by 25% (see Fig. S1 in the supplemental material).

A second set of genes shared by all reutericyclin-resistant L. reuteri strains but absent in other L. reuteri genomes (see Table S1 in the supplemental material) includes components of ABC transporters, with a possible contribution to reutericyclin transport and/or immunity (L. reuteri TMW1.656 genome, nucleotides 50140 to 52314). However, the deletion of one of the three ORFs (L. reuteri TMW1.656 genome, locus tag HQ33_09475) in L. reuteri TMW1.656 or L. reuteri TMW1.656ΔrtcN to generate L. reuteri TMW1.656ΔABC or L. reuteri TMW1.656ΔrtcNΔABC did not affect reutericyclin biosynthesis or immunity (Table 5), indicating that this gene cluster was not related to reutericyclin metabolism. Moreover, reutericyclin resistance was fully eliminated by deletion of rtcT, making a major contribution of other gene products to reutericyclin resistance unlikely.

Gene expression.

To determine a potential regulation mechanism, the expression levels of the gene cluster in the wild type L. reuteri strain TMW1.656 and its mutants were quantified (Fig. 3). Deletion of rtcN resulted in reduced expression of genes in the cluster, suggesting a positive-feedback mechanism of reutericyclin production, which has been reported for other antibiotics (28). The expression of rtcP, which encodes the activator enzyme for RtcK and RtcN, was significantly reduced by deletion of rtcRS, supporting the hypothesis that the putative regulatory proteins regulate reutericyclin synthesis in the wild-type strain.

FIG 3.

Quantification of gene expression levels in L. reuteri TMW1.656ΔrtcN (filled bars) and L. reuteri TMW1.656ΔrtcNΔrtcRS (shaded bars) relative to L. reuteri TMW1.656. The asterisk indicates that expression levels differ significantly between the two mutant strains (P < 0.05).

Homologues of reutericyclin biosynthesis genes in the genomes of other lactic acid bacteria.

The GC content of the genomic island harboring genes coding for reutericyclin production, genes coding for reutericyclin resistance, and regulatory genes was 5 to 10% lower than the GC content of the overall genome of L. reuteri LTH2584, TMW1.112, TMW 1.656, or LTH5448 (Fig. 4), suggesting that the island may have been acquired by lateral gene transfer. The island was inserted at two ORFs that encode a transposase and a tRNA, respectively, which are common locations for horizontally acquired genes (29). Moreover, L. reuteri LTH2584, TMW1.112, and TMW1.656, as well as the reutericyclin-producing L. reuteri strain TMW1.106, belong to the rodent-adapted lineage III of L. reuteri (11, 15, 30), while L. reuteri LTH5448 belongs to the rodent-adapted lineage I (30). The presence of the reutericyclin genomic island in a few phylogenetically unrelated strains of the species L. reuteri that were isolated from the same site further supports the hypothesis that the genomic island was transferred horizontally among these strains.

FIG 4.

Genomic location and GC content of the genomic island containing reutericyclin biosynthesis genes. Shown are the sequence identities of the reutericyclin-producing L. reuteri strains TMW1.112, TMW1.656, and LTH5448 to L. reuteri LTH2584. The closely related rodent lineage III strain L. reuteri 100-23 is included as a reference.

To gain insight into the evolution and potentially the origin of the gene cluster, homologues of RtcK, RtcN, RtcA, RtcB, and RtcC were retrieved from the nr database in NCBI, and phylogenetic trees were inferred (Fig. 5; see also Fig. S2 in the supplemental material). The gene cluster for reutericyclin synthesis has two phylogenetically distinct components: the PKS/NRPS component, which is related to bacitracin synthesis in the Gram-positive Bacillus spp., and the PhlABC component, related to 2,4-DAPG synthesis in the Gram-negative bacterium Pseudomonas fluorescens. Analyses of currently available genome and protein sequences indicate that combined activity of a PKS/NRPS system and PhlABC is restricted to the order Lactobacillales. The combination of these two enzyme systems thus may provide a unique solution to antibiotic synthesis within the lactic acid bacteria. Homologues of RtcK and RtcN were absent in all other genome-sequenced strains of L. reuteri but present in L. plantarum and Streptococcus mutans and in members of the Bacillales (Fig. 5A; see also Fig. S2A); RtcN is 48% identical to the corresponding proteins in S. mutans. Homologues of RtcA, RtcB, and RtcC were also identified in L. plantarum and Streptococcus mutans; RtcA is 57% identical to the corresponding proteins in S. mutans. More-distant relatives were found in Clostridium species and Pseudomonas spp. (Fig. 5B; see also Fig. S2B and C). The distinct phylogenetic trees for RtcA, RtcB, and RtcC, on the one hand, and RtcK and RtcN, on the other, imply that these two parts of the genomic island have different phylogenetic origins. The combination of RtcK, RtcN, RtcA, RtcB, and RtcC is found only in members of the Lactobacillales and most frequently in strains of S. mutans (Fig. 5; see also Fig. S2). Moreover, the organization of genes coding for RtcK RtcN, RtcA, RtcB, and RtcC homologues in S. mutans and L. plantarum was similar to their organization in L. reuteri (see Fig. S3 in the supplemental material). The low protein identities between PKS/NRPS systems in S. mutans, L. plantarum, and L. reuteri, however, argue against direct transfer between these three species.

