Regulation of Mannitol Metabolism in Enterococcus faecalis and Association with par_EF0409 Toxin-Antitoxin Locus Function

ABSTRACT

The par_EF0409 type I toxin-antitoxin locus is situated between genes for two paralogous mannitol family phosphoenolpyruvate phosphotransferase systems (PTSs). In order to address the possibility that par_EF0409 function was associated with sugar metabolism, genetic and phenotypic analyses were performed on the flanking genes. It was found that the genes were transcribed as two operons: the downstream operon essential for mannitol transport and metabolism and the upstream operon performing a regulatory function. In addition to genes for the PTS components, the upstream operon harbors a gene similar to mtlR, the key regulator of mannitol metabolism in other Gram-positive bacteria. We confirmed that this gene is essential for the regulation of the downstream operon and identified putative phosphorylation sites required for carbon catabolite repression and mannitol-specific regulation. Genomic comparisons revealed that this dual-operon organization of mannitol utilization genes is uncommon in enterococci and that the association with a toxin-antitoxin system is unique to Enterococcus faecalis. Finally, we consider possible links between par_EF0409 function and mannitol utilization.

IMPORTANCE Enterococcus faecalis is both a common member of the human gut microbiota and an opportunistic pathogen. Its evolutionary success is partially due to its metabolic flexibility, in particular its ability to import and metabolize a wide variety of sugars. While a large number of phosphoenolpyruvate phosphotransferase sugar transport systems have been identified in the E. faecalis genome bioinformatically, the specificity and regulation of most of these systems remain undetermined. Here, we characterize a complex system of two operons flanking a type I toxin-antitoxin system required for the transport and metabolism of the common dietary sugar mannitol. We also determine the phylogenetic distribution of mannitol utilization genes in the enterococcal genus and discuss the significance of the association with toxin-antitoxin systems.

INTRODUCTION

Enterococci are notorious for their metabolic versatility and intrinsic resistance to inhospitable environments. One aspect of this versatility is the ability to transport and utilize a variety of carbohydrates as carbon sources. At least 13 sugars are metabolized by all enterococci, and over 30 are used by at least two species (1). The import of most of these sugars is accomplished by phosphoenolpyruvate phosphotransferase systems (PTSs) that couple uptake with the phosphorylation of the substrate. Enterococcus faecalis strain OG1RF encodes 39 predicted PTSs based on the NCBI genome annotation, but the targets for most of these transporters have yet to be determined experimentally. In addition, genome sequencing indicates that a number of PTSs are encoded on mobile genetic elements (MGEs), leading to extensive within-species variability and a possible association with virulence (2). Indeed, PTSs have been associated with the increased onset of colitis in a mouse model (3), increased killing of macrophages (3), enhanced stress resistance (4), endocarditis and biofilm formation (5), and murine intestinal colonization (6) in E. faecalis and E. faecium.

Mannitol is an abundant sugar found naturally in plants; is used in food as an artificial sweetener, particularly in candies, dried fruits, and chewing gum; and is poorly absorbed by the intestines (7). Thus, enterococci are very likely to find it as a common carbon source in the intestinal tract where it resides. An E. faecalis-encoded mannitol-specific PTS component, designated mtlF, and an associated mannitol-1-phosphate dehydrogenase gene, mtlD, were originally cloned and characterized by Fischer et al. in 1991 (8). However, sequencing of the OG1RF genome revealed a complex organization of two paralogous PTSs: an upstream cluster that also included a putative mannitol-responsive positive transcriptional regulator, mtlR, and a downstream cluster that included mtlD (Fig. 1). It was the latter cluster that had been previously cloned by Fischer et al. Between the two clusters is a type I toxin-antitoxin (TA-1) system, designated par_EF0409 (9), related to par_pAD1 originally described on the E. faecalis pheromone-inducible conjugative plasmid pAD1 (10). Toxin-antitoxin (TA) systems are ubiquitous on bacterial MGEs where they function to facilitate stable inheritance by a postsegregational killing (PSK) mechanism (11, 12). They are classified based on the nature and mechanism of action of their antitoxins; the antitoxins of TA-1 systems are regulatory small RNAs (sRNA) that bind to and prevent the translation of the toxin mRNA (13–15). The antitoxin sRNA is less stable than the toxin mRNA, so plasmid loss leads to the selective degradation of the sRNA and translation of the toxin, thereby inhibiting the growth of plasmid-free segregants. pAD1-borne par_pAD1 was shown to function in just such a manner (16–18). The explosion of genomic sequencing in bacteria revealed that paralogs of many plasmid-encoded TA systems were also present on bacterial chromosomes (19). The functions of these chromosomally encoded systems are controversial and mostly unknown (15, 20, 21). The function of par_EF0409 is similarly obscure.

FIG 1.

(A) Genetic and transcriptional organization of the mtlA operon, par_EF0409 locus, and mtlA2 operon. Large arrows depict the position and direction of the transcription of the relevant genes color-coded for function. Blue, PTS transport; orange, metabolism; green, regulation; red, par_EF0409 TA-1. Current NCBI OG1RF numerical gene designations are shown above the genes, while the orthologous gene designations based on sequence homology are shown below. Small inverted triangles above each gene indicate transposon insertion sites. Broken arrows represent promoters for the mtlA and mtlA2 operons with an expanded sequence showing transcription start sites identified by 5′ RACE. Predicted MtlR and CcpA binding sites are in green and red, respectively. Translational start sites of mtlA2 and fst are marked with a filled triangle. Small red arrows at the base of each gene represent the primer sites used for RT-PCR to determine operon organization. The numbers under the small arrows correspond to the primer sequences in Table S1 in the supplemental material. (B) Alignment of predicted MtlR binding and cre sites. The MtlR binding site is compared to the previously determined B. subtilis MtlR binding site (28) and cre site as determined for the E. faecalis eutS gene (29).

