In Pseudomonas aeruginosa three type VI secretion systems (T6SSs) coexist, called H1- to H3-T6SSs. Several T6SS components are proposed to be part of a macromolecular complex resembling the bacteriophage tail. The T6SS protein HsiE1 (TagJ) is unique to the H1-T6SS and absent from the H2- and H3-T6SSs. We demonstrate that HsiE1 interacts with a predicted N-terminal α-helix in HsiB1 (TssB) thus forming a novel subcomplex of the T6SS. HsiB1 is homologous to the Vibrio choleraeVipA component, which contributes to the formation of a bacteriophage tail sheath-like structure. We show that the interaction between HsiE1 and HsiB1 is specific and does not occur between HsiE1 and HsiB2. Proteins of the TssB family encoded in T6SS clusters lacking a gene encoding a TagJ-like component are often devoid of the predicted N-terminal helical region, which suggests co-evolution. We observe that a synthetic peptide corresponding to the N-terminal 20 amino acids of HsiB1 interacts with purified HsiE1 protein. This interaction is a common feature to other bacterial T6SSs that display a TagJ homologue as shown here with Serratia marcescens. We further show that hsiE1 is a non-essential gene for the T6SS and suggest that HsiE1 may modulate incorporation of HsiB1 into the T6SS.
Pathogenic bacteria use a broad arsenal of virulence factors, including protein secretion systems, which contribute to bacterial survival and successful establishment of acute as well as chronic infections (Cossart and Sansonetti, 2004; Merrell and Falkow, 2004). In Gram-negative bacteria six different types of protein secretion systems have been described (Saier, 2006; Durand et al., 2009; Bleves et al., 2010; Holland, 2010; Rego et al., 2010; Filloux, 2011). The type VI secretion system (T6SS) has recently been identified in many Gram-negative bacteria (Bingle et al., 2008; Shrivastava and Mande, 2008; Boyer et al., 2009) and intensely studied since. T6SSs have been shown to play a role in virulence and eukaryotic cell interactions (Ma et al., 2009; Ma and Mekalanos, 2010; Suarez et al., 2010), but have also been implicated in inter-bacterial interactions (Hood et al., 2010; MacIntyre et al., 2010; Murdoch et al., 2011; Russell et al., 2011; Zheng et al., 2011) or biofilm formation (Aschtgen et al., 2008). Pseudomonas aeruginosa possesses three T6SSs, namely the H1- to H3-T6SSs (Filloux et al., 2008; Barret et al., 2011), with the H1-T6SS involved in the secretion of three bacterial toxins, i.e. Tse1, Tse2 and Tse3 (Hood et al., 2010). The mode of action of the cytosolic toxin Tse2 is currently unknown although it was shown to be toxic when expressed in bacteria or eukaryotic cells (Hood et al., 2010). Recently, Tse2 was shown to induce quiescence rather than cell death in cells devoid of Tsi2 (Li et al., 2012). In contrast, the amidase Tse1 and the muramidase Tse3 are injected into the periplasm of target bacteria in a cell contact-dependent manner leading to cell death (Russell et al., 2011). These enzymes degrade the peptidoglycan of target bacteria leading to lysis of the recipient cells, hence providing P. aeruginosa with a growth advantage (Russell et al., 2011; Benz et al., 2012; Chou et al., 2012; Ding et al., 2012; Zhang et al., 2012). The T6SS could thus ensure P. aeruginosa survival in complex multi-microbial infections as they occur in the lungs of cystic fibrosis patients (Rogers et al., 2010; Sibley and Surette, 2011). The H1-T6SS of P. aeruginosa has also been linked to chronic infection (Potvin et al., 2003; Mougous et al., 2006).
The T6SSs are usually not expressed in laboratory conditions but induced in vivo, while the bacteria colonize the host (Das et al., 2000; Parsons and Heffron, 2005; de Bruin et al., 2007; Mattinen et al., 2007). In P. aeruginosa the H1-T6SS can be induced in vitro by introducing a mutation into the retS gene or by overexpressing the ladS gene (Goodman et al., 2004; Mougous et al., 2006; Ventre et al., 2006; Moscoso et al., 2011). RetS and LadS are two hybrid sensor kinases that belong to a sophisticated regulatory cascade together with the GacS/GacA two-component system, the post-translational repressor RsmA and the two small regulatory RNAs, RsmY and RsmZ (Goodman et al., 2004; 2009; Ventre et al., 2006; Lapouge et al., 2008; Brencic and Lory, 2009; Brencic et al., 2009; Bordi et al., 2010). RetS and LadS have antagonistic functions and further to their regulatory impact on T6SS expression also control the production of biofilm determinants such as the Pel and Psl polysaccharides (Friedman and Kolter, 2004; Vasseur et al., 2005). This observation supports the idea of the T6SS being associated with chronic infection, since in these infections the biofilm lifestyle is favoured (Costerton, 2001).
The number of genes, ranging from 15 to more than 20, and the genetic organization of T6SS gene clusters vary between different bacteria and between T6SS clusters within the same bacterium (Das and Chaudhuri, 2003; Bingle et al., 2008; Boyer et al., 2009; Barret et al., 2011). However, a set of conserved T6SS core components can be identified (Boyer et al., 2009). Among these, some components are clearly associated with the inner membrane (IcmF and DotU) while others are anchored in the outer membrane (Lip in P. aeruginosa). Two of the most studied T6SS components are the VgrG and Hcp proteins, which were presented initially as T6SS substrates (Mougous et al., 2006; Pukatzki et al., 2007). Yet, the discovery of a structural resemblance between VgrG proteins and the T4 bacteriophage tail spike components gp5/gp27, together with the structural similarity of Hcp multimers to the tail tube composed of gp19, led to the idea of a T6SS mechanism resembling bacteriophage DNA delivery (Leiman et al., 2009; 2010). It is now proposed that the T6SS forms an inverted puncturing device, perforating the bacterial cell envelope from the inside to the outside in order to secrete effectors (Pukatzki et al., 2009; Cascales and Cambillau, 2012). In the case of the P. aeruginosa H1-T6SS, the identified effectors are Tse1-3, with Tse1 and Tse3 being injected into the periplasm and Tse2 being transported to the cytosol of target bacterial cells (Russell et al., 2011; Li et al., 2012). Another of the T6SS core genes is clpV, which encodes a protein from the AAA+ ATPase family. In Vibrio cholerae ClpV specifically interacts with the VipA/VipB complex, called HsiB/HsiC (TssB/TssC) in P. aeruginosa, modulating the assembly of tubular complexes formed by these two components (Bonemann et al., 2009; 2010; Filloux, 2009; Pietrosiuk et al., 2011; Basler and Mekalanos, 2012; Basler et al., 2012). This VipA/VipB tubular structure resembles the bacteriophage tail sheath (Bonemann et al., 2010; Basler et al., 2012), a macromolecular structure that is composed of gp18 subunits (Aksyuk et al., 2009). The contractile sheath surrounds the bacteriophage tail tube and drives DNA injection upon contraction. In the T6SS, the VipA/VipB complex may thus surround the Hcp tube (Bonemann et al., 2010; Basler et al., 2012). Finally, the gp25 protein of the T4 page has been proposed to be similar to another T6SS component, HsiF (TssE) in P. aeruginosa (Lossi et al., 2011).
