Identification and Characterization of an Antibacterial Type VI Secretion System in the Carbapenem-Resistant Strain Klebsiella pneumoniae HS11286

The type VI secretion system (T6SS) is a class of sophisticated cell contact-dependent apparatus with anti-eukaryotic or anti-bacterial function. Klebsiella pneumoniae is one of the most common bacterial pathogens with resistance to the carbapenem antibiotics. However, little is known about the antibacterial T6SS in K. pneumoniae. Using core-component protein searches, we identified a putative T6SS gene cluster on the chromosome of the carbapenemase-producing K. pneumoniae (CRKP) strain HS11286. Intraspecies and interspecies competition assays revealed an antibacterial function of the HS11286 T6SS. The phospholipase Tle1KP was found to be an effector protein that is transferred by T6SS. The overexpression of this effector gene in the periplasm caused severe growth inhibition of Escherichia coli. A sub-inhibitory concentration of β-lactam antibiotics stimulated the expression and secretion of the HS11286 T6SS and enhanced T6SS-dependent killing. It suggested that the antibiotics might be an impact factor for the T6SS secretion and antibacterial activity.


INTRODUCTION
The type VI secretion system (T6SS) is structurally related to the cell-puncturing device of tailed bacteriophages and functions as a contractile injection machinery that perforates eukaryotic and prokaryotic target membranes (Pukatzki et al., 2006;Kapitein and Mogk, 2013;Alteri and Mobley, 2016). T6SS gene cluster contains the core component genes and variable regions that encode effector and immunity proteins (Zhang et al., 2012; see the comprehensive review of Russell et al., 2014). The VipA and VipB proteins form a dynamic tubular sheath that switches between extended and contracted states within the bacterial cytosol (Basler and Mekalanos, 2012). Upon contraction of the cytoplasmic VipA/VipB tube, the hemolysin-coregulated protein (Hcp) capped with valine-glycine repeat protein G (VgrG) were ejected from the bacterium (Bönemann et al., 2009;Leiman et al., 2009;Basler and Mekalanos, 2012). ClpV (a class of ATPases) is required to disassemble the contracted VipA/VipB sheath (Bönemann et al., 2009;Basler and Mekalanos, 2012). Two core proteins, Hcp and VgrG, have frequently been used as markers to investigate the T6SS translocation of secreted effectors (Pukatzki et al., 2006Li et al., 2016).
The bacterial species Klebsiella pneumoniae is an increasingly important human pathogen. Dramatic increases in the levels of multidrug resistance associated with this species pose an emerging global problem (Broberg et al., 2014), particularly for carbapenemase-producing K. pneumoniae (CRKP; Tzouvelekis et al., 2012;Cubero et al., 2015). The in silico analysis showed that the putative T6SS gene clusters were present in the complete genome sequence of K. pneumoniae NTUH-K2044, MG78578, and Kp342 (Sarris et al., 2011). Furthermore, a T6SS in the hypervirulent K. pneumoniae strain Kp52.145 with the K2 capsular serotype transfers the effector phospholipase D1 (Tle5, PLD1), which has been reported as a novel virulence factor (Lery et al., 2014). However, little is known about the antibacterial function of T6SS in K. pneumoniae (Hood et al., 2017). A recent finding suggested that the rice pathogenic bacterium Acidovorax avenae strain RS-1 exposure to the β-lactam antibiotics enhanced the virulence of T6SS (Li et al., 2016). Thus, a better understanding of the role of T6SS in CRKP under antibiotic stress is warranted.
We have recently reported the complete genome of K. pneumoniae HS11286, an ST11, carbapenemase (KPC)-2-producing clinical isolate collected in 2011 from the sputum specimen of an inpatient in Shanghai, China (Liu et al., 2012). K. pneumoniae ST11 is a dominant KPC-producing clone in China and is closely related to the worldwide-dominant CRKP clone ST258. In this study, we annotated an entire T6SS gene cluster of K. pneumoniae HS11286. The anti-bacterial function of this T6SS was subsequently investigated with bacterial competition assays and overexpression of the effector gene tle1 KP in E. coli.

