SEARCH

SEARCH BY CITATION

Keywords:

  • axetxe;
  • protein interaction specificity;
  • toxin–antitoxin;
  • yefMyoeB

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

Toxin–antitoxin complexes are ubiquitous in bacteria. The specificity of interactions between toxins and antitoxins from homologous but non-interacting systems was investigated. Based on molecular modeling, selected amino acid residues were changed to assess which positions were crucial in the specificity of toxin–antitoxin interaction in the related Axe–Txe and YefM–YoeB complexes. No cross-interactions between wild-type proteins were detected. However, a single amino acid substitution that converts a Txe-specific residue to a YoeB-specific residue reduced, but did not abolish, Txe interaction with the Axe antitoxin. Interestingly, this alteration (Txe-Asp83Tyr) promoted functional interactions between Txe and the YefM antitoxin. The interactions between Txe-Asp83Tyr and YefM were sufficiently strong to abolish Txe toxicity and to allow effective corepression by YefM-Txe-Asp83Tyr of the promoter from which yefMyoeB is expressed. We conclude that Asp83 in Txe is crucial for the specificity of toxin–antitoxin interactions in the Axe–Txe complex and that swapping this residue for the equivalent residue in YoeB relaxes the specificity of the toxin–antitoxin interaction.


Abbreviations
CyaA

adenylate cyclase

TA

toxin–antitoxin

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

Toxin–antitoxin (TA) systems are two-component modules comprising genes coding for a toxin and a cognate antitoxin. TA complexes are abundant in bacteria [1]. These modules can be located on either plasmids or chromosomes. TAs encoded by the former are responsible for stable plasmid maintenance in bacterial cell lines [2]. Chromosomal TAs instead are involved in responses to various stress conditions, ensure genomic stability, function as anti-addiction modules or may act only as selfish genetic entities [3-7].

TA cassettes have been classified into five types (I–V) based on evolutionary, genetic and functional relationships [8]. The modules coding for type II TA systems usually consist of a pair of genes forming an operon. The first gene of the operon encodes a labile protein antitoxin which is degraded more rapidly by cellular proteases than the more stable protein toxin encoded by the second gene. The antitoxin binds to the toxin molecule and inhibits its biochemical activity.

Toxins of type II systems have been divided into 12 superfamilies based on sequence similarity [9]. These toxin superfamilies do not share evolutionary relationships and originate from distinct ancestors. Type II toxins also can be sub-divided into six groups according to structural homology [10]. The toxins within each group often show only limited sequence similarity, despite having a common fold [10]. Type II toxins predominantly are translation inhibitors, but there are examples of toxins that interfere with peptidoglycan synthesis and DNA gyrase inhibitors.

Antitoxins of type II systems have been classified into 20 superfamilies [9]. In general, type II antitoxins are composed of two distinct domains, a DNA-binding region and a toxin-binding domain. DNA-binding motifs among antitoxins can assume helix–turn–helix, ribbon–helix–helix or non-canonical folds, whereas the toxin-binding domain is less structured [6]. However, the toxin interaction domain becomes better folded upon binding to the toxin which renders the antitoxin less unstable compared with free antitoxin.

Toxins and antitoxins from different families can associate genetically and form hybrid systems [9, 11]. Sequence analyses indicate that TAs are linked by complex evolutionary relationships reflecting swapping of functional domains between different TA families [12]. One such example is the yefMyoeB locus encoded by the Escherichia coli chromosome. YefM (83 amino acids) is an antitoxin from the Phd superfamily and the YoeB toxin (84 amino acids) is a member of the ParE/RelE superfamily. The canonical association would be phddoc and relBrelE [13]. YoeB acts as a ribonuclease cleaving RNA associated with ribosomes [14]. Structures of free YoeB and of the YefM–YoeB complex have been determined and conformational changes upon formation of the complex were proposed [15]. The YefM–YoeB complex forms a heterotrimer that contains one YoeB monomer and two YefM subunits. YefM belongs to a unique group of DNA-binding proteins with a distinctive fold typified as the Phd superfamily, whereas YoeB forms a compact globular structure consisting of a five-stranded β sheet and two α helices and is related structurally to the RelE superfamily of toxins [15]. The N-terminal segments of the YefM monomers form a symmetrical dimer. One of the YefM C-terminal segments exclusively binds the YoeB monomer, whereas the other YefM C-terminus remains structurally disordered [15].

The plasmid-encoded Axe–Txe complex of Enterococcus faecium is homologous to YefM–YoeB [13]. Txe is a positively charged toxin consisting of 85 amino acid residues. Axe is an 89 amino acid long antitoxin, being negatively charged and able to neutralize the biological activity of the toxin [13]. In its free form, Txe, like YoeB, is a specific ribosome-dependent endoribonuclease that cleaves mRNA molecules three bases downstream of an AUG start codon [16]. Molecular modeling of Axe and Txe proteins showed extensive structural similarity between YefM and Axe antitoxins, as well as between YoeB and Txe toxins [17].

