|Home | About | Journals | Submit | Contact Us | Français|
Bacteria are constantly challenged by bacteriophage (phage) infection and have developed multiple adaptive resistance mechanisms. These mechanisms include the abortive infection systems, which promote “altruistic suicide” of an infected cell, protecting the clonal population. A cryptic plasmid of Erwinia carotovora subsp. atroseptica, pECA1039, has been shown to encode an abortive infection system. This highly effective system is active across multiple genera of gram-negative bacteria and against a spectrum of phages. Designated ToxIN, this two-component abortive infection system acts as a toxin-antitoxin module. ToxIN is the first member of a new type III class of protein-RNA toxin-antitoxin modules, of which there are multiple homologues cross-genera. We characterized in more detail the abortive infection phenotype of ToxIN using a suite of Erwinia phages and performed mutagenesis of the ToxI and ToxN components. We determined the minimal ToxI RNA sequence in the native operon that is both necessary and sufficient for abortive infection and to counteract the toxicity of ToxN. Furthermore, site-directed mutagenesis of ToxN revealed key conserved amino acids in this defining member of the new group of toxic proteins. The mechanism of phage activation of the ToxIN system was investigated and was shown to have no effect on the levels of the ToxN protein. Finally, evidence of negative autoregulation of the toxIN operon, a common feature of toxin-antitoxin systems, is presented. This work on the components of the ToxIN system suggests that there is very tight toxin regulation prior to suicide activation by incoming phage.
Interactions between bacteria and their natural parasites, bacteriophages (phage), have global-scale effects (42). Although the vast majority of the phage infections, which occur at a rate of 1025 infections per s (26), are overlooked by humans, en masse they affect environmental nutrient cycling (18) and have long been known to be vital to the spread and continued diversity of microbial genes (11). A tiny proportion of this activity can directly affect our everyday activities; the lysis of bacteria following phage infection has potential medical benefits, such as use in phage therapy (30), or can be economically damaging, as it is in cases of bacterial fermentation failure (for instance, in the dairy industry ).
Gram-positive lactococcal strains used in dairy fermentation have been shown to naturally harbor multiple phage resistance mechanisms (16). These mechanisms can be broadly classed as systems which (i) prevent phage adsorption, (ii) interfere with phage DNA injection, (iii) restrict unmodified DNA, and (iv) induce abortive infection. There is also an increasing amount of research that focuses on new systems that use clustered regularly interspaced short palindromic repeats to mediate phage resistance (3). Clustered regularly interspaced short palindromic repeats and associated proteins, although widespread in archaea and bacteria (39), have not been identified yet in lactococcal strains (23).
The abortive infection (Abi) systems induce cell death upon phage infection and often rely on a toxic protein to cause “altruistic cell suicide” in the infected host (16). Although Abi systems have been studied predominantly using lactococcal systems, because of their potential economic importance (8) they have been identified in some gram-negative species, such as Escherichia coli, Vibrio cholerae, Shigella dysenteriae, and Erwinia carotovora (9, 14, 36, 38). The prr and lit systems of E. coli have been studied at the molecular level, and their mode of action and mode of activation by incoming phage have been identified (2, 37, 38). In contrast, lactococcal Abi systems have been characterized mainly by the range of phages actively aborted and the scale of these effects, and the Abi systems have been grouped based on general modes of action (8, 12). More recently, research has begun to identify more specific lactococcal Abi activities at the molecular level (12, 17) and has revealed phage activation of two such Abi systems (6, 21).
An Abi system was identified on plasmid pECA1039, which was isolated from a strain of the phytopathogen E. carotovora subsp. atroseptica (14). Designated ToxIN, this two-component Abi system operates as a novel protein-RNA toxin-antitoxin (TA) system to abort phage infection in multiple gram-negative bacteria. The toxic activity of the ToxN protein was inhibited by ToxI RNA, which consists of 5.5 direct repeats of 36 nucleotides. It is now recognized that TA loci, which were originally characterized as “plasmid addiction” modules (43), are widely distributed in the chromosomes of archaea and bacteria (19) and in phage genomes, such as that of the extrachromosomal prophage P1 (27). As a result, the precise biological role of TA systems is under debate (29). It is clear, however, that they can be effective phage resistance systems, as is the case for toxIN in E. carotovora subsp. atroseptica (14) and hok/sok and mazEF in E. coli (22, 33). Previously characterized TA systems operate with both components interacting as either RNAs (e.g., hok/sok) (type I) or proteins (e.g., MazE and MazF) (type II). In this study, a mutagenesis approach was used to further characterize the ToxI and ToxN components of the new (type III) protein-RNA TA Abi system. The regulation of the operon and the mode of phage activation were also examined.
