Klebsiella pneumoniae type VI secretion system-mediated microbial competition is PhoPQ controlled and reactive oxygen species dependent

Klebsiella pneumoniae is recognized as an urgent threat to human health due to the increasing isolation of multidrug resistant strains. Hypervirulent strains are a major concern due to their ability to cause life-threating infections in healthy hosts. The type VI secretion system (T6SS) is widely implicated in microbial antagonism, and it mediates interactions with host eukaryotic cells in some cases. In silico search for genes orthologous to T6SS component genes and T6SS effector genes across 700 K. pneumoniae genomes shows extensive diversity in T6SS genes across the K. pneumoniae species. Temperature, oxygen tension, pH, osmolarity, iron levels, and NaCl regulate the expression of the T6SS encoded by a hypervirulent K. pneumoniae strain. Polymyxins and human defensin 3 also increase the activity of the T6SS. A screen for regulators governing T6SS uncover the correlation between the transcription of the T6SS and the ability to kill E. coli prey. Whereas H-NS represses the T6SS, PhoPQ, PmrAB, Hfq, Fur, RpoS and RpoN positively regulate the T6SS. K. pneumoniae T6SS mediates intra and inter species bacterial competition. This antagonism is only evident when the prey possesses an active T6SS. The PhoPQ two component system governs the activation of K. pneumoniae T6SS in bacterial competitions. Mechanistically, PhoQ periplasmic domain, and the acid patch within, is essential to activate K. pneumoniae T6SS. Klebsiella T6SS also mediates anti-fungal competition. We have delineated the contribution of each of the individual VgrGs in microbial competition and identified VgrG4 as a T6SS effector. The DUF2345 domain of VgrG4 is sufficient to intoxicate bacteria and yeast. ROS generation mediates the antibacterial effects of VgrG4, and the antitoxin Sel1E protects against the toxic activity of VgrG4. Our findings provide a better understanding of the regulation of the T6SS in bacterial competitions, and place ROS as an early event in microbial competition.


23
Short tile: Klebsiella pneumoniae T6SS for antimicrobial warfare The type VI secretion system (T6SS) is a bacterial nanomachinery that delivers substrates 71 into a target cell in an one-step process.The system was first described as a cluster of impaired in 72 nitrogen fixation (imp) genes in Rhizobium leguminosarum [1]. Shortly after, Rao and co-workers We sought to investigate the diversity of the T6SS locus in different K. pneumoniae strains. 162 To obtain a global picture of the T6SS in K. pneumoniae, we downloaded 700 K. pneumoniae 163 genomes from NCBI, selected to represent the broad phylogenetic structure of the species. Using 164 the SecRet6 database we searched for genes orthologous to T6SS component genes (Fig 2A) and

176
The T6SS of different pathogens has been shown to mediate killing of other bacteria. To 177 determine whether Kp52145 exhibits T6SS-mediated antibacterial activity, we used strain E. coli 290 activity of Kp52145 T6SS. We tested human β-defensin 3 (HBD3) because the levels of this peptide 291 increase several fold during human pneumonia [35]. A sub lethal concentration of HBD3 also 292 triggered a significant decrease in the recovery of E. coli when co-cultured with Kp51245 but not 293 when the clpV mutant was tested (Fig 5B), indicating that the increased Kp52145-mediated killing is 294 T6SS-dependent.

295
Based on the knowledge that antimicrobial peptides activate the PhoPQ system and that

304
T6SS resulting in increased T6SS-dependent antimicrobial activity. Therefore, we hypothesized that 305 antimicrobial peptides-induced T6SS-dependent killing should be abrogated in the phoQ mutant 306 background. Indeed, this was the case (Fig 5A-B). Complementation of the phoQ mutant restored 307 the killing potential of the phoQ mutant to wild-type levels (Fig 5A-B). with an active T6SS (Fig 7D-F). In contrast, we observed no up-regulation of the fusion when

345
Kp52145 was incubated with any of the T6SS mutants (Fig 7D-F). To connect PhoPQ up-regulation 346 with that of Kp52145 T6SS, we assessed whether co-incubation with a prey with an active T6SS will     (Fig 8A), and A. baumannii 370 ATCC17978 ( Fig 8B). In contrast, complementation of the phoQ mutant with a construct in which 371 phoQ lacked the periplasmic domain, phoQ 45-190, did not restore luciferase levels (Fig 8A-B). Klebsiella was grown in LB pH6 and LB NaCl was not observed in any of the phoQ variants (S10B Fig).