Despite these similarities to several T4 bacteriophage components, the role and function of most T6SS core components remain unknown and the mode of assembly and composition of the T6SS apparatus remain to be elucidated. A systematic analysis of T6SS components has been performed in Edwardsiella tarda demonstrating that 13 out of the 16 predicted T6SS components are required for secretion of EvpP (Zheng and Leung, 2007). More recently, it was shown in V. cholerae that 12 proteins constitute the T6SS apparatus (Zheng et al., 2011).
In this study, we characterize the HsiE1 component of the H1-T6SS, a protein found in 30% of all described T6SS and within P. aeruginosa unique to the H1-T6SS as it is not found in the P. aeruginosa H2- and H3-T6SSs. We identify a novel subcomplex formed by HsiE1 (TagJ) and HsiB1 (TssB), a homologue of VipA in V. cholerae. We identify a predicted α-helix at the N-terminus of HsiB1 that is essential for this interaction. Additionally, we show that SMA2270 (TagJ-like component) and SMA2258 (TssB-like component) of Serratia marcescens form a complex in vitro, strongly suggesting that this interaction is likely to be conserved in all T6SSs that contain a TagJ-like component. This study thus offers further insight into the assembly of the T6SS by highlighting specific but subtle interactions that are not found in all T6SSs but might be central to the development of an optimized T6SS mechanism.
The hsiE1 gene is part of the hsiA1–vgrG1a operon
The H1-T6SS gene cluster is organized in two putative transcriptional units, i.e. hsiA1–vgrG1a and fha1–tagQ (Fig. 1A). It has been hypothesized that several core T6SS components whose function is unknown form a supramolecular complex resembling the bacteriophage baseplate (Leiman et al., 2009; 2010). Among these putative baseplate components is HsiF1 (Lossi et al., 2011). Upstream of the hsiF1 gene is the hsiE1 gene whose homologue is not found in the H2- and H3-T6SS clusters (Filloux et al., 2008). To evaluate whether hsiE1 is part of a hsiA1–vgrG1a operon overlapping RT-PCR was performed. The data presented confirmed that genes from hsiA1 to vgrG1a are transcribed as one single polycistronic mRNA, whereas fha1 is part of a distinct and divergent transcription unit (Fig. 1B).
HsiE1 interacts with HsiB1
The hsiE1 gene is found in only one out of three P. aeruginosa T6SS gene clusters and is conserved in only 30% of the known T6SSs (Boyer et al., 2009). Despite not being a core component, we analysed whether HsiE1 could still be part of a larger complex together with other T6SS proteins. We systematically tested protein–protein interactions between HsiE1 and other Hsi proteins encoded within the same operon, using the bacterial two-hybrid (BTH) system (Fig. 2). In the BTH assay positive interactions are characterized by dark red colonies on MacConkey agar plates and interactions can be quantified by measurement of β-galactosidase activity. The leucine zipper protein (zip), which readily dimerizes, is used as a positive control (Karimova et al., 1998). The target Hsi proteins were cloned in the pUT18C and pKT25 plasmids as described in Experimental procedures (Table 1). The two-hybrid screen revealed that HsiE1 strongly interacts with HsiB1 yielding a β-galactosidase activity of 2012 ± 352 Miller units compared to 2262 ± 129 Miller units for the positive control (zip/zip) and 288 ± 22 Miller units for the negative control consisting of co-transformants containing the cloning vectors pUT18C and pKT25 only (T25/T18) (Fig. 2). HsiE1 did not interact strongly with any other Hsi component or Hcp1 in this assay (Fig. 2).
Fusion of hsiA1 to cya gene T25 fragment in pKT25, KmR
Fusion of hsiB1 to cya gene T25 fragment in pKT25, KmR
Fusion of hsiC1 to cya gene T25 fragment in pKT25, KmR
Fusion hsiE1 to cya gene T25 fragment in pKT25, KmR
Fusion hsiE1 to cya gene T18 fragment in pUT18C, ApR
Fusion of hsiF1 to cya gene T18 fragment in pUT18C, ApR
Fusion of hsiG1 to cya gene T18 fragment in pUT18C, ApR
Fusion of hsiH1 to cya gene T18 fragment in pUT18C, ApR
Fusion of hcp1 to cya gene T18 fragment in pUT18C, ApR
Fusion of hsiB2 to cya gene T25 fragment in pKT25, KmR
Fusion of gene fragment encoding HsiB1Δ1–15 to cya gene T25 fragment in pKT25, KmR
HsiB1 and HsiE1 form a stable complex
To further confirm the interaction between HsiE1 and HsiB1, we constructed the plasmid pACYC-E1B1 (Table 1), which expresses histidine-tagged HsiE1 (His–HsiE1, predicted molecular weight 30 kDa) and untagged HsiB1 (predicted molecular weight 18.8 kDa). The plasmid was transformed into Escherichia coli B834 (DE3) and gene expression induced by addition of isopropyl β-D-thiogalactoside (IPTG). His–HsiE1 was purified by immobilized nickel affinity chromatography and eluted fractions were analysed for the presence of both HsiB1 and HsiE1 by SDS-PAGE, Coomassie staining and immunoblotting (Fig. 3A and B). The eluted fractions contained two bands with a molecular weight of approximately 19 kDa and 30 kDa, as visualized by Coomassie staining (Fig. 3A, left panel). These two bands were identified as HsiB1 (Fig. 3B, left panel) and HsiE1 (Fig. 3B, right panel) by immunoblotting using polyclonal antibodies directed against HsiB1 and HsiE1 respectively. Furthermore, the identity of these bands was verified by peptide-mass fingerprinting using trypsin digestion and a MALDI mass spectrometer (PNAC, Cambridge, UK) (data not shown). As an additional control we also tested whether untagged HsiB1 expressed from pACYC-B1 could bind to the Ni+ column in the absence of His–HsiE1. As shown in Fig. S1 the majority of HsiB1 was recovered in the flow-through in these conditions and only minimal amounts were found in the eluted fractions, although undetectable by Coomassie staining. In conclusion, HsiE1 can be co-purified with HsiB1 demonstrating that HsiB1 and HsiE1 form a stable complex.