In this report, we characterize the complex organization of the mannitol utilization genes and determine their functional roles. We demonstrate that (i) the genes are arranged in two operons that flank par_EF0409; (ii) the PTS encoded in the downstream operon is the primary transport system for mannitol metabolism; (iii) the putative transcriptional regulator MtlR positively regulates the downstream genes, and phosphorylation sites shown to be critical for regulation in orthologs are important for its function; and (iv) the PTS encoded by the upstream operon plays an auxiliary regulatory role, presumably by modulating MtlR function. We also present several lines of circumstantial evidence that the function of par_EF0409 relates in some as-yet-undefined way to the regulation of mannitol metabolism.

RESULTS

Functional analysis and transcriptional organization of the mtlA and mtlA2 operons.

In order to evaluate the functional roles of the putative mtlA and mtlA2 operons, transposon insertion mutants in mtlA, mtlR, mtlF, mtlA2, and mtlF2 (Fig. 1) were obtained from the ordered transposon mutant library produced by the laboratory of Gary Dunny (22, 23). All transposon insertion strains grew normally on glucose (M9 medium with yeast extract-glucose [M9YEG]) but showed various levels of defects in growth on mannitol (M9 medium with yeast extract-mannitol [M9YEM]), supporting a role for these genes in mannitol metabolism. Insertions in mtlR, mtlA2, and mtlF2 failed to grow on mannitol, while those in mtlA showed slightly reduced growth (Fig. 2A), and insertions in mtlF had no effect (not shown).

FIG 2.

(A) Growth of mutants with transposon insertions in the mtlA and mtlA2 operons on M9YEG (left) and M9YEM (right). Clockwise from the top right are OG1RF (WT), mtlF2 (2088 and 2328), mtlR (8259), mtlA2 (0258), and mtlA (5429). (B) Complementation of OG1RF mutants with ectopic mtlR. Cells were grown on M9YEG (left) and M9YEM (right). A, OG1RF(pCIE); B, OG1RF(pCIE::mtlR); C, OG1RF ΔmtlR(pCIE); D, OG1RF ΔmtlR(pCIE::mtlR); E, OG1RF ΔMBS(pCIE); F, OG1RF ΔMBS(pCIE::mtlR).

The genetic organization of the relevant genes suggested transcription as two operons: mtlA-mtlR-mtlF (mtlA operon) and mtlA2-mtlF2-mtlD (mtlA2 operon) (Fig. 1). Promoters for the putative operons were identified by 5′ rapid amplification of cDNA ends (RACE) using primers within mtlA and mtlA2 (see Table S1 in the supplemental material) on RNA purified from cells grown in either glucose or mannitol. Each primer produced a single product (Fig. 3), and the transcriptional start sites identified along with the putative promoter sequences are shown in Fig. 1A. Identical transcription start sites were mapped in glucose- and mannitol-grown cells, but substantially more product was obtained for mtlA2 in mannitol-grown cultures (Fig. 3). 5′ RACE performed with primers within mtlR, mtlF, mtlF2, and mtlD produced no detectable products (data not shown), suggesting that these genes were transcribed from upstream promoters. To confirm the operon structure, reverse transcription-PCR (RT-PCR) was performed on the same RNA samples with primers straddling the junction between the genes of interest (Fig. 1 and Table S1). Products of the appropriate size were found between all genes suspected of belonging to either the mtlA or mtlA2 operon but not in controls lacking reverse transcriptase (Fig. S1 and S2). Interestingly, a product was also obtained with primers flanking the mtlF and RNA II_EF0409 junction, suggesting that in addition to transcription from its own promoter, transcriptional readthrough from the upstream operon might affect RNA II_EF0409 antitoxin levels. The pattern of transcriptome sequencing (RNA-seq) reads was also consistent with an mtlARF and mtlA2F2D organization (data not shown). As shown in Fig. 1, the mtlA2 and RNA I_EF0409 divergent promoters are separated by only 144 nucleotides (nt) and most likely contain at least one transcription factor binding site (see below), suggesting the possible coordination of expression.

FIG 3.

5′-RACE results with primers specific for mtlA (A) and mtlA2 (A2) in glucose and mannitol. Molecular weight markers (M) are on the left of each gel. Transcriptional start sites determined by sequencing these products are shown in Fig. 1A.

Fischer et al. (8) previously suggested that mtlF (mtlF2 here) and mtlD were cotranscribed based on the overlap of termination and initiation codons and the presence of an intrinsic terminator downstream of mtlD. However, their sequence showed an ∼60-nt gap between mtlA2 and mtlF2, leaving in question the relationship between these two genes. A comparison of their sequences with the NCBI OG1RF sequence revealed several discrepancies, most critically an insertion of a T at position 309 of the sequence reported by Fisher et al. (NCBI OG1RF nucleotide 311902) that changes the reading frame and results in the premature termination of mtlA2. In the NCBI sequence, only 14 nt separate the 3′ end of mtlA2 from the 5′ end of mtlF2, and there is no intrinsic terminator between, consistent with the operon organization proposed here.

