Regulatory crosstalk between type I and type II toxin-antitoxin systems in the human pathogen Enterococcus faecalis

Abstract

We discovered a chromosomal locus containing 2 toxin-antitoxin modules (TAs) with an antisense transcriptional organization in the E. faecalis clinical isolate V583. These TAs are homologous to the type I txpA-ratA system and the type II mazEF, respectively. We have shown that the putative MazF is toxic for E. coli and triggers RNA degradation, and its cognate antitoxin MazE counteracts toxicity. The second module, adjacent to mazEF, expresses a toxin predicted to belong to the TxpA type I family found in Firmicutes, and the antisense RNA antidote, RatA. Genomic analysis indicates that the cis-association of mazEF and txpA-ratA modules has been favored during evolution, suggesting a selective advantage for this TA organization in the E. faecalis species. We showed regulatory interplays between the 2 modules, involving transcription control and RNA stability. Remarkably, our data reveal that MazE and MazEF have a dual transcriptional activity: they act as autorepressors and activate ratA transcription, most likely in a direct manner. RatA controls txpA RNA levels through stability. Our data suggest a pivotal role of MazEF in the coordinated expression of mazEF and txpA-ratA modules in V583. To our knowledge, this is the first report describing a crosstalk between type I and II TAs.

Keywords:

• antisense RNAs
• Enterococcus faecalis
• RNA stability
• toxin-antitoxin systems
• transcription activation

Introduction

The Gram positive bacterium Enterococcus faecalis is a natural resident of the gastrointestinal tract that has emerged as a major cause of nosocomial infections over the last 30 years. The dual life style of E. faecalis relies primarily on remarkable capacities to face and survive in harsh environmental conditions.^1-3 We previously reported a peculiar chromosomal locus (coined hereafter TAI-II locus) containing 2 adjacent putative toxin-antitoxin (TA) modules, organized in an antisense orientation and predicted to belong to type I (txpA-ratA) and type II (mazEF) families in the clinical E. faecalis isolate V583 (Fig. 1A).⁴ The so-called txpA-ratA module was shown to encode a TA distantly related to txpA-ratA in B. subtilis.^5,6 Bacterial TAs are non-essential genetic elements co-expressing a stable toxin protein that inhibits cell growth, and the antidote, an unstable cognate antitoxin. Originally described as systems ensuring inheritance of ‘plasmid addiction’ in the absence of selective pressure, TAs have also been found abundantly in bacterial chromosomes. Remarkably, the number of TAs varies from an isolate to another within the same species, and throughout the bacterial kingdom.^5,7-9 Such a variability remains poorly understood, but the impact of TAs on bacterial growth and physiological adjustments in response to environmental changes has been demonstrated, including antibiotic resistance, biofilm development, stress adaptation, virulence and persistence during chronic infections.^10-15

Figure 1. Crosstalk regulation at the TAI-II locus between mazEF and txpA-ratA modules. (A). Gene organization and transcripts expressed by the TAI-II locus in the E. faecalis V583 isolate. (B). DNA regions used to overexpress various portions of the TAI-II locus. (C). Northern blots showing steady-state RNA levels expressed by genes mazEF, txpA, and ratA. Each lane indicates the portion of the TAI-II locus present as multicopy in addition to the chromosomal copy, as described in panel B. ‘CS’: V583 strain carrying the empty vector. Transcripts detected are mentioned on the right side of the panel. Exposure times were optimized for each panel for visualization. (D). Diagram showing fold changes of RNA levels normalized to levels in CS considered as equal to 1. p-values from statistical analysis (Student t test) are reported as ‘*’: p ≤ 0.05; ‘**’: p ≤ 0.01; ‘***’: p ≤0 .001. (E). Steady state RNA levels originating from the TAI-II locus in strains carrying an increased copy number of drm sequences.

TA modules have been classified in 5 types depending on the nature of the antitoxin and the mechanism used to counteract the toxin effects.^16-19 Type I TA (TA-I) comprises an antisense non-coding RNA gene that pairs with the mRNA encoding toxin and prevents translation or activates degradation.^5,20-23 TA-I toxins, such as TxpA, are generally small hydrophobic proteins forming pores in membranes.^5,24-27 Type II TA systems (TA-II) are usually made up by a bicistronic operon encoding the antitoxin protein that interacts with the toxin and inactivates its deleterious effects. Generally, TA-II toxin and antitoxin repress their own transcription via ‘conditional cooperativity’, a process relying on the DNA-binding properties of the antitoxin, the affinity between the toxin and the antitoxin, and their limited abundance.^28-30 Different activities have been characterized among TA-II toxins, and most of them inhibit translation acting as endoribonucleases (endoRNases), including MazF.^7,12,31,32 Although TAs are key players in adjusting physiology during adaptation processes, how bacteria coordinate their expression individually and as a whole set remain poorly understood.^15,33-37 In particular, a limited number of studies have focused on putative crosstalks between TAs. For instance in E. coli, the TA-II endoribonuclease toxin MsqR was shown to enhance a selective degradation of the ghoST mRNA in favor of the type V TA toxin GhoT, by direct cleavage within the ghoS sequence and, a complex regulatory interplay has been unveiled between several TA-II modules.^38,39 These observations suggest that together TAs may mediate concerted physiological responses in the cell.

