Structure and activity of AbiQ, a lactococcal endoribonuclease belonging to the type III toxin–antitoxin system

Authors

  • Julie E. Samson,

    1. Département de biochimie, de microbiologie et de bio-informatique, Faculté des sciences et génie, Université Laval, Québec, Canada
    Search for more papers by this author
  • Silvia Spinelli,

    1. Architecture et Fonction des Macromolécules Biologiques, UMR 6098, CNRS – Aix-Marseille University, MARSEILLE Cedex 09, France
    Search for more papers by this author
  • Christian Cambillau,

    1. Architecture et Fonction des Macromolécules Biologiques, UMR 6098, CNRS – Aix-Marseille University, MARSEILLE Cedex 09, France
    Search for more papers by this author
  • Sylvain Moineau

    Corresponding author
    • Département de biochimie, de microbiologie et de bio-informatique, Faculté des sciences et génie, Université Laval, Québec, Canada
    Search for more papers by this author

For correspondence. E-mail Sylvain.Moineau@bcm.ulaval.ca; Tel. (+1) 418 656 3712; Fax (+1) 418 656 2861.

Summary

AbiQ is a phage resistance mechanism found on a native plasmid of Lactococcus lactis that abort virulent phage infections. In this study, we experimentally demonstrate that AbiQ belongs to the recently described type III toxin–antitoxin systems. When overexpressed, the AbiQ protein (ABIQ) is toxic and causes bacterial death in a bacteriostatic manner. Northern and Western blot experiments revealed that the abiQ gene is transcribed and translated constitutively, and its expression is not activated by a phage product. ABIQ is an endoribonuclease that specifically cleaves its cognate antitoxin RNA molecule in vivo. The crystal structure of ABIQ was solved and site-directed mutagenesis identified key amino acids for its anti-phage and/or its RNase function. The AbiQ system is the first lactococcal abortive infection system characterized to date at a structural level.

Introduction

In most environments, bacteria and phages are engaged in a seemingly endless arms race. Phages evolve and adapt to improve their fitness when facing new hosts. This phenomenon is highlighted by the diversity in phage genomes. Conversely, bacteria have developed a plethora of strategies to cope with phage infections (Labrie et al., 2010). The nature of this predator–prey interaction shapes microbial populations by maintaining their equilibrium and coexistence. Studying phage–host relation is of primary importance to, among others, industrial applications such as food fermentations and possibly controls our own microbiome.

Lactococcus lactis is a Gram-positive lactic acid bacterium used to produce an array of fermented dairy products. This microorganism, which is generally recognized as safe, has also been leveraged to produce proteins of medical or industrial interest (Mierau and Kleerebezem, 2005). As with any microbes, this bacterium can be infected by virulent phages present in their environment (Moineau and Lévesque, 2005). Thus, economically relevant bacterial cultures can be lost due to phage lysis. To circumvent this problem, it is possible to use or to construct bacterial strains resistant to phage infection (Garneau and Moineau, 2011). Natural anti-phage systems are divided into several groups based on the phage multiplication step they affect: inhibition of phage adsorption, inhibition of phage DNA ejection, cleavage of nucleic acids (e.g. CRISPR-Cas and Restriction-Modification), and abortion of phage infection (Abi) (Labrie et al., 2010). Abi is a general term that groups all defence systems that acts after phage DNA ejection and before cell lysis as well as leading to bacterial cell death, thereby limiting phage spreading.

Lactococcus lactis has been a productive model to discover abortive mechanisms as 23 different lactococcal Abi systems have been reported (Chopin et al., 2005; Durmaz and Klaenhammer, 2007; Holubova and Josephsen, 2007; Haaber et al., 2008). Usually, the resistance phenotype is mediated by a single host gene, except for AbiE, AbiG, AbiL, AbiT and AbiU, which are encoded by two genes, AbiR, encoded by possibly four genes and AbiS, composed of a specific DNA structure (Dai et al., 2001; Josephsen and Neve, 2004; Chopin et al., 2005; Yang et al., 2006). Most of these genes have been found on plasmids (Chopin et al., 2005). Their mode of action has been poorly characterized due to their diversity and to the lack of in-depth knowledge of lactococcal phage biology, although some progress has been made lately. Nonetheless, the mode of action of lactococcal Abi systems appears to be highly diverse and system-dependent.

For example, AbiD1 was shown to be activated by a phage protein and to interfere with a RuvC-like endonuclease to inhibit phage proliferation (Bidnenko et al., 1998, 2009). On the other hand, the reverse transcriptase-related protein AbiK mediates phage resistance by polymerizing an untemplated DNA (Fortier et al., 2005; Wang et al., 2011). Lactococcal AbiP, a protein anchored to the membrane by its N-terminal domain, has been shown to arrest phage DNA replication and to inhibit the switch off of phage early transcripts (Domingues et al., 2004, 2008). AbiV blocks the activation of late gene transcription probably by inhibition of general translation (Haaber et al., 2010). AbiZ may interact with the phage holin to cause premature cell lysis (Durmaz and Klaenhammer, 2007).

The gene coding for AbiQ system was isolated from the L. lactis native plasmid pSRQ900 (Emond et al., 1998). This protein of 172 amino acids (20.3 kDa) is positively charged with a predicted pI of 9.6. Phage infection aborted by the AbiQ system leads to the accumulation of non-mature forms of the viral DNA within the host (Emond et al., 1998). Recent bioinformatic analyses predicted that the AbiQ mechanism is related to another Abi mechanism named ToxIN also found on a plasmid but of Pectobacterium atrosepticum (Fineran et al., 2009). ToxIN is also a type III toxin–antitoxin (TA) system. TA systems have also other roles including among others plasmid maintenance, response to environmental stresses, persister cells and programmed bacterial death (Magnuson, 2007; Yamaguchi et al., 2011). They are composed of a toxic protein that kills bacteria if it is not neutralized by an antitoxin. The nature of the neutralizing relationship defines the type of TA system (Yamaguchi et al., 2011). Type I systems are composed of an RNA antitoxin that inactivates the mRNA of the toxin; type II systems involve interaction of the protein antitoxin with the toxic protein; and type III systems involve an antitoxic RNA molecule interacting with its toxic protein. ToxIN is the only type III TA characterized to date and its role as an abortive phage infection mechanism has been demonstrated (Fineran et al., 2009).