Ethics and Consent
In this study, sampling collection of patient sputum is a routine hospital procedure. As such, verbal informed consent was obtained from the volunteer. The bacterial sample and data sheet were anonymized. This study protocol, including the verbally informed consent procedure, was approved by the ethics committee of the School of Life Sciences & Biotechnology, Shanghai Jiao Tong University, China.

Bacterial Strains and Growth Conditions
The strains, plasmids, and primers used in this study are listed in Supplementary Tables 1-3, respectively. K. pneumoniae strain HS11286 was used as the sample strain in all experiments in this study unless otherwise noted in the figure legend. E. coli strain DH10B (streptomycin resistant) and E. coli strain BL21 were used for cloning and expressing the target gene, respectively. Unless stated otherwise, bacteria were grown in Luria-Bertani (LB) broth at 37 • C with shaking motion (220 rpm). The antibiotic concentrations used were 100 µg/mL ampicillin, 25 µg/mL chloramphenicol, 100 µg/mL streptomycin, 50 µg/mL kanamycin, 200 µg/mL hygromycin, and 50 µg/mL apramycin. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added at a final concentration of 0.5 mM to induce the expression of the T7 promoter.

Identification of T6SS Loci in K. pneumoniae Strain HS11286
The nucleotide sequence and annotation of the completely sequenced K. pneumoniae HS11286 chromosome were downloaded from the NCBI RefSeq Project (accession number NC_016845.1). The putative T6SS gene clusters were identified and aligned by using the web-based tool VRprofile (http:// bioinfo-mml.sjtu.edu.cn/VRprofile/; Li et al., 2017). Briefly, it performs the Hidden Markov Model (HMM)-based detection of T6SS core components encoded by K. pneumoniae chromosome sequences. The annotated T6SS core components are searched against the VRprofile-specific HMM profiles. Significant hits within a defined gene distance are subsequently grouped to detect putative T6SS gene clusters. The minimum number of the colocalized T6SS core components is five to exclude false positives. Besides, VRprofile also employs BLAST searches to examine the rearranged architecture of a given T6SS gene cluster among multiple K. pneumoniae genomes under investigation. It aids identification of putative effectors and immunity protein genes inserted within or around the T6SS core components.

Bacterial Competition Assays
Bacterial strains were grown overnight to stationary phase and suspended in LB medium. We washed strains with 10 mM MgSO 4 and diluted to an optical density at OD 600 of 0.2 before mixing at a ratio of 1:1 (attacker: prey). Cells of each strain were spotted on a sterile 0.22 µm filter (Millipore) on LB solid medium (casein tryptone 10 g/L; yeast extract 5 g/L; NaCl 5 g/L; agar 3%). Competitions were incubated for 5 h at 37 • C. Each group of bacteria was harvested, and serial dilutions were grown in selective culture medium containing 200 µg/mL hygromycin. Cells of each strain were grown overnight. The CFU per milliliter of the surviving prey strain was measured by counting single clones. The prey strain was the tle1 KP tli1 KP ::hph of K. pneumoniae HS11286 carrying the hygromycin-resistance gene (Supplementary Table 1). The fatality rate was calculated as follows: where CFU competition is the difference between the CFU of the surviving prey strain ( tle1 KP tli1 KP ::hph) that was coincubated with attacker strain vipA and the CFU of the same prey strain when co-incubated with wild type attacker strain. The CFU control is the CFU of the prey strain ( tle1 KP tli1 KP ::hph) without growth competition by any attacker strain.

Mutagenesis
For gene mutation, the gene was first replaced by an FRT siteflanking hph cassette (Supplementary Figure 1) via lambda red recombination, resulting in an intermediate strain. Then, the hph cassette was eliminated via Flp-FRT recombination to obtain a markerless in-frame indel mutant strain (Supplementary Figure  2). In-frame deletion of a gene was performed as described previously (Hoang et al., 1998;Chaveroche et al., 2000;Bi et al., 2015).