In this work, we demonstrated that open reading frames at TAI-II locus in the chromosome of E. faecalis V583 strain feature TA-I and TA-II module properties, and characterized the regulatory interplay between mazEF and txpA-ratA modules, where MazEF ensures a pivotal role in the coordinated expression of the 2 modules.

Results

Chromosomal organization of mazEF and txpA-ratA modules within the E. faecalis species

In the E. faecalis V583 isolate, the mazEF/txpA-ratA locus (TAI-II locus) encompasses ∼1,500 nucleotides (nt) on the chromosome and expresses several transcripts (Fig. 1A).⁴ The mazEF module is transcribed from 2 promoters providing mazEF1 (∼1,200 nt) and mazEF2 (˜800 nt) RNAs. The txpA-ratA system consists of the toxin encoding gene, txpA (ef3263 whose transcript was originally named ref40, ∼290 nt), and a non-coding asRNA transcribed in the same direction as mazEF, named RatA (˜300 nt, originally named Ref46). Our previous 3′ RNA end mappings have shown the antisense organization of mRNAs encoding the MazF and TxpA toxins.^4,40 Two additional transcripts, Ref5J and Ref45 were previously reported.⁴ Because none of them was detected by Northern blot in the diverse strains and growth conditions used here, they will be not considered. Moreover, the TAI-II locus features a 167 bp long direct repeat motif, ‘drm1’ and ‘drm2’, flanking the mazEF module. drm1 and drm2 contain the transcription start sites (TSSs), the −10 boxes and extend 5 nucleotides upstream of the mazEF1 and ratA promoters, respectively (Fig. 1A; Fig. S1), indicating possible common regulatory elements. Together, antisense organization and drms strongly suggest a regulatory crosstalk between the TA-II (mazEF) and the TA-I (txpA-ratA) modules.

Searches in the V583 genome show that mazEF and txpA sequences are unique at both nucleotide and protein levels. Across the E. faecalis species, the occurrences of each module vary, and no correlation with the phylogeny of strains or virulence traits is observed. Out of 65 assembled genomic sequences from independent isolates, publicly available from the NCBI Genome Database, only 2 strains do not carry homologous sequences to any module indicating a possible selective advantage for the presence of these TA systems (Table 1, and data not shown). Half of the strains contain the txpA-ratA system alone and only a few bear mazEF alone (∼8%), suggesting that the presence of mazEF in the absence of txpA-ratA may not be favored (Table 1). About a third of the strains contain both modules, and remarkably both modules always map to the same locus and form a cis organization identical to the one described in the V583 strain, including drm sequences (Fig. 1A). This observation suggests a functional advantage provided by the proximity of the modules and reinforces our hypothesis of a crosstalk between mazEF and txpA-ratA.

Table 1. Distribution of mazEF and txpA-ratA modules in E. faecalis species. Numbers provided take into account only the independent isolates available from NCBI database whose genome have been assembled. ‘%’: percentage within the 65 assembled genomes (February 2013). ‘cis’: organization of the TAI-II locus (Fig. 1A). ‘trans’: mazEF and txpA-ratA modules found at different loci

TA modul	Strains (out of 65)s	%
None	2	3.07
mazEF	5	7.69
txpA-ratA^*	34	52.3
mazEF / txpA-ratA (cis)^#	24	36.9
mazEF / txpA-ratA (trans)	0	0

* The ‘Symbioflor’ strain contains 2 txpA-ratA modules directly repeated.

^#: The ‘D32’ strain contains the TAI-II locus, an additional copy of the txpA-ratA modules directly repeated with the one at the TAI-II, and 3 copies of the drm sequence.

TxpA and MazF of E. faecalis are toxins

First, we show that MazF (121 amino acids) and TxpA (27 aa) are genuine toxins. The corresponding ORFs were expressed in E. coli under control of the inducible promoter P_Bad. In contrast to TxpA from Bacillus subtilis which is not toxic in E. coli and does not induce lysis,²⁷ the expression of TxpA from E. faecalis V583 provoked cell death without lysis (Fig. S2 and data not shown), showing the toxic effect of the protein. This data fully agrees with the previous report on TxpA from E. faecalis.⁵ Similarly, the induction of MazF from E. faecalis V583 leads to a total growth inhibition which is reverted when MazE (76 aa) is co-expressed from the inducible promoter P_tac, indicating the toxic effect of MazF and the antitoxin activity of MazE (Fig. S2).