A type III locus contains an antitoxin composed of a tandem array of nucleotide direct repeats followed by a toxin-encoding gene. Palindromic repeat sequences that form a hairpin structure are found between the TA components and act as a transcriptional terminator that regulates the levels of antitoxin RNA versus toxin transcripts (Fineran et al., 2009). This genetic organization is also found in the AbiQ system. Specifically, a promoter region is followed by 2.8 repeats of 35 nucleotides, a rho-independent terminator, a consensus ribosome binding site (RBS) sequence and the abiQ gene. Comparative analyses showed that the ABIQ protein shares 31% sequence identity with ToxN. However, the sequences of the antitoxins antiQ and toxI are quite different as is the number of repeats since toxI contains 5.5 repeats of 36 nucleotides. The ToxN protein has been recently crystallized with its cognate antitoxin RNA ToxI, which revealed a heterohexameric triangular assembly of three ToxN proteins interspersed by three 36 nt ToxI RNA pseudoknots. It was also shown that ToxN is a ribonuclease (Blower et al., 2011). A recent survey based on structure similarity searches combined with protein comparison showed that type III TA systems are likely diverse and abundant in the microbial world (Blower et al., 2012).

The aim of this study was to characterize the AbiQ system through biochemical and structural analyses to demonstrate the TA nature of this phage resistance mechanism. To this end, we have solved its 3D structure and identify amino acids important for catalysis and RNA binding. Biochemical experiments demonstrated the endoribonuclease activity of ABIQ and its ability to cleave its cognate antitoxin RNA.

Results

AbiQ is a toxin–antitoxin system

In order to demonstrate that AbiQ is a toxin–antitoxin system, we first tried to separately clone the two components of the AbiQ system (antiQ repeats and abiQ gene) into two different vectors under control of the AbiQ native promoter in both L. lactis and Escherichia coli. No clone containing the abiQ gene alone was obtained, even when we tried to introduce the abiQ gene on a plasmid into a strain harbouring the putative antitoxin sequence. Second, the lactococcal nisin-inducible expression system NICE was used in an attempt to better control ABIQ protein expression (de Ruyter et al., 1996). The abiQ gene was cloned and placed under the control of the inducible nisA promoter in the expression vector pNZ8010 and successfully transformed into the laboratory strain L. lactis MG1363. A plasmid containing the intact abiQ gene was thus obtained and an attempt was made to transfer this plasmid into the host strain for nisin regulated gene expression L. lactis NZ9000. Again, plasmids obtained from chloramphenicol-resistant transformants contained major deletions or mutations in the abiQ sequence, suggesting that this gene is highly toxic even when expressed at a low level.

To address this difficulty, the vector pNZ8010-abiQ was co-transformed with the vector pTRKH2 (O'Sullivan and Klaenhammer, 1993) containing the antitoxin sequence (pTRKH2-antiQ) into L. lactis NZ9000. Chloramphenicol-resistant transformants without modification in abiQ were finally obtained and grown. Growth was then followed, post-nisin induction, by measuring the optical density at 600 nm for 4 h. The bacterial growth of the ABIQ-induced cells essentially stopped when nisin was added compared with the controls without abiQ that continue to grow (Fig. 1A). This confirmed that ABIQ was toxic to L. lactis. To determine if this effect was bacteriostatic or bactericidal, the induced cells were collected and transferred into fresh media without nisin and bacterial growth was measured by following the OD600 for 7 h. The recovery period of the abiQ-containing bacteria was longer than the controls without abiQ, but the bacteria began to grow again, indicating that the AbiQ system acts in a bacteriostatic manner (Fig. 1B).

Figure 1.

Effect of ABIQ overexpression on bacterial growth.

A. L. lactis NZ9000 with different combinations of vectors/recombinant plasmids was grown to an OD600 of 0.2 prior to induction with 5 ng ml−1 nisin. The graph represents the post-induction growth of the different cultures over 4 h.

B. Second growth curve of the induced culture. The 4 h induced bacteria were transferred into fresh medium without nisin. The growth was followed for 7 h.

Symbols: diamonds, NZ9000+pTRKH2+pNZ8010; circles, NZ9000+pTRKH2-antiQ+pNZ8010; triangles, NZ9000+pTRKH2-antiQ+pNZ8010-abiQ.

The abiQ toxin gene is transcribed and translated constitutively during phage infection

The laboratory phage-sensitive strain L. lactis IL1403 with or without the AbiQ system cloned in the pNZ123 vector (de Vos, 1987) was infected with the virulent phage P008 to study AbiQ expression. Samples were taken from non-infected cells and at various time intervals following phage addition. RNA or proteins were extracted as indicated in Experimental procedures. After electrophoresis in an agarose gel, a 32 nt oligonucleotide DNA probe complementary to a sequence in the middle of the abiQ gene was used to detect the specific RNA transcript in Northern blot experiments. In parallel, the protein extracts were migrated in a polyacrylamide gel and the ABIQ-6 His-tagged protein was detected using anti-His antibodies in Western blot experiments. Of note the presence of the His-tag did not interfere with AbiQ anti-phage activity as the EOP (10−5) of phage P008 was similar on strains carrying tagged and non-tagged ABIQ protein.