E. carotovora subsp. atroseptica strain SCRI 1043 (4) and E. coli DH5α (Gibco/BRL) were grown at 25°C and 37°C, respectively, in Luria broth (LB) (5 g liter−1 yeast extract, 10 g liter−1 Bacto tryptone, 5 g liter−1 NaCl) at 300 rpm or on LB agar (LBA) containing 1.5% (wt/vol) agar. Bacterial growth (optical density at 600 nm [OD600]) was measured with a Unicam Heλios spectrophotometer. When required, LB was supplemented with antibiotics at the following final concentrations: kanamycin, 50 μg ml−1; spectinomycin, 50 μg ml−1; ampicillin, 100 μg ml−1; and tetracycline, 35 μg ml−1. When required, d-glucose (Glc) was used at a concentration of 0.2% (wt/vol), l-arabinose (l-Ara) was used at a concentration of 0.1% (wt/vol), and IPTG (isopropyl-β-d-thiogalactopyranoside) was used at a concentration of 1 mM, unless otherwise stated.
E. carotovora subsp. atroseptica SCRI 1043 phages B1, B5, and B24 were kindly provided by Arild Sletten (Bioforsk, Aas, Norway). Mutants of phages A2 and M1 were isolated in this study. All other phages have been described previously (41). Efficiencies of plaque formation (EOP) were calculated after overnight incubation of phages on LBA lawns of the bacterial host by dividing the phage titer on the test host by the phage titer on the control host. The LBA used contained either 0.35% or 0.5% (wt/vol) agar (see Table Table33 and Fig. Fig.11 and and3).3). All experiments were performed at least in duplicate, and where applicable, the data were expressed as means ± standard deviations.
Molecular biology techniques and sequencing were performed as previously described (15). All plasmids were verified by DNA sequencing and are shown in Table Table1.1. DNA oligonucleotides used in this study are shown in Table Table2.2. DNA sequence data were analyzed using GCG (Genetics Computer Group, University of Wisconsin).
Analysis of the ToxN protein sequence was performed using Compute pI/Mw as part of the ExPASy server (http://www.expasy.ch/tools/pi_tool.html) and SignalP 3.0 (5; http://www.cbs.dtu.dk/services/SignalP/). In silico structural analysis was performed using a FUGUE (35) server (http://tardis.nibio.go.jp/fugue/prfsearch.html) and the HMM SAM (24) servers (http://compbio.soe.ucsc.edu/sam.html). Multiple-sequence alignment of proteins was performed using Clustal W2 (25; http://www.ebi.ac.uk/Tools/clustalw2/index.html), and secondary structure predictions were made using psiPRED (32; http://bioinf.cs.ucl.ac.uk/psipred/) and jPRED (10; http://www.compbio.dundee.ac.uk/www-jpred/). Figure Figure22 was generated by manually combining outputs of psiPRED and CLC Main Workbench 5.
In vitro transposon mutagenesis of plasmid pECA1039 with EZ::TN <NotI/KAN-3> was performed as described previously (14).
A series of toxI deletion mutations in the native toxIN locus were constructed to assess the redundancy of toxI stoichiometry with respect to phage infection. A complete toxIN locus under its natural promoter was cloned previously into pBR322 (7), resulting in plasmid pTA46 (14). The pTA46 plasmid therefore contains all 5.5 repeats constituting the toxI sequence. While the toxN site-directed mutants were being made (see below), mutations in two clones resulted in deletions in toxI upstream of mutant versions of the toxN gene. This was assumed to have occurred due to slippage between repetitive sequences in toxI during rounds of PCR amplification. The 4.5- and 3.5-repeat sequences were then cloned adjacent to a wild-type toxN gene using an overlap extension PCR strategy. First, the 4.5- and 3.5-repeat toxI sequences were amplified using primers MJ7 and PF146 and the mutant toxI clones as the template DNA. Next, the wild-type toxN gene was amplified from pTA46 with primers PF145 and MJ11. The toxI and toxN PCR products were then mixed and used as a template in an overlap extension PCR with external primers MJ7 and MJ11. The resulting product was digested with EcoRI and HindIII and cloned into pBR322 cut with the same enzymes. The resulting plasmids were pTA69 (toxI [4.5 repeats] toxN) and pTA63 (toxI [3.5 repeats] toxN). To generate a plasmid with 2.5 toxI repeats, PCR was performed with primers PF167 and MJ11. Next, the native promoter region was cloned with primers MJ7 and PF162. The products of these two reactions were then used as a template in an overlap PCR using primers MJ7 and MJ11, which generated plasmid pTA75 (toxI [2.5 repeats] toxN).