Since TagJ-like proteins are found in 30% of all known T6SSs, we asked if the TagJ/TssB interaction was a specific feature of the H1-T6SS of P. aeruginosa or more likely a general feature of T6SSs that encode a TagJ-like component. In order to address this question, we coexpressed S. marcescens T6SS components SMA2270 (TagJ-like protein) and SMA2258 (TssB-like component) from pSC156 in E. coli B834 cells. This plasmid expresses histidine-tagged SMA2270 and untagged SMA2258 from individual T7 promoters. Histidine-tagged SMA2270 was then purified using Ni+ affinity chromatography and eluted fractions were analysed for the presence of both SMA2270 and SMA2258 by SDS-PAGE and Coomassie staining (Fig. 3C). Two major bands were observed in the eluted fractions that were identified as SMA2270 and SMA2258 by peptide-mass fingerprinting using trypsin digestion and a MALDI mass spectrometer (PNAC, Cambridge, UK) (data not shown) demonstrating that SMA2270 and SMA2258 also form a stable complex in vitro. This finding thus suggests that the TagJ/TssB interaction is a general feature of T6SSs that encode a TagJ-like component.
HsiB1 but not HsiE1 is essential for H1-T6SS function
HsiB1 is predicted to be an essential T6SS component, while the importance of HsiE1 in T6SS function has not been described to date. Individual gene deletions in hsiB1 and hsiE1 were analysed in the PAKΔretS strain (Table 2). In PAKΔretS expression of the H1-T6SS gene cluster is constitutive (Goodman et al., 2004; Mougous et al., 2006) and several proteins are secreted in a T6SS-dependent manner, i.e. Hcp1, VgrG1a, VgrG1c and the toxins Tse1, Tse2 and Tse3 (Mougous et al., 2006; Hood et al., 2010; Hachani et al., 2011). Hence, we monitored the impact of hsiE1 or hsiB1 gene deletion on secretion levels of Hcp1, VgrG1a/c and Tse3 compared to those of the parental strain PAKΔretS. As a negative control secretion levels were compared to those of a hsiF1 mutant (PAKΔretSΔhsiF1), which has previously been shown to be affected in T6SS-mediated secretion (Lossi et al., 2011). RNA polymerase (RNAP) was monitored in both whole-cell lysates and culture supernatant to ensure that no cell lysis had occurred (Fig. 4A and B, bottom panel). Deletion of hsiE1 had no detectable effect on the T6SS-mediated secretion of VgrG1a/c (Fig. 4A, top panel), Hcp1 (Fig. 4A, second panel) or Tse3 (Fig. 4A, third panel) suggesting that hsiE1 is a non-essential gene for T6SS function. In contrast, deletion of hsiB1 clearly reduced secretion of VgrG1a and VgrG1c (Fig. 4A, top panel), Hcp1 (Fig. 4A, second panel), as well as the effector protein Tse3 (Fig. 4A, third panel) when compared to secretion levels obtained with the parental strain demonstrating that HsiB1 is an essential component for the H1-T6SS. The secretion level of VgrG1a/c, Hcp1 and Tse3 was restored to those of the parental strain when hsiB1 was reintroduced into the chromosome of the hsiB1 mutant using mini-CTX-B1 (Fig. 4B; see Experimental procedures).
As shown previously, P. aeruginosa is able to kill other Gram-negative bacteria, such as E. coli, in a H1-T6SS-dependent manner due to the activity of the Tse1-3 toxins (Hood et al., 2010; Russell et al., 2011). This is a very sensitive method to assess H1-T6SS function and we tested whether the hsiE1 and hsiB1 mutants were attenuated in this process (see Experimental procedures). Briefly, we used an E. coli strain harbouring a plasmid encoding the β-galactosidase enzyme (pCR2.1), which gave blue colonies on Luria–Bertani (LB) plates supplemented with X-gal. When a H1-T6SS-proficient P. aeruginosa strain (PAKΔretS) was incubated on solid media together with the E. coli strain, hardly any blue colonies were observed when a dilution series of the mixed bacterial culture was spotted on the appropriate medium (Fig. 4C). In contrast, when E. coli was incubated with a P. aeruginosa strain lacking an active H1-T6SS (PAK wild type) little to no E. coli killing was observed as seen by the significant number of detectable blue colonies (Fig. 4C). Importantly, the hsiE1 mutant (PAKΔretSΔhsiE1) was still able to kill E. coli as efficiently as PAKΔretS, whereas the hsiB1 mutant (PAKΔretSΔhsiB1) shows a clear attenuation (Fig. 4C) in line with the observation that the hsiE1 mutant is not affected in protein secretion, while the hsiB1 mutant is, as demonstrated above (Fig. 4A). The differences in detectable blue colonies were quantified (Fig. 4D) as described in Experimental procedures. We thus confirmed that the hsiE1 gene is not essential for H1-T6SS function whereas the hsiB1 gene is essential.
