Secretin channel-interactors prevent antibiotic influx during type IV pili assembly in Pseudomonas aeruginosa

Type IV pili (T4P) are important virulence factors involved in host attachment and other aspects of bacterial pathogenesis. In Gram-negative bacteria, the T4P filament is polymerized from pilin subunits at the platform complex in the inner membrane (IM) and exits the outer membrane (OM) through the OM secretin channel. Although essential for T4P assembly and function, the OM secretin complexes can potentially impair the permeability barrier function of the OM and allow the entry of antibiotics and other toxic molecules. The mechanism by which Gram-negative bacteria prevent secretin-mediated OM leakage is currently not well understood. Here, we report a discovery of SlkA and SlkB (PA5122 and PA5123) that prevent permeation of several classes of antibiotics through the secretin channel of Pseudomonas aeruginosa type IV pili. We found these periplasmic proteins interact with the OM secretin complex and prevent toxic molecules from entering through the channel when there is a problem in the assembly of the T4P IM subcomplexes or when docking between the OM and IM complexes is defective. Thus, our results indicate that the secretin channel-interacting proteins play an important role in maintaining the OM permeability barrier, suggesting they may be attractive targets for potentiators that sensitize Gram-negative pathogens to antibiotics that are normally ineffective at penetrating the OM.

To gain insight into the cellular function of the Slk proteins, we performed a genetic selection 165 for suppressors of the erythromycin sensitivity phenotype of mutants lacking these proteins.

166
The ΔslkAB strain was mutagenized by random insertion of a mariner transposon, and the were reported to significantly lower PilQ protein levels and inhibit PilQ multimer assembly (Ayers et al, 2009). PilF is an OM lipoprotein known as a pilotin that promotes localization pilMNOPQ also suppressed the erythromycin sensitivity of the ΔslkAB strain, but pilQ expression alone was sufficient for abolishing the suppression, indicating that mutations sensitivity of slkAB mutants might be related to the twitching motility defect of our original parent strain.
To investigate the relationship between erythromycin sensitivity and the twitching motility defect, we looked for mutations that can affect twitching motility by resequencing the genomes of the two PAO1 strains using the PAO1 reference genome (NC_002516) (Stover We next sought to determine if other mutations that cause a defect in T4P assembly also result in antibiotic permeation through the PilQ secretin in the ΔslkAB strain. The T4P

Loss of TsaP also causes PilQ-mediated antibiotic diffusion in the absence of Slk
proteins assembly of the OM secretin channel such as pilF and fimV were identified as suppressors of the erythromycin susceptibility along with pilQ mutations (Fig. 2A). TsaP is a component of  OM permeability defect in the absence of Slk proteins. Accordingly, the ΔtsaP ΔslkAB mutant 283 became susceptible to erythromycin, and the erythromycin sensitivity was suppressed by 284 pilQ mutation (Fig. 3E). An alternative explanation for the permeability defect of the ΔtsaP

285
ΔslkAB strain seemed to be that TsaP and the Slk proteins function redundantly for 286 preventing permeation through the T4P secretin complex. However, unlike the ΔslkAB ΔpilC 287 double mutant, the ΔtsaP ΔpilC double mutant did not exhibit a noticeable increase in erythromycin susceptibility (Fig. S7), suggesting that TsaP does not have a major role in preventing permeation through the PilQ secretin channel but is instead important to prevent 290 the formation of disengaged pores.

292
Slk proteins interact with the PilQ secretin complex.

293
The above results led us to hypothesize that the Slk proteins prevent antibiotic permeation proteins in the PAO1 tw strain and its T4P mutant derivatives. To monitor SlkA localization, we 299 used a SlkA-mScarlet fusion protein produced from its native chromosomal locus. As SlkB might interfere with SlkA-mScarlet localization by competing for interaction with the PilQ 301 channel, we deleted slkB when we introduced slkA-mScarlet fusion at the native locus. As

302
for SlkB localization, slkB-mScarlet was expressed from an ectopic locus in the chromosome

303
(the Tn7 integration site) in the ΔslkAB strain background to avoid competition between SlkB-mScarlet and untagged Slk proteins. SlkA-mScarlet and SlkB-mScarlet are partially 305 functional in that the strains expressing the fusion proteins do not show a permeability defect 306 when combined with ΔpilC that is as severe as the ΔpilC ΔslkAB strain (Fig. S9).
the IM complex assembles with the secretin channel. Overall, the localization pattern of Slk proteins in wild-type and T4P mutant strains is consistent with our hypothesis that Slk 316 proteins interact with the PilQ secretin channel to prevent the diffusion through the OM when the IM complex is not properly assembled with the secretin channel.

332
The effect of a pscJ mutation was examined in the strains that overproduce the master 333 regulator ExsA to induce T3SS expression. However, we did not observe increased 334 sensitivity of these strains to macrolides compared with that of the PAO1 tw and its ΔslkAB with the T4PS.

Discussion
The cell envelope of Gram-negative pathogens functions as an effective permeability barrier against antibiotics because the OM efficiently restricts the influx of various toxic compounds 341 and efflux pumps expel the toxic molecules from the periplasm as well as from the cytoplasm.

356
Although secretin channels form large pores to accommodate protein substrates, diffusion 357 through the channels is thought to be minimized by the gate structure of the channels that  independent support for the specific interaction between SlkA and the PilQ secretin channel.

396
Taken together, all the available data are consistent with the idea that the Slk proteins 397 function as a plug that help seal the secretin channel to maintain the OM permeability barrier 398 until the T4PS is fully assembled (Fig. 5).

399
SlkAB homologues are only found in gamma-proteobacteria and not widely conserved the diffusion of hydrophobic drugs. However, hydrophobicity did not seem to be a general increased susceptibility of P.aeruginosa to several classes of drugs with different properties.
Thus, although the loss of Slk protein function combined with defects in T4P assembly 421 clearly creates a PilQ-dependent OM permeability defect, the precise chemical properties 422 that promote permeation through the secretin channel remain to be determined.

423
Understanding these and other chemical properties that promote OM permeation will be promising avenue for developing effective strategies that broaden the spectrum of several approved antibiotics to include activity against Gram-negative infections.

Materials and Methods
medium (Cheng et al, 1970). The strains and plasmids that were used for the study are listed 440 in SI Appendix, Table S3 and Table S4. Detailed procedures for strain and plasmid 441 construction are also provided in SI Appendix.

475
We thank the members of Cho lab at Sungkyunkwan University for helpful comments and

696
The LB and LB agar contained 1 mM IPTG to induce pilQ expression.

Media, Bacterial Strains, and Plasmids
pOKP10 -To construct a plasmid for deletion of slkA (PA5122), a 700bp region upstream of pEXG2 using BamHI and XbaI.  Transformation of P. aeruginosa strains All transformations of P. aeruginosa strains were performed by electroporation using OKP20 and OKP21 -Strains that express pilQ-mScarlet sandwich fusion at the native pilQ 502 locus in the PAO1 tw and HJP1 strain backgrounds were generated by using the procedure 503 described above after transferring pOKP35 by conjugation.

505
OKP23 -pilC was deleted in PAO1 tw by using the procedure described above with pOKP42.   To generate the phylogenetic tree, the amino acids sequence of PA5122 and PA5123 were 687 input into BLASTp and searched against the NCBI "non redundant" (nr) database with an e-688 value cutoff of 1e -6 for each protein (Pruitt et al, 2005