Western Blot Analysis
An Hcp antibody was produced by B&M Biotech (Beijing, China). Before immune injection into rabbits, we checked whether the rabbit serum gave an immune response to the whole protein extract of HS11286 at the target protein length to avoid a false-positive response. Rabbits were grown in specific-pathogenfree (SPF) conditions to ensure a non-specific immune response. Upon receiving the antibody, we first tested its specificity by blotting against Hcp protein. The specificity of antibody was shown in Supplementary Figure 3. For western blot, proteins were resolved on a precast 15% SDS/PAGE gel and transferred to a PVDF membrane (Millipore) by electrophoresis. The membrane was then blocked with 5% skimmed milk for 1 h at room temperature and incubated with primary antibodies (MBL Biotech) at 4 • C overnight. The membrane was washed three times with TBST buffer (50 mM Tris, 150 mM NaCl, 0.05% Tween 20, pH: 7.6) and incubated with an HRP-conjugated secondary antibody (Pierce) for 1 h at room temperature. Signals were detected using DAB (diaminobenzidine, Dako Denmark A/S) solution.

E. coli Toxicity Assays
The wild type and point mutants of tle1 KP were sub-cloned into the pET-22b vector containing N-terminal PelB signal peptide sequences. The plasmid pET28a was used to express tle1 KP in the cytoplasm. A single colony harboring the expression plasmid was grown in LB medium at 37 • C. After overnight culture to the stationary phase and suspended in LB medium, we diluted strains to an optical density at OD 600 of 0.2. Then the cells were serially diluted in 10-fold steps and grown in select culture medium with 0.05 mM IPTG and 50 µg/mL ampicillin (or 200 µg/mL kanamycin). The plates were prepared for imaging and plate counts after an additional 20 h incubation at 37 • C. The CFU per milliliter of the surviving prey strain was measured by counting single clones (Hu et al., 2014).

Statistical Analysis
Data are shown as mean ± SD. The t-test and one-way ANOVA were used to compare continuous variables. A P < 0.05 was considered statistically significant. Data were analyzed with the R package.

RESULTS
Identification of the T6SS Gene Cluster in K. pneumoniae Strain HS11286 We detected an entire 23-gene T6SS cluster (KPHS_22970-23190) on the chromosome of K. pneumoniae strain HS11286 (Supplementary Table 4). This T6SS consisted of 12 core components (Figure 1). Interestingly, we found a 4.7 kb insertion region that contained KPHS_23060-23110, which is located between the core component genes icmF and vgrG. There is a sequencing mistake in the insertion region. According to Sanger sequencing of the PCR amplification, a homopolymer error within KPHS_23100 was made by 454 sequencing. This error resulted in a frame-shift mutation. After correcting the sequencing mistake, the region from KPHS_23100 to KPHS_23110 was combined into a new protein-coding gene that we named KPHS_23105 (Supplementary Figure 4). This gene coded for a putative effector of T6SS that we called Tle1 KP . This effector contains a conserved alpha/beta hydrolase domain (DUF2235) like the reported T6SS effector Tle1 in P. aeruginosa (Hu et al., 2014). A tandem array of four genes (KPHS_23060-23090) were predicted to code for immunity proteins Tli1 KP that were related to Tle1 KP . These genes exhibited 94% nucleotide sequence identities to each other. In addition, VRprofile typed 254 putative T6SS gene clusters in the 107 completely sequenced K. pneumoniae genomes (including HS11286), of which, 42 of the gene clusters code for homologous effectors of Tle1 KP (numbers 1-42 in Supplementary Figure 5).