HsiE1 interacts with HsiB1 but not with HsiB2
The P. aeruginosa genome contains two additional T6SS gene clusters, namely H2-T6SS and H3-T6SS, which both encode a HsiB1 homologue, HsiB2 and HsiB3, respectively, but do not encode a HsiE1 homologue. Two-hybrid analysis revealed that even though HsiE1 strongly interacts with HsiB1, i.e. β-galactosidase activity of 2682 ± 53 Miller units (Fig. 5A) and dark red colony formation on MacConkey agar (Fig. 5B), no interaction was detected between HsiB2 and HsiE1 as demonstrated by a low level of β-galactosidase activity of 306 ± 16 Miller units and white colony formation on MacConkey agar plates (Fig. 5). This further suggests that in P. aeruginosa the interaction between HsiB1 and HsiE1 is specific.
Identification of a predicted N-terminal α-helix in HsiB1 absent in HsiB2 and HsiB3
HsiE1 belongs to the ImpE family (Bladergroen et al., 2003) of T6SS components and was also recently named type VI-associated gene J (TagJ) (Hood et al., 2010). As discussed above, HsiE1 did not interact with HsiB2 (PA1657) encoded by the H2-T6SS gene cluster (Fig. 5), suggesting that a specific feature of HsiB1 is required for the interaction between HsiB1 and HsiE1.
Secondary structure predictions using the amino acid sequence of HsiB1 (PA0083), HsiB2 (PA1657) and HsiB3 (PA2365) were carried out using the secondary structure prediction tool ‘PSIPRED’ (McGuffin et al., 2000) (Fig. 6A). This analysis revealed the presence of a putative α-helical region at the N-terminus of HsiB1 (amino acids 1–20), which was not predicted for HsiB2 or HsiB3 (Fig. 6A). Moreover, analysis of the secondary structure predictions of several members of the TssB family (HsiB homologues) in other well-studied T6SSs revealed a bias towards the prediction of an α-helical region in the TssB homologue especially when associated with a TagJ homologue; i.e. both corresponding genes are found within the same T6SS cluster (summarized in Table 3). For example, VipA (V. cholerae) and IglA (Francisella tularensis) do not show a predicted helical region at their N-terminus and both clusters do not encode a TagJ-like protein (Table 3) (Bonemann et al., 2009; Broms et al., 2009). On the other hand, ImpB (Rhizobium leguminosarum), STM0273 (Salmonella Typhimurium) and SMA2258 (S. marcescens) all contain a predicted N-terminal helical region and are encoded in T6SS clusters encoding a TagJ homologue, namely ImpE (pRL120471), SciE (STM0270) and SMA2270 (summarized in Table 3) (Folkesson et al., 2002; Bladergroen et al., 2003).
Table 3. List of HsiB/TssB homologues and HsiE/TagJ homologues in various known T6SSs
N-terminal helical region (HsiB1 homologue)
PA0083 (HsiB1 = TssB1)
PA1657 (HsiB2 = TssB2)
PA2365 (HsiB3 = TssB3)
To investigate the evolutionary relationship between the presence of a TagJ-like protein and the presence of a predicted N-terminal helical region in TssB homologues, phylogenetic analysis was carried out and is summarized in Table 3 and Fig. 6B. The phylogenetic tree shows that T6SS gene clusters lacking (shown in green) or containing (shown in red) a TagJ homologue mostly cluster at opposite ends of the phylogenetic tree supporting an evolutionary link between the predicted α-helix in the TssB homologue and the presence of a TagJ-like protein (Fig. 6B). However, there are several exceptions as demonstrated by the presence of a predicted N-terminal α-helical region in the protein y3675 of Yersinia pestis, for which the respective cluster does not contain any TagJ-like protein. Possible explanations are that those systems might have contained a TagJ-like protein previously, which has been lost during evolution, or that a TagJ-like protein is encoded elsewhere on the genome or that secondary structure predictions by the ‘PSIPRED’ algorithm were incorrect in some cases. Two peculiar cases are Rhizobium and Agrobacterium. Both have a TagJ homologue and a TssB homologue containing a predicted N-terminal helix, but are located on isolated branches of the phylogenetic tree.
Conserved N-terminal region of HsiB1 is essential for the interaction with HsiE1
Overall, the general trend observed in the phylogenetic tree described previously (Fig. 6B) supports the idea that the predicted N-terminal α-helical region of TssB-like proteins is required for the interaction with TagJ-like proteins. To investigate whether this region is indeed involved in this interaction, the gene fragment encoding HsiB1Δ1–15 (HsiB1 lacking the 15 N-terminal residues) was amplified from genomic DNA of P. aeruginosa and cloned into pKT25 and pUT18C. The resulting constructs were then used in a two-hybrid assay to analyse the impact of the lack of the first 15 amino acids (aa) of HsiB1 on its interaction with HsiE1. Co-transformants of HsiB1 and HsiE1 formed dark red colonies on MacConkey agar plates (β-galactosidase activity 1948 ± 64 Miller units) similar to the positive zip/zip control (β-galactosidase activity 1998 ± 98 Miller units) (Fig. 7A and B). The loss of red colony formation on MacConkey agar plates for HsiB1Δ1–15/HsiE1 co-transformants (β-galactosidase activity 294 ± 25 Miller units) indicates a loss of interaction in the absence of the first 15 N-terminal amino acids of HsiB1. No interaction was observed no matter if HsiB1Δ1–15 was expressed as a T18 or T25 fusion protein and paired with T25-HsiE1 or T18-HsiE1 respectively (data not shown). The negative control, T25/T18, formed white colonies (β-galactosidase activity 201 ± 19 Miller units) similar to HsiB1Δ1–15/HsiE1 co-transformants. We further checked that the chimera between T18 and HsiB1, HsiB1Δ1–15 or HsiE1 was produced using an antibody directed against the T18 subunit. HsiE1 was produced at higher levels than HsiB1, whereas HsiB1Δ1–15 was produced in similar amounts as full-length HsiB1 (Fig. 7C). As expected T18-HsiB1Δ1–15 migration was slightly faster than T18-HsiB1 due to the lack of the first 15 amino acids of HsiB1 (Fig. 7C).