To resolve questions concerning the potential polar effects of the transposon insertions, in-frame deletions of mtlA, mtlR, mtlF, and mtlA2 were constructed. Growth curves of these constructs were performed in M9YEG and M9YEM. The growth of the ΔmtlR mutant was significantly impaired in M9YEM compared to the wild type (WT) (Fig. 4). Indeed, the growth of the ΔmtlR mutant on M9YEM was indistinguishable from the growth on M9YE without added sugar (Fig. S3), suggesting that the mutant was unable to use mannitol for growth. The residual growth observed was likely due to the utilization of other nutrients in the yeast extract and Casamino Acids in M9YEM. In contrast, the ΔmtlA2 mutant continued to grow slowly after the growth of the ΔmtlR mutant had ceased (Fig. 4). The growth of the ΔmtlR mutant was also slightly reduced in M9YEG, while the ΔmtlA2 mutation had no effect on growth in this medium. The growth of the ΔmtlA mutant was indistinguishable from that of the WT in both M9YEG and M9YEM (data not shown). These results suggested that mtlA2-mtlF2 encoded the EIICBA PTS components essential for mannitol-specific transport and that mtlR encoded a positive regulator of the mtlA2 operon.

FIG 4.

Growth of in-frame deletion mutants of mtlR and mtlA2 on glucose (M9YEG) and mannitol (M9YEM). Results are averages from three independent experiments, and standard deviations are shown as error bars. The error at later time points was so small that the error bars are not visible beyond the marker. OD 600, optical density at 600 nm.

Regulation of the mtlA and mtlA2 operons in response to mannitol.

In order to investigate the transcriptional response to mannitol, RNA-seq was performed on WT cells grown in M9YEG and M9YEM. The three genes of the mtlA2 operon were all induced >100-fold when grown in M9YEM and were by far the most highly induced genes during growth on mannitol (Table 1; see also the supplemental material). The levels of mtlA operon expression were not significantly different on glucose and mannitol. The results with mtlA2 (Table 2) and mtlA (data not shown) were confirmed by quantitative RT-PCR (qRT-PCR).

TABLE 1.

Impact of growth in mannitol and deletion of mtlR in glucose on the expression of genes in the mtlA and mtlA2 operons

Gene	Fold changeb
	WT in mannitol vs glucose	ΔmtlR mutant vs WT in glucose
mtlA	0.9	0.7
mtlR	0.7	NA
mtlF	1.4	1.4
mtlA2	311a	0.2a
mtlF2	257a	0.2
mtlD	131a	0.4

^aDifferences determined to be statistically significant.

^bNA, not applicable. Since the mtlR gene was deleted in this strain, no reads were obtained.

TABLE 2.

mtlA2 levels in the mtlA operon and MtlR binding site mutants

Strain	Mean expression level ± SD
	Glucose	Mannitol
WT	0.228 ± 0.046	2.603 ± 0.047
ΔmtlA	0.771 ± 0.015	2.549 ± 0.059
ΔmtlF	0.495 ± 0.026	2.527 ± 0.033
ΔmtlR	0.059 ± 0.017	0.078 ± 0.013
ΔMBSa	0.087 ± 0.014	0.109 ± 0.006

^aMBS, mannitol binding site.

RNA-seq identified 217 other genes significantly altered in expression >2-fold: 147 induced and 70 repressed (see the supplemental material). Among the most strongly affected genes were several that would be expected to impact the flow of carbon through the glycolytic cycle, including glucosamine 6-phosphate deaminase, fructose bisphosphatase, and acetaldehyde CoA alcohol dehydrogenase (all induced over 10-fold); the putative glycerol dehydrogenase/dihydroxyacetone kinase operon (induced over 6-fold); and beta-phosphoglucomutase (repressed approximately 22-fold). In addition, a number of ABC and sugar transport systems were moderately induced, perhaps because of the release of carbon catabolite repression (CCR). PTSs orthologous to those for trehalose and mannose were among the most highly repressed genes. Since trehalose has been shown to induce beta-phosphoglucomutase in other organisms (24, 25), the expression of these two genes may be linked.

The paralogous PTS encoded in the mtlA operon was not essential for or induced by growth in mannitol, but the location of the mtlR gene within this operon suggested that it performed a regulatory function. The mtlR gene encodes a member of the BglG family of regulatory proteins and is most similar to the subfamily that acts as transcriptional activators (26). The growth defect of the ΔmtlR mutant in M9YEM suggested that the gene encoded a transcriptional activator of the mtlA2 operon. Consistent with this hypothesis, qRT-PCR experiments showed that the ΔmtlR mutant negated the mannitol induction of mtlA2 (Table 2). Unexpectedly, the ΔmtlR mutant also reduced mtlA2 expression in M9YEG; mtlA2 expression was reduced 4- and 33-fold in M9YEG and M9YEM, respectively, compared to the WT control (Table 2). Because the ΔmtlR mutant also showed a growth defect in M9YEG, RNA-seq was performed to determine if the transcription of any other genes was altered during growth in glucose. As expected, reductions were observed in all three genes of the mtlA2 operon compared to the WT (Table 1), although only the change in mtlA2 levels was significant because of the low basal levels of expression of mtlF2 and mtlD on M9YEG. Only two other genes showed a significant change in expression of over 2-fold in the ΔmtlR strain grown in M9YEG relative to the WT: dltX, involved in the alanylation of teichoic acids, and an orphan EIIC PTS component, OG1RF_RS01210. The expression of both was increased approximately 2.5-fold in the ΔmtlR strain. It is possible that the mtlR deletion resulted in posttranscriptional changes, e.g., protein phosphorylation, that affected growth in glucose.

Complementation of the ΔmtlR mutant with an ectopic copy of mtlR in the pCIE expression vector (9) confirmed the role of MtlR as an activator of mtlA2 transcription. Complementation restored the growth of ΔmtlR on M9YEM and WT mtlA2 mRNA levels on both M9YEG and M9YEM (Fig. 2B and Table 3). Complementation was complete without cCF10 induction of the P_Q promoter, indicating that basal levels of transcription produced sufficient MtlR for a response to mannitol. P_Q induction produced only a small increase in mtlA2 expression in M9YEM. In contrast, cCF10 induction of mtlR expression in M9YEG resulted in a 6-fold-higher transcription level of mtlA2 than in the absence of inducing cCF10 (Table 3). The qRT-PCR results showed that mtlR expression in the absence of cCF10 induction was 4-fold higher than that from its natural chromosomal context in OG1RF(pCIE) in both M9YEG and M9YEM media; cCF10 induction led to a further 3- to 4-fold induction on both media (Table 4). Therefore, increased mtlA2 expression in mannitol was not due to a difference in mtlR levels but presumably due to mechanisms of CCR in glucose.