Then, we tested whether MazF from E. faecalis was able to trigger RNA degradation. We analyzed the stability of lpp and ompA transcripts in E. coli in a time course experiment when the toxin was induced (Fig. 2A). lpp and ompA RNAs are targets for MazF in E. coli.⁴¹ and are commonly used as indicators of the RNase activity of TA-II toxins. MazF from E. coli was used as control for this set of experiments. In comparison to the vector control, both lpp and ompA mRNAs show a decreased abundance during the time course, indicating degradation (Fig. 2A). Although both toxins inhibit growth of E. coli similarly (Fig. 2B), MazF from E. coli induces degradation more efficiently than MazF from E. faecalis. The consensus sequences recognized by TA encoded endoRNases vary from one species to another, and it may that the abundance of sequences recognized by the 2 toxins in lpp and ompA is different.¹⁷ However, it is also possible that the RNA mediated degradation of lpp and ompA by MazF from E. faecalis is indirect and a consequence of the toxic effect of the protein in E. coli.

Figure 2. MazF and TxpA from E. faecalis are toxins. (A). MazF from E. faecalis induces RNA degradation in E. coli. Time after induction is mentioned at the top of each lane. Transcripts probed are indicated on the right side of the panel. (B). Growth curve showing the toxic effect of MazF from E. faecalis when expressed in E. coli. MazF from E. coli was used as control.

Trans and cis regulatory crosstalks between mazEF and txpA-ratA modules

To investigate potential regulatory interplays between mazEF and txpA-ratA modules, we overexpressed each module separately or together in a cis-association, in addition to the TAI-II chromosomal locus (Fig. 1B). As readout, we used RNA levels expressed from each gene, and thereby we investigated: i) the effects of one module cloned on a plasmid (˜10 molcules per cell).⁴² on the expression of the other present at its chromosomal locus (i.e. trans-effect), and ii) the changes in the expression of one module cloned on a plasmid in comparison to its expression at the chromosomal locus (i.e., cis-effect). We verified that the V583 wild type strain (wt) and the wt strain harboring the empty vector (referred to as control strain (CS)) showed the same abundance for each detected transcript and grew at an equivalent rate (not shown). We also verified by direct extraction that the plasmid copy number was similar for all V583-derivative strains used (not shown).

txpA-ratA activates mazEF

In comparison to CS, the strain harboring the ectopic txpA-ratA module shows an increased abundance of txpA and RatA RNAs, as expected due to higher copy number (Fig. 1C, lane ‘txpA-ratA’). However, the magnitude of the increase differs between transcripts: ∼3.2- and ∼2-fold for txpA and RatA, respectively. Such an increase leads to an expression pattern in favor of txpA, in contrast to the reverse pattern observed in CS where the RatA signal is higher than txpA (Fig. 1D). This result suggests a regulatory process encoded outside the module that affects the txpA/RatA RNA ratio.

The overexpression of txpA-ratA further causes increased mazEF1 and slightly decreased mazEF2 transcript levels expressed from the chromosome (Fig. 1C, D): globally, the ectopic module trans-activates directly or indirectly mazEF expression (∼1.6-fold, mazEF1 and mazEF2 considered together).

mazEF activates ratA

In comparison to CS, the overexpression of the mazEF module imposes major changes in the RNA pattern (Fig. 1C). Beside the expected increase in mazEF RNA levels (∼2.4-fold), remarkably, RatA is also increased (∼2.4-fold). Therefore, mazEF must (directly or indirectly) trans-activate ratA at its chromosomal locus (Fig. 1C, D). In addition and although average values did not show significant variations for the txpA signal in comparison to CS, we did notice a reproducible slight decrease of the toxin RNA in the mazEF strain (Figs. 1C and 3).

Crosstalk between modules

To validate the effects of TA modules on each other, we overexpressed the entire TAI-II locus and compared RNA levels to those detected with constructs expressing each module individually. In comparison to the mazEF strain, the presence of txpA-ratA in the TAI-II construct, allows a higher expression of mazEF (∼5-fold relative to CS), (Fig. 1C, D, lanes ‘mazEF’ and ‘TAI-II’). This result further supports the hypothesis that the region encompassing the txpA-ratA module enhances mazEF expression.

In comparison to the txpA-ratA strain, the presence of mazEF in cis mildly increases the txpA level (≤25%) and more significantly RatA (≥50%); in comparison to CS, transcript levels increase ∼4- and ∼3.1-fold, respectively (Fig. 1C, D). Indeed, mazEF restores the RNA pattern observed in the CS strain where the RatA signal is higher than the txpA signal; a reverse pattern to the one observed in the txpA-ratA strain where signals are in favor of the toxin (Fig. 1C). This further supports our conclusion that mazEF activates ratA.