The HsiB1 homologue VipA (V. cholerae) has been shown to form a complex with the HsiC1 homologue VipB (TssC family) (Bonemann et al., 2009). This interaction has been described in several other bacteria encoding T6SSs, such as F. tularensis (IglA/IglB) (Broms et al., 2009) or Burkholderia cenocepacia (BscK/BscL) (Aubert et al., 2010), and was also shown for HsiB2/HsiC2 and HsiB3/HsiC3 encoded by the H2- and H3-T6SSs of P. aeruginosa respectively (Broms et al., 2009). Here, we confirmed that HsiB1 interacts with HsiC1 (dark red colonies on MacConkey agar, β-galactosidase activity 2257 ± 43 Miller units) (Fig. 7A and B). The production of the T18-HsiC1 chimera was tested as described above using the T18 antibody (Fig. 7C). Importantly, HsiB1Δ1–15 was still able to interact with HsiC1 to the same extent as the full-length HsiB1 (B1/C1: β-galactosidase activity 2257 ± 43 Miller units, B1Δ1–15/C1 2438 ± 184 Miller units) indicating that truncation of HsiB1 did not affect folding, which could have impaired interaction between HsiE1 and HsiB1Δ1–15 (Fig. 7A and B). In summary, these observations confirm our hypothesis that the N-terminal region of HsiB1 is essential for the HsiB1–HsiE1 interaction.
Solution NMR studies show that HsiE1 interacts with a peptide corresponding to the first 20 amino acids of HsiB1
Bacterial two-hybrid analysis showed that the predicted N-terminal α-helical region of HsiB1 is essential for the interaction between HsiB1 and HsiE1. To confirm this interaction recombinant GST–HsiE1 was expressed from pGEX-E1 (Table 1) in E. coli B834 (DE3) and purified by affinity chromatography. After removal of the GST tag by thrombin cleavage, the purified recombinant HsiE1 was analysed by 1d 1H NMR spectroscopy (Fig. 8A). The spectrum of HsiE1 reveals a well-folded monomeric domain, characterized by a number of ring-current shifted methyl groups and dispersed amide region. To confirm the direct interaction between HsiE1 and the predicted N-terminal α-helical region of HsiB1, increasing amounts of a peptide comprising residues 2–20 of HsiB1 (GSTTSSQKFIARNRAPRVQ) were added into the sample of HsiE1, and the variation in the spectrum was monitored using 1d 1H NMR spectroscopy (Fig. 8A). Clearly, incremental addition of up to ∼ 8 molar equivalents of peptide caused a number of shift changes in the spectrum of HsiE1 indicative of binding (Fig. 8A, shift indicated by arrows). Fitting the change in chemical shift of perturbed signals with increasing peptide concentration to the standard equation for a saturation isotherm gave an approximate Kd of 90 μM (Fig. 8B). To confirm the specificity, we repeated the titration of HsiE1 using both the N-terminal HsiB1 peptide and a scrambled version (QGVSRTPTIASRSNQRKAF), and performed 2d 1H-15N NMR spectra to resolve signal overlap in the 1d spectrum (Fig. 8C). For the HsiB1 peptide, a number of peak shifts arise from peptide interaction, consistent with a localized binding site on HsiE1 (Fig. 8C). In contrast, the scrambled peptide does not show any sign of interaction even at 10-fold excess (Fig. 8C). The apparently low affinity of binding to HsiB1 (90 μM) may be an artefact of using an isolated 20 residues-long peptide and may thus not reflect accurately the in vivo affinity for full-length HsiB1. However, the above confirm that HsiE1 binds specifically the first 20 amino acids of HsiB1.
The T6SS is a molecular machine involved in protein secretion in Gram-negative bacteria but yet little is known about its assembly. However, the structure of individual T6SS components has been elucidated such as the outer membrane lipoprotein (Lip/TssJ) (Felisberto-Rodrigues et al., 2011; Rao et al., 2011) or the inner membrane protein DotU/TssL (Durand et al., 2012). The system is composed of core components, i.e. 13 proteins that are conserved in both pathogenic and non-pathogenic bacteria (Boyer et al., 2009). Some genes are not conserved in all T6SS clusters, such as hsiE1/tagJ from P. aeruginosa, which is only found in about 30% of the T6SS clusters identified to date (Boyer et al., 2009). Interestingly, hsiE1 is found in only one out of the three P. aeruginosa T6SS clusters (Filloux et al., 2008), namely the H1-T6SS, indicating that occurrence of hsiE1/tagJ in T6SSs is not species-specific but rather system-specific. In this study, we showed that the hsiE1 gene is part of an operon ranging from PA0082 to PA0091 including hsiA1-C1, hcp1, hsiE1-H1, clpV1 and vgrG1a genes (Fig. 2B). While hcp1, clpV1, vgrG1a and their homologues have been studied intensely in P. aeruginosa and other Gram-negative bacteria, little information is available on the function of the Hsi proteins. HsiF1 belongs to the gp25-like protein family, which is named after the baseplate wedge protein gp25 of the T4 bacteriophage (Kostyuchenko et al., 2003; Leiman et al., 2010; Yap et al., 2010), and was proposed to have lysozyme activity (Szewczyk et al., 1986). We recently showed that HsiF1 is a cytosolic protein with no lysozyme activity and is likely to be a structural component of the H1-T6SS (Lossi et al., 2011). In V. cholerae, homologues of HsiB1 (VipA) and HsiC1 (VipB) have been shown to form a complex structurally resembling the T4 bacteriophage tail sheath (gp18) (Bonemann et al., 2009; Basler et al., 2012). During phage infection, the contraction of the tail sheath that surrounds the gp19 tail tube (Hcp-like) is central to the DNA injection mechanism (Takeda et al., 1990; 2004; Aksyuk et al., 2009). In the case of the phage sheath gp18 it is unclear how contraction is initiated and what the driving force to the process is (Kostyuchenko et al., 2005; Aksyuk et al., 2009). Some studies suggested that GTPase activity could be associated with the sheath protein (Serysheva et al., 1992) but no nucleotide-binding fold is identified in gp18 (Aksyuk et al., 2009). In the case of the V. cholerae T6SS, it was shown that in vitro the AAA+ ATPase ClpV severs the VipA/VipB sheath-like structure (Bonemann et al., 2009). More recently, time-lapse video studies using GFP-tagged VipA clearly showed that the tail sheath-like structures are highly dynamic, and that in a clpV mutant the tagged VipA is found in static structures (Basler et al., 2012).