Irrespective of the role of TA systems, interactions between toxin and antitoxin factors must be highly specific to perform their biological function precisely. Specificity is crucial to avoid any coincidental reactions with other TA complexes that might potentially disturb the balance between toxicity and non-toxicity. Altering this balance potentially would lead to accidental death of the host cell, a process that is biologically unacceptable and which, in fact, has been eliminated during TA evolution [18]. Apart from being important features of bacterial physiology, TA complexes are of possible use in the battle against bacterial pathogens. There is considerable interest in designing molecules that compete effectively for the interface between toxin and antitoxin, or which destabilize the complex in other ways and promote toxin activation artificially [19, 20]. Thus, TA systems could serve as targets for new antibacterial drugs. However, knowledge about the specificity of TA interactions needs to be extended [21-24]. Williams and Hergenrother [24] highlighted certain key problems regarding the discovery of an artificial toxin activator. One of the current limitations is the lack of sufficient data concerning the amino acid residues that define ‘hotspots’ between toxin and antitoxin, which makes it difficult to design molecules that specifically target this interaction. Therefore, the aim of this work was to learn about mechanism(s) of the high specificity of interactions between toxin and antitoxin proteins. We assumed that different TA systems which are similar in sequences and structures but are deficient in cross-interactions between their elements should be appropriate models in such studies.

Phylogenetic studies revealed that distantly related TA systems generally show an absence of cross-talk, whereas homologs that are sufficiently related are able to cross-interact [3]. Because of the strict sequence and predicted structure similarity between Axe–Txe and YefM–YoeB complexes combined with data reporting only weak cross-interactions between the two systems [13], we aimed here to identify the determinant(s) of this specificity. A single amino acid substitution in the Txe toxin not only made this protein susceptible to the inhibitory action of YefM antitoxin, but also allowed formation of an active repressor–corepressor complex of the yefMyoeB promoter.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

Specificity of protein–protein interactions within Axe–Txe and YefM–YoeB systems

Although the YefM and Axe antitoxins are ~ 25% identical and YoeB and Txe toxins share ~ 50% sequence identity, the proteins have diverged sufficiently to associate productively only with the cognate partners in vivo [13]. Thus, we aimed to test the specificity of interactions between toxins and antitoxins from these homologous complexes. A bacterial two-hybrid system based on reconstituted adenylate cyclase (CyaA) activity was employed (BACTH Two-Hybrid System; Euromedex, Souffelweyersheim, France). Gene fusions coding for hybrid proteins composed of toxin or antitoxin and one of the two CyaA subunits were constructed in various combinations: toxin and antitoxin genes were cloned in both low (pKT25 and pKTN25) and high (pUT18 and pUT18C) copy number plasmids, either as 5′ (pUT18 and pKTN25) or 3′ (pUT18C and pKT25) fusions to the genes encoding CyaA subunits (Table 1). Examination of colony color on plates containing X-gal (data not shown) and measurement of β-galactosidase activity in planktonic cultures were used to assess protein–protein interactions. Interactions were detected only between native partners, i.e. Axe with Txe and YefM with YoeB (Figs 1 and S1). These results suggested that Axe–Txe and YefM–YoeB interactions are specific, and no cross-interactions occur. However, it is necessary to note that, even for the natural partners, the interactions could be detected only in selected two-hybrid constructs (Fig. S1A). As comparatively strong interactions are required to observe significant β-galactosidase activity in the two-hybrid assay, the relatively large sizes of CyaA subunits (18 and 25 kDa) relative to toxins and antitoxins (9–10 kDa) may interfere with precise interactions between hybrid proteins. The data also suggest that a proper stoichiometric relationship between toxin and antitoxin is important to form a stable complex [15]. The only pair of two-hybrid plasmids showing interactions for both Axe–Txe and YefM–YoeB was that in which the antitoxin gene was cloned in the pKNT25 vector and the toxin gene was located on pUT18C (Figs 1 and S1). These variants were chosen for further experiments (Fig. S1, rectangles).