MazE and MazEF coordinate gene expression at the TAI-II locus

In E. coli, MazE has been described as a ‘milder’ transcription repressor of the mazEF promoters than MazEF.⁴³ Thus, we hypothesized that MazE and MazEF proteins in E. faecalis would have similar properties as in E. coli. Therewith, the mazEF-dependent activation of ratA could be due to the transcriptional activity of MazE and MazEF, since promoters P_mazEF1 and P_ratA share common drm sequences (Fig. 1A; Fig. S1). Based on this hypothesis and in comparison to CS, we predicted that the overexpression of MazE in trans would lead to: i) increased levels of mazEF mRNAs since the competition to bind mazEF DNA promoters between the overexpressed MazE and the chromosomally encoded MazEF would be in favor of MazE, a milder repressor than MazEF, and ii) activation of ratA at its chromosomal locus. To test our predictions, MazE was overexpressed using an identical construct to the one used for MazEF except that approximately 2-third of the C-terminal part of MazF were omitted (Fig. 1C, lane ‘mazE’). In this strain, we observed: i) increased levels of mazEF mRNAs expressed from the chromosome (∼3.3-fold), indicating that most likely in E. faecalis and E. coli, MazE and MazEF feature similar transcriptional regulatory properties with respect to mazEF promoters;⁴³ ii) increased RatA abundance expressed from the chromosome (∼2.1-fold), showing that MazE directly or indirectly trans-activates ratA, as previously concluded with the mazEF construct; iii) the increased level of RatA is accompanied by a decreased abundance of txpA (˜0.7-fold), (Fig. 1C, D).

We further confirmed the trans-activation of ratA by the mazEF module by introducing the constructs overexpressing MazEF and MazE into the E. faecalis strain OG1RF, an isolate lacking mazEF, but carrying the txpA-ratA module and a single copy of the drm sequence (Fig. S1).⁴⁴ In OG1RF, the ectopic expression of MazE and MazEF amplified the effects observed in V583: RatA levels increased more than 20-fold and were accompanied by decreased txpA levels (Fig. 3). This data shows that the txpA pattern follows RatA abundance in an inverse related manner and changes in the expression of the antitoxin RatA affects levels of txpA mRNA (Figs. 1B and 3). Although signals for txpA mRNA are not visible in some lanes presented Figure 3, they were still detected when films were overexposed, suggesting that the abundance of txpA does not depend solely on RatA level (not shown). V583 and OG1RF are 2 distant isolates within the E. faecalis species and share ∼75% of their genome,⁴⁴ suggesting that the trans-activation of ratA by mazEF may be direct. The efficacy of ratA activation by mazE and mazEF in the 2 E. faecalis strains is different, and differences in drm sequences may partly account for this (Fig. S1).

Figure 3. Effects of mazE and mazEF on txpA-ratA expression in E. faecalis strains OG1RF and V583. For OG1RF, ‘CS’ strain corresponds to the parental strain carrying the plasmid vector. Legend otherwise as in Fig. 1C.

The direct repeat motifs are DNA regulatory elements

P_mazEF1 and P_ratA share common sequences, drm1 and drm2, respectively, encompassing the −10 box and the 5′ transcribed region (Fig. S1). These data prompted us to test if drm sequences were required for the mazEF-mediated effects. For the experimental strategy, we assumed that TAI-II would conform general features of TA systems, i.e. TAs act in a stochastic manner based on low abundance of the encoded gene products, mRNA and proteins.^12,13,45 If so, MazEF protein complexes would be present in limiting amount and the increased copy number of target sequences should titrate away MazEF from chromosomal drms. Each drm was cloned on the multicopy plasmid and transcripts expressed from the chromosomal TAI-II locus were analyzed. In comparison to CS, a higher copy number of drm1 or drm2 increases the amount of mazEF transcripts (1.6- and fold2-, respectively, Fig. 1D, E). Thus, each drm is able to counteract MazEF autorepression, indicating that this regulatory effect requires the drm and most likely MazEF binds to these DNA sequences. A higher effect was observed for drm2, in comparison to drm1, suggesting a better affinity of MazEF for drm2 (Fig. 1E). We also noticed different mazEF RNA levels in strains ‘drm2’ and ‘txpA-ratA’ which also bears drm2 as multicopy (Fig. 1C, E), suggesting that sequences in the vicinity of drm2 may modulate the effect of MazEF. In comparison to CS, no significant variations of RatA and txpA levels were observed, in agreements with the model of MazEF being titrated from P_ratA on the chromosome by the drm in trans, thereby precluding MazEF-dependent activation of ratA (Fig. 1E). This data also indicates that independently of MazEF, P_ratA endows a sufficient activity to counteract txpA expression as observed when the txpA-ratA module is overexpressed alone (Fig. 1C, lane ‘txpA-ratA’).

Hereinabove, we observed the mazEF activation due to the increased copy number txpA-ratA (Fig. 1C). To verify whether such an activation could also be due to gene-encoded products from txpA-ratA, in addition to the increased copy number of drm2 sequence (Fig. 1B), we abolished the transcription initiation of txpA and ratA genes via mutagenesis of the −10 promoter boxes within the TAI-II construct.⁴⁶ Two consecutive constructs were made: a simple mutant for the toxin, txpA, and then, a double txpA ratA mutant (Fig. 4A). For each construct, RNA levels were determined by Northern blot (Fig. 4B). In the simple (‘-10_txpA’) and in the double mutant (‘-10_txpARatA’), mazEF mRNA levels remain unchanged when compared to the TAI-II strain. This data indicates that neither the expression process (i.e., transcription, per se), nor txpA- nor ratA-encoded products affect mazEF expression, but rather that the DNA sequence of txpA-ratA is responsible for the mazEF activation, most likely the drm2 sequence.