379
In good agreement with our previous results demonstrating that the PhoPQ system controls the 380 activation of the T6SS in bacterial competitions, only the PhoQ-FLAG tagged recovered the 381 reduction of K. pneumoniae ATCC43817, A. baumannii ATCC17978 preys upon co-incubation with 382 phoQ mutant (Fig 8C-D). Complementation with phoQ  and phoQ GNNNNAQ did not restore

383
Kp52145 T6SS-mediated killing of the T6SS active preys (Fig 8C-D). Western blot analysis 384 confirmed that the PhoQ FLAG tagged proteins were expressed by Kp52145 (S10C Fig). 385 Collectively, these findings revealed that the PhoQ periplasmic domain, and the acid patch 386 within, is essential to activate K. pneumoniae T6SS upon incubation with T6SS active competitors,

387
and by the last-line antibiotic colistin.

388
Contribution of K. pneumoniae VgrGs to bacterial killing.

389
Our bioinformatics analysis revealed that Kp52145 encodes three VgrGs, making relevant to 390 investigate the relative contribution of each of them to bacterial killing under different conditions. We

410
We next sought to determine the relative contribution of each VgrG in inter, and intra bacterial 411 species killing. Co-culture of the vgrG1 mutant with K. pneumoniae ATCC43816 (Fig 9A), A.

412
baumannii ATCC17978 (Fig 9B), and B. cenocepacia K56-2 ( Fig 9C) resulted in no killing of any the 413 preys. Interestingly, we observed no increase in the recovery of any of the preys when co-incubated 414 with the vgrG2 mutant. In contrast, co-incubation of the preys with vrgG4 mutant resulted in an 415 increase in the recovery of Klebsiella (Fig 9A), Acinetobacter ( Fig 9B) and Burkholderia ( Fig 9C).

416
Complementation of vgrG1 and vgrG4 mutants restored Kp52145-dependent killing of the preys (Fig

420
To determine whether Kp52145 T6SS also mediates antifungal competition, we co-incubated

421
Kp52145 with the clinically important fungal pathogen Candida albicans and the model yeast

422
Saccharomyces cerevisiae. Co-culture of Kp52145 with both organisms resulted in a significant 423 decrease in the recovery of yeast and fungal cells (Fig 10A and B). When the experiments were 424 performed using the Klebsiella clpV mutant the recovery of Saccharomyces and Candida was not 425 affected, demonstrating that the observed inhibition is T6SS dependent (Fig 10A-B).

426
We then investigated which of the VgrGs may play a role in Kp52145-dependent antifungal 427 activity. Quantitative killing assays revealed an increase in C. albicans recovery when co-cultured 428 with either vgrG2 or vgrG4 mutants ( Fig 10A). Complementation of these mutants resulted in a 429 decrease in the fungi recovered ( Fig 10A), demonstrating that VgrG2 and VgrG4 are necessary for

430
Kp52145-induced killing of C. albicans. VgrG4 was also necessary for Kp52145-mediated killing of 431 S. cerevisiae because we observed an increase in the recovery of yeast when co-cultured with the 432 vgrG4 mutant (Fig 10B). In contrast, no increase in recovery was observed when S. cerevisiae was 433 co-incubated with either vgrG1 or vgrG2 mutants. Collectively, these findings highlight a role for

435
K. pneumoniae T6SS mutants are attenuated in the Galleria mellonella infection model.

436
The G. mellonella infection model is well established to assess the virulence of K. into the larvae resulted in no mortality ( Fig 11A). 90% of the larvae infected with 10 6 CFUs of the 442 clpV and tssB1 mutants survived after 120 h. In contrast, only 15% of the larvae challenged with 10 6

443
CFUs of Kp52145 survived ( Fig 11A). Supplementary Figure 12 shows that infection with 444 ATCC43816 and NTUH-K2044 tssB mutants also resulted in decrease mortality as compared to the 445 wild-type strains, demonstrating that the contribution of the T6SS to virulence in G. mellonella is not 446 dependent on the K. pneumoniae strain background.

447
We next sought to determine the relative contribution of Kp52145 VgrGs to virulence. 80%

448
of the larvae infected with 10 6 CFUs of the vgrG1 and vgr2 mutants survived whereas there was no 449 difference in survival between larvae challenged with the vgr4 mutant and the wild type ( Fig 11B).

450
There were no differences in the mortality triggered by any of the double vgrG mutants, and all of 451 them killed only 25% of the challenged larvae after 120 h ( Fig 11C). Infection with the triple mutant 452 resulted in 100% survival of the infected larvae ( Fig 11C).