Table 1. Plasmids used in this study.
PlasmidDescriptionReference
pREG531pFH450 derivative plasmid containing axetxe cassette, used for PCR amplification of this module, CmR [13]
pBAD33Plasmid for cloning genes under arabinose promoter, ori p15A, CmR [43]
pBAD33YoeBDerivative of pBAD33 with yoeB gene cloned under arabinose promoter, CmRThis study
pBAD33TxeDerivative of pBAD33 with txe gene cloned under arabinose promoter, CmRThis study
pBAD33TxeD52NDerivative of pBAD33 with mutated txe gene cloned under arabinose promoter, CmRThis study
pBAD33TxeK56FDerivative of pBAD33 with mutated txe gene cloned under arabinose promoter, CmRThis study
pBAD33TxeR70ADerivative of pBAD33 with mutated txe gene cloned under arabinose promoter, CmRThis study
pBAD33TxeF77LDerivative of pBAD33 with mutated txe gene cloned under arabinose promoter, CmRThis study
pBAD33TxeD83YDerivative of pBAD33 with mutated txe gene cloned under arabinose promoter, CmRThis study
pBAD24Plasmid for cloning genes under arabinose promoter, ori ColE1, AmpR [43]
pBAD24kanRAs pBAD24 but KanRThis study
pBAD24YefMDerivative of pBAD24 with yefM gene cloned under arabinose promoter, KanRThis study
pBAD24AxeDerivative of pBAD24 with axe gene cloned under arabinose promoter, KanRThis study
pKK223-3Vector for cloning genes under IPTG-inducible ptac promoter, ori ColE1, AmpRPL-Pharmacia Biochemicals Inc., Milwaukee, WI, USA
pKKYefMDerivative of pKK223-3 vector with yefM gene under ptac promoter, KanRThis study
pKT25Plasmid for two-hybrid system for cloning T25 on N-terminus of protein, ori p15A, KanREuromedex BACTH, Souffelweyersheim, France
pKNT25Plasmid for two-hybrid system for cloning T25 on C-terminus of protein, ori p15A, KanREuromedex BACTH, Souffelweyersheim, France
pKTzipPlasmid for positive control in two-hybrid system, zinc finger linked to T25, ori p15A, KanREuromedex BACTH, Souffelweyersheim, France
pUT18CPlasmid for two-hybrid system for cloning T18 on N-terminus of protein, ori ColE1, AmpREuromedex BACTH, Souffelweyersheim, France
pUT18CzipPlasmid for positive control in two-hybrid system, zinc finger linked to T18C, ori ColE1, AmpREuromedex BACTH, Souffelweyersheim, France
pUT18Plasmid for two-hybrid system for cloning T18 on C-terminus of protein, ori ColE1, AmpREuromedex BACTH, Souffelweyersheim, France
pUT18YefMPlasmid pUT18 with yefM gene, AmpRThis study
pUT18YoeBPlasmid pUT18 with yoeB gene, AmpRThis study
pUT18AxePlasmid pUT18 with axe gene, AmpRThis study
pUT18TxePlasmid pUT18 with txe gene, AmpRThis study
pUT18CYefMPlasmid pUT18C with yefM gene, AmpRThis study
pUT18CYoeBPlasmid pUT18C with yoeB gene, AmpRThis study
pUT18CAxePlasmid pUT18C with axe gene, AmpRThis study
pUT18CTxePlasmid pUT18C with txe gene, AmpRThis study
pKT25YefMPlasmid pKT25 with yefM gene, KanRThis study
pKT25YoeBPlasmid pKT25 with yoeB gene, KanRThis study
pKT25AxePlasmid pKT25 with axe gene, KanRThis study
pKT25TxePlasmid pKT25 with txe gene, KanRThis study
pKNT25YefMPlasmid pKNT25 with yefM gene, KanRThis study
pKNT25YoeBPlasmid pKNT25 with yoeB gene, KanRThis study
pKNT25AxePlasmid pKNT25 with axe gene, KanRThis study
pKNT25TxePlasmid pKNT25 with txe gene, KanRThis study
pUT18CTxeD52NPlasmid pUT18C with mutated txe gene, AmpRThis study
pUT18CTxeK56FPlasmid pUT18C with mutated txe gene, AmpRThis study
pUT18CTxeR70APlasmid pUT18C with mutated txe gene, AmpRThis study
pUT18CTxeF77LPlasmid pUT18C with mutated txe gene, AmpRThis study
pUT18CTxeD83YPlasmid pUT18C with mutated txe gene, AmpRThis study
pBBRluxVector for generating transcriptional fusion to lux, CmR [44]
pBBRlux_yyyefMyoeB promoter-operator region cloned between SpeI–BamHI sites in pBBRluxThis study
image

Figure 1. Interaction of Axe–Txe and YefM–YoeB proteins with cognate and non-cognate partners in a bacterial two-hybrid system. The antitoxin genes cloned into the pKNT25 vector and the toxin genes located on pUT18C were co-introduced in pairs into the BTH101 reporter strain. β-galactosidase activity was measured in overnight cultures which were grown in the presence of 0.5 mm IPTG. K+ and K− denote positive (pKTzip-pUT18Czip) and negative (pKTzip-pUT18C) controls, respectively. Results are averages of at least three independent experiments with error bars indicating SD.