Figure 4. Effects of txpA and ratA expression on mazEF. (A). Single (‘-10_txpA’) and double (‘-10_txpARatA’) −10 box promoter mutants for txpA and ratA genes. (B). Effects of ‘-10_txpA’ and ‘-10_txpARatA’ mutations on transcripts expressed at the TAI-II locus. Legend otherwise as in Fig. 1C.

Further, in the simple mutant (‘-10_txpA’), txpA level was similar to the CS strain, as expected since the plasmid no longer produces the mRNA encoding toxin (Fig. 4B). Remarkably, in that strain RatA abundance is higher than in the TAI-II strain and hence we expect a decreased amount of txpA expressed from the chromosome. However, the txpA amount detected in the simple mutant and in the CS strain are similar (Fig. 4B), suggesting that even if RatA and txpA RNAs are fully complementary (Fig. 1A), they would not affect the abundance of each other when expressed at different loci. In the double mutant (‘-10_txpARatA’), RatA and txpA are both expressed from the chromosome. In comparison to CS, we observed an increased RatA abundance in agreement with our previous conclusion that the increased copy number of mazEF in trans activates ratA expression from the chromosome (Figs. 1C and 4B). Yet, we observed an increased amount of txpA that we don't fully understand and additional regulatory processes may occur.

All together, these data show that the txpA-ratA module activates mazEF expression via the drm2 sequence encompassing the −10 region of P_ratA.

TAI-II cross-regulation via RNA synthesis and stability

To evaluate the contribution of transcription and RNA stability to changes observed in RNA levels in strains overexpressing differents parts of the TAI-II locus (Fig. 1C), we measured the half-lives of mazEF, RatA and txpA RNAs. Northern blots were performed on total RNAs extracted from bacteria submitted to kinetic experiments after rifampicin treatments; an antibiotic that blocks transcription initiation (Fig. 5; Fig. S3).⁴⁷ In the CS strain, the 3 transcripts show distinct half-lives: ∼1, ∼2.5 and ∼11 min for mazEF, RatA and txpA, respectively.

Figure 5. Stability measurements of transcripts expressed by E. faecalis V583 strains CS, mazEF, txpA-ratA and TAI-II. Fig.S3 shows a representative experiment of RNA signals detected. For each transcript, half-life was estimated as the time required for 50% decrease of its initial amount.

Although mazEF is less stable than RatA and txpA, its stability appears to be independent of the constructs expressed in trans, indicating that variations in RNA levels are mainly due to transcriptional changes (Fig. 5).

RatA shows similar stabilities in CS, mazEF and TAI-II strains (from 2.5 to 2.9 min; Fig. 5; Fig. S3). In the mazEF strain, where RatA abundance increases compared to CS, RatA is expressed solely from its chromosomal locus, demonstrating that the mazEF-dependent activation of ratA occurs at the transcriptional level (Fig. 1C). In contrast, in the txpA-ratA strain and in comparison to the other strains, RatA half-life decreases ∼30% (∼1.8 min; Fig. 5; Fig. S3) and the ratio of txpA/RatA signals is in favor of txpA. This data indicates that the increased txpA level observed at steady state is accompanied by RatA destabilization (Fig. 1C).

txpA stability shows a more complex pattern. In comparison to CS, in strains harboring any module or the entire TAI-II locus, txpA is less stable (Fig. 5). To verify that such an effect was specific to the inserts cloned into the vector, we verified that txpA and RatA RNA levels were equivalent in the CS strain harboring the vector (pJIM2246) and a strain harboring the same vector containing a gene used as negative control (data not shown). The gene used as control was ref25C, encoding an sRNA.⁴ Thus, this observation indicates that txpA stability is sensitive to the ratio between the txpA RNA and the other encoded products at the TAI-II locus, i.e. RatA and MazEF. When txpA is expressed only from its chromosomal locus, i.e., in CS and mazEF strains, its stability drops near to 2-fold in the later strain (from 11 min in CS to 7 min; Fig. 5) where RatA level is increased via the mazEF-dependent transcription activation. This data indicates that txpA destabilization is related to increased RatA abundance. When txpA is overexpressed, in strains txpA-ratA and TAI-II, the RNA ratio of txpA/RatA is different: it is in favor of txpA in txpA-ratA strain in comparison to TAI-II strain where RatA is more abundant (Fig. 1C). In these backgrounds, txpA half-life decreases by ∼30% (from 8.5 min. to 6 min), while RatA half-life increases by ∼40% (from 1.8 min. to 2.6 min) (Fig. 5; Fig. S3). Thus, stabilities of txpA and RatA RNAs are anti-correlated and modulated by the expression level of the respective other transcript (Figs. 1C and 3).

Both mazEF and txpA-ratA are required for H2O2 sensitivity

The systematic cis-association of mazEF and txpA-ratA modules in the E. faecalis species (Table 1) prompted us to test if these TAs could co-exist separately in a ‘trans-association’. We were able to gather the 2 modules cloned in 2 different plasmids in the wild type V583 strain, and all plasmid combinations were viable (not shown). However, empty vectors together were unstable in V583 which didn't allow us to characterize further phenotypes provided by the trans-association of TAs. Yet, this set of constructs strongly suggests that the trans-association of mazEF and txpA-ratA is viable, and most likely that the cis-organization found at the TAI-II locus has been favored during evolution of the E. faecalis species.