453
Altogether, these data provide evidence demonstrating that K. pneumoniae T6SS is essential 454 for virulence in the G. mellonella infection model. Our results also revealed that Klebsiella 455 pathogenicity in Galleria is combinatorial with all three vgrGs required to result in the overall virulence 456 phenotype.

457
VgrG4 is a toxin inhibited by the antitoxin Sel1E.

458
Bioinformatics analysis of VgrG4 revealed that it encodes a C-terminus containing a 459 DUF2345 domain of unknown function ( Fig 12A). The presence of C-terminal extensions in VgrGs 460 is considered an indication of an effector function. Therefore, we sought to determine whether VgrG4

461
is an antibacterial toxin. VgrG4 was cloned into pBAD30 to control the expression of vgrG4 by 462 arabinose. We also cloned vgrG1 and vgr2 into pBAD30 to assess whether expression of any of the 464 and the toxicity of the proteins upon induction assessed by plating. Figure 12B shows that expression 465 of VgrG1 and VgrG2 in E. coli had no impact on E. coli recovery. In stark contrast, induction of VgrG4 466 resulted in an 80% decrease in E. coli recovery ( Fig 12B). To assess the relative importance of 467 VgrG4 domains for the antibacterial effect, we constructed truncated variants of VgrG4 and 468 investigated whether they retain the antibacterial effect. Whereas VgrG4 1-517 derivative did not affect 469 E. coli recovery, VgrG4 570-899, containing the DUF2345 domain, was sufficient to exert an antibacterial 470 effect ( Fig 12B).

471
We sought to determine whether VgrG4 will also exert a toxic effect when expressing in

483
In T6SS, protection against kin cells is conferred by the production of immunity proteins that 484 inhibit their cognate toxins. Immunity proteins are normally adjacent to the cognate toxin.

485
Downstream of vgr4 there are 5 genes of unknown function encoding sel-1 repeats (Fig 1). To 486 determine whether any of these genes assures protection against VgrG4, the genes were expressed 487 from the pMMB207 plasmid in E. coli harbouring pBADVgrG4. Induction of Sel1D by IPTG did not 488 alleviate VgrG4-triggered toxicity ( Fig 12D). In contrast, induction of Sel1E abrogated VgrG4-489 dependent killing ( Fig 12D). Furthermore, Sel1E also conferred protein against the toxicity triggered 490 by VgrG4 570-899 containing DUF2345 domain ( Fig 12D). Altogether, these results indicate that Sel1E 491 is the immunity protein that protects against the toxic activity of VgrG4.

499
Homology modelling is a common method used for predicting the 3D structures of proteins 500 based on the known structure of a related protein [40] and was used to model the 3D structure of K.

524
(residues 550-645). After the end of the C-term part of the model there is a proline-rich region in the 525 VgrG4 sequence ( Fig 13A) and since prolines are unfavoured residues in beta strands (24), it is 526 unlikely that the beta helical spike structure could continue. We were unable to model the remaining 527 C-terminal residues (704-899) of VgrG4, and thus the folding of this region remains unknown. Figure   528 13C depicts the monomeric 3D modelling of VgrG4 with the two regions of beta strands circled.  14A). Thus, the I-TASSER models suggest that the N-terminal end of Sel1E shares the typical fold 549 of SLR-proteins while the structure of the C-terminal region could not be accurately determined due 550 to a lack of homologous structures. Next, we used the ConSurf server [52] to predict exposed residues with potential for protein-552 protein interactions (Fig 14A) and calculated the electrostatic potential of the protein surface for the 553 high confident N-terminal region of the Sel1E model ( Fig 14C). As our experiments show that SEl1D 554 is not the immunity protein for VgrG4 (Fig 12D), we compared Sel1E and Sel1D and identified two 555 regions with significant differences in the SelE and Sel1D sequences (Fig 14A, green asterisk). The 556 first one is an acidic patch formed by Asp230 and Glu251 and several Ser and Thr residues (Fig   557  14C, top), whereas the second one is more positively charged due to the two Arg residues (Fig 14C,   558 bottom  (Fig 15A). Interestingly, neither VgrG1 nor VgrG2 573 induced the expression of any of the VgrG4-upregulated genes (Fig 15A), indicating that VgrG4-

591
Having previously established that Sel1E protects against VgrG4-mediated toxicity, we

605
However, these studies have also revealed significant differences between bacteria species, making 606 necessary to investigate different bacterial models to obtain a better picture of the intricacies of the 607 system. In this study, we provide a comprehensive characterization of K. pneumoniae T6SS, demonstrating its role in microbial competition and uncovering new aspects on how bacteria regulate 609 T6SS-mediated antagonism.