TABLE 3.

mtlA2 levels in OG1RF ΔmtlR complemented strains

Plasmid	Mean expression level ± SDb
	Glucose	Mannitol
pCIE vector only	0.032 ± 0.005	0.060 ± 0.009
pCIE::mtlR+	0.225 ± 0.088	2.842 ± 0.064
pCIE::mtlR+ plus cCF10a	1.354 ± 0.120	2.762 ± 0.427
pCIE::mtlR H334A	0.035 ± 0.006	ND
pCIE::mtlR H334D	0.109 ± 0.007	1.306 ± 0.101
pCIE::mtlR H392A	0.029 ± 0.002	ND
pCIE::mtlR S412A	0.613 ± 0.132	3.035 ± 0.135
pCIE::mtlR H586A	1.914 ± 0.126	ND
pCIE::mtlR H586D	1.434 ± 0.085	ND

^acCF10 was added to the culture at a concentration of 5 ng/mL 1 h prior to harvest. cCF10 was not added to any other samples.

^bND, not determined.

TABLE 4.

mtlR levels in OG1RF ΔmtlR complemented strains

Plasmid	Mean expression level ± SD
	Glucose	Mannitol
pCIE vector only	0.244 ± 0.034	0.287 ± 0.057
pCIE::mtlR+	0.917 ± 0.137	1.240 ± 0.344
pCIE::mtlR+ plus cCF10a	3.510 ± 0.464	3.817 ± 0.561

^acCF10 was added to the culture at a concentration of 5 ng/mL 1 h prior to harvest.

The Bacillus subtilis MtlR protein is a multidomain transcriptional activator whose function is regulated by phosphorylation at multiple residues (27). It consists of an N-terminal DNA binding domain, two consecutive PTS regulatory domains (PRDs) (PRD1 and PRD2), an EIIB^Gat-like domain, and an EIIA^Mtl-like domain (Fig. 5). The PRD2 domain performs a CCR function by sensing the phosphorylation state of the general PTS protein Hpr. In the presence of a preferred carbon source, Hpr dephosphorylates two histidine residues in PRD2, which inhibits MtlR function. In contrast, the dephosphorylation of an EIIB^Gat-like domain cysteine residue and an EIIA^Mtl-like domain histidine residue by the mannitol-specific EIIA and EIIB components, respectively, activates MtlR in response to the presence and transport of mannitol. In the case of B. subtilis MtlR, the phosphorylation state of the cysteine residue is most important. E. faecalis MtlR is 28% identical and 51% similar to B. subtilis MtlR and shows homology through all five domains. In particular, all of the relevant histidine residues are conserved in the PRDs and the EIIA^Mtl-like domain (Fig. 5). However, in E. faecalis, a potentially phosphorylatable serine residue replaces the B. subtilis MtlR EIIB^Gat-like domain cysteine residue. To determine if these residues are functionally conserved, nonphosphorylatable alanine and phosphomimetic aspartic acid replacements were made in residues analogous to those previously examined in B. subtilis MtlR. The mutants were then used to complement the ΔmtlR strain from pCIE without cCF10 induction. The results are shown in Table 3. Alanine replacement of the PRD2 histidine residues (H334A and H392A) inactivated MtlR, as was observed in B. subtilis, suggesting that it performs a similar CCR function. Substitution with aspartic acid at H334 (H334D), however, did not lead to constitutive activity, suggesting either that both histidines in E. faecalis MtlR must be phosphorylated or that the aspartic acid is not sufficiently phosphomimetic in this context. In support of the former interpretation, growth on mannitol led to the transcriptional activation of mtlA2 by the H334D MtlR mutant, presumably due to the phosphorylation of H392. The replacement of the EIIB^Gat-like domain serine with alanine (S412A) resulted in a partial derepression of mtlA2 in glucose but did not have an effect as dramatic as that of the cysteine-to-alanine substitution in B. subtilis MtlR. The S412A mutant could be further activated by growth in mannitol, indicating the importance of other residues. The replacement of the EIIA^Mtl-like domain histidine with either alanine or aspartic acid (H586A and H586D, respectively) resulted in the full activation of mtlA2 expression on glucose. Constitutive activation was expected for the alanine mutation, but the aspartic acid substitution in B. subtilis MtlR resulted in the inactivation of the protein. It is possible that aspartic acid is not sufficiently phosphomimetic in the E. faecalis MtlR context.

FIG 5.

Comparison of B. subtilis and E. faecalis MtlR. The B. subtilis domain structure with key phosphorylatable residues indicated in color is shown above (reprinted from reference 26 with permission). E. faecalis has a similar domain structure, with the analogous phosphorylatable residues and surrounding homology shown on the left. The effects on mutations in these residues are compared in the text on the right.

As shown in Fig. 1, a putative MtlR binding site was identified upstream of the mtlA2 promoter based on homology to the known sequence in B. subtilis (28). An OG1RF strain with a scrambled sequence in this site phenocopied the ΔmtlR mutant. Thus, the MtlR binding site mutant was unable to grow on M9YEM (Fig. 2B) and showed reduced levels of the mtlA2 transcript compared to the WT on M9YEG (Table 2). The provision of excess MtlR ectopically did not complement the binding site mutation (Fig. 2B).