Generally, TAs affect cell growth in response to sudden environmental changes that cause an imbalance in favor of the toxin.¹⁹ We searched for growth-related phenotypes using the set of V583-derivative strains presented Figure 1B. Several stress conditions were tested, including an array of bacteriostatic and bacteriocidal antibiotics, temperature variations, long term survival (not shown) and an oxidative stress by direct treatment with hydrogen peroxide (H₂O₂). Among those, only the H₂O₂ treatment showed a specific effect. As control, we verified that all untreated strains have a comparable growth rate and plating efficiency in the conditions used (Fig. 6, left panel). When cells were treated with H₂O₂, no significant differences were observed between CS, mazEF and txpA-ratA strains, in contrast the strain overexpressing TAI-II displayed more than 100-fold growth defect (Fig. 6, right panel). Since neither the overexpression of mazEF nor txpA-ratA conferred H₂O₂ sensitivity, we concluded that both modules, mazEF and txpA-ratA, were required for the phenotype observed indicating that together they may confer particular traits to the strain.

Figure 6. Sensitivity to H₂O₂ of E. faecalis V583 strains overexpressing mazEF, txpA-ratA and TAI-II locus. Names of strains indicated between panels.

Discussion

Gene products encoded by the TAI-II locus

The TAI-II locus of E. faecalis V583 expresses 4 detectable transcripts, including 3 mRNAs, mazEF1, mazEF2 and txpA, and a non-protein coding sRNA, RatA. We show here that the annotated ORF ef3262 encodes a toxin predicted to belong to the MazF family, and ef3261 encodes a protein that counteracts the deleterious effects of MazF, demonstrating that the so-called MazE is the cognate antitoxin. MazF enhances degradation of lpp and ompA RNAs in E. coli suggesting that it may act as an endoRNase, as described for other members of the MazF family in diverse bacterial species.¹⁷ TxpA was previously identified as a putative peptide targeting the membrane and is found in several related Gram positive species. RatA, its cognate antitoxin was characterized as an asRNA.^4,5

Regulatory model at the TAI-II locus

The cis-organization of mazEF and txpA-ratA modules, ensures a coordinated expression of the TAI-II locus featuring i) controls previously known for individual TAs in other species,^12,48-50 and ii) a regulatory interplay relying on drm sequences and the mazEF-dependent transcriptional activation of ratA (Fig. 7). To our knowledge, this is the first demonstration of a regulatory crosstalk between modules of type-I and -II families.

Figure 7. Regulatory model of the TA crosstalk at the TAI-II locus in E. faecalis V583. The mazEF-encoded product, MazEF, is able to act as an autorepressor and activates ratA transcription by using drm sequences, drm1 and drm2, respectively. MazEF is present in limiting amounts and would act preferentially on drm2, i.e., activates ratA. ratA encodes the antitoxin ncRNA RatA which decreases the stability of the encoding toxin txpA mRNA. By ‘utilizing’ the produced MazEF, indirectly, the txpA-ratA module activates mazEF transcription; a process resulting in novel synthesis of MazEF which then blocks mazEF expression. MazEF proteins and drm sequences at P_mazEF1 and P_ratA promoters are key players of the regulatory crosstalk between the TA-I and TA-II modules at the TAI-II locus. Arrows indicate activation and bars repression processes. Transcriptional and posttranscriptional controls are indicated by full and dashed lines, respectively.

mazEF expression is auto-repressed. Although, we have not confirmed the direct binding of MazE and MazEF to mazEF promoters, our genetic data agree with a model where mazEF is auto-repressed⁴³ and demonstrate that mazE- and mazEF-dependent effects require the drm DNA sequence, including drm2 contained within the ratA promoter region.

txpA transcript levels change in a RatA-dependent manner, and RNA stability is a key factor regulating the expression of the toxin. We hypothesize that txpA and RatA form an RNA duplex substrate for degradation, similarly to the homonymous system in B. subtilis that proved to be recognized and degraded by RNase III, a double-strand-specific RNase.⁵¹ However, in E. fecalis, the control of txpA levels is not strictly and only dependent on RatA. The drastically increased abundance of RatA in V583 and OG1RF strains due to the ectopic expression of MazE and MazEF reduces txpA levels, yet some a certain amount of txpA remains detectable, suggesting that a fraction of txpA molecules is refractory to RatA effects. This is consistent with RNA patterns observed in stability measurements where a significant amount of txpA remains after RatA has been wiped out. Indeed, this is further in favor of a txpA/RatA duplex formation. We assume that half-lives determined within the first minutes after rifampicin treatments result from 2 decaying kinetics: an unstable txpA/RatA duplex and a more stable ‘free’ form of txpA. This is also consistent with reported changes of txpA stability in genetic backgrounds that modify the amounts of partners forming the txpA/RatA duplex since they impact the ratio [duplexed to RatA/free form] of txpA. Furthermore and interestingly, in the −10_txpA mutant, the txpA level which is solely due to the chromosomal transcription is not affected by the increased RatA level from the vector (i.e. in trans). This indicates that txpA and RatA RNAs most likely interact during their synthesis. Folding constraints may prevent the formation of a duplex efficient for degradation once transcripts have been synthesized. Whether or not the fraction of free txpA mRNA is functional and expresses the TxpA toxin is unknown but translation regulation processes may apply as suggested for other toxins of the TA-I family.²¹