The function of HsiE1 or of a TagJ-like protein has not been studied in any bacterial species and in this study we showed that deletion of the hsiE1 gene has no effect on secretion of Hcp1, VgrG1a, VgrG1c and the H1-T6SS effector Tse3 (Fig. 4A). These data thus suggest that HsiE1 is an accessory component of the T6SS, although the question needs to be addressed on whether HsiE1 may have a subtle although non-essential function.
One way of addressing the above question is to test whether HsiE1 efficiently interacts with any core component of the H1-T6SS. Two-hybrid analysis showed that HsiE1 forms a complex with HsiB1. Such interaction is not unique to P. aeruginosa and we showed that the S. marcescens homologues do also tightly interact together suggesting that this is a general feature in T6SSs that involve a TagJ-like protein such as HsiE1. The role of this complex is unclear, but one may speculate that by interacting with HsiB1, HsiE1 contributes to modulate the assembly of a putative HsiB1/HsiC1 sheath-like structure similar to the previously characterized VipA/VipB tubules (Bonemann et al., 2009; Basler et al., 2012). This may come in addition to the predicted role of ClpV1 in the dynamic of a HsiB1C1 complex. A NCBI BLAST analysis revealed that the N-terminal region of HsiE1 (aa 1–150) contains tetratricopeptide repeat (TPR) domains, which are known to play a role in protein–protein interactions and in the assembly of multi-protein complexes (Goebl and Yanagida, 1991; Das et al., 1998; Blatch and Lassle, 1999). In bacteria, examples of TPR-containing proteins associated with supramolecular complex assembly are MamA in the magnetosome of magnetotactic bacteria (Yamamoto et al., 2010) or PilW in Neisseria meningitidis type IV pili (Trindade et al., 2008). TPR-containing proteins have also been proposed to act as co-chaperones for Hsp90 or Hsp70 chaperones. It is interesting to note that ClpV is an AAA+ ATPase member of the Hsp100/Clp chaperone family that plays a central role in protein quality control network in bacteria (Schlieker et al., 2005). As previously indicated, since in V. cholerae ClpV has been shown to interact with the HsiB1C1-like complex VipA/VipB (Bonemann et al., 2009; Basler et al., 2012), it is a possibility that in P. aeruginosa HsiE1 acts as a co-chaperone for ClpV1 and targets the ATPase to the complex via an interaction with HsiB1. In other words HsiE1 could be an adaptor protein between the HsiB1C1 complex and ClpV1. In this respect, it was recently suggested that the V. cholerae ClpV is not involved in the dynamic polymerization/contraction of the tail sheath-like structure, but rather involved in the recycling of VipA and VipB proteins from the contracted tail sheath. Importantly, this process was considered as a non-essential process for T6SS function in V. cholerae (Zheng et al., 2011; Basler et al., 2012).
Whatever the role of HsiE1 in T6SS assembly is, we demonstrated that the interaction between HsiE1 and HsiB1 seems to be specific as no interaction between HsiE1 and HsiB2, the HsiB-like component of the H2-T6SS of P. aeruginosa, was observed. We identified a predicted α-helical region at the N-terminus of HsiB1, which was not found in the secondary structure of HsiB2 and HsiB3. We assessed the role of this region in the interaction with HsiE1 and showed both genetically, using the BTH assay, and biochemically, using NMR and a purified peptide corresponding to the first 19 amino acids of HsiB1, that this region is essential for the HsiB1E1 interaction. TPR domains such as those predicted for HsiE1 typically contain 34 amino acids with the number of repeats in one protein varying from 1 to 19. It is generally proposed that several tandem repeats can generate a right-handed helical structure with an amphipathic channel that is thought to accommodate an alpha-helix of a target protein (D'Andrea and Regan, 2003) supporting the hypothesis that the predicted N-terminal α-helix of HsiB1 is adequate for the interaction with HsiE1. The specific interaction between HsiB1 and HsiE1 is highlighted by phylogenetic analysis that shows that HsiB-like proteins (TssB family) display a predicted N-terminal helix mostly when they are associated with a HsiE-like protein (TagJ family).
In summary, we have shown that the T6SS component HsiE1 is not essential for a functional H1-T6SS, while HsiB1 is. HsiE1 interacts with the core-component HsiB1 that is likely to have evolved to provide a specific site of interaction at its N-terminus. This interaction thus appears not crucial for the H1-T6SS function but might fine-tune the assembly of the system. The role of HsiE1 in the T6SS remains unknown, but it might be involved in fine-tuning and optimization of the T6SS assembly, for example, by stimulating the interaction between ClpV1 and the HsiB1C1 complex. Alternatively, it might be involved in HsiB1 targeting to, or HsiB1 de-association from, the HsiB1C1 complex, providing an additional control on the dynamic of the contractile sheath-like structure (Basler et al., 2012). Further experiments will aim at investigating these hypotheses and reconstituting the T6SS supramolecular assembly.
Bacterial strains and growth conditions
Bacterial strains used in this study are described in Table 2. P. aeruginosa strains were grown in tryptone soy broth supplemented with antibiotics where appropriate (tetracyclin 15 μg ml−1). E. coli strains were grown in LB broth supplemented with antibiotics where appropriate (streptomycin 50 μg ml−1, ampicillin 50–100 μg ml−1, kanamycin 50 μg ml−1, chloramphenicol 30 μg ml−1, tetracyclin 15 μg ml−1).
All plasmids and oligonucleotides used in this study are listed in Tables 1, S1 and S2 respectively. All constructs were confirmed by sequencing (GATC, Germany).
The plasmid mini-CTX-B1 was constructed as follows. The region encoding the lac promoter was amplified from pCR2.1 (Invitrogen) using primers 1267/1268 adding restriction sites for subcloning into mini-CTX-1. The hsiB1 gene was amplified from genomic DNA using primers 1269/1270 and ligated into mini-CTX-1 downstream of the newly engineered lac promoter and resulting in the complementing plasmid mini-CTX-B1. The plasmid was conjugated and integrated at the att site into the chromosome of PAKΔretSΔhsiB1.