As shown in Fig. 2A, abiQ was constitutively transcribed in the non-infected cells and decreased over time during phage infection, starting 20 min post infection. A similar RNA profile was observed when a probe targeting the bacterial recA gene was used, suggesting that the decrease was a general phenomenon associated with the abortive system and cell death (Fig. 2A). The size of the unique transcript (∼ 700 nt) corresponded to transcription of the complete operon and was in accordance with the genetic analyses (702 nt). On the other hand, the ABIQ protein was present in similar quantity during the entire sampling period (Fig. 2A). This indicated that ABIQ was stable in the cell, which is a characteristic of toxin TA proteins (Yamaguchi et al., 2011). The expression of the antitoxin is presented below.

Figure 2.

Northern and Western blots to detect abiQ and antiQ RNA or ABIQ protein in L. lactis during a time-course infection with phage P008. Samples were taken from a non-infected (NI) culture and at 0, 10, 20 and 30 min. The genetic organization of the AbiQ construction used during the infection is represented under the corresponding gels. Black line below the AbiQ organization corresponds to the relative position of the probes used to detect abiQ or antiQ RNA. The diamonds represent the repeats (antiQ), the white arrows the abiQ gene, the black arrows the putative promoter and the hairpins the putative rho-independent terminators.

A. Northern blots using a probe targeting abiQ (abiQ transcription) or recA (recA transcription). Western blots using anti-His antibodies (Rockland) that recognized the ABIQ-tagged protein (translation).

B. Northern blots using a probe targeting antiQ. In the antiQ genetic organization at the right of the gels, the diamonds correspond to the repeat transcribed according to the transcript size.

ABIQ is an RNase that cleaves its antitoxin in vivo

The fate of the antitoxin RNA molecule was also analysed in a similar manner, except that an additional strain was used, namely a strain containing only antiQ. Total RNA from non-infected or P008-infected cells containing only the antitoxin or the complete AbiQ system were migrated in a polyacrylamide gel and transferred to nylon membranes. Northern blots were performed using one repeat of the antiQ sequence as a probe. As shown in Fig. 2B, the antiQ was constitutively produced in non-infected cells and in phage-infected cells. Moreover, the expression of the antiQ was constant over time in phage-infected cells, in sharp contrast with abiQ transcripts (Fig. 2), suggesting that the antiQ RNA molecules are either more produced or more stable than the abiQ mRNA. Of note, phage products did not cleave the long antitoxin molecule since one unique band (approximately 120 nt) corresponding to the entire repeat region is detected during the phage infection of cells containing only the antitoxin (Fig. 2B).

Interestingly, in cells containing the complete AbiQ system (antiQ and abiQ), the antitoxin was cleaved in vivo resulting in multiple bands on the gel (Fig. 2B). This indicates that ABIQ specifically cuts the antitoxin or activate an RNase activity in the bacteria, possibly within the repeats as it was also shown for ToxIN (Blower et al., 2011). The antitoxin was also cleaved at similar levels throughout the infection without significant differences in phage-infected cells compared with the non-infected control, confirming the stability of the antitoxin molecules and/or the ABIQ protein. Altogether, these data strongly suggest that ABIQ is an endoribonuclease or activates a bacterial RNase.

Crystal structure of ABIQ

The structure of ABIQ was solved by molecular replacement using the structure of ToxN as starting model (2XDD) (Blower et al., 2011) with which it shares 31% sequence identity (Fig. 3A). The polypeptide chain of ABIQ could be traced between residues Met1 and Lys168, with only a few residues not visible in the electron density maps, residues Lys54–Gly56 and Ser169–Gln172, which are probably disordered (Fig. 3B). The protein is composed of a central six-stranded anti-parallel β-sheet with connectivity 6-1-2-3-5-4. Six α-helices surround the β-sheet, with three of them covering one face of the sheet (α1, α5, α6) and two covering the opposite face (α3, α4). The last one, α2, sits on the side of the sheet (Fig. 3B). As mentioned above, the electron map density at position 51, at the end of strand 3 (Fig. 3B) clearly indicates the presence of a Leu residue, instead of a Ser for the native enzyme (Fig. S1).

Figure 3.

The ABIQ sequence and crystal structure.

A. ABIQ sequence aligned with the sequence of ToxN with Multalin (Corpet, 1988; Gouet et al., 2003). The conserved residues are with a red background. The ABIQ secondary structure is represented above its sequence. Important residues for ABIQ are identified by red marks and for ToxN by blue marks. Triangles identify the nucleophiles positions. Blue circles pinpoint the ToxN residues important for catalysis, and plus signs, those important for RNA binding. The mutated sites of ABIQ are also indicated: ‘=’ means no effect on the RNA cleavage, and ‘−’ means RNA cleavage decreased based on results presented in Fig. 6B.

B. Stereo view of the overall fold of ABIQ in ribbon representation and rainbow colouring (Nt→Ct: blue to red). The beta strands are numbered from β1 to β6 and the helices from α1 to α6. The segment not visible in the electron density map (Lys54–Gly56) is identified by a grey line.

Contrary to ABIQ, ToxN was solved in complex with its antitoxin RNA ToxI (Blower et al., 2011). They form a 3 + 3 hexamer with a ToxN at each edge of the triangle made by one repeat pseudoknot structure of the antitoxin (Fig. 4A). When superposing ABIQ to one of the ToxN molecule (Fig. 4B), we notice an overall excellent fit, since the r.m.s.d. deviation between Cα atoms is only 1.5 Å. Deviations occur mainly in loops in which residues of ABIQ are inserted or deleted compared with ToxN (Fig. 3A), namely loops around ABIQ residues Tyr27–Asn29, Glu59–Lys66, Asp106–Lys108 and Asp145–Tyr148 (Fig. 4B). Noteworthy, the loop Glu59–Lys66 is just after the active-site nucleophile (Ser51) and is in part disordered. It is possible that this loop needs its cognate antitoxin RNA to become ordered, as in the ToxIN complex.

Figure 4.

Comparison of ABIQ and ToxN.