An overlap extension PCR site-directed mutagenesis (SDM) strategy was used to construct 12 alanine substitution mutants and a single glutamate substitution mutant of ToxN. First, the toxIN locus was amplified by PCR using primers MJ7 and MJ11 and cloned into pBR322 cut with EcoRI and EagI, generating plasmid pMJ1. Next, left and right arms of mutant constructs were generated by PCR with pMJ1 as the template using primer MJ7 and a specific reverse primer (left arm) and a specific forward primer and primer 322Mut (right arm). The appropriate left and right PCR products were then used as the template in a further round of PCR with primers MJ7 and 322Mut, and the product was digested with EcoRI and BamHI and ligated to pMJ1 previously digested with the same enzymes. The resulting site-directed mutant plasmids and the specific primers used for their construction are as follows: for Y10A, pMJX1 and primers Y10Ar and Y10Af; for K20A, pMJX2 and primers K20Ar and K20Af; for K20E, pMJX3 and primers K20Er and K20Ef; for F35A, pTA64 and primers F35Ar and F35Af; for G37A, pMJX5 and primers G37Ar and G37Af; for Y47A, pMJX6 and primers Y47Ar and Y47Af; for S53A, pMJX7 and primers S53Ar and S53Af; for S64A pMJX8 and primers S64Ar and S64Af; for K87A, pMJX9 and primers K87Ar and K87Af; for K116A, pMJX10 and primers K116Ar and K116Af; for G142A, pMJX11 and primers G142Ar and G142Af; for C148A, pMJX12 and primers C148Ar and C148Af; and for E154A, pTA65 and primers E154Ar and E154Af.
To create FLAG-tagged versions of the toxIN site-directed mutants, primers MJ7 and MJ13 were used to amplify the site-directed mutations with addition of a C-terminal FLAG tag. The products were digested with EcoRI and HindIII and ligated into pBR322 digested with the same enzymes. The resulting FLAG-tagged SDM plasmids were as follows: for Y10A, pTA53; for K20A, pMJX2-F; for K20E, pMJX3-F; for F35A, pTA67; for G37A, pTA55; for Y47A, pMJX6-F; for S53A, pTA56; for S64A, pTA57; for K87A, pTA58; for K116A, pTA59; for G142A, pTA60; for C148A, pTA61; and for E154A, pTA68. In addition, a toxIN-FLAG wild-type control was constructed in the same manner using pMJ1 as the template, yielding pMJ4 (toxIN-FLAG).
The stability of the FLAG-tagged SDM ToxN proteins in E. carotovora subsp. atroseptica SCRI 1043 was determined by Western blotting, as described previously (14).
To test the toxicity of the site-directed mutant ToxN proteins, expression vectors were constructed. The untagged SDM plasmids (see above) were used as templates in PCRs with primers PF137 and MJ13, which resulted in a C-terminal FLAG tag being added to the ToxN open reading frame (ORF) in each case. The products were digested with EcoRI and HindIII and cloned into pBAD30 (20) cut with the same enzymes. The resulting inducible ToxN-FLAG SDM plasmids were as follows: for Y10A, pTRB39; for K20A, pTRB47; for K20E, pTRB40; for F35A, pTRB48; for G37A, pTRB53; for Y47A, pTRB42; for S53A, pTRB49; for S64A, pTRB50; for K87A, pTRB44; for K116A, pTRB51; for G142A, pTRB52; for C148A, pTRB45; and for E154A, pTRB46.
Cultures of DH5α containing ToxN expression plasmids, grown overnight with ampicillin and Glc, were used to inoculate 25 ml LB supplemented with ampicillin and Glc in 250-ml flasks. Each culture was grown to an OD600 of ~1, and a 500-μl sample was pelleted by centrifugation at 16,200 × g for 5 min at 4°C. The pellet was resuspended and serially diluted in phosphate-buffered saline. Dilution series were plated on LBA containing ampicillin and l-Ara. Colony counts were obtained after overnight incubation at 37°C. All experiments were performed in triplicate, and data were expressed as means ± standard deviations.