610
Our in silico analysis revealed a significant diversity in how the T6SS is organized in K.

682
The antifungal activity may result beneficial for Klebsiella given that fungi and Klebsiella colonize 683 common sites such as the gut, and the respiratory system although it is also possible that this 684 fungicidal action may be also important in the environment.

685
We have uncovered that K. pneumoniae has a preference for different VgrGs depending the 686 conditions. VgrG1 is essential in all conditions tested except antifungal competition, VgrG2 is 687 required for antibacterial competition in LB pH6 , and antifungal competition, whereas VgrG4 is needed 688 for antibacterial competition in LB NaCl , and to intoxicate kin, non-kin and fungi. Based on the in silico 689 analysis shown in Figure 1, and the accepted notion that specific effectors are absolutely dependent

755
String test

756
A loop was used to lift a portion of a colony from a fresh LB agar plate of K. pneumoniae.

757
The formation of viscous strings of >5 mm in length indicates a positive string test [73]. The string 758 test was done on nine randomly chosen colonies from three independent plates.

759
Bioinformatics mapping of the T6SS clusters

760
The unidentified genes within the T6SS clusters underwent a manual annotation process  Marker-less mutation in K. pneumoniae PCR primers used to construct the mutants were designed using the whole genome 782 sequence of Kp52145 (GenBank Accession No. FO834906.1). The primer pairs (Table S1) were 783 used in separate PCR reactions to amplify ~1000-bp fragments flanking the gene to be mutated.

784
BamHI restriction sites internal to these flanking regions were incorporated at the end of each 785 amplicon. Purified fragments were then polymerised and amplified as a single PCR amplicon using 786 the primers gene_UPFWD and gene_DWNRVS. These amplicons were cloned into pGEM-T Easy  (Table S1).

849
To determine luciferase activity of bacteria grown in solid media, reporter strains were 850 incubated with preys as described in the antibacterial completion assay for 6 hours. The mixture was 851 then recovered with 1 ml of PB buffer in a microcentrifuge tube. Bacteria were collected by 852 centrifugation (13,300 rpm, 2 mins), washed with 1 ml of PB buffer, and adjusted to an OD 600 of 1.0.

853
One hundred microlitres were used to assess luminescence. All strains were tested in quadruplicate 854 from three independent cultures.

855
Site-directed mutagenesis of phoQ 856 The phoPQ promoter and coding region including a FLAG-tag was amplified by PCR using pUC18R6KT7mini-Cm--PhoPphoQ GNNNNAQ _FLAG pUC18R6KT7mini-Cm containing the promoter and genes of phoP and phoQ; phoQ has a C terminal FLAG tag and the acidic patch EQGDDSE has been mutated to GNNNNAQ This Study a. Burkholderia cenocepacia research and referral repository for Canadian CF Clinics.
(C) Growth on SD (glucose) and SG (galactose) plates of serial dilutions of wild-type YPH499 yeast cells harbouring the indicated plasmids, and therefore expressing under the control of the GAL1 promoter the corresponding proteins on galactose-based medium. Image is representative of three independent experiments. (D) Recovery of E. coli cells following induction of the Ara promoter of the pBAD plasmid with arabinose (0.1%), and the tac promoter of the pMMB207 plasmid with IPTG (1 mM) for 60 min. #, results are significantly different (P < 0.0001 [two-tailed t test]) from the results for bacteria harbouring pBAD30 and pMMB207 without induction; n.s., not significant differences. The data are presented as means ± the standard deviations (n = 3).  (A) Sequence alignment between Sel1D and Sel1E showing the residues (green) that are both variable and predicted to be exposed by ConSurf.
(B) MODFOLD residue accuracy prediction for the Sel1E I-TASSER model. The colour scheme for the accuracy (Å): blue (high accuracy) through green, yellow and orange to red (low accuracy).
(C) Electrostatic surface potentials for Sel1E. The close-ups show the variable residues (green; marked by an asterisks in (A), which contribute to the unique electrostatic potential surface of Sel1E, see text for more details. Due to the low confidence of the C-terminal part of the model, only the region of high confidence was used for the comparison. The electrostatic surfaces were calculated with the APBS tool (Adaptive Poisson-Boltzmann Solver) in PyMOL and the colour ranges from -7 to 7.