MazE and the MazEF complex have been described in several bacterial species to bind directly to promoter regions of their encoding genes and repress transcription initiation¹² and we have demonstrated here that the mazEF-dependent activation of ratA occurs at the transcriptional level. Although we cannot rule out the possibility that mazEF may repress a transcriptional repressor of ratA, our observations indicate that this activation is most likely direct. In V583, P_mazEF1 and P_ratA share significant homology, drm1 and drm2, and surrounding sequence differences may account for the opposite effect of MazE and MazEF on each promoter. DNA palindromes that encompass the TSS region and the minimal promoter sequence (i.e., −10 and −35 boxes, and the spacer), have been shown to be recognized and bound by MazE and MazEF.^43,52-54 We sought for palindromes within promoter sequences of P_mazEF1, P_mazEF2, P_ratA, P_ratA.OG (the ratA promoter in OG1RF) and indeed all 4 promoters feature major inverted repeats flanking TSSs (Fig. S4), suggesting that these DNA elements could be targeted by the MazEF complex. However, that P_mazEF1 and P_ratA.OG have identical sequences within the portion containing the inverted repeats, thus these structural elements alone cannot be responsible for the positive and negative MazEF-dependent transcriptional effects observed on P_mazEF1, P_ratA and P_ratA.OG. In addition, both drm1 and drm2 overcome mazEF-dependent effects on mazEF repression and ratA activation. This demonstrates that mazEF-mediated control requires drm sequences to function, which supports the hypothesis of a limited number of active MazE and MazEF complexes in the cell. This agrees with oscillatory properties of TA-II loci that enable bistable genetic switches.^12,45,55,56 Recent studies have demonstrated that TA-II systems can affect positively or negatively the expression of other TA-II modules.^39,57,58 However in these studies, only changes in steady-state RNA levels have been shown, and RNA synthesis (transcription) and RNA stability effects cannot be distinguished. In our experiments, direct RNA stability measurements rule out stability effects of MazE and MazEF on RatA, since the RatA half-life is independent of MazE and MazEF levels.

Thus, our current regulatory model of TAI-II is based on drm sequences recognized by MazEF, which is present in limiting amounts and with a dual transcriptional activity. Our data indicate that MazEF may act preferentially on drm2 (within P_ratA) as compared to drm1 (within P_mazEF1). If so, the MazEF-dependent activation of ratA prevents mazEF-autorepression and reduces txpA RNA stability. This ‘unconventional’ use of MazEF as activator of ratA, enables on its turn MazEF synthesis which results in mazEF transcriptional repression. Indeed, the crosstalk between mazEF and txpA-ratA modules coordinates their negative autoregulation (Fig. 7). As consequence, such a regulatory loop enables higher steady state RNA levels to be expressed and maintained for each module in comparison to a ‘standard’ isolated autorepression. This is substantiated by the significant difference of txpA and RatA RNA levels observed in V583 and OG1RF isolates, and supports the hypothesis that if individually mazEF and txpA-ratA modules offer selective advantages, their regulatory crosstalk can provide strains with novel fitness capacities. A few reports show that TA-II modules coordinate their expression by acting on their respective RNA levels or by direct interactions between non-cognate toxins and antitoxins.^39,58 Modeling such interactions suggests that interplays between TAs offer better chances to create heterogeneous populations.⁵⁹ Thus, it is tempting to speculate that functional and regulatory crosstalks between TAs, associated to genomic plasticity may provide the E. faecalis species with capacities to conquer habitats and adapt life styles.

Materials and Methods

Strains, plasmids and primers

Bacterial strains, plasmids and primers used in this study are listed Table S1. Parental E. coli K12 strains were DJ624Δara,⁶⁰ MG1655Δ10,⁶¹ and DH5α (Invitrogen), and E. faecalis strains were V583, and OG1RF.^44,62 Plasmid pCRII-TOPO (Invitrogen) was used for direct cloning of amplicons. Clones were screened with oligos FR.T7 and FR.SP6 (Table S1). Sequences of inserts were verified and cloned into the shuttle vector pJIM2246 at the EcoRI restriction site.⁴²

Direct repeated sequences, drm1 and drm2, were specifically amplified from plasmid constructs harboring mazEF and txpA-ratA modules alone, and then cloned into the vector pJIM2246. Punctual mutants of the −10 promoter boxes for ratA and txpA were obtained directly from the plasmid bearing TAI-II (Fig. 1B) and using the ‘Q5-Site-Directed mutagenesis kit’ following the manufacturer's protocol (New England Biolabs). The simple mutant ‘-10_txpA’ corresponds to changes of the −10 box TATACT of the txpA promoter (P_txpA) into CCCGAC. The double mutant ‘−10_txpAratA’ was obtained by using the ‘−10_txpA’ simple mutant construct as template and changing the −10 sequence TATAAT of the ratA promoter (P_ratA) into CCCGAC. Constructs were verified by sequencing the entire TAI-II locus. For each mutant, 2 independent plasmids were used to transform E. faecalis V583 and to perform independent experiments.