A. Superimposition of ABIQ (ribbon with rainbow colouring) on ToxN monomer 1 (beige) belonging to the ToxN3–ToxI3 hexameric complex. The RNA is in light green sticks representation.

B. Close-up view of ABIQ–ToxN superposition, with the diverging loops identified (same colours as in A).

Because of the similarity between ABIQ and ToxN, we could use the ToxI–ToxN structure as a template to model the putative ABIQ/antiQ complex. The resulting in silico chimera ABIQ–ToxI displays very interesting features (Fig. 5). The Ser51 is close to the RNA cleavage site; its OH moiety is at 3.0 Å from the O2′ atom of the 2′–3′ cyclic phosphate. This group is formed between the backbone phosphate group from the RNA molecule and the 2′O of the ribose of adenine 32. In the ToxIN structure, it probably results from the nucleophilic serine on the 2′O of the ribose of adenine 32, which in turn performs the nucleophilic attack on the P=O group and leads to cyclization (Neubauer et al., 2009; Blower et al., 2011). A similar mechanism can be suggested for ABIQ. Also most of the RNA bases fit well inside the active-site crevice, i.e. bases uracil 22 to adenine 32 at the 5′ end and −3 to −1 at the 3′ end. In contrast, bases cytosine 0 to adenine 1 at the 3′ end clash with ABIQ, indicating that AbiQ antitoxin RNA should follow a different path of 4 residues after the cleavage site 3′ end. More remote favourable interactions are also established with bases adenine 6 – uracil 7 and uracils 16–17.

Figure 5.

The in silico generated chimeric ABIQ–ToxI complex.

A. ABIQ in ribbon representation and rainbow colouring and ToxI in light green sticks. Note the position of the cleaved bases (adenine −3, adenine 32), and the closeness of Ser51 (modelled on the Xray determined position of Leu51) to the cyclic phosphate of base adenine 32.

B. View of the surface of the above mentioned ABIQ–ToxI complex. ABIQ surface is coloured beige. Note that the RNA position fits well the ABIQ surface, except for bases cytosine 0 to adenine 1 (black arrow).

Mutagenesis of key amino acids to inactivate ABIQ activities

Based on the structure, different amino acids probably important for the activities of ABIQ were identified (Fig. 6A). Site-directed mutagenesis of these amino acids was accomplished and the activity of the mutated proteins was measured by evaluating the phage P008 EOP. RNA cleavage of the antitoxin by ABIQ was also determined in vivo using Northern blot analysis (Fig. 6B). For all the mutants, except Lys55Ala and Arg69Glu, the anti-phage activity of mutated ABIQ protein was completely abolished (EOP of 1.0) (Fig. 6A). Thus, the amino acids Tyr27, Pro49, Ser51, Ser52, Lys54, Lys60 and Arg67 are important for the phage resistance phenotype. Conversely, only the mutation Ser51Leu completely abolished the RNA cleavage of the antitoxin by ABIQ (Fig. 6B). This was expected since this serine is likely the nucleophilic residue. Interestingly, the amino acid substitution Ser51Thr still cleaved the antitoxin molecule (Fig. 6B). This threonine residue is the amino acid found in the ToxN protein active site (Blower et al., 2011). Partial cleavage of the antitoxin was observed for the mutations Pro49Ala, Ser52Ala and Arg67Glu when compared with the ABIQ wild-type profile since there was an accumulation of the larger transcripts and a decrease in the amount of smaller transcripts. For the other mutations, even if the anti-phage activity was abolished, the proteins retain the capacity to cleave their antitoxins.

Figure 6.

RNase assay to determine the AbiQ activity against its cognate antitoxin.

A. Site-directed mutations in the ABIQ protein and phage P008 EOP evaluation.

B. Northern blot to visualize the antitoxin (antiQ) cleavage profile achieved by the RNase activity of wild-type and mutant ABIQ. The probe consists of a single repeat of the antitoxin.

C. Effect of the phage P008 on the antitoxin cleavage by the ABIQ mutant Ser52Ala during a time-course infection.

To verify the effect of the antitoxin cleavage when an ABIQ-mutant strain was infected with phage P008, a time-course infection with a strain containing ABIQ wild type and a strain containing the Ser52Ala mutation was performed (Fig. 6C). Similar profiles were obtained with an accumulation of the larger transcripts in the mutant ABIQ compared with the wild-type control.

Discussion

Toxin–antitoxin systems are widespread in bacteria and have been demonstrated to play key roles in regulating growth during stresses such as phage infection. In the latter, this regulation is achieved by the death of the phage-infected TA-containing bacteria for the survival of the whole population. Here, we have shown through a series of experiments that the lactococcal AbiQ mechanism belongs to type III TA system and that ABIQ is an endoribonuclease. The operon is composed of 2.8 direct repeats followed by a gene coding for a toxic protein similar to ToxN, the sole type III TA characterized before this study. First, It was not possible to clone abiQ alone under its native or inducible promoter, indirectly demonstrating the toxic nature of the protein. Then, when overexpressed in bacteria containing its cognate antitoxin (antiQ), the ABIQ protein inhibited the bacterial growth in a bacteriostatic manner, a phenotype associated with TA systems.

The AbiQ system is encoded on a low-copy native plasmid pSRQ900 (10.8 kb) found in a raw milk L. lactis isolate (Emond et al., 1998). It is well documented that this bacterium contains several plasmids that provide additional activities such as lactose utilization, proteolysis, and of interest here, phage resistance (Duckworth et al., 1981; Klaenhammer, 1987; Boucher et al., 2001). In addition, it is tempting to speculate that AbiQ plays a role in plasmid maintenance to assure transfer of the plasmids to the daughter cells and maintain key functions for growth in the phage-containing milk environment. A similar phenomenon was observed in E. coli. The hok/sok system (type I TA) serves in plasmid R1 maintenance and in defence against phage T4 (Pecota and Wood, 1996). These functions were also suggested for the ToxIN system (Fineran et al., 2009).