Following inoculation using overnight cultures, 25-ml cultures of E. carotovora subsp. atroseptica(pMJ4) and E. carotovora subsp. atroseptica(pTA46) were grown to an OD600 of ~1 in 250-ml flasks. Preinfection samples were removed, and the E. carotovora subsp. atroseptica(pMJ4) cultures were split into two 10-ml cultures in new 250-ml flasks. These cultures were then infected at a multiplicity of infection of 1 with each phage. A wild-type phage and the corresponding escape phage were used to infect 10-ml cultures derived from the same starter culture. At 5, 10, 20, and 40 min postinfection, samples were taken and the OD600 were determined. Each sample consisted of 1 ml of culture and was centrifuged at 16,200 × g for 5 min at 4°C to pellet the bacteria, and the supernatant was removed before the pellet was flash-frozen. The pellets were analyzed to determine the presence of FLAG-tagged protein by Western blotting performed as described previously (14).
Autoregulation assays using pRW50-based (28) promoter-lacZ fusion vectors were performed as described previously (15). To assess expression of the toxIN-lacZ promoter probe construct when there was overexpression of ToxI and ToxN components, β-galactosidase activity was measured in strains of E. coli DH5α(pTA104) (PtoxIN-lacZ) (14) carrying additional plasmids. For ToxN, plasmids pTA50 (ParaBAD toxN-FS control) and pTA49 (ParaBAD toxN) were used, and for ToxI, pTA100 (vector control) and pTA75 (Ptac toxI) were used, together with appropriate antibiotics and IPTG and l-Ara for induction. To determine the effects of ToxI or ToxN on toxIN promoter activity in the native operon context, β-galactosidase assays were performed with E. coli DH5α carrying plasmid pTA104, pTA106, pTA119, or pTA120. Plasmids pTA104, pTA106, and pTA119 were generated previously (14). Plasmid pTA120 was constructed by amplifying the toxIN operon and promoter with primers PF186 and PF220, digesting the product with EcoRI and HindIII, and ligating the product into pRW50 previously cut with the same enzymes.
ToxIN has been shown previously to be an Abi system in multiple genera of gram-negative bacteria that is active against several phages (14). To more fully characterize the ToxIN Abi phenotype, we investigated the system using a suite of Erwinia phages isolated using E. carotovora subsp. atroseptica SCRI 1043 (41) (some phages were provided by A. Sletten). We performed EOP assays with 25 Erwinia phages active against E. carotovora subsp. atroseptica SCRI 1043(pTA46) (toxIN), using E. carotovora subsp. atroseptica SCRI 1043(pTA47) (toxIN with a frameshift mutation in toxN) as a control (Table (Table3).3). Conservatively, for 15 of the 25 phages the EOP was reduced by ToxIN. Surprisingly, in one case (B24) there was a slight increase in the EOP on E. carotovora subsp. atroseptica SCRI 1043(pTA46).
During these assays, spontaneous “escape” mutants of Α2 and Μ1 were isolated from plaques formed on E. carotovora subsp. atroseptica SCRI 1043(pTA46) at frequencies of approximately 10−6 and 10−5, respectively. The escape phenotype was shown to be heritable by performing EOP assays following at least two passages of plaque-purified phage through wild-type E. carotovora subsp. atroseptica SCRI 1043 (Table (Table3).3). The escape mutations did not restore the EOP for the phages to 1, and there was still a modest reduction in plaque size on E. carotovora subsp. atroseptica SCRI 1043(pTA46) compared to the plaque size on E. carotovora subsp. atroseptica SCRI 1043(pTA47) (Table (Table3).3). Therefore, in each case the “escape” was significant although not complete.