Download figure to PowerPoint

In their free forms, both Txe and YoeB cause strong growth inhibition of E. coli cultures. This inhibition is diminished upon interaction with the cognate antitoxin [13]. Accordingly, severe inhibition of growth was observed when expression of either toxin gene was induced from the strictly regulated arabinose promoter (Fig. 2A,B). Simultaneous expression of the cognate antitoxin gene abolished this toxic effect. Importantly, this phenomenon was observed only when production of either Axe–Txe or YefM–YoeB pairs occurred in the same cell. No cross-interactions between non-cognate pairs, i.e. Axe–YoeB and YefM–Txe, were detected (Fig. 2A,B). These findings agree with previous observations that growth inhibition by the homologous YoeB and Txe toxins is relieved only weakly, if at all, by Axe and YefM, respectively [13]. Thus, despite the homology between the systems, Axe–Txe and YefM–YoeB interactions are highly specific.

image

Figure 2. Axe–Txe and YefM–YoeB protein–protein interactions in toxicity rescue experiments. E. coli SC301467 cells harboring derivatives of pBAD33 vector (for toxins) and pBAD24 (for antitoxins) were grown at 37 °C in the presence of 1% l-arabinose. Absorbance readings at 600 nm were taken at 60 min intervals. (A) YefM, but not Axe, is able to counteract the toxicity of YoeB. (B) Toxicity of Txe is inhibited by Axe, but not YefM. Results are typical of three independent experiments.

Download figure to PowerPoint

Differences in TA-binding domains between Axe–Txe and YefM–YoeB complexes

Based on amino acid sequence and structural similarity, YefM and Axe are included in the Phd antitoxin superfamily, whereas YoeB and Txe toxins are homologs in the RelE superfamily. The antitoxins exhibit more sequence divergence than the toxins: YefM shows ~ 25% identity overall with Axe [13]; Txe and YoeB are ~ 50% identical. The structures of YefM–YoeB and free YoeB proteins have been determined [15]. Models for Axe and Txe mainly showing electrostatic potentials have been described [17], but bona fide structures are not available. The levels of structural similarity and difference between the two complexes were assessed here by molecular modeling. Tertiary fold-recognition was performed along with protein structure modeling by homology modeling using the ‘Frankenstein's monster’ approach [25, 26]. Models were built based on the sequence alignment between Axe–Txe and the template crystal structure of the YefM–YoeB complex (PDB 2A6Q) [15]. The assessment of proq scores for Axe and Txe single molecules was not high; however, since these proteins exist as a complex, such results were acceptable. Moreover, to obtain detailed information about interactions between Axe and Txe, a protein–protein docking procedure, using haddock software, was performed [27]. Amino acids on the interface of Axe (residues 51, 54, 58, 61, 64) and Txe (residues 64, 66, 82, 83, 84, 85) were selected as active. Protein–protein docking was performed with default settings that work well for average complexes (Fig. S2).

Results of molecular modeling and protein–protein docking show a high similarity between Axe–Txe and YefM–YoeB structures (Fig. 3A). Analysis of electrostatic potential and differences in amino acid properties on the protein surfaces showed some disparities that might potentially be responsible for specificities of interactions between Txe and YoeB and their partner antitoxins. The surface in YefM that binds YoeB is formed by the L-shaped turn between the acidic α-helix H3 and the neutral α-helix H4 which hooks the S2–S3 loop of YoeB [15]. In this loop, the major role is played by Leu48, His50, Asn51 and Leu52 which stabilize the L-shaped structure of the antitoxin [15]. These four amino acids are conserved in Txe, apart from Asn51 which is replaced by Asp52. Asn51 in YoeB interacts with Tyr53 of YefM within the disordered C-terminus. Tyr53 is replaced by Arg54 in Axe. As asparagine and aspartic acid have different physicochemical properties, Asn51 and Asp52 in YoeB and Txe, respectively, were chosen for further analysis as described below.

image

Figure 3. Molecular modeling of Axe–Txe and YefM–YoeB. (A) Model of structural similarity between Axe–Txe and YefM–YoeB protein complexes. Axe, light blue; YefM, dark green; Txe, dark blue; YoeB, light green. (B) YefM–YoeB dimer structure based on [15]. YefM, dark green; YoeB, green. Amino acids in YoeB potentially responsible for TA specificity are highlighted. Protein models are displayed with pymol software. (C) Alignment of amino acid sequences of YoeB and Txe proteins. Blue, black and red colors correspond to identical, similar and different amino acids, respectively. Black stars indicate residues chosen for swapping from YoeB into Txe.

Download figure to PowerPoint

YoeB presents a concave surface created by twisted β sheets S3 and S4 which expose hydrophobic residues: Phe55 (Lys56 in Txe), Val67 (Ile68 in Txe), Leu76 (Phe77 in Txe) and Ala78 (Tyr79 in Txe) [15]. These residues, via intermolecular van der Waals interactions, bind the amphipathic α-helix H4 in YefM antitoxin, notably residues Leu64 (Ile65 in Axe), Met65 (Arg66 in Axe), Ile68 (Asp69 in Axe) and Leu71 (Phe72 in Axe) [15]. The residues described that differ between YoeB and Txe toxins and between YefM and Axe antitoxins may be implicated in the contact specificity of these complexes. Thus, from these residues Phe55 and Leu76 were chosen for further analyses. Upon binding to the YefM homodimer, certain mechanistic conformational changes occur in the three C-terminal residues (Tyr82, His83 and Tyr84) of the YoeB toxin [15]. This region builds the catalytic center of the YoeB ribonuclease. The two most C-terminal amino acids are conserved in both toxins, but Tyr82 in YoeB is replaced by Asp83 in Txe. This residue was also chosen for analysis in further tests. The final residue selected in YoeB was Ala69 (Arg70 in Txe) which lies in close proximity to the contact site with the antitoxin. All residues chosen for mutational analysis are depicted in Fig. 3B,C.