To shed further light on its activities, we started by evaluating the transcription and translation of the abiQ gene. The transcription of abiQ is constitutive and the size of the transcript corresponds to the size predicted by bioinformatics analysis for the complete operon, including the repeats (antiQ). This result means that the terminator located between the antitoxin repeats and abiQ gene is leaky, and that no additional promoter is present since only one RNA band is detected. It also confirms that the toxin and antitoxin were co-transcribed as observed for the ToxIN system (Fineran et al., 2009).

During a time-course infection with virulent phage P008, the amount of abiQ transcripts decreased rapidly 20 min post infection. This finding could be attributed to the death of the bacteria, the RNase activity of ABIQ or the low stability of the mRNA molecule. The first hypothesis is the most plausible, since a similar pattern of transcription was also observed for the host recA gene. Interestingly, the antiQ RNA was produced at a similar level throughout the phage infection and was not cut by any phage product. This small RNA was probably very stable or constantly produced. This is rather surprising since antitoxin molecules are usually more labile than their cognate toxin in TA systems. The ABIQ protein was translated constitutively and its concentration was constant during the phage infection. This protein either was always produced or was stable in L. lactis. In any case, this is consistent with the constant presence of the antitoxin in the cell, as it is needed to protect it from the toxic effect of ABIQ. The ToxN-FLAG-tagged protein level during a time-course experiments with P. atrosepticum phages φA2 and φM1 was also constant, similarly to the ABIQ protein level reported here (Blower et al., 2009). Thus, stable protein expression even during a phage infection is probably a general phenomenon for type III TA systems. However, a similar time-course phage infection combined with Northern blot experiments to detect toxI- and toxN-specific transcripts as yet to be run with the ToxIN system but would be of interest to determine whether the above is a general observation for type III TA systems.

Because it only kills the phage-infected bacteria, ABIQ endoribonuclease must be somehow activated or made available by a phage component. In vivo experiments presented here suggest that by cleaving its antitoxin, ABIQ is sequestered and not available to attack its bacterial targets. However, when the system is stressed through a phage infection, the ratio of toxin–antitoxin may become unbalanced or the antitoxin may be somehow titrated out (Fineran et al., 2009). We can eliminate a few hypotheses with the data presented here. First, the antitoxin antiQ was not cleaved by a phage protein nor was abiQ. In addition, there is no activation of the abiQ transcription and ABIQ translation, indicating that the phage does not play a role in increasing the amount of ABIQ in the cells. However, if a phage protein or product binds to antiQ to sequester the antitoxin, free ABIQ could act on its cellular targets leading to cell death. On the other hand, ABIQ could bind directly to a phage product, which will change its activity and direct it towards its bacterial targets. The phage AbiQ's activator component is likely expressed late during the phage infection, which would explain why the phage DNA is replicated but not maturated in the capsid as reported previously (Emond et al., 1998).

The ABIQ 3D structure was determined without its antitoxin and was found to be very similar to the ToxN structure recently solved. Moreover, the ToxI–ToxN binding seems to be in a large part compatible to a putative antiQ–ABIQ binding. Despite the fact that the sequence of antiQ and toxI are non-homologous and the number of repetition in the antitoxin operon sequences are different (2.8 repeats for antiQ and 5.5 repeats for toxI) (Fineran et al., 2009), the toxI structure has revealed a pseudoknot corresponding approximately to one repetition of the complete toxI-coding sequence that could be also predicted for antiQ by bioinformatics analyses (supplementary Fig. S2, Blower et al., 2011, 2012). However, contrary to toxI, the antiQ pseudoknot is formed in the middle of a repetition and not at the boundaries of the repeats (supplementary Fig. S2, Reeder and Giegerich, 2004; Blower et al., 2012).

Based on the structure and on the complex model, we constructed site-directed mutations likely to impact RNase and anti-phage activities of AbiQ (Fig. 6). Surprisingly, only one mutation, Ser51Leu, completely abolished both functions. This mutated form of ABIQ could be overexpressed in E. coli without significant impact on the bacterial growth, demonstrating the importance of the RNase activity on the cell toxicity activity of AbiQ. This amino acid Ser51 is the best candidate to perform the nucleophilic attack on the cleavable RNA phosphate (for a review on nucleases mechanism see Yang, 2011). This is in contrast with ToxN, for which the residue Ser53, similar to ABIQ Ser52, has been proposed to perform the nucleophilic attack. Noteworthy, ToxN mutation Ser53Ala did not completely abolish the endonuclease activity, in sharp contrast with our results for ABIQ mutation Ser51Leu (Blower et al., 2011). Interestingly, when ABIQ Ser51 was replaced by a threonine, the residue found in a similar position in ToxN, the RNase activity of ABIQ was still functional but its anti-phage activity was abolished.

Of note, the Tyr27 residue appears to be important only for the anti-phage mechanism since, when mutated, the AbiQ anti-phage activity is abolished while its RNase activity is still functional, suggesting that both activities are not entirely related. Residues Pro49 and Ser52 mutations diminished both ABIQ activities since there was an accumulation of the large transcript and cells were phage-sensitive. Pro49 is two positions before the nucleophilic Ser51 residue and probably imposing the proper conformation of the catalytic machinery. It is also conserved in ToxN. Ser52 was also shown to be important for catalysis. Its role however is difficult to assign, since the OH moiety is turned opposite to the RNA. A probable explanation is that it might move upon RNA binding, as probably does the loop Lys54–Asn57, which is disordered in the absence of RNA antitoxin in our structure. In this context, Ser52 may very well play the role as one of the oxyanion stabilizer.