FIG 5.

Unrooted phylogenetic trees of L. reuteri RtcN (A) and RtcA (B) and related proteins. The trees were simplified to represent NRPS only from genomes that also harbor PKS homologous to RtcK and to represent RtcA only from genomes that also harbor homologues to RtcB and RtcC. This simplification did not alter the topology of the trees showing sequences of S. mutans, L. plantarum, and L. reuteri (data not shown). The number at each branch point represents the percentage of bootstrap support calculated from 1,000 replicates. Only bootstrap values above 50% are shown.

DISCUSSION

This study identified the genetic determinants of reutericyclin biosynthesis, regulation, and immunity through an approach combining comparative genomics, bioinformatics analysis, and the characterization of null-mutants. Moreover, data mining of bacterial genome sequences demonstrated that putative reutericyclin-biosynthetic operons are present only in four strains of L. reuteri, one strain of L. plantarum, and several strains of Streptococcus mutans. The reutericyclin-producing strains of L. reuteri persisted in an industrial sourdough over 10 years of continuous propagation (15), but reutericyclin is not a common trait of sourdough organisms. Reutericyclin thus provides an interesting model of the role of antibiotic production in microbial ecology and evolution.

Proposed pathway for reutericyclin biosynthesis.

A genomic island was identified in reutericyclin-resistant strains of L. reuteri; two open reading frames of this genomic island were disrupted in the reutericyclin-resistant but non-reutericyclin-producing strain L. reuteri LTH5448. Mutagenesis of four genes identified genetic determinants of reutericyclin biosynthesis and reutericyclin resistance. The annotations of the genes involved in reutericyclin biosynthesis and the characterization of reutericyclin production by isogenic mutants of L. reuteri TMW1.656, in combination with the chemical structure of reutericyclin (2), allow the suggestion of a putative biosynthetic pathway for reutericyclin (Fig. 6). The assembly of the tetramic acid core structure by RtcN and RtcK can be inferred from the well-characterized activity of bacterial NRPS and PKS. Because the N-acylation and RtcABC-mediated acetylation of reutericyclin are exceptional among tetramic acids, the prediction of the function of the corresponding biosynthetic enzymes is plausible but more speculative (Fig. 6).

FIG 6.

Proposed pathway for reutericyclin biosynthesis. The A domain of RtcN selects for d-leucine (2), but RtcN has no epimerase domain. Most Gram-positive bacteria are capable of synthesizing d-Ala and d-Glu as building blocks for peptidoglycan (43), but the synthesis of other d-amino acids is less common. An isoleucine 2-epimerase with leucine epimerase activity has been characterized in lactobacilli (44); a corresponding epimerase is conserved in all four strains of L. reuteri used in this study (WP_011953381.1), and strains of L. reuteri have been reported to produce d-leucine (45). It is therefore highly likely that the isoleucine 2-epimerase homologues are responsible for d-leucine synthesis in L. reuteri. d-Amino acids can be directly activated by bacterial NRPS (46). Once activated, the leucine aminoacyl adenylate is likely loaded onto the phosphopantetheinyl arm of the RtcN thiolation (T) domain. Then the C domain of RtcN may acylate the amino group of the NRPS-bound leucine with ACP- or coenzyme A-activated 2-decenoic acid. The reutericyclin genomic island does not code for enzymes related to lipid metabolism; the C₁₀ fatty acid in reutericyclin thus likely originates from the general metabolism. Cross talk between polyketide synthases and fatty acid synthases has been reported in Escherichia coli (47). In bacteria, the 2-decenoic acid component can be synthesized by β-hydroxydecanoyl-ACP dehydratase (FabA) or a more common enzyme, hydroxyacyl-ACP dehydratase (FabZ). Two fabZ homologues are present in the genomes of reutericyclin producers (L. reuteri TMW1.656 genome locus tags HQ33_05470 and HQ33_05510). Sequence analysis of RtcK suggests the presence of KS, ACP, and TE domains. However, no acyltransferase (AT) domain is found. The amino acid sequence between the KS and ACP domains shows homology to the region N-terminal to AT domains found in other bacterial PKS. This is similar to what is found in the AT domain-less type I PKS, where the remnant of the AT domain sequence acts as a docking domain for a distinct AT enzyme (48). We propose that the KS domain of RtcK catalyzes a decarboxylative Claisen condensation. Once the enzyme-bound malonate decarboxylates, the resulting enolate attacks the leucine thioester loaded on the T domain of RtcN. The TE domain of RtcK may then catalyze the cyclization and offloading of the reutericyclin precursor via the nucleophilic attack of the amide nitrogen on the thioester carbonyl. Finally, we propose that the resulting tetramic acid is acetylated by RtcABC, thus forming reutericyclin. RtcABC show sequence homology to PhlABC, proteins involved in the biosynthesis of 2,4-diacetylphloroglucinol (26). Biosynthetic studies suggest that PhlABC are responsible for acetylating phloroglucinol, yielding 2-acetylphloroglucinol and 2,4-diacetylphloroglucinol (49). This acetylation is highly analogous to the predicted role in reutericyclin biosynthesis.