The plasmids pGEX-B1 and pGEX-E1 were constructed as follows. The hsiB1 and hsiE1 genes were amplified by PCR from genomic DNA of P. aeruginosa PAO1 using primer pairs 379/382 and 170/574 respectively. PCR products were cloned into pGEX-6p1 resulting in the recombinant plasmid pGEX-E1, expressing HsiE1 with an N-terminal GST tag, and pGEX-B1, expressing HsiB1 with an N-terminal GST tag respectively.
For coexpression of HsiB1 and HsiE1, hsiB1 and hsiE1 were amplified from genomic DNA of P. aeruginosa PAO1 using primer pairs 381/382 and 170/269. The PCR products were cloned into pACYC-duet vector using BamHI/HindIII to insert hsiE1 into the first multiple cloning site and NdeI/EcoRV to insert hsiB1 into the second multiple cloning site yielding pACYC-E1B1. The cloned genes are under the control of two separate T7 promoters.
For coexpression of SMA2258 and SMA2270 of S. marcescens, SMA2258 and SMA2270 were amplified from genomic DNA of S. marcescens using primer pairs SC2424/SC2425 and SC2422/SC2423, respectively, adding appropriate restriction sites. The PCR products were cloned into pACYC-duet vector using BamHI/HindIII to insert SMA2270 into the first multiple cloning site and NdeI/XhoI to insert SMA2258 into the second multiple cloning site yielding pSC156. The cloned genes are under the control of two separate T7 promoters.
Deletion mutants in P. aeruginosa were constructed using the suicide vector pKNG101. DNA regions upstream and downstream of the gene of interest (GOF) were amplified from P. aeruginosa PAK genomic DNA using oligonucleotide pairs listed in Table S1. The resulting overlapping PCR products, which were then merged by PCR, were cloned into pCR2.1 and subcloned into pKNG101 using restriction sites ApaI and BamHI yielding the suicide vector pKNG-Δgene-of-interest (GOF).
All plasmids for BTH analysis were constructed as follows. The GOF was amplified from P. aeruginosa PAO1 genomic DNA adding restriction sites XbaI and KpnI. The resulting PCR product was ligated into BTH plasmids pKT25 and pUT18C leading to plasmids expressing in-frame fusions of the protein of interest to the T25 or T18 subunit of adenylate cyclase respectively.
Antibodies and reagents
Polyclonal anti-VgrG1a and peptide antibody against Hcp1 have been described previously (Hachani et al., 2011) and were used at a dilution of 1:1000. Monoclonal RNAP antibody (W0023, Neoclone), directed against the beta subunit of RNA polymerase, monoclonal anti-β-lactamase antibody (#15720, QED Bioscience), polyclonal anti-HsiE1 and anti-HsiB1 antibodies (produced by Eurogentec) were used at a dilution of 1:1000. Polyclonal anti-peptide antibody against Tse3 was described previously and has been used at a dilution of 1:250 after adsorbing with P. aeruginosa PAK whole-cell lysate (Hachani et al., 2011). The monoclonal antibody used to detect the T18 subunit of adenylate cyclase (anti-CyaAm sc-13582) was purchased from Santa Cruz Biotechnology and used at a dilution of 1:5000. Primary antibodies were incubated with nitrocellulose membranes (Whatman) for 1–2 h followed by 45 min incubation at room temperature (RT) with secondary antibody as appropriate [goat anti-rabbit HRP (A6154, Sigma) or rabbit anti-mouse HRP (A 9044, Sigma)] at a dilution of 1:5000. Western blots were developed using SuperSignal West Pico Chemiluminescent Substrate (Pierce) and Las3000 Fuji Imager.
Construction of deletion mutants in P. aeruginosa
Pseudomonas aeruginosa deletion mutants were constructed as described previously (Vasseur et al., 2005) using the suicide plasmid pKNG101 (Herrero et al., 1990; Kaniga et al., 1991). To create PAKΔretSΔgene-of-interest (GOF), the plasmid pKNG-ΔGOF was constructed as described above, maintained in the E. coli strain CC118 λpir and mobilized in P. aeruginosa PAKΔretS using E. coli 1047 carrying the conjugative plasmid pRK2013 (Figurski and Helinski, 1979). Clones, in which double recombination events occurred, resulting in the deletion of GOF, were selected on sucrose plates as previously described (Kaniga et al., 1991; Vasseur et al., 2005). Deletion of GOF was verified by PCR.
Production of polyclonal antibodies
Expression vectors pGEX-B1 or pGEX-E1 were transformed into E. coli B834 (DE3) for expression and purification of fusion proteins. Protein production was induced for 16 h at 18°C in the presence of 0.5 mM IPTG. Cell pellets were collected by centrifugation and resuspended in lysis buffer (50 mM HEPES, 200 mM NaCl, pH 7.2, supplemented with 5–10 mM DTT, 2 mM EDTA and complete® protease inhibitor cocktail tablets) before cells were lysed by passage through a French press. Supernatants were cleared by ultracentrifugation at 18 000 g for 45 min at 4°C and applied to Glutathione Sepharose™ High Performance resin packed into a XK 16/20 column (GE Healthcare) at a rate of 0.1–0.2 ml min−1. After washing the column with 40 column volumes of wash buffer (50 mM HEPES, 200 mM NaCl, pH 7.2, supplemented with 5–10 mM DTT), the GST tag was cleaved using PreScission protease [35 μl (ml resin)−1] overnight at 4°C. Eluted proteins were dialysed against 10 mM Tris (pH 8.0) and injected into rabbits for antibody production (Eurogentec).
Co-purification of HsiB1 and HsiE1 or SMA2258 and SMA2270
Expression vectors pACYC-E1B1 and pACYC-SMA2270/2258 were transformed into E. coli B834 (DE3) for production and purification of recombinant proteins. Cells were grown at 37°C to an OD600 of 0.6 and expression was subsequently induced using 1 mM IPTG (Sigma) for 16 h at 20°C. Cells were pelleted at 4000 g for 15 min at 4°C and lysed by passage through a French press in lysis buffer (50 mM HEPES, 200 mM NaCl, 5 mM DTT, 2 mM EDTA, supplemented with Complete® protease inhibitor cocktail). The cell extract was clarified at 40 000 g for 45 min at 4°C and supernatants were applied to Ni+–NTA columns (Hi-Trap, GE Healthcare) for purification of His–HsiE1 in complex with HsiB1 (no tag) by Ni+–NTA affinity chromatography using imidazole gradient. Eluted fractions were analysed by SDS-PAGE, Coomassie staining and immunoblotting for the presence of HsiB1 and HsiE1. In case of SMA2270 and SMA2258 eluted fractions were analysed by SDS-PAGE and Coomassie staining only. These results were also confirmed by peptide mass fingerprinting using trypsin digestion and a MALDI mass spectrometer (PNAC, Cambridge, UK).