As part of the characterization of toxIN as a TA locus, it was previously demonstrated that a single ToxI antitoxin repeat could protect E. coli from ToxN toxicity, when both proteins were expressed artificially from foreign promoters on separate plasmids (14). This indicated that at least one ToxI repeat RNA, when it was highly expressed, was required and was sufficient to protect E. coli from ToxN. However, the functional redundancy of the ToxI RNA in the native toxIN operon in terms of both protection from ToxN toxicity and phage infection was unknown. First, in vitro random transposon mutagenesis was performed for the native pECA1039 plasmid. Seven mutants with unique insertional mutations in the repeats were isolated, and their resistance to Α2 in E. carotovora subsp. atroseptica SCRI 1043 was compared with the phage resistance of a no-vector control, a mutant with a transposon insertion that did not affect toxIN function (pECA1039-Km12), and a negative control toxN transposon mutant (pECA1039-Km23) (Fig. (Fig.1A).1A). The control plasmid pECA1039-Km12 provided phage resistance (EOP, 3 × 10−4), whereas the transposon mutation in toxN eliminated resistance. Six independent transposon mutations in the second, fourth, and fifth repeats reduced phage resistance (Fig. (Fig.1A).1A). However, an insertion in the sixth (half) repeat resulted in a ~10- to 100-fold increase in resistance to Α2 (Fig. (Fig.1A).1A). Therefore, transposon insertions throughout the toxI antitoxin sequence interfered with the function, whereas one insertion improved phage resistance.
To investigate the redundancy in the toxI repeats for phage resistance more subtly, a deletion analysis was performed. Plasmids were constructed that contained the toxIN operon under its native promoter with full-length toxI (5.5 repeats) and different numbers of repeats (4.5, 3.5, and 2.5 repeats). The initial 5′ toxI repeat has a single base alteration compared to the subsequent four full-length repeats; the deleted repeats were repeats in this array of four exact toxI repeats. Attempts were made to construct plasmids with 1.5 and 0.5 repeats; however, this could be done only in an operon with toxN-FS (frameshifted toxN) and not in an operon with toxN, suggesting that 1.5 and 0.5 repeats were not sufficient to protect E. coli from toxN toxicity under the growth conditions used (data not shown). In E. carotovora subsp. atroseptica SCRI 1043, plasmids with 5.5, 4.5, 3.5, and 2.5 toxI repeats could all function with toxN to provide resistance to phage Α2 (Fig. (Fig.1B).1B). Therefore, 2.5 toxI antitoxin repeats were both necessary and sufficient in the native toxIN operon context to provide a functioning phage resistance mechanism with toxN and counteract ToxN toxicity.
The toxN gene encodes a 171-amino-acid protein with a predicted molecular mass of 19.7 kDa and a pI of 5.98 that is not predicted to contain any signal sequence. Database searches showed that ToxN is the defining member of a novel family of toxins (14) that includes the previously identified Abi protein AbiQ (13). This family, which currently consists of approximately 20 members, has no structural homologues in the databases or any predicted characterized domains, as ascertained by FUGUE (35) and HMM SAM (24) analysis. To begin an investigation of the mode of action of toxicity and the Abi phenotype, a ClustalW2 alignment of ToxN homologues was constructed (Fig. (Fig.2).2). From this alignment it was possible to select residues of interest to mutate using SDM. Predictions of secondary structure were also made using the ToxN sequence and the ToxN homologue alignment obtained. The consensus output (Fig. (Fig.2)2) shows a predicted secondary structure with mixed alpha and beta characteristics, including two extended alpha-helices at the C terminus.
An SDM procedure was used to obtain 13 different ToxN amino acid substitution mutants in the native toxIN operon context. The abilities of these plasmids to confer resistance to A2 in E. carotovora subsp. atroseptica SCRI 1043 were assessed (Fig. (Fig.3A).3A). Mutations Y10A, K20E, F35A, G37A, Y47A, S53A, K116A, and E154A all eliminated ToxN function, whereas K20A, S64A, K87A, G142A, and C148A mutants could still abort A2.
To allow monitoring of the levels of wild-type and mutant ToxN proteins, a FLAG tag was introduced into all SDM plasmids. It was observed that an N-terminal tag reduced wild-type ToxN protein expression compared to the expression with a C-terminal tag (data not shown), so all tags were introduced at the C terminus. The biological function of the tagged ToxN variants was first tested using A2 (Fig. (Fig.3A).3A). The same functional trend was observed for all of the FLAG-tagged SDM mutants, although there were some quantitative differences. The EOP due to ToxN-FLAG compared to ToxN was investigated further using phages A2 and M1 and the corresponding escape phages A2a and M1b. The EOP of A2 and M1 were decreased a further 10- to 100-fold by ToxN-FLAG compared to the results obtained with untagged ToxN. Furthermore, the escape phage M1b was not affected by ToxN-FLAG, but the EOP for the escape phage A2a was 1 × 10−3. These data imply that there may be multiple mutational routes that can confer different levels of immunity to ToxN, while the C-terminal FLAG tag with ToxN enhances the Abi phenotype of ToxIN.