Another potential candidate for a CCR role in regulating the mtlA2 operon is catabolite control protein A (CcpA). CcpA functions to repress genes involved in the metabolism of nonpreferred carbon sources in the presence of glucose. Indeed, a ccpA-deleted strain (kindly provided by the Garson laboratory) showed elevated expression of mtlA2 in M9YEG and an even greater level of derepression in M9YE with 90% glucose and 10% mannitol (Table 5). Repression was restored in a complemented strain. A putative CcpA binding site (CRE) was identified, based on sequence homology to the CRE site of the E. faecalis eutS gene (29), between the −10 and −35 promoter elements of the mtlA2 operon (Fig. 1). Unfortunately, mutants constructed in this element eliminated mtlA2 expression, so a more detailed examination of the sequence will be required to confirm its function.

TABLE 5.

mtlA2 levels in ΔccpA and complemented strains

Strain	Mean expression level ± SD
	Glucose	90% glucose + 10% mannitol
WT	0.228 ± 0.046	0.959 ± 0.082
ΔccpA	0.314 ± 0.023	2.092 ± 0.075
ΔccpA::ccpA+	0.187 ± 0.006	1.024 ± 0.092

With the function of the mtlA2 operon and the mtlR gene established, the question of the purpose of the mtlAF paralogous PTS remained. In both B. subtilis and Lactococcus lactis, MtlR is negatively regulated by phosphorylated MtlF, and mtlF mutations lead to the derepression of the mannitol transporter in glucose. To investigate if the E. faecalis mtlAF PTS might perform a similar regulatory function, in-frame deletions of the two components were tested for their impact on mtlA2 expression. Consistent with a negative regulatory role, the deletions resulted in a 2- to 3-fold increase in mtlA2 expression in M9YEG compared to the WT (Table 2). The deletions had no effect on mtlA2 expression in M9YEM.

Between the operons, what is par_EF0409 doing?

In most published genomes of the Firmicutes, the genes for mannitol transport, regulation, and metabolism are located in a single operon in the order mtlA-mtlR-mtlF-mtlD. Variations include the independent transcription of mtlD in Lactococcus lactis and the separation of mtlR from the rest of the operon in Bacillus subtilis (28, 30). Using the phylogeny of Zhong et al. (31), we searched for mtlA, mtlA2, and mtlD sequences in enterococcal species by NCBI BLASTn. Twenty-five of the 34 species in the dendrogram lacked homologs of all three genes and therefore are unlikely to be able to utilize mannitol as a carbon source. Of the nine remaining species, Enterococcus gilvus, Enterococcus avium, and Enterococcus raffinosus carry an mtlA-mtlR-mtlF-mtlD operon. The remaining six species (E. faecalis, E. faecium, Enterococcus mundtii, Enterococcus thailandicus, Enterococcus casseliflavus, and Enterococcus gallinarum) carry both an mtlA-mtlR-mtlF operon and an mtlA2-mtlF2-mtlD operon but with significant variability in spacing and intervening genes. The par_EF0409 TA-1 locus is unique to E. faecalis, and none of the other species have homologs of par Fst toxins or evidence of a TA locus. E. mundtii is the most divergent, with the two operons being separated by 14.5 kbp and multiple genes. In E. faecium, the operons are about 2.3 kbp apart, and the ATP binding and permease components of an ABC transporter are encoded in the gap and transcribed in the direction opposite that of the mannitol operons. The closely related species E. casseliflavus and E. gallinarum have similar interoperon gaps of 313 nt, significantly shorter than the E. faecalis 611-nt gap. The gap includes two inverted repeats that show characteristics of intrinsic terminators but no direct repeats or open reading frames. E. thailandicus has an interoperon gap of 544 nt and a complex combination of direct and inverted repeats but with different spacing than par-related TA systems and no apparent open reading frame.

Whether the E. faecalis par_EF0409 TA locus plays a unique adaptive role or is merely an evolutionary accident, e.g., the remnant of a long-lost MGE, is not clear. To examine whether there might be a link between par_EF0409 function and mannitol metabolism, we compared the toxicities of Fst_EF0409 in M9YEG and M9YEM. As shown in Fig. 6, cells grown in M9YEM were approximately 10-fold more sensitive to toxin expression than cells grown in M9YEG. In M9YEG, the induction of Fst_EF0409 at 2 and 5 ng/mL cCF10 from the P_Q promoter of pCIE showed complete inhibition of growth in mannitol but substantially less inhibition in glucose (Fig. 6A). Partial inhibition of growth was observed in mannitol at cCF10 levels of 0.5 ng/mL, suggesting that the toxin is approximately 10 times more toxic in mannitol than in glucose. A similar increase in sensitivity on mannitol was observed when cellular viability was assessed (data not shown).

FIG 6.

Toxicity of Fst_EF0409 TA-1 toxin in cells grown in M9YEG versus M9YEM. Fst_EF0409 was expressed from the cCF10-responsive P_Q promoter of pCIE. G, glucose; M, mannitol. Numbers are the concentrations of cCF10 added in nanograms per milliliter. (A) Comparison of growth in glucose and mannitol at 0, 2, and 5 ng/mL cCF10. (B) Effects of 0, 0.2, 0.5, 1, and 2 ng/mL cCF10 on cells grown in mannitol.