Partial RNA cleavage and no anti-phage activity have also been obtained for the Arg67Glu which are likely to be involved in RNA binding and, thus, have an important role in stabilization of the Michaelis complex. Lysines close to the active site could bring positive charges important for oxyanion stabilization. On the other hand, ABIQ mutations Lys54Ala, Lys55Ala and Lys60Ala showed no appreciable difference in antiQ RNA cleavage, thereby they probably do not play a role as members of oxyanion stabilizing residues, while this role was fulfilled by the homologous ToxN Lys55. Conversely, Lys54Ala and Lys60Ala eliminate the anti-phage activity and these residues play a key role in phage resistance. Therefore, while the Xray data made it possible to significantly rationalize the nucleophilic and antiQ RNA binding features of ABIQ, such a progress could not be reached concerning its anti-phage activity. Studies are underway to address this.

In conclusion, we have determined that the lactococcal AbiQ anti-phage mechanism is a type III TA system. This is the first TA described in L. lactis and the second type III TA characterized to date. Similarities as well as differences with the other studied type III TA system (ToxIN) were highlighted indicating, among others, that additional studies are needed on other type III TA systems to propose a general model. Elucidating the anti-phage activity should provide new insights into methods to control phage populations. From an application perspective, besides the advantage of phage resistance for industrial bacterial strains, AbiQ could serve as natural selection markers for the development of food-grade plasmids.

Experimental procedures

Bacterial strains and phages

The bacterial strains and phages used in this study are listed in Table 1. lactis strains were grown at 30°C in M17 broth (Oxoid) supplemented with 0.5% glucose (GM17). When necessary, chloramphenicol or erythromycin (5 μg ml−1) was added to the media for plasmid maintenance. For phage P008 propagation, L. lactis IL1403 was grown to an optical density at 600 nm (OD600) of 0.2 before the addition of 104 phages and 10 mM CaCl2. The culture was incubated until complete bacterial cell lysis had occurred and the resulting lysate was filtered using a 0.45 μm syringe filter. The efficiency of plaquing (EOP) was measured by dividing the titre of the phage on the AbiQ+ strain by the titre of the phage on the AbiQ strain. To obtain a high phage titre, one litre of phage lysate was separated on a discontinuous caesium chloride gradient (Sambrook and Russell, 2001). E. coli strains were grown at 37°C in Luria–Bertani (LB) or Brain Heart Infusion (BHI) broths. Chloramphenicol (34 μg ml−1), ampicillin (100 μg ml−1), erythromycin (150 μg ml−1) or kanamycin (25 μg ml−1) was added to the medium as needed.

Table 1. Bacterial strains, phage and plasmids used in this study
Bacterial strains, phage or plasmidsRelevant characteristicsReferences and sources
Escherichia coli  
MG1655F λ ilvG- rfb-50 rph-1Blattner et al. (1997)
NEB 10 BetaaraD139 Δ(ara-leu)7697 fhuA lacX74 galK (ϕ80 Δ(lacZ)M15) mcrA galU recA1 endA1 nupG rpsL (StrR) Δ(mrr-hsdRMS-mcrBC)New England Biolabs
RosettaF ompT hsdSB(rB mB) gal dcm (DE3) pLysSRARE2 (CamR)Novagen
XL1-BluerecA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F′ proAB lacIqZΔM15 Tn10 (Tetr)]Stratagene
Lactococcus lactis  
IL1403Laboratory strain, plasmid free, host of phage P008Chopin et al. (1984)
MG1363Laboratory strain, plasmid free, cloning hostWegmann et al. (2007)
NZ9000MG1363 derivative, inducible strain for the NICE systemKuipers et al. (1998)
Phage  
P008Phage infecting L. lactis IL1403, 936 group, sensitive to AbiQMahony et al. (2006)
Plasmids  
pDEST17Gateway IPTG-inducible expression vector of E. coli, 6.4 kb, AmpRInvitrogen
pDONR201Gateway intermediate vector, 4.5 kb, KmRInvitrogen
pNZ123Shuttle vector L. lactisE. coli, 2.5 kb, CmRde Vos (1987)
pNZ123-AbiQAbiQ complete system cloned in pNZ123 at EcoRI site, CmRThis study
pNZ123-AbiQ6HAbiQ complete system cloned in pNZ123 at EcoRI site, C-terminal 6His-tagged protein, CmRThis study
pNZ123-antiQAntitoxin region cloned in pNZ123 at EcoRI site, CmRThis study
pNZ8010NICE-inducible expression vector in L. lactis, CmRde Ruyter et al. (1996)
pSRQ928pNZ123 + 2.2 kb fragment containing AbiQ, CmREmond et al. (1998)
pTRKH2High-copy-number vector, EmR, 6.9 kbO'Sullivan and Klaenhammer (1993)
pTRKH2-antiQAntitoxin region cloned in pTRKH2 at EcoRV–BamHI site, EmRThis study

Plasmids, primers and DNA manipulations

The plasmids used in this study are described in Table 1 and the primers in supplementary material (Table S1). E. coli plasmids were prepared using Qiagen plasmid purification kits as indicated by the manufacturer. L. lactis plasmid DNA was obtained with Qiagen plasmid purification kits with the following modifications. To increase bacterial lysis, cells were treated by adding 30 mg ml−1 lysozyme directly to the P1 buffer. The samples were incubated at 37°C for 20 min before continuing with the normal protocol. Restriction endonuclease (Roche), Taq DNA polymerase (Invitrogen), Pwo DNA polymerase (Roche), Antarctic Phosphatase (New England Biolabs) and T4 DNA ligase (Invitrogen) enzymes were used according the manufacturer's instructions. Cloning procedures were carried out as described elsewhere (Sambrook and Russell, 2001) and with primers listed in Table S1 (supplementary material). L. lactis strains were electro-transformed using a Gene Pulser II apparatus (Holo and Nes, 1989). E. coli transformation was performed by thermal treatment. Clones were confirmed by sequencing of the inserts at the Plateforme de séquençage et de génotypage des génomes of the CHUL centre. The antibiotic selective pressure was maintained throughout the study to ensure maintenance of the plasmids in the various cells.