T6SS Secretion Activity Is Induced by Meropenem and Ceftazidime
To characterize the predicted T6SS, we first determined the expression of the T6SS core component and effectors in the logarithmic and stationary phases of K. pneumoniae grown in LB medium. The marker genes that we used were gapA, vipA, clpV, hcp, and tle1 KP , which encode the housekeeping protein GapA, a VipA sheath, an ATPase, a hallmark effector and the Tle1 KP effector, respectively. The semi-quantitative PCR assays showed that there was no detectable band of the four marker genes (besides the housekeeping gapA) in either the logarithmic or the stationary phase (Figure 2A), suggesting that T6SS was inactive when cultured in LB broth. We also observed that T6SS was inactive in the M9 medium culture conditions (Supplementary Figure 7). Previously, it was reported that T6SS could be induced under certain conditions,  Table 4), the effector gene in red, and the immunity protein genes in blue. such as temperature, pH, and the presence of chitin or with antibiotics (Cerith et al., 2013;Borgeaud et al., 2015). Accordingly, we tested whether the expression of T6SS in strain HS11286 could be induced with a sub-inhibitory concentration of β-lactam antibiotics (meropenem and ceftazidime). As shown in Figure 2B, vipA, clpV, hcp, and tle1 KP exhibited strong bright bands in the presence of a sub-inhibitory concentration of meropenem (4 mg/L) or ceftazidime (32 mg/L) compared to control conditions without antibiotics. The determination of sub-inhibitory concentrations of the antibiotics is shown in Supplementary Figure 8. To further explore whether the T6SS effectors could be secreted in K. pneumoniae under sub-inhibitory antibiotic stress, we produced a vipA mutant that cannot secrete T6SS effectors due to the lack of the VipA sheath element. Then, we examined Hcp, one of the reported T6SS effectors, in both the cellular and supernatant fractions by immunoblotting. As shown in Figure 2C, Hcp was absent in both the cellular and supernatant fractions of the wild type and vipA mutant when cultured with LB broth. In contrast, Hcp was easily detected in the cellular fraction of both the wild type and vipA mutant under β-lactam antibiotic stimulation, indicating that the T6SS of HS11286 could be produced upon exposure to a sub-inhibitory concentration of β-lactam antibiotics. We also observed that Hcp was present only in the supernatant of the wild type strain but not in vipA mutant (Figure 2C), suggesting that the secretory function of T6SS occurred under antibiotic stimulation. Also, we identified the specificity of antibiotics by using apramycin (Supplementary Figure 9) and got the similar results. It was to say that not only β-lactam antibiotic can induce the secretion of T6SS.

Tle1 KP and Tli1 KP Are Organized into an Effector-Immunity Protein Pair
Bioinformatics analysis predicted that Tle1 KP in HS11286 is an effector of T6SS. To test this, we examined Tle1 KP expression in the supernatant of both the wild type and the T6SS-apparatusdeletion mutant ( vipA) treated with ceftazidime. The results showed that Tle1 KP was present in both the cytosol and supernatant fractions of the wild type strain with antibiotic treatment but was absent in the supernatant of the vipA mutant strain, indicating that Tle1 KP is a T6SS effector (Figure 3A).
The toxicity of Tle1 KP was predicted to be neutralized by the upstream cognate immune protein Tli1 KP . T6SS-dependent killing (dueling) was performed to confirm this prediction, which uses co-incubation of attacker and prey bacterial strains to show whether a predicted immune protein will protect the prey from killing. The attacker and prey strains are co-cultured on agar, and the survival strains are quantified. The double deletion mutant lacking the effector and immune protein pair (strain tle1 KP tli1 KP -V) was used as the prey that could be  Table 1). All strains carried the vector control plasmid pBAD33-Apra. Each group has three replications. *P < 0.05. susceptible to T6SS. The wild type and tle1 KP -V mutant were the attackers. As shown in Figure 3B, the WT-V (Supplementary Table 1) caused a statistically significant impairment/killing of the tle1 KP tli1 KP -V mutant, but the tle1 KP -V mutant did not. Additionally, when the prey ( tle1 KP tli1 KP -V) was complemented with tli1 KP , the inhibition by the attacker WT-V was same to the attacker tle1 KP -V. These results show that the Tli1 KP protein counteracts the toxicity of effector Tle1 KP . Hence, Tle1 KP and Tli1 KP represent a cognate effector and immunity protein pair.