One potential function for par_EF0409 might be to suppress recombination between homologous regions of the mtlA and mtlA2 genes. If such recombination occurred, par_EF0409 would be on the excised DNA (Fig. S4 and S5). Its loss during the outgrowth of recombined strains would trigger TA-1 PSK function. A comparison of the mtlA and mtlA2 sequences revealed several regions of significant homology at which recombination could occur (Fig. S6). To detect recombinants, primers were synthesized complementary to regions of nonhomology flanking the homologous regions. PCR was then performed, and products were examined for the predicted size of the recombinants. In addition, primers were synthesized to amplify across the junction of the predicted excised circle. Products of the appropriate size were detected with both primer sets using either WT OG1RF or an isogenic strain deleted for par_EF0409 (data not shown). PCR products were cloned into pGEM-T Easy and sequenced to identify the junction sites. The majority of recombination events occurred within a 54-nt region with 85% identity between nucleotides 306486 and 306539 in mtlA and between nucleotides 311329 and 311382 in mtlA2 encoding the conserved P-loop motif in the IIB domain. In particular, a 5′-ATGGGGGC-3′ motif within this region was a hot spot for recombination. The same bias was observed in the WT and mutant strains and in predicted chromosomal and excised-circle recombinants.

While these results are suggestive of a link between TA function and the function of the surrounding genes, this interpretation has several caveats. As with any overexpression study, the relevance to genes in their normal context must be considered. Under all conditions examined so far, the antitoxin RNA is produced in large molar excess over the toxin message (9). The results from this study suggest that this is also true for mannitol. RNA-seq experiments showed that the toxin-encoding message RNA I_EF0409 (designated OG1RF_RS01630 in the supplemental material) is barely detectable in both M9YEG and M9YEM. RNA I_EF0409 was detectable by qRT-PCR and Northern blotting, but the levels were low and variable, and no significant differences between growth media were observed (data not shown). As shown in Fig. 7, RNA II_EF0409 antitoxin levels were reduced in M9YEM in stationary phase relative to M9YEG in Northern blot analyses. However, log-phase cultures showed no difference in antitoxin levels. Therefore, although the overexpression data suggest that mannitol-grown cells may be more sensitive to the toxin than glucose-grown cells, conditions under which the expression of RNA I_EF0409 from its natural position would be high enough to escape RNA II_EF0409-mediated repression have yet to be identified.

FIG 7.

RNA II_EF0409 levels in glucose- and mannitol-grown cells. 16S rRNA was used as a loading control. G, M9YEG grown; M, M9YEM grown. Duplicate samples were performed with RNA from independently grown cultures. (A) Stationary phase. (B) Log phase.

DISCUSSION

In this report, we define the genetic organization and functional roles of paralogous mannitol family PTSs in E. faecalis strain OG1RF. The relevant genes are organized into two operons flanking the par_EF0409 TA-1 system. The downstream mtlA2 operon is highly induced by mannitol and encodes the essential transport apparatus, mtlA2F2, and metabolic enzyme, mtlD, for mannitol utilization. This operon was partially described previously (8). The upstream mtlA operon is constitutively transcribed and carries genes essential for the proper regulation of the mtlA2 operon. This includes the mtlR gene, which is homologous to mannitol-responsive positive transcriptional regulators in other organisms and was shown to be essential for the mannitol-dependent induction of the mtlA2 operon. Mutations in mtlR also showed reduced basal transcription of mtlA2 and a slightly reduced growth rate in glucose. The PTS encoded by mtlAF is not essential for growth in mannitol but is required for the full repression of the mtlA2 operon in its absence. While the mtlAF genes show the greatest homology to mannitol transporters, we cannot rule out at this time the possibility that mtlAF transports some other sugar.

The E. faecalis mtlR gene encodes a protein closely related to the well-studied B. subtilis mannitol PTS regulator mtlR. The effects of mutations in phosphorylatable residues of E. faecalis MtlR were generally consistent with those previously observed in B. subtilis MtlR (28), with some subtle differences. Thus, alanine substitutions of the two histidine residues in the E. faecalis MtlR PRD2 domain resulted in the loss of mtlA2 operon induction in mannitol, supporting a CCR function of this domain. However, unlike in B. subtilis, substitution of the phosphomimetic aspartic acid residue at H334 (analogous to H342) did not result in constitutive activation, suggesting either that the aspartic acid is not sufficiently phosphomimetic in E. faecalis MtlR or that both PRD2 histidines must be phosphorylated. Alanine substitutions in phosphorylatable residues in the E. faecalis MtlR EIIB^Gat-like and EIIA^Mtl-like domains resulted in the constitutive expression of the mtlA2 operon, suggesting that these two residues are dephosphorylated by the PTS in the presence of mannitol, as in B. subtilis MtlR, to induce the mtlA2 operon. In B. subtilis, the C419 residue in the EIIB^Gat-like domain plays the predominant role. In E. faecalis, this residue is replaced by S412, and the H586 residue in the EIIA^Mtl-like domain appears to play the predominant role. One caveat to the interpretation of these results is that complementation from the pCIE vector resulted in a 3- to 4-fold-higher expression level of mtlR mRNA than from the gene in the chromosome (Table 4). It is unclear what effect this difference in transcript dosage might have. A putative MtlR binding site showing substantial sequence similarity to that for B. subtilis MtlR was found upstream of the mtlA2 promoter, and mutation of this site prevented the mannitol-dependent transcriptional activation of the mtlA2 gene, supporting its role as the site of MtlR action.

CcpA and the MtlAF PTS transporter appear to play secondary roles in repressing the mtlA2 operon in the presence of glucose and in the absence of mannitol, respectively. CcpA also plays a secondary role to the phosphorylation of MtlR PRD2 in the repression of mannitol metabolic genes during CCR in other Gram-positive species (32). A putative CRE site for CcpA binding was identified between the −10 and −35 boxes of the mtlA2 promoter based on homology to a CRE site in the E. faecalis eut operon (29), but we were unable to confirm that this site was essential for CcpA function. MtlAF presumably functions through MtlR phosphorylation, but the rationale for having a second transporter system is unclear. In other systems, a single transporter appears to be sufficient for this mode of regulation. It is possible that CcpA and/or MtlAF may play greater roles under different culture conditions, e.g., different combinations of carbon sources or synthetic versus semisynthetic media.