Bacterial cell toxicity assay

First, a PCR product of the region containing the antitoxin antiQ was cloned into the vector pTRKH2 at the EcoRV–BamHI restriction sites using the vector pSRQ928 (Emond et al., 1998) as a template. This construct was obtained in E. coli XL1-Blue and then introduced into L. lactis NZ9000. Second, a PCR product of the abiQ gene, using pRSQ928 as a template, was cloned into the nisin-inducible vector pNZ8010 at the BamHI–PstI restriction sites (de Ruyter et al., 1996). The resulting plasmid was transformed into L. lactis MG1363 to avoid toxic production of the protein and then into L. lactis NZ9000 or NZ9000+pTRKH2-antiQ for protein expression. Plasmids pTRKH2 and pNZ8010 have compatible replicons and different antibiotic selection markers to assure their maintenance. For the toxicity assay, 10 ml of inoculated GM17 media containing 1% of the different strains were incubated at 30°C to reach an OD600 of 0.2 prior to induction with 5 ng ml−1 nisin. The OD600 was followed for 4 h after induction. To verify whether AbiQ was bacteriostatic, the induced cultures were centrifuged and resuspended in 10 ml of fresh GM17 media with or without 5 ng ml−1 nisin. Bacterial growth was followed for 7 h at 30°C. This bacterial toxicity experiment was repeated twice.

Time-course phage infection

A PCR product containing the AbiQ system was cloned into the pNZ123 vector at the EcoRI restriction site using pSRQ928 as template. The vector pNZ123 is a small high-copy-number plasmid in L. lactis and can be selected using chloramphenicol. The construction pNZ123-AbiQ relied on the native promoter for the expression of the AbiQ-operon (Emond et al., 1998; Boucher et al., 2001). To enable protein detection by Western blot, primers were designed to include six histidines and thereby create a fusion protein. These plasmids were transformed into L. lactis IL1403 and we performed a time-course infection with the virulent lactococcal phage P008. Fifty millilitres of GM17 was inoculated (1%) with IL1403 + AbiQ and grown at 30°C to an OD600 of 0.5. The cells were centrifuged at 8000 r.p.m. for 5 min at room temperature in a Beckman GSA rotor and resuspended in 6 ml of fresh media. One ml of culture was removed (non-infected), while 10 mM CaCl2 and phages at a multiplicity of infection (moi) of 5 were added to the remaining 5 ml of culture. One millilitre of samples were withdrawn at different times after the beginning of the infection (0, 10, 20, 30 and 40 min), centrifuged 1 min at high speed and frozen at −80°C.

Northern experiments

Total RNA from the time-course phage infection (1 ml) was extracted from phage-free and phage-infected L. lactis cells using Trizol reagent (Invitrogen) with the addition of a lysozyme pre-treatment (60 mg ml−1 for 10 min at 37°C) to increase lysis of these Gram-positive cells. RNA samples were treated with DNase I (Roche) at 37°C for 20 min to eliminate residual DNA and protected with RNase inhibitor (Roche). RNA concentration was determined using a NanoDrop 2000. Five micrograms of RNA was loaded on a 1% formaldehyde-agarose denaturating gel or on a 10% polyacrylamide/8 M urea gel, subjected to electrophoresis and transferred to a nylon membrane (Sambrook and Russell, 2001). Northern blot experiments were performed using a 32P radiolabelled (Perkin-Elmer) DNA probe complementary to the abiQ gene (5′-GGGGTATTAATTCGCTGTCAGGAACTGGAATC-3′) or the antiQ antitoxin (5′-GCTCCAATTTTATCAATTCCAACTATGGCTTGGATA-3′) (Fortier et al., 2006).

Western experiments

To extract total proteins, 1 ml of the phage time-course infected bacteria were centrifuged and the pellets resuspended in 200 μl of lysis buffer [10 mM Tris-HCl pH 8.0, 0.3% SDS, 1 mM EDTA, 60 mM DTT and protease inhibitor (Roche)]. Cells were lysed by sonication using a Sonifier W-350 apparatus (6 × 45 s, output control 3, duty cycle 80%, hold). The suspension was centrifuged at 17 000 g for 30 min at 4°C, the supernatants were recovered and protein concentrations were determined using the Bradford standard method (Bio-Rad) with bovine serum albumin as standard. Samples (1 μg) were subjected to electrophoresis in 4–15% tris-glycine polyacrylamide pre-cast gels (Bio-Rad) and electrotransferred to polyvinyldene fluoride membranes (PALL Corporation). The membranes were blocked overnight at 4°C with 5% (w/v) non-fat dry milk in phosphate-buffered saline supplemented with 0.1% Tween 20 (PBS-T). Western blot was performed for 1 h using the antibody IgG anti-6His (Rockland) resuspended in the blocking buffer at a concentration of 1:1000 followed by four 10 min washes in PBS-T buffer. Then, membranes were treated for 1 h with the second antibody, anti-IgG-HRP, diluted in the blocking buffer at a concentration of 1:10 000. The membranes were washed four times for 10 min each in PBS-T with a final 10 min equilibration with PBS buffer. Detection was done by exposing the membranes to Kodak chemiluminescent films using the Amersham ECL Plus reagent as indicated by the manufacturer (GE Healthcare).