Substitutions of Txe-specific residues by YoeB-specific residues neither abolishes toxicity entirely nor abolishes effective interactions with the Axe antitoxin

On the basis of the molecular modeling described above, five candidate amino acids in the Txe toxin were targeted which might be implicated in the specificity of antitoxin interactions in Axe–Txe and YefM–YoeB. Site-directed mutagenesis was used to generate Txe derivatives bearing Asp52Asn, Lys56Phe, Arg70Ala, Phe77Leu or Asp83Tyr substitutions in which the indicated residues in Txe were swapped to those present in YoeB. We then tested whether any of these substitutions affected Txe toxicity, antitoxicity by Axe, or promoted cross-interactions between the variant Txe toxins and the native YefM antitoxin.

All changes introduced into the Txe protein impaired, to various extents, the interaction with Axe in two-hybrid assays (Fig. 4A). The most pronounced effects were observed when Txe-Asp52Asn, Txe-Phe77Leu and Txe-Asp83Tyr variants were tested. Nevertheless, all mutant Txe proteins interacted more strongly with Axe than did YoeB indicating that the changes did not entirely abolish interaction with the cognate antitoxin. No interactions between the Txe variants and YefM were evident in the two-hybrid system (Fig. 4B). However, as previous results suggested that the biological toxicity assay may be more sensitive than the two-hybrid system in testing interactions between elements of the TA systems (Figs 1 and 2), the growth rates of strains producing the different Txe variants and wild-type YefM were tested as described above. None of the Asp52Asn, Lys56Phe, Arg70Ala, Phe77Leu or Asp83Tyr substitutions impaired the toxic activity of Txe in vivo (Fig. 5A). Interestingly, the toxicity of all of the Txe variants and of wild-type Txe were counteracted equally effectively by Axe, producing indistinguishable growth patterns (Fig. 5B). These data indicate that the Asp52Asn, Lys56Phe, Arg70Ala, Phe77Leu and Asp83Tyr substitutions in Txe neither abolished toxicity nor abolished effective interactions with the Axe antitoxin.

image

Figure 4. The ability of Txe mutants to interact with Axe (A) or YefM (B) measured in a bacterial two-hybrid system by the β-galactosidase assay. Derivatives of vectors pUT18C (for toxins) and pKT25N (for antitoxins) were co-introduced into the BTH101 reporter strain. β-galactosidase activity was measured in overnight cultures which were grown in the presence of 0.5 mm IPTG. Results are averages of at least three independent experiments with error bars indicating SD.

Download figure to PowerPoint

image

Figure 5. The ability of Txe mutants (A) to express toxicity, (B) to interact with Axe and (C) to interact with YefM in toxicity rescue experiments. E. coli SC301467 cells harboring derivatives of pBAD33 vector (for toxins) and pBAD24 (for antitoxins) were grown at 37 °C in the presence of 1% l-arabinose. Absorbance readings at 600 nm were taken at 30 min intervals. Results are typical of at least two independent experiments.

Download figure to PowerPoint

The Asp83Tyr substitution in Txe allows for functional interaction with both Axe and YefM antitoxins

YefM does not counteract Txe toxicity (Fig. 2). Similarly, Txe derivatives bearing the Asp52Asn, Lys56Phe, Arg70Ala or Phe77Leu substitutions that retained toxicity as outlined above were not neutralized by YefM. In contrast, toxicity of the Txe variant bearing the Asp83Tyr change in which a Txe-specific residue was altered to a YoeB-specific amino acid was significantly impaired by YefM (Fig. 5C). As Txe-Asp83Tyr is also neutralized effectively by Axe (Fig. 5B), this result suggests that this position in Txe and the corresponding position in YoeB are crucial for specificity of Axe–Txe and YefM–YoeB interactions.