HS11286 T6SS-Dependent Killing within Bacterial Populations
We performed the competition assays to investigate the antibacterial activity of the K. pneumoniae HS11286 T6SS. The wild type, the T6SS-apparatus-deletion mutant ( vipA), and the transferred effector deletion mutant ( tle1 KP ) were employed as the attacker strains. The effector and immune protein gene double deletion mutant strain ( tle1 KP tli1 KP ) served as the prey strain. As shown in Figure 4A, the killing activity of the deletion mutant lacking the T6SS apparatus or the effector was lower than that of the wild type (The complementary of Tle1 KP in Supplementary Figure 11). The survival of the prey was greater with the vipA mutant than with tle1 KP , which indicated the T6SS might contain another toxic effector. Also, the fatality rate of the antibiotics-free group (the control) is nearly two-fold less than that of the meropenem or ceftazidime group (Figure 4B). It indicated that the T6SS killing activated by cell-to-cell contact might become stronger under antibiotics stress.
We also confirmed the T6SS-mediated interspecies competition between K. pneumoniae and E. coli. We utilized E. coli DH10B as the prey strain. The HS11286 wild type, the vipA mutant, and the tle1 KP mutant were employed as the attacker strains. The attacker strains were individually mixed with the prey and grown on LB agar, and the prey survival was quantified after co-incubation. As expected, the wild type exhibited the strongest killing activity compared to the vipA mutant and the tle1 KP mutant (Supplementary Figure 12). tle1 KP tli1 KP on LB medium and then viability on selective medium was determined. tle1 KP tli1 KP was the prey. (B) T6SS-dependent killing activity under antibiotic stress. The attacker strains (WT and vipA) were individually mixed with the prey strain ( tle1 KP tli1 KP ) on different antibiotic media and then titered for viable counts on selective media. The fatality rate was calculated based on viable cell count of the prey with or without competence. CFU competition /CFU control . The sub-inhibitory concentrations of meropenem (4 mg/L) and ceftazidime (32 mg/L). Control was added (ddH 2 O). Each group in (A,B) has three replications. **P < 0.01; ***P < 0.001.

Overexpression of Tle1 KP in the Periplasm Was Lethal in E. coli
The Tle1 family effector member from P. aeruginosa (Tle1 PA ) can hydrolyze cell membranes (Hu et al., 2014). To confirm The growth of E. coli BL21 containing tle1 KP pET22b and tle1 KP pET28a on LB agar induced by IPTG. BL21 with an empty vector was included as a control. pET22b contains a periplasm-targeting signal. Ten-fold serial dilutions of overnight culture were grown on LA media plates with selective culture medium. (B) Amino acid sequence alignment for Tle1 KP and the Tle1 PA of P. aeruginosa (Hu et al., 2014). Red triangles highlight the characterized active sites of phospholipase Tle1 PA . (C) The 3D structure and active site of Tle1 KP predicted by Phrye2 based on the reported structure of Tle1 PA (PDB entry 405p). (D) Identification of the active sites of Tle1 KP . According to the active sites shown in (C), E. coli BL21 harboring pET22b expressing wild type and its point mutants in the periplasm were grown in agar plates. The cells were prepared with serial 10-fold dilutions on LA medium and induced by IPTG. Each group in (A,D) has three replications. The CFU per milliliter of the surviving prey strain was measured by counting single clones. ***P < 0.001. the membrane-targeting activity of Tle1 KP , this effector was produced in E. coli strain BL21 and designed to target to the periplasm (Hu et al., 2014). As expected, the presence of Tle1 KP in the periplasm inhibited the growth of E. coli with significant difference (Figure 5A, P = 0.0004), suggesting that Tle1 KP is an active membrane-targeting phospholipase effector in K. pneumoniae. Meanwhile, no decrease in viability was observed after overexpressing tle1 KP in the cytoplasm (Figure 5A), suggesting that Tle1 KP is not toxic in the cytoplasm.
In addition, the structure and active domain of Tle1 PA have been reported recently (Hu et al., 2014). The protein sequence alignment between Tle1 KP and Tle1 PA exhibited 43% identities. According to homology modeling, Tle1 KP also contains a catalytic triad (S248-D297-H375), like the one that has been characterized in Tle1 PA (Figures 5B,C). Thus, we constructed three tle1 KP mutant strains (S248A, D297A, and H375A) by sitedirected mutagenesis for a toxicity assay, as described above. Only the D297A mutant showed an impaired antibacterial effect ( Figure 5D, P = 0.0003), suggesting that D297 is indispensable in catalytic activity. In Tle1 PA , site-directed mutations that alter each of these three amino acids lost activity, which implies that Tle1 KP and Tle1 PA have different active centers that may have different functions.