1d 1H and 2d 1H-15N spectra of recombinant HsiE1 (purified as described in the section polyclonal antibodies) were acquired at 298 K on Bruker AvanceII 800 and AvanceIII 600 spectrometers equipped with a TXI/TCI cryoprobes respectively. Samples of HsiE1 were prepared in appropriate buffer (50 mM HEPES, 200 mM NaCl, pH 7.0). For titration, 10 mM peptide (GSTTSSQKFIARNRAPRVQ, Eurogentec) was dissolved in identical buffer and added incrementally, recording a 1d spectrum after each addition. Molar ratios of peptide were: 0, 0.2, 0.5, 0.8, 1.2, 2.2, 4.2 and 8.2 equivalents. The dissociation constant was fitted from the change in frequency of three resolved signals with increasing peptide concentration to the standard equation for a 1:1 saturation isotherm. Fitting was performed in Excel 2007 (Microsoft Corporation) using the Solver module. For comparison, 2d 1H-15N HSQC spectra were also recorded using the above peptide and a scrambled peptide (QGVSRTPTIASRSNQRKAF, Eurogentec) at approximately 10-fold excess.
Preparation of supernatant from P. aeruginosa culture
Pseudomonas aeruginosa strains were grown in tryptone soy broth overnight, subcultured to an OD600 of 0.1 and grown to early stationary phase at 37°C under agitation. Cells were separated from culture supernatants by centrifugation at 4000 g at 4°C. Cells were directly resuspended in 1 × Laemmli buffer (Laemmli, 1970). Concentrated P. aeruginosa culture supernatant (10 ×) was prepared as follows. Proteins of the culture supernatant were precipitated using 6 M trichloroacetic acid (Sigma) at a final trichloroacetic acid concentration of 10%. Protein pellets were washed in 90% acetone, dried and suspended in 1 × Laemmli buffer for analysis by SDS-PAGE. Samples were boiled at 95°C for 10 min before SDS-PAGE.
Bacterial competition assay
Overnight cultures of indicated P. aeruginosa strains were incubated with overnight cultures of equivalent bacterial numbers of E. coli containing the plasmid pCR2.1 (carrying the lacZ gene) on LB agar for 5 h at 37°C. Subsequently, patches of bacteria were recovered, resuspended in LB broth and dilution series ranging from 100 to 10−7 were plated in triplicate on LB supplemented with 100 μg ml−1 5-bromo-4-chloro-indolyl-β-D-galactopyranoside (X-gal, Invitrogen) allowing for colorimetric detection of lacZ-positive E. coli. For semi-quantitative analysis of the amount of surviving E. coli, the β-galactosidase activity of each patch of the 10−2 dilution was measured using ortho-nitrophenyl-β-galactoside (ONPG, Sigma) as a substrate as described previously (Miller, 1992). Arbitrary units representing survival of E. coli were calculated using a modified formula used to calculate Miller units of β-galactosidase activity (Miller, 1992) and assuming equal β-galactosidase activity in any E. coli cell carrying the pCR2.1–lacZ vector.
Bacterial two-hybrid assay
Protein–protein interactions were analysed using the BTH system as described previously (Karimova et al., 1998). In brief, DNA fragments encoding the protein of interest were amplified by PCR, adding appropriate restriction sites, using P. aeruginosa PAK genomic DNA. DNA fragments encoding the proteins or protein domains of interest were cloned into plasmids pKT25 and pUT18C, which each encode for complementary fragments of the adenylate cyclase enzyme, as previously described (Karimova et al., 1998), resulting in constructs expressing N-terminal fusion of the protein of interest with the T25 or T18 subunit of adenylate cyclase. Recombinant pKT25 and pUT18C plasmids were transformed simultaneously into the E. coli DHM1 strain, which is devoid of adenylate cyclase, and transformants were spotted onto MacConkey agar plates (Difco) supplemented with 1% maltose, in presence of 100 mg ml−1 ampicillin, 50 mg ml−1 kanamycin and 1 mM IPTG. Positive interactions were identified as red colonies on MacConkey agar plates after 48 h incubation at 30°C followed by 48 h incubation at RT. The positive controls used in the study were pUT18C or pKT25 derivatives encoding the leucine zipper from GCN4, which readily dimerizes.
For quantitative analysis of BTH interactions, β-galactosidase activity of co-transformants scraped from MacConkey plates was measured as described previously and activity was calculated in Miller units (Miller, 1992).
PAKΔretS cultures were grown to late exponential phase before RNA was extracted from equivalent bacterial numbers using RNeasy Mini Kit (Qiagen) according to the manufacturer's manual. Subsequently, RNA samples were treated with TurboDNase (Life Technologies) and 200 ng of RNA were used to prepare cDNA using SuperScript™ II Reverse Transcriptase (Invitrogen) and random hexamer primers (Amersham Biosciences) as previously described (Burr et al., 2006). The resulting cDNA was used as a template for non-quantitative RT-PCR. Non-quantitative RT-PCR was performed using Taq polymerase (Invitrogen) and primers (Eurogentec) spanning genes of interest as shown in Table S1. As a positive control genes were also amplified from P. aeruginosa genomic DNA (PureLink™ Genomic DNA Mini Kit, Life Technologies) using identical primer sets.
We thank Abderrahman Hachani for help with the bacterial competition assay. Alain Filloux is supported by the Royal Society, Wellcome Trust Grant WT091939 and MRC Grant G0800171/ID86344. Nadine Lossi and Rana Dajani are supported by the MRC Grant G0800171/ID86344. Eleni Manoli is supported by the Wellcome Trust Grant WT091939.