The YefM–YoeB complex represses expression of its cognate promoter at the transcriptional level [28, 29]. As Txe-Asp83Tyr is neutralized by YoeB in vivo, the ability of this variant to function together with YefM as a corepressor was tested. Hence, a fusion of the promoter with the lux operon was constructed and luminescence was measured in cultures of bacteria expressing genes coding for the YefM antitoxin and either YoeB or Txe-Asp83Tyr. For these experiments, a plasmid carrying yefM under the isopropyl thio-β-d-galactoside (IPTG) inducible ptac promoter in a pKK223-3 derivative was used, whereas the toxins were produced from the arabinose-inducible pBAD promoter in pBAD33 derivatives. It was necessary to trigger transcription of toxin and antitoxin genes with different inducers since a proper stoichiometric relationship is required for repression activity by toxin and antitoxin proteins which is impossible with an expression system in which both are induced with arabinose (B. Kędzierska, unpublished data). The values obtained in the experiments without any repression (vector pBBRlux bearing the yefMyoeB promoter co-introduced with pBAD33 and pKK223-3 plasmids) were considered as ‘no repression’, and presented values (fold repression) reflect the efficiency of luminescence in the tested systems (where YefM alone or together with YoeB or TxeAsp83Tyr was present in pBAD33 and pKK223-3 plasmids) relative to that of the control (i.e. without any repression). The YefM–YoeB complex repressed the activity of the tested promoter ~ 110-fold (Fig. 6). As described previously, YefM alone was a significantly weaker repressor, causing ~ 30-fold repression [28, 29]. Significantly, the combination of YefM with Txe-Asp83Tyr was ~ 50% as effective as the wild-type YefM–YoeB complex in repressing the yefMyoeB promoter (Fig. 6). It was not feasible to use wild-type Txe as a control in these experiments as it is not neutralized by YefM and therefore remains an active endoribonuclease. In summary, the repression data corroborate the finding in toxicity and antitoxicity assays that the Asp83Tyr substitution in Txe allows for effective interaction of this toxin with the YefM antitoxin.

image

Figure 6. Txe-Asp83Tyr acts as a corepressor of the yefMyoeB promoter in the complex with YefM antitoxin. Transcriptional fusion of the yefMyoeB promoter to the luxCDABE operon in pBBRlux plasmid was co-introduced into E. coli SC301467 along with pKKYefM and pBAD33YoeB or pBAD33TxeD83Y. Cultures were grown until A600 ~ 0.1 and then 0.1 mm IPTG and 1% l-arabinose were added. Luminescence in relative luminescence units was measured after 1 h of induction. Fold repression denotes scores obtained by dividing the results of probes where only the yefMyoeB promoter was present by the results of experiments where repressors were overexpressed from plasmids, in addition. Results are averages of three independent experiments with error bars indicating SD.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

To achieve their biological aim, interactions between components of TA systems must be highly specific. It is crucial to avoid non-specific cross-interactions between different TA complexes and with other cellular factors. Moreover, in addition to being interesting subjects for studies of the biological significance of toxins or models for investigating specificity of protein–protein interactions, TA complexes are potential targets for new antimicrobial agents. In particular, ligands that compete effectively and specifically for the interface between toxin and antitoxin or which destabilize the complex in other ways to promote toxin activation may be important agents in the ongoing battle against antibiotic-resistant infectious bacteria. EDF quorum-sensing peptides are examples of naturally occurring ligands that compete for the TA interface [30, 31]. Synthetic peptides that inhibit TA interactions in a type II complex have also been described [20]. Thus, understanding the specific modes of interaction between proteins within TA modules will guide the design of novel peptides and small molecules with antibacterial potential.

In this report, we demonstrated that a single amino acid substitution, Asp83Tyr, in Txe allows this toxin to interact with the YefM antitoxin that belongs to a different, but homologous, TA system. These interactions were sufficiently effective to protect E. coli cells against toxin-mediated growth inhibition and to act with the non-cognate antitoxin as a transcriptional repressor of the yefMyoeB promoter. The Txe-Asp83Tyr derivative retained toxicity in vivo where it was also counteracted by the cognate antitoxin, Axe. The Asp83Tyr substitution in Txe did not make this toxin as effective as YoeB in interactions with YefM, although the association was sufficiently strong to support functions in vivo. Nevertheless, the Asp83 residue in Txe and the Tyr residue at the corresponding position in YoeB appear to be crucial for specificity of interactions between these toxins and their partner antitoxins. Overall, the results reveal that the interaction specificity in related but independent TA complexes indeed depends on amino acid residues at the toxin–antitoxin interface.

Tyr82 of YoeB, which corresponds to Asp83 in Txe, interacts with Tyr53 of YefM, and the stacking of aromatic rings in these side chains is stabilized by hydrogen bonds from each hydroxyl group to carbonyl groups of the opposing backbone atoms [15] (Fig. 7A). Therefore the Asp83Tyr substitution in Txe may make this toxin sufficiently similar to YoeB to permit effective interactions with YefM. Nevertheless, interactions of Axe with Txe may differ in important ways. The positions corresponding to Tyr82 in YoeB and Tyr53 in YefM are the negatively charged Asp83 in Txe and positively charged Arg54 in Axe, respectively. Thus, the different nature of crucial physicochemical interactions at these positions – ionic interactions in Axe–Txe versus stacking of aromatic rings in YefM–YoeB – may be responsible at least in part for the lack of substantial cross-interactions between these systems (Fig. 7).

image

Figure 7. Conformation of the toxin active site when bound to the antitoxin in the case of E. coli YefM–YoeB (A) and E. faecium Axe–Txe (B) obtained by superimposition of modeled subunits of Axe and Txe into the crystal structure of YefM–YoeB [15]. Key interactions within the Axe–Txe complex, obtained by the haddock method, with indicated hydrogen bond formation between depicted residues, are shown in (C). Side chains of residues participating in interactions are shown as sticks and labeled.