To test if the expression and/or stability of the ToxN proteins had been affected by the mutations, Western blot analysis of uninfected E. carotovora subsp. atroseptica strains was performed. This analysis showed that all mutant proteins were expressed at levels comparable to that of wild-type ToxN-FLAG, except for ToxN(Y47A), which was present at a lower level (Fig. (Fig.3B).3B). Together, these data identified conserved residues important for the function of ToxN in phage resistance.
To investigate the effect of the SDM substitutions on the second phenotype of ToxN (toxicity), cultures of DH5α containing the relevant expressive vectors were grown to an OD600 of ~1 under repression conditions and then plated under inducing conditions to obtain viable counts. The toxicity of C-terminally FLAG-tagged ToxN wild-type and mutant proteins was compared to the toxicity of vector-only and nontagged ToxN controls (Fig. (Fig.3C3C).
In all cases, the quantitative data exhibited the same trend as the EOP data (Fig. (Fig.3A).3A). Where a substitution removed Abi activity, it also eliminated toxicity. This suggests that the two phenotypes may be inextricably linked.
As ToxN-FLAG is detectable in an uninfected cell (Fig. (Fig.3B),3B), we investigated the effect of phage infection on the relative cellular levels of ToxN-FLAG. ToxN-FLAG levels were monitored before and after phage infection. Cultures of E. carotovora subsp. atroseptica SCRI 1043(pMJ4) (toxIN-FLAG) were infected with either phages A2 and M1 or the escape mutants A2a and M1b. Samples were taken at multiple time points postinfection, and the last sample was taken 40 min postinfection, which was chosen because it preceded the burst of A2 and M1 observed at 45 to 50 min after infection of wild-type E. carotovora subsp. atroseptica SCRI 1043 (data not shown). Samples were assessed by Western blotting to determine FLAG levels (Fig. (Fig.4A).4A). ToxN-FLAG was detectable prior to infection, and the level did not increase during infection.
As we were able to visualize ToxN-FLAG in uninfected bacteria, we became interested in whether this product regulated transcription from the toxIN promoter. TA operons are tightly regulated, and repression of their promoters is elicited by TA complexes. In some cases the antitoxin alone is sufficient for partial repression of the TA operon, but the TA complex is required for full inhibition of expression (19). To test if the toxIN operon was autoregulated, the expression of a low-copy-number toxIN promoter lacZ fusion was assessed, and the effects of overexpressing ToxI or both ToxI and ToxN were measured (Fig. (Fig.4B4B and and4E).4E). ToxI had no effect on expression of the toxIN promoter, whereas a combination of ToxI and ToxN led to an approximately twofold reduction in promoter activity (Fig. (Fig.4B).4B). The impact of ToxN alone could not be assessed due to its growth-inhibiting effects on E. coli. To examine this result in a more native context, transcriptional activity was measured for plasmids that contained lacZ fusions after the toxIN promoter (pTA104), after toxI (pTA119), before toxN (pTA106), and after toxN (pTA120). This analysis confirmed that the presence of ToxI had no effect on PtoxIN activity (Fig. (Fig.4C4C and and4E).4E). Furthermore, production of both ToxI and ToxN (pTA120) completely eliminated detectable PtoxIN activity compared to the ToxI-expressing control (pTA106) (Fig. (Fig.4D4D and and4E).4E). Note that the difference in expression between ToxI-expressing vectors pTA119 and pTA106 is due to the presence of a terminator between toxI and toxN in pTA106 (14). Therefore, ToxI and ToxN together can repress the expression of a toxIN promoter fusion, suggesting that this operon is autoregulated in a manner similar to that of other TA operons.
ToxIN is a highly effective Abi system which has devastating effects on the EOP of some phages (Table (Table3).3). However, some phages are only moderately affected, and others appear to either avoid or, inexplicably in one case (B24), benefit from the system. Twenty-two of the 25 Erwinia phages (all but the B phages) have been characterized and grouped according to plaque morphology, structural morphology, genomic DNA restriction digestion profiles, and host range (40). The groups do not correlate with the observed susceptibility to abortive infection by ToxIN. By isolating “escape” ToxIN-resistant mutants of phages A2 and M1, we may be able to understand the range of ToxIN activity better. It is known that some Abi systems are activated by a specific phage product (6, 21), and specific phage mutations provide resistance to other Abi systems (17). As we did not observe gross genomic changes in the escape mutants using EcoRV restriction digestion profiles (data not shown) and as the other characteristics of the Erwinia phages discussed above (40) cannot be used to define ToxIN susceptibility, it is possible that small differences, even a single point mutation(s), may provide resistance to ToxIN. If so, the wide range of independently isolated phages sensitive to ToxIN may contain a common product capable of activating ToxIN. Whether the actual method of activation is conserved and whether the mode is direct or indirect remain to be resolved. As we were able to isolate escape mutants of only a minority of ToxIN-sensitive phages, it is possible that the method of activation is more diverse. In order to answer these questions, sequencing of the wild-type and multiple independent escape phages will be undertaken to look for any common themes in the mode(s) of escape.