Proteins encoded by open reading frames (ORFs) annotated in the E. faecalis V583 chromosome as ef3261, ef3262, ef3263, re-named in this study mazE, mazF and txpA, respectively, were expressed under control of inducible promoters in the K12 E. coli strain DJ624Δara. ef3262 and ef3263 were cloned into pBAD33 under control of P_BAD, an arabinose-inducible promoter, and ef3261 was cloned into pKK223-3 under control of P_tac, an IPTG-inducible promoter (Table S1).

Bacterial growth

E. coli derivative strains were grown in Luria Broth medium (LB, Invitrogen) at 37°C in aerated conditions at 160 rpm. To test the toxicity of MazF (Ef3262) and TxpA (Ef3263) in E. coli, pBAD33 derivative plasmids were transformed into the DJ624Δara strain. TxpA was used as a control. Colonies of derivative strains were diluted in 10⁻² M MgSO4 and survival was estimated by spotting 10 µl of serial dilutions from 1 to 10⁻⁶ on solid M9 minimum medium containing the appropriate antibiotics and inducers.

To assay the anti-MazF activity of MazE, the pKK223-3-MazE construct was introduced into the strain harboring the plasmid pBAD33-MazF and induced by 10 µM IPTG. Plates were incubated at 37°C for 16 hours before reading. To avoid interference, the endoribonuclease activity of MazF from E. faecalis V583 was evaluated in the MG1655Δ10 E. coli strain, a K12 strain deleted for 10 TA systems encoding endoRNases, MG1655Δ10.⁶¹ Total RNA was recovered after 0, 15, 30 and 60 minutes of induction of the toxin.

E. faecalis strains were grown in brain heart infusion medium (BHI, Difco) at 37°C in static conditions as previously described.⁴ Assays using hydrogen peroxide (H₂O₂) treatments were performed as follows. Bacteria were grown to an OD₆₀₀ ranged from 0.5 to 0.55 and recovered by 3 min centrifugation at 13 000 rpm, and then resuspended in the same volume of 0.9% NaCl solution with and without (negative control) 50 mM hydrogen peroxide (H₂O₂). Cells were kept at room temperature for 1 hour, washed with 1 volume 0.9% NaCl and recovered in 1 volume 0.9% NaCl. Bacterial survival was estimated by spotting 4 µl of serial dilutions from 1 to 10⁻⁷ on BHI plates. Plates were incubated at 37°C for 16 hours before reading.

When required, for E. faecalis V583 and OG1RF derivative strains, chloramphenicol and tetracycline were added to BHI media at final concentration of 10 and 5 µg/ml, respectively. For E. coli recipient strains, ampicillin, chloramphenicol and tetracycline were added to media when required at final concentrations of 100 µg/ml, 20 µg/ml and 12.5 µg/ml, respectively.

RNA preparation, Northern blots and signal quantification

Preparation of total RNA from E. faecalis strains and Northern blots were performed as previously described.⁶³ Specific DNA oligonucleotides utilized as probe are listed Table S1. Northern blots shown are representative data obtained with at least 3 independent RNA preparations. For all figures shown, 5 µg of total RNA were loaded per lane. To measure RNA stability, bacteria were grown as described above, and at an OD₆₀₀ ranged between 0.50 and 0.55, they were treated by rifampin at 300 µg/ml as previously described.⁶⁴ Sensitivity to rifampin was confirmed by monitoring growth. Values provided are averages obtained from at least 4 independent RNA preparations tested. Half-lifes were calculated as the time necessary to obtain half of the amount detected at 0 min.

Signals were quantified using the ImageJ program.⁶⁵ Variations of RNA levels between strains were estimated by taking ‘pixel values’ of the control strain (‘CS’) as reference, i.e. equal to 1. For each transcript, gel exposures were adapted to visualize signals in each strain. Although, statistical values show the significance of quantifications, strong signals were ‘overexposed’ and corresponding pixel values were possibly underestimated.

BLAST searches

Searches for highly homologous sequences to mazE, mazF, ratA and txpA at DNA and proteins levels within assembled E. faecalis genomes were performed on February 2013.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.

Acknowledgments

We acknowledge Tatiana Rochat for comments on the manuscript and our colleagues C. Archambaud, S. Furlan, S Gaubert and L. Rigottier-Gois for constructive comments and materials.

Additional information

Funding

This research was supported by INRA, UMR1319 Micalis and grant ANR-12-BSV6-0008 (ReadRNA) from the ‘Agence Nationale pour la Recherche’.