Download figure to PowerPoint

Interestingly, investigation of YefM–YoeB from E. coli and Streptococcus pneumoniae revealed a lack of cross-interactions between toxins and antitoxins from these systems [17]. Comparison of protein sequences indicates that YefM–YoeB from S. pneumoniae is more similar to E. faecium Axe–Txe than to YefM–YoeB from E. coli. The toxins exhibit 55% identity whereas the antitoxins share as much as 46% identical amino acid residues (Fig. 8). Notably, residues Asp83 in Txe and Arg54 in Axe are conserved at corresponding positions in S. pneumoniae YefM and YoeB, respectively, compared with the Tyr residues at the equivalent positions in E. coli YefM and YoeB. Therefore, it seems plausible that cross-interactions will occur between toxins and antitoxins from Axe–Txe and S. pneumoniae YefM–YoeB. Another line of evidence supporting our hypothesis about the important role of Asp83 in Txe and Tyr82 in YoeB in directing interaction specificity with the cognate antitoxins comes from observations that YefM of Mycobacterium tuberculosis can neutralize toxicity caused by YoeB of E. coli [32]. The mycobacterial proteins show high identity – 42% for antitoxin and 46% for the toxin – to E. coli YefM–YoeB proteins. Moreover, the mycobacterial proteins possess Tyr residues at the positions corresponding to Tyr53 in YefM and Tyr82 in YoeB of E. coli (Fig. 8).

image

Figure 8. Alignments of selected members of the YefM (A) and YoeB (B) families of proteins using clustalw2. Ec, E. coli; Mt, M. tuberculosis; Spn, S. pneumoniae; Ef, E. faecium. Ovals indicate residues crucial for YoeB toxicity (dotted line) and important for YefM–YoeB interactions which cause changes in the toxin active site. Black ovals indicate identical or similar amino acids in all members, whereas red ovals highlight dissimilar amino acids.

Download figure to PowerPoint

The data with E. coli, enterococcal, streptococcal and mycobacterial YefM–YoeB homologs are consistent with a key role in toxin–antitoxin interaction for residues at positions equivalent to Tyr53 in YefM and Tyr82 in YoeB of E. coli. In contrast, YefM and YoeB homologs from E. coli and Staphylococcus aureus both possess the critical tyrosine residues at corresponding positions, but E. coli YefM is unable to neutralize the toxicity of staphylococcal YoeB [33]. However, these experiments were performed with a C-terminal His-tag on the YefM antitoxin. It is known that the C-terminal domain of YefM is engaged in contacts with the toxin and a tag at this position could potentially influence these interactions, especially with the non-cognate toxin for which the contact sites might not be optimal. Moreover, the data for the E. coli -S. aureus experiments were based on bacterial growth on solid medium. When relatively weak interactions occur, residual toxicity accumulated overnight could kill the cells. Additionally, staphylococcal YoeB is four amino acids longer at its N-terminus than E. coli YoeB. This and differences in other segments of the primary sequence may be responsible for the observed lack of cross-interaction.

In summary, upon formation of the complex, YoeB residues necessary for endoribonuclease activity are hidden by binding of the YefM antitoxin. Residues Glu46, Arg65, His83 and Tyr84 are required for YoeB toxicity and are conserved in YoeB homologs (Fig. 8, black circles) [15]. A mechanistic conformational change of the three C-terminal amino acids of YoeB (Tyr82, His83, Tyr84) takes place upon YefM binding which destabilizes a canonical conformation in the catalytic center of the endoribonuclease [15]. The analyses of amino acid sequence and the complex structures suggest that the mechanisms of inhibition of YoeB homologs by YefM homologs, including Axe–Txe, are highly similar despite changes in amino acid composition at some positions. Tyr82 is present in E. coli and mycobacterial YoeB, but is replaced by an Asp residue in Txe and streptococcal YoeB (Fig. 8, red circle). Residues in YefM that are engaged in conformational change at the catalytic site of YoeB are depicted in Fig. 7 and shown as circles in Fig. 8 (Glu50, Tyr53, Ser57, Asn60 and Arg63). Among these, Tyr53 and Asn60 differ between Axe and YefM. The only pair of residues that interact with each other and possess pronounced physicochemical differences are Tyr82 of YoeB and Tyr53 of YefM. These residues are critical for the specificity of homologous but non-interacting toxins and antitoxins which harbor dissimilar residues at the corresponding positions.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

Strains

E. coli DH5α and XL1-Blue were used for plasmid construction, strain SC301467 [34] for luminescence assays and toxicity rescue experiments, and strain BTH101 (Euromedex, Souffelweyersheim, France) for two-hybrid assays. Bacteria were grown in Luria-Bertani (LB) medium. Ampicillin, kanamycin and chloramphenicol were added to final concentrations of 100, 50 and 34 μg·mL−1, respectively, when required.