The distribution of mannitol utilization genes on the enterococcal family tree is sparse and spotty, and the operon organization is highly variable. This suggests that the relevant genes were acquired by horizontal gene transfer independently on multiple occasions. Most, but not all, enterococcal species that harbor genes for mannitol transport and metabolism have two presumptive operons similar to the E. faecalis mtlA and mtlA2 operons, but the interoperon distance varies from 14.5 kbp and 313 bp. Only E. faecalis encodes a discernible TA system within this intervening region. The origin and function of the par_EF0409 TA-1 system remain unclear, but several pieces of circumstantial evidence suggest that its function may be coordinated with that of surrounding genes. First, as previously reported (33), mutation of an efflux pump located 2 kbp downstream of mtlD results in hypersensitivity to Fst_EF0409, the par_EF0409 toxin. Located between the mtlA2 operon and the genes for the efflux system are genes annotated as NAD binding proteins that could conceivably be involved in recycling the NAD⁺ cofactor required for MtlD function. Second, transcription initiated from the mtlA promoter appears to read through into RNA II_EF0409, the par_EF0409 antitoxin, potentially impacting its expression and/or interaction with RNA I_EF0409, the toxin-encoding mRNA. Third, mannitol-grown cells are hypersensitive to Fst_EF0409 expression. Fourth, homologous sequences in mtlA and mtlA2 are prone to recombination, potentially excising the intervening DNA. Since the excised DNA would include the par_EF0409 locus, the TA function might prevent its loss. Why this would be advantageous to E. faecalis but not to those species with the two operons separated by similar interoperon distances is not clear. In addition, hypersensitivity in the efflux system mutant and in mannitol-grown cells was detected only using ectopic overexpression of the toxin. In the natural context, under all growth conditions examined so far, antitoxin is produced in substantial molar excess to toxin mRNA, so it is unclear when the cell would be exposed to sufficient toxin to observe an effect. The data presented here provide essential context for further efforts to determine the function of par_EF0409 as well as to examine the evolution of carbohydrate metabolism in the enterococci.

MATERIALS AND METHODS

Bacterial strains, media, and growth conditions.

E. faecalis strain OG1RF was used for all experiments in this study (34, 35). An OG1RF strain deleted for par_EF0409, OG1RF Δpar_EF0409 (9), was also used in recombination studies. Transposon mutant strains were obtained from the ordered library constructed in OG1RF by the Gary Dunny laboratory at the University of Minnesota (22). The locations of transposon inserts based on the NCBI OG1RF genome sequence are as follows: mtlA (305429), mtlR (308259), mtlF (309542), mtlA2 (310258), and mtlF2 (312088 and 312328). For growth curves and RNA preparation, E. faecalis cultures were grown in M9YE base (36) supplemented with either 20 mM glucose or mannitol. E. faecalis liquid cultures were grown at 37°C with shaking at 25 rpm. Agar plates were prepared by the addition of 17 g of agar (Research Products International, Mt. Prospect, IL, USA) per L of medium. Plates were incubated overnight at 37°C. Chloramphenicol (Cm) (Sigma-Aldrich, St. Louis, MO, USA) was added for plasmid selection where appropriate. Cultures were grown overnight with 25 μg Cm per mL and diluted to 1% or 2% in fresh medium with 10 μg Cm per mL. Expression from pCIE was established by the addition of the peptide pheromone cCF10 (H-LVTLVFV-OH) (Mimotopes, Clayton, Australia) at the concentrations indicated for individual experiments, as previously described (9). An equal volume of the solvent (dimethylformamide) was added for uninduced controls. Plasmid DNA was introduced into E. faecalis by electroporation as previously described (37). Escherichia coli strain DH5α (New England BioLabs, Ipswich, MA, USA) was used for subcloning procedures and plasmid purification. E. coli cultures were grown in Luria-Bertani (38) medium with the following antibiotics at the indicated concentrations where appropriate: ampicillin (Amp) at 100 μg/mL and Cm at 25 μg/mL. E. coli liquid cultures were grown at 37°C with shaking at 250 rpm.

Genetic manipulations.

In-frame, markerless deletions of mtlA, mtlR, mtlF, and mtlA2 were constructed in OG1RF by allelic exchange using vector pJH086 (39) as previously described (33, 40). The mutant alleles were synthesized by Blue Heron Biotech LLC (Bothell, WA, USA) and contained the first 5 and last 5 codons of the corresponding gene and approximately 900 bp of flanking DNA. The construct was synthesized with PaeI and SmaI restriction sites at each end, and these enzymes were used to subclone the fragment from the commercially provided pUCminusMCS vector into pJH086. After a postligation cut with BamHI, plasmids were purified and introduced into competent DH5α cells with selection for Cm and growth at 30°C. Restriction enzymes and DNA ligase were obtained from New England BioLabs (Ipswich, MA, USA). Plasmids were purified by using a QIAprep spin miniprep kit (Qiagen, Valencia, CA, USA), checked for the appropriate restriction pattern, and introduced into E. faecalis cells for allelic exchange. Recombinants were screened by colony PCR (41), using primers flanking the desired deletion (see Table S1 in the supplemental material). PCR products showing the appropriate size for the deletion were then sequenced to ensure that no spurious mutations were obtained. All primers were produced by Integrated DNA Technologies (IDT) (Coralville, IA, USA). All DNA sequencing was performed by Eurofins Genomics LLC (Louisville, KY, USA).