Reutericyclin is likely exported by the transporter protein RtcT. Both MFS and ABC transporters have been reported as export or immunity mechanisms for polyketides and/or nonribosomal peptides (5, 9); this study demonstrated that deletion of RtcT eliminated resistance to reutericyclin, while deletion of a putative ABC transporter that is also unique to reutericyclin-resistant strains did not affect reutericyclin production or reduce reutericyclin resistance.

Reutericyclin resistance is regulated by the TetR family regulator RtcR, RtcS, or both. TetR regulators are regulatory proteins with helix-turn-helix motifs that can act as either transcriptional repressors or activators. RtcR and RtcS do not have identical sequences and thus cannot form homodimers as reported for other TetR regulators (31). Whether they function alone or individually remains to be determined.

NRPS/PKS-mediated antibiotic production in lactic acid bacteria and lateral gene transfer.

In bacteria, the number of genes is generally proportional to genome size. This relation also holds for specific gene classes, but PKS/NRPS genes are an exception. Below a genome size of 3 Mbp, PKS/NRPS genes are rare, and a linear correlation between genome size and PKS/NRPS gene content could be established only for genomes of >5 Mb (9). In the order Lactobacillales, where genomes are generally small, the presence of PKS/NRPS is exceptional, and corresponding genes have been annotated only in Lactobacillus plantarum (32), Streptococcus thermophilus (33), and S. mutans (34, 35). To our knowledge, the only PKS/NRPS of any lactic acid bacterium that has been functionally characterized is a NRPS/PKS in S. mutans (34). The NRPS/PKS system characterized in S. mutans synthesizes a pigment involved in oxidative stress tolerance (34), and the NRPS and PKS are 30 and 67% identical to the enzymes in L. reuteri. A PKS/NRPS with higher homology to RtcN and RtcK (Fig. 4) was annotated as a putative bacitracin synthesis cluster (35). To our knowledge, NRPS/PKS-mediated production of antibiotics has not been described in S. mutans.

Homologous gene clusters that are comparably organized and share a common evolutionary origin were found in S. mutans and L. plantarum. These related gene clusters in streptococci are located on putative conjugative transposons (35), providing a potential vehicle for gene transfer between lactic acid bacteria. S. mutans has been suggested to exchange genetic information with food-borne bacteria, including lactobacilli, as they pass through the oral cavity (36, 37). L. plantarum adapts to a variety of ecological niches by the acquisition of “lifestyle cassettes,” genomic islands appropriate to niche requirements (38).

Ecological role of reutericyclin production for sourdough lactobacilli.

Lactobacilli in sourdoughs that are maintained by continuous propagation are characterized by high growth rates, acid resistance, and small genomes, and they generally lack the ability to produce antimicrobial compounds, such as bacteriocins or reuterin (1, 12, 39, 40). The production of antimicrobial compounds and the maintenance of producer immunity diverts energy from metabolic functions that support rapid growth and thus may reduce the competitiveness of bacteriocin-producing organisms (40, 41). This was confirmed by the observation that isogenic but reutericyclin-negative mutants of L. reuteri TMW1.656 grow faster than L. reuteri TMW1.656 (see Fig. S1 in the supplemental material). However, reutericyclin production provides a competitive advantage to sourdough lactobacilli, and reutericyclin-producing strains persisted in an industrial sourdough for at least 10 years of continuous propagation (15). This suggests that reutericyclin production provided at least a temporary advantage over non-reutericyclin-producing strains.

Sourdough L. reuteri strains are of intestinal origin (30) and may have acquired the reutericyclin gene cluster before or after transfer to sourdough. Because the reutericyclin gene cluster has been identified in only a few strains of lactobacilli (Fig. 4), reutericyclin production does not seem to be beneficial in intestinal or sourdough ecosystems. It is tempting to speculate that the reutericyclin cluster was acquired by a rodent L. reuteri strain and that this novel trait, although energetically unfavorable, was specifically selected for in the SER sourdough (15), in which it was maintained over 10 years of continuous propagation. L. reuteri LTH5448, which harbors an inactivated version of the cluster, was isolated from a different dough in the same facility 15 years after the first reutericyclin-producing strain was isolated, suggesting that the cluster became obsolete under different ecological conditions. Therefore, the reutericyclin genomic island identified here appears to represent a classic example of laterally transferred DNA, which is constantly sampled by bacteria but is retained only if it provides a major advantage in a specific niche, and is eliminated if it fails to provide a meaningful function (42). Because sourdough represents a straightforward experimental system, the reutericyclin genomic island can serve as a tractable model for the study of the role of acquired DNA in the ecology and evolution of a bacterial species.