Plasmids and oligonucleotides

Oligonucleotides and plasmids used in this study are listed in Table S1 and Table 1, respectively.

Tertiary structure prediction and modeling

Tertiary fold-recognition was carried out via the genesilico metaserver gateway (http://genesilico.pl/meta2/) [35]. Homology modeling was carried out using the ‘Frankenstein's monster’ approach [25, 26]. Models were built with swiss-model [36-38] based on the sequence alignment between Axe, Txe and the template crystal structure of YefM–YoeB complex (PDB 2A6Q) obtained from various fold-recognition servers. The preliminary models were assessed by the metamqap [39] and proq [40] methods to predict their accuracy. Fragments for which there was no template were refined de novo using refiner [41]. Final models were evaluated as potentially ‘good’ by the proq and metamqap methods. proq assessment scores for Axe chain C (Predicted LGscore 1.508, Predicted MaxSub 0.184), Axe chain D (Predicted LGscore 1.440, Predicted MaxSub 0.190) and Txe chain F (Predicted LGscore 1.748, Predicted MaxSub 0.267) are indicated in parentheses.

haddock [27] (High Ambiguity Driven Biomolecular Docking) was used to show a simulation of Axe–Txe protein–protein docking; 165 structures were clustered in 17 clusters, which represents 82.5% of the water-refined haddock generated models. The statistics of the best cluster are shown below. Obtained haddock scores were −128.2 ± 7.6; cluster size 9, rmsd from the overall lowest-energy structure 1.1 ± 0.6; van der Waals energy −31.2 ± 7.0; electrostatic energy −524.5 ± 37.6; desolvation energy 7.5 ± 6.0; restraints violation energy 3.9 ± 1.69; buried surface area 1833.0 ± 163.5; Z score −1.9.

The graphic representation of protein structures was performed with pymol software (PyMOL Molecular Graphics System, Version 1.2r3pre; Schrödinger LLC).

Protein sequence analysis

Alignments of selected members of the YefM and YoeB families of proteins were performed using clustalw2.

Bacterial two-hybrid system assay

The BACTH Two-Hybrid System (Euromedex, Souffelweyersheim, France) was used to establish cross-interactions between YefM–YoeB and Axe–Txe proteins and to assess the ability of Txe mutants to interact with cognate and non-cognate antitoxins. Genes were amplified by PCR using specific primer pairs (Table S1) and E. coli genomic DNA or pREG531 plasmid DNA templates for YefM–YoeB and Axe–Txe, respectively. PCR fragments were cloned into the pKT25, pKT25N, pUT18 and pUTC18 to produce recombinant plasmids (Table 1). Combinations of plasmid pairs were used to co-transform the BTH101 reporter strain, and bacteria were plated onto screening medium containing 40 μg·mL−1 X-gal, ampicillin (100 μg·mL−1), kanamycin (50 μg·mL−1) and 0.5 mm IPTG. Following 48 h incubation at 30 °C, colonies were evaluated for blue or white color and then inoculated into LB with appropriate antibiotics for β-galactosidase assays. Assays were performed using cells permeabilized with chloroform and SDS as described by Miller [42]. Co-transformants containing vectors pKT25-zip and pUT18C-zip or pKT25-zip and pUT18C were used as positive and negative controls, respectively.

Toxicity rescue experiments

E. coli SC301467 overnight cultures containing pBAD24 or pBAD33 derivatives were diluted 100-fold into fresh LB broth containing the appropriate antibiotics and 1% l-arabinose. Subsequent growth was monitored by taking absorbance readings at 600 nm at half-hourly or hourly intervals.

Promoter fusion studies and bioluminescence assays

Strain SC301467 harboring pBBRlux-amp with the lux operon under transcriptional control of the yefMyoeB promoter and derivatives of pKK223-3 and pBAD33 plasmids carrying antitoxin and toxin, respectively, were used. Overnight cultures carrying recombinant plasmids were diluted (1 : 100) into fresh LB medium with appropriate antibiotics, grown until A600 ~ 0.1 and induced with 0.1 mm IPTG and 1% l-arabinose for 1 h. Luminescence of 200 μL of cell culture was measured in a luminometer (Berthold Technologies, Junior). Results in relative light units were divided by the optical density (A600) of the cultures.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

We are grateful to Anna Czerwoniec from VitaInSilica Sp. z o. o. for performing the Axe–Txe protein model and figures showing protein structures, and to Aleksandra Sikora for a gift of pBBRlux plasmid. This work was supported by the Polish Ministry of Science and Higher Education (project grant no. N N301 251936 to BK).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Materials and methods
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
febs12517-sup-0001-FigS1-S2-TableS1.zipapplication/ZIP325K

Fig. S1. Axe–Txe and YefM–YoeB protein–protein interactions in the bacterial two-hybrid system.

Fig. S2. Axe–Txe protein–protein interactions showed by docking procedure haddock.

Table S1. Oligonucleotides used in this study.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.