Mutations in the putative MtlR and CcpA (CRE) binding sites were constructed in a manner similar to that for the deletions with the exception that synthesized sequences contained scrambled binding site sequences flanked by approximately 900 bp for recombination. The MtlR binding site sequence was changed from TTTGGCACAATGGTTTGTGCCAA to ACATGGCTCGATTAGATGTATGTC, and the CRE site was changed from GGAAAGC to GACAGGA. Allelic exchange was carried out as described above. In the case of the MtlR binding site mutants, roughly half of the isolated recombinants showed a growth defect on M9YEM plates. These recombinants were confirmed to contain the desired mutation by PCR using flanking primers SMBS-F and SMBS-R (Table S1) and sequencing. The recombinants containing the scrambled CRE site were identified by PCR using the SCRE-detect primer that binds to the scrambled sequence and the SCRE-R primer flanking the region. The presence of the scrambled CRE sequence was confirmed by sequencing the product amplified using the SCRE-F and SCRE-R primers (Table S1).

For complementation studies, a promoterless version of the mtlR gene was amplified by PCR from OG1RF genomic DNA purified as previously described (41), using the primers indicated in Table S1. The amplified product and pCIE were digested with BamHI and SphI. After a postligation cut with SalI, plasmids were purified and introduced into competent DH5α cells with selection for Cm and growth at 37°C. This construct, designated pCIE::mtlR⁺, was then used as the template for the introduction of the relevant MtlR mutations by GenScript (Piscataway, NJ, USA). These constructs were then introduced into OG1RF ΔmtlR by electroporation.

RNA purification and analysis.

RNA samples were prepared for all purposes as previously described (9). 5′ rapid amplification of cDNA ends (RACE) was performed using the Invitrogen (Carlsbad, CA) 5′-RACE system for rapid amplification of cDNA ends, version 2.0, according to the manufacturer’s instructions, with modifications as described previously (42). Quantitative reverse transcription-PCR (qRT-PCR) was performed as previously described (9). All data represent the averages from at least three biological replicates. Standard deviations were calculated in Microsoft Excel. The statistical analyses included one-way analysis of variance (ANOVA) with a Tukey multiple-comparison post hoc test. All differences discussed in the text had P values of <0.01.

Northern blot analyses were performed as previously described (43), with oligonucleotide probes specific for RNA I_EF0409, RNA II_EF0409, and E. faecalis 16S rRNA as a loading control, as described previously (9). RT-PCR for the identification of polycistronic RNAs was performed as previously described (9), with the primer pairs shown in Table S1. The PCR temperature cycling conditions were as follows: an initial denaturation step for 1 min at 94°C, followed by 30 standard cycles of denaturation at 94°C for 30 s, primer annealing for 30 s at 55°C, and primer extension at 72°C for 1 min. The last cycle was followed by a 10-min incubation at the primer extension temperature of 72°C. The sizes of the products were checked by running them on an agarose gel. Products were sequenced by Eurofins Genomics LLC (Louisville, KY, USA) using the gene-specific primers (Table S1) to ensure that they were not spurious products from some other region.

RNA-seq was performed by the University of Nebraska DNA Sequencing Core at the University of Nebraska Medical Center in Omaha, NE. Libraries were constructed from 300 ng of total RNA from samples provided by our laboratory using TruSeq stranded total RNA library prep (Illumina, Inc., San Diego, CA) according to the recommended protocol. In the step in which the total RNA is incubated with Ribozero to remove ribosomal transcripts, 5 μL of a bacterial Ribozero solution was used since these cultures do not have contaminating eukaryotic RNA. The resultant libraries from the individual samples were multiplexed and subjected to 75-bp paired-read sequencing to generate approximately 60 million pairs of reads per sample using a HighOutput 150-cycle flow cell on an Illumina NextSeq500 sequencer in the UNMC Genomics Core facility. The original fastq format reads were trimmed by the fqtrim tool (https://ccb.jhu.edu/software/fqtrim) to remove adapters, terminal unknown bases (N’s), and low-quality 3′ regions (Phred score of <30). The trimmed fastq files were processed by FastQC (44). The reference genome for Enterococcus faecalis OG1RF was downloaded from Ensembl. The trimmed fastq files were mapped to Enterococcus faecalis by CLC Genomics Workbench 12 for RNA-seq analyses. For statistical analysis, each gene’s read counts were modeled by a separate generalized linear model (GLM), assuming that the read counts follow a negative binomial distribution, and were normalized based on transcripts per million (TPM). The Wald test was used for statistical analysis of the two-group comparisons. The false discovery rates (FDRs) and Bonferroni-adjusted P values were also provided to adjust for multiple-testing problems. Fold changes were calculated from the GLM, which corrects for differences in library size between the samples and the effects of confounding factors.

Detection of mtlA-mtlA2 genomic recombinants.

Genomic DNA was purified as previously described (41). Primers were designed to read across regions of homology between mtlA and mtlA2 at which recombination was predicted to occur (Table S1). Depending on the combination of primers used for PCR, products would be produced from genomic recombinants (5′-mtlA + mtlA2-3′), the predicted excised circle (5′-mtlA2 + mtlA-3′), or the unrecombined genes (5′-mtlA + mtlA-3′ and 5′-mtlA2 + mtlA2-3′). PCR was performed using Kapa Biosystems (Wilmington, MA, USA) KAPA2G fast PCR mix according to the manufacturer’s instructions. PCR products were cloned into pGEM-T Easy and transformed into DH5α competent cells (Promega). Plasmid DNA was purified using the Qiagen (Valencia, CA, USA) midiprep kit, and DNA was sequenced at Eurofins Genomics.

Data availability.

Raw and processed files are available at the GEO repository under accession number GSE195652.