DISCUSSION
In this study, we identified and characterized a new T6SS with the antibacterial effector Tle1 KP in K. pneumoniae HS11286. This T6SS is silent in basic medium without an induction signal. It follows that expression and assembly of this structure would be tightly regulated. In some cases, there is evidence for transcriptional regulation of the T6SS via quorum sensing (Gueguen et al., 2013;Salomon et al., 2013), biofilm formation (Aubert et al., 2008;Hood et al., 2010), iron limitation (Brunet et al., 2011;Chakraborty et al., 2011), and temperature variation (Salomon et al., 2013), which may react to stress responses (Gueguen et al., 2013). In P. aeruginosa, kanamycin stimulated T6SS expression but did not affect the cognate effector secretion (Cerith et al., 2013). But in K. pneumoniae HS11286, a subinhibitory concentration of the β-lactam antibiotics could induce both T6SS expression and secretion of the effectors Hcp and Tle1 KP .
Furthermore, antibiotics also enhanced the antibacterial activity of the HS11286 T6SS. We propose that antibiotics induce the expression and secretion of T6SS and might make the attacker T6SS + strains more aggressive in the competition for growth. A previous report (Li et al., 2016) has shown a similar phenomenon in Acidovorax avenae subsp. avenae (Aaa) strain RS-1. Exposure of RS-1 to ampicillin alters the virulence, colonization capacity, composition of extracellular polymeric substances and secretion of the T6SS effector Hcp. Thus, for CRKP HS11286 in the clinical practice, antibiotics may not inhibit its proliferation but instead may induce the activity of T6SS, making HS11286 more aggressive (Supplementary Figure 13). Under antibiotic stress, CRKP HS1186 thus dominates the growth superiority comparing to the T6SS − /multidrug resistant strain.
There are five known Tle family T6SS effectors (Tle1-5; Lu et al., 2014), and the newly identified effector encoded within the T6SS gene cluster of K. pneumoniae HS11286 belongs to Tle1. Tle1-Tle4 families contain a conserved G-X-S-X-G motif, and Tle5 features a conserved H-X-K-X-X-X-X-D motif (Durand et al., 2014;Russell et al., 2014). Members of Tle1, Tle2, and Tle5 had been experimentally confirmed to possess phospholipase A2, A1, and D activities, respectively (Durand et al., 2014). After being injected into their periplasmic space by the T6SS, Tle1 of Burkholderia thailandensis and Tle2 of V. cholerae can hydrolyze the membrane phospholipids of neighboring cells, causing an increase in cellular permeability (Russell et al., 2013;Hu et al., 2014).According to homologous model, Tle1 KP of K. pneumoniae HS11286 contains a Tle1 family conserved motif G-X-S-X-G and exhibits periplasmic activity. But the active motif of Tle1 KP may not be the same as Tle1 PA , as shown in our point mutations results. And the amino acid sequence identities between reported Tle1 PA and Tle1 KP were just 43%, also indicating that they might have different active amino acids. Similarly, Tle4 of P. aeruginosa possesses an unusual pentapeptide motif T-X-S-X-G (Lu et al., 2014), different from the canonical hydrolases with the (G-X-S-X-G) motif (Durand et al., 2014;Russell et al., 2014).
In conclusion, the anti-bacterial function of the T6SS of K. pneumoniae HS11286 was confirmed with the intraspecies and interspecies competition assays. Overexpression of the effector gene tle1 KP in the periplasm caused severe growth retardation of E. coli. The results also indicated that the antibiotics could be an important factor for the T6SS secretion and antibacterial activity. To our knowledge, this is the first report about the antibacterial function of T6SS in the K. pneumoniae. This information might deepen our understanding of the T6SS-carrying CRKP under antibiotic treatments.

AUTHOR CONTRIBUTIONS
HO and YY conceived and designed the experiments. LL and XL performed the experiments. LL, XL, MY, JL, HO, and YY analyzed the data. MY, JL, YY, HO, and ZD contributed reagents/materials/analysis tools. LL, YY and HO wrote the paper.