The toxI gene and analogues of this gene (14) are characterized as DNA sequences containing multiple consecutive sequence repeats followed by a terminator structure and an ORF homologous to toxN, rather than by any specific DNA sequence. As such, it is difficult to predict any kind of toxI consensus sequence, so to mutate toxI and investigate its role, we chose to use random transposon mutagenesis of the native plasmid pECA1039. Insertions in the full-length repeats eliminated Abi activity, but an insertion in the half-repeat at the 3′ end of ToxI enhanced the effect compared to controls (Fig. (Fig.1A).1A). Although this finding cannot be explained yet, it could have been due to transcriptional readthrough from either the internal kanamycin resistance cassette or Tn5 itself. Whatever the cause, this result can be construed as highlighting a necessary balance of operon transcription through toxI into toxN to mediate abortive infection. Furthermore, dissection of the locus showed that a minimum of 2.5 ToxI repeats is necessary to produce a functional “native” Abi system (Fig. (Fig.1B).1B). Examination of the ToxI analogues (14) showed that ToxI itself is unusual as the majority of analogues have only 2 or 3 full repeats and the ToxN homologue AbiQ is effective as an Abi system with only 2.8 repeats in the cognate toxI gene (13). It is possible that the sensitivity of the system depends on the number of ToxI repeats. The redundancy suggests that the number of repeats may have increased via a strand slippage mechanism during DNA replication.
The ToxN homologues shown in Fig. Fig.22 represent the best protein BLAST (1) results, although they reflect levels of amino acid identity ranging only from 28% to 39%. The homologues shown are found on both chromosomes and plasmids, which reflects the widespread nature of other TA systems (19). Whereas secondary structure predictions indicate that the ToxN family consists of cytosolic proteins with mixed alpha-helix and beta-sheet characteristics, the lack of structural homologues prevents any robust conclusions from being made.
The ToxN residues picked for mutagenesis include many residues that are highly conserved in multiple ToxN homologues and other residues that are not highly conserved (Fig. (Fig.2).2). In many cases, mutating the highly conserved residues produced a clear phenotype. For instance, the Y10A, G37A, Y47A, S53A, and E154A mutations are all mutations in conserved residues that eliminated both ToxN activities assayed, highlighting the correlation between ToxN mutations that eliminate Abi activity and ToxN mutations that prevent toxicity. In contrast, mutation of another conserved residue, C148, had no effect on the phenotype. This mutation is also important as, although ToxN has no clear signal sequence, it provides more evidence that further rules out disulfide bond formation between the two cysteines of ToxN as a process that is essential for ToxN activity. Through mutation of less-well-conserved residues, we explored the possibility of defining important residues, such as F35 and K116, and the production of unexpected phenotypes. With ToxN(K87A), for instance, we observed that the Abi phenotype was partially reduced (for the FLAG-tagged version [Fig. [Fig.3A]),3A]), but protein expression and toxicity appeared to be unaffected (Fig. (Fig.3B3B and and3C).3C). When the colonies containing ToxN(K87A) were examined, they were found to be much smaller than vector-only control colonies (data not shown). Thus, it can be suggested that the toxic phenotype was also partially reduced in this case and that ToxN(K87A) represents an intermediate mutant form of ToxN. Our SDM study therefore showed that amino acids Y10, K20, F35, G37, Y47, S53, K87 (see above), K116, and E154 are essential for the phage resistance and toxic activities of the ToxN protein. Furthermore, the expression of the mutant proteins was not altered compared to that of wild-type ToxN, as assessed by Western blotting, except for the Y47A mutant. However, SDM, rather than identifying specific key residues, may have led to incorrect folding of the final mutant protein, resulting in removal of ToxN activity. Therefore, to understand fully the importance of these residues in ToxN function, it will be necessary to acquire structural data for a member of the novel toxin family. As ToxN is inhibited by ToxI RNA, it seems reasonable to hypothesize that ToxN can bind RNA. Key ToxN residues that have been identified may be involved either in this binding or in the toxic activity of ToxN. Importantly, if ToxN can indeed bind RNA, we need to establish whether this activity is directly linked to toxicity, as multiple toxins of TA systems act as RNases (19, 34).