ABIQ protein purification

The abiQ gene was amplified by PCR using pSRQ928 as template and primers, which added sequences necessary for the Gateway technology (Invitrogen). This amplicon was cloned into the pDONR201 vector and transferred into the expression vector pDEST17 using the lambda recombinase Gateway enzymes. This construction added 6 histidines to the N-terminus to allow purification of the protein by affinity chromatography in an E. coli strong expression inducible system. Plasmids were subcloned into E. coli NEB 10-beta and transformed into the expression strain Rosetta 2 (DE3) pLysS. Potential clones were screened by sequencing the insert and one clone containing the mutation Ser51Leu was used for protein purification since a wild-type clone was never isolated despite numerous attempts. Two litres of culture were grown in Turbo broth (AthenaES) at 37°C to an OD600 of 0.6. Protein expression was induced by adding 0.5 mM IPTG, and the culture incubated overnight at 25°C. The culture was harvested by centrifugation at 4000 g for 10 min, resuspended in 50 ml of lysis buffer (50 mM Tris, 300 mM NaCl, 10 mM imidazole, pH 8.0) supplemented with 0.25 mg ml−1 lysozyme, 10 μg ml−1 DNase, 20 mM MgSO4 and EDTA-free antiproteases (Roche), and frozen at −80°C. After thawing on ice, the protein lysate was sonicated (output control 4, duty cycle 80%, 3 × 45 s), cleared at 12 000 r.p.m. for 30 min at 4°C, and filtered using a 0.45 μm syringe filter. The protein was purified by nickel affinity chromatography using a 5 ml His-Trap column on a fast protein liquid chromatography apparatus (AKTA, GE Healthcare) with a stepwise gradient of imidazole. The protein was then subjected to gel filtration using a Superdex 200 column in the following buffer: 10 mM Tris, 300 mM NaCl, pH 8.0. The concentration of the protein was determined using a NanoDrop 1000 with the absorbance at 280 corrected for the difference in absorption coefficient due to amino acid composition of the protein monomer with ProtParam tool (web.expasy.org/protparam/).

Crystallography

Optimal crystallography conditions were screened using the commercial kits MDL, Wizard 2, Stura and JSCG+. A volume of 100 nl of the kits was mixed to 100–300 nl of ABIQ at 4.6 mg ml−1 using a Cartesian nano-dispensing robot (Sulzenbacher et al., 2002). The protein crystallized by mixing 200 nl of protein at 4.6 mg ml−1 with 100 nl of 1 M Na/K tartrate as precipitant, and in 100 mM Mes, pH 6.0 buffer, at 20°C. Crystals suitable for X-ray diffraction were obtained after 2 months. A crystal was cryo-cooled with glycerol and a data set was collected at 2.16 Å resolution at the SOLEIL Proxima1 beamline. After processing the data sets using the xds program (Kabsch, 2010), the scaling was performed with XSCALE (Table 2). Crystals belong to the P212121 space group with cell dimensions a = 50.1 Å, b = 54.5 Å and c = 64.2 Å. The Vm calculated for one ABIQ molecule in the asymmetric unit was 2.08 Å3 Da−1, corresponding to a solvent volume of 39%. The structure was solved by molecular replacement with MOLREP (Vagin and Teplyakov, 2010) in CCP4 (Collaborative Computational Project, Number 4, 1994). Structure refinement was performed with AutoBUSTER (Blanc et al., 2004) alternated with model rebuilding using COOT (Emsley et al., 2010), leading to R/Rfree values of 19.6% and 22.3%, respectively, and all residues in the preferred or allowed regions of the Ramachandran plot (Table 2). Figures were made with Pymol (DeLano). The co-ordinates have been deposited at the PDB with the entry number 4GLK. The in silico chimera model of ABIQ–ToxI was obtained by performing a rigid-body fitting of the ABIQ structure onto the model of ToxN in the ToxNI structure (2XDD) (Blower et al., 2011). Since the structure of the antiQ RNA is still unknown, the ToxI RNA structure was kept in the in silico chimera model.

Table 2. Data collection and refinement statistics
Data collection/refinementSoleil Proxima 1a
  1. aParentheses refer to the highest resolution bin.
Data collection 
Wavelength (Å)0.9791
Resolution limits (Å)45.0–2.16 (2.22–2.16)
Rmerge (%)9.0 (61.0)
No. of observations86620 (6333)
No. unique reflections9856 (753)
Mean((I)/sd(I))17.1 (3.7)
Completeness (%)99.7 (99.6)
Multiplicity13.7/13.1
Refinement 
Resolution (Å)27.7–2.16 (2.42–2.16)
No. of reflections9880 (2734)
No. protein/water/glycerol atoms1351/70/18
No. test set reflections694 (173)
Rwork/Rfree (%)19.6/22.3 (23.3/31.8)
r.m.s.d. bonds (Å)/angles (°)0.09/1.18
B-wilson/B-average35.5/37.4
Ramachandran plot97.5/2.5
Preferred/allowed %

Site-directed mutagenesis and RNase assay

The plasmid pNZ123-AbiQ isolated from E. coli MG1655 was used as a template for the PCR reactions. Amplification of the plasmid was achieved using primers designed to create specific substitutions in AbiQ protein. The residual template plasmid was removed using DpnI (NEB) that cleaves methylated DNA. An aliquot of the digested PCR product was transformed into E. coli XL1-Blue and sequenced-confirmed plasmids were transferred into L. lactis IL1403. The plasmid constructs in L. lactis were sequenced-confirmed. To verify if the antitoxin was cleaved by ABIQ, 10 ml of bacterial culture was grown in GM17 media to an OD600 of 0.5. After centrifugation, bacterial pellets were frozen at −80°C. Extraction of the RNA, migration on polyacrylamide gel, and Northern assays were performed as described above using antiQ probe.

Acknowledgements

We would like to thank Barbara-Ann Conway for editorial assistance and Maxime Bélanger for technical assistance. This work was supported by the Natural Sciences and Engineering Research Council of Canada (Strategic programme) as well as the Ministère du Développement économique, de l'Innovation et de l'Exportation, Programme de soutien à la recherche: Programme de soutien à des initiatives internationales de recherche et d'innovation. J.E.S. is the recipient of a scholarship from the Fonds Québécois de Recherche sur la Nature et les Technologies. S.M. holds a Tier 1 Canada Research Chair in Bacteriophages.

Ancillary