A key mechanistic question in understanding Abi systems is, how do phages trigger these systems? In the case of Abi systems such as the abiA, abiB, abiD1, and abiG systems, no increase in transcription could be observed following phage infection (6). It has been shown that the translational efficiency of abiD1 mRNA increases following phage infection (6). This method of regulation may apply to the other Abi systems mentioned above. In the case of ToxN, however, a C-terminal FLAG tag allowed visualization of the protein in uninfected cells, and the results showed that there was no discernible change following phage infection (Fig. (Fig.4A).4A). Therefore, unlike the findings for the Abi systems described above, regulation by phage infection appears to occur posttranslationally. This regulation matches the regulation of protein-protein TA systems such as PhD-Doc, in which the proteinaceous Doc toxin is kept inactive by the antitoxic PhD protein (19). Degradation of PhD allows Doc toxicity to take effect. In the case of ToxIN, however, we have shown that ToxN is inhibited by an RNA, ToxI, which may or may not occur through direct binding (14). We hypothesize that by interfering with cellular processes such as bacterial transcription or by degrading host DNA, phage infection leads to an imbalance in the TA components. It is predicted that ToxI is rapidly degraded and the more stable ToxN protein remains in the cell and has a toxic, cell-inhibiting effect, in much the same way that plasmid loss or stress responses lead to triggering of TA loci (19, 43).
TA systems autoregulate at the transcriptional level (19). Using specific regulatory sequences of DNA, they control activity at the operon promoter, usually as a TA complex, although some antitoxins, such as ParD, can act alone to control transcription (19). We tested whether ToxIN is regulated at the transcriptional level by using defined lacZ fusions. ToxI alone did not regulate the toxIN promoter, but when it was in a complex with ToxN, transcription from PtoxIN was reduced (Fig. (Fig.4B4B and and4D).4D). We were not able to determine whether ToxN alone mediates the repression, as the toxicity of uncomplexed ToxN prevents the necessary experiments from being performed. Analysis of the toxIN promoter region for inverted or tandem repeats failed to identify any obvious potential regulator binding sites near the previously predicted −10 and −35 regions, so the binding site of ToxIN is unknown. Prior to phage infection and during cell division, it may be necessary for cells to negatively autoregulate PtoxIN in order to correctly control the levels of the components to prevent aberrant activation and toxicity by ToxN. The lack of a detectable increase in the ToxN protein concentration following phage infection (Fig. (Fig.4A)4A) suggests that the autoregulation is not an important factor in the Abi “cell suicide” cascade.
We have shown that ToxIN acts as an effective Abi system and that this phenotype relies on a delicate interplay between ToxI and ToxN. These two components regulate their own levels, allowing an appropriate response following phage infection. Furthermore, the number of ToxI repeats was shown to be a factor in the phage resistance mechanism and perhaps is an added level of stoichiometric control. A direct phage interaction may be required for activation. Sequencing of the isolated “escape” phages should allow exploration of this hypothesis, together with an accurate comparison between the observed AbiQ activity and ToxN. AbiQ acts at a late stage, following phage DNA replication, to abort infection (13). ToxN aborts infection through toxicity to the host cell, but the stage in relation to the phage morphogenesis program is not known. However, phage DNA accumulation is similar in cells with and without ToxN, suggesting that this system may be a late-acting Abi system (data not shown). The SDM analysis determined a number of key amino acids essential for functionality in the system. Further use of the characterized ToxN mutant versions may result in a greater understanding of how ToxN is inhibited by ToxI and how ToxN mediates toxicity as part of this new type III class of TA systems.
This study was supported by the Biotechnology and Biological Sciences Research Council and was performed under Department for Environment, Food and Rural Affairs plant health license PHL 177B/5951. T.B. was also supported by a Collaborative Award in Science and Engineering studentship from UCB-Celltech Ltd.
We thank Arild Sletten, Bioforsk, Aas, Norway, for his very generous provision of phages B1, B5, and B24. We also thank members of the Salmond and Welch groups, particularly Ian Foulds, for invaluable technical support and discussions.
Published ahead of print on 24 July 2009.