A global genomic approach uncovers novel components for twitching motility-mediated biofilm expansion in Pseudomonas aeruginosa

Pseudomonas aeruginosa is an extremely successful pathogen able to cause both acute and chronic infections in a range of hosts, utilizing a diverse arsenal of cell-associated and secreted virulence factors. A major cell-associated virulence factor, the Type IV pilus (T4P), is required for epithelial cell adherence and mediates a form of surface translocation termed twitching motility, which is necessary to establish a mature biofilm and actively expand these biofilms. P. aeruginosa twitching motility-mediated biofilm expansion is a coordinated, multicellular behaviour, allowing cells to rapidly colonize surfaces, including implanted medical devices. Although at least 44 proteins are known to be involved in the biogenesis, assembly and regulation of the T4P, with additional regulatory components and pathways implicated, it is unclear how these components and pathways interact to control these processes. In the current study, we used a global genomics-based random-mutagenesis technique, transposon directed insertion-site sequencing (TraDIS), coupled with a physical segregation approach, to identify all genes implicated in twitching motility-mediated biofilm expansion in P. aeruginosa. Our approach allowed identification of both known and novel genes, providing new insight into the complex molecular network that regulates this process in P. aeruginosa. Additionally, our data suggest that the flagellum-associated gene products have a differential effect on twitching motility, based on whether components are intra- or extracellular. Overall the success of our TraDIS approach supports the use of this global genomic technique for investigating virulence genes in bacterial pathogens.

Proteins were resolved in 8%, 10%, 12% or 15% gels using the Mini-PROTEAN system 160 (Bio-Rad) and transferred to Nitrocellulose membrane (GE Healthcare) by electrophoresis. 161 Membranes were blocked in 5% milk (Sigma) before incubation with primary antibodies. plates were recovered as a pool, resuspended in LB, pelleted by centrifugation (10,000 g, 10 193 min, 4 °C), and then resuspended in LB plus glycerol (15 % (v/v)) and stored at -80 °C. The 194 protocol was repeated on a large scale until ~2 million mutants were obtained. 195

TraDIS assay with mutant pool 196
The transposon mutant library pool was diluted 1:10 into 9 mL CAMHB in 10 197 separate 50 mL Falcon tubes which were covered with aeroseal to facilitate aeration within 9 poured and allowed to set overnight at room temperature. The following morning the agar 200 was flipped into a larger petri dish to expose the smooth underside set against the petri dish 201 base which promotes rapid twitching motility-mediated biofilm expansion (53). 1.5 mL of 202 overnight growth of the pooled transposon mutant library was pelleted by centrifugation 203 (10,000 g, 3 min, 4 °C), and the whole pellet then spotted into the centre of the flipped agar 204 plate. This was repeated for all 10 overnight cultures and performed in triplicate (i.e. a total 205 of 30 plates). All plates were incubated under humid conditions at 37 °C for 65 h. To harvest 206 mutants based upon their ability to undergo twitching motility-mediated biofilm expansion 207 mutants were harvested from the inner, non-twitching zone and from the outer, active-208 twitching motility zone (see Figure S1A) for all 3 replicates. The cells from the inner and 209 from the outer zones were harvested separately for all 3 replicates by resuspension in 5 mL 210 LB, followed by centrifugation (10,000 g, 10 min, 4 °C), to pellet the cells. The supernatant 211 was discarded and the cells used for genomic DNA extraction. 212

Generation of DNA sequencing libraries and library sequencing 227
TraDIS was performed using the method described in Barquist et al., (2016 accessions of each sample are ERS427191-3 for the non-twitching cells, ERS427194-6 for 245 the twitching cells and ERS427197-9 for the base library without selection. 246

Results 252
Confirmation of genes known to be involved in twitching motility 253 To identify genes involved in twitching motility-mediated biofilm expansion we 254 generated a high-density random transposon mutant library in P. aeruginosa PA14 using 255 conjugation of a Tn5 minipro vector and gentamicin selection. We determined that this 256 library consisted of 310,000 unique Tn5 mutants by sequencing DNA from 10 9 cells from the 257 raw base library, in duplicate, without selection. 258 Approximately 10 9 cells from an overnight culture of the pool of transposon mutants 259 were concentrated and inoculated as a central spot on top of an inverted agar plate. These 260 were incubated for 65 h at 37 °C under humid conditions to allow a twitching motility-261 mediated surface biofilm to form. An inverted agar plate was used to expose the smooth 262 underside of the moist, set agar, which facilitates rapid twitching motility-mediated biofilm 263 expansion and discourages other forms of motility (53). This colony biofilm assay was 264 favoured over the subsurface twitching motility assay (53) as this assay allows a much greater 265 number of cells to be recovered, thus allowing sufficient amounts of genomic DNA to be 266 extracted for downstream sequencing. Transposon mutants were separated based upon their 267 ability to expand via twitching motility, away from the site of inoculation, with cells being 268 harvested from the inner, non-twitching section of the colony biofilm, and the outer, actively 269 expanding edge ( Figure S1A). The outer and inner zones from 10 plates were combined to 270 form each replicate, and 3 replicates were performed over different days. Genomic DNA was 271 extracted from both combined pools of mutants then separately sequenced to determine the 272 number of insertions per gene, using a TraDIS approach, as described previously (42). The 1 2 mutant pools were compared as described previously (42) and using a cut off of log 2 FC=4 275 and a Q-value of <0.01 to identify genes with differential insertion levels during twitching 276 motility-mediated biofilm expansion. This revealed 942 genes as having a putative role in 277 twitching motility-mediated biofilm formation: 82 genes with increased insertions and 860 278 with decreased insertions (Table S1). 42 of the 44 genes known to be involved in twitching 279 motility-mediated biofilm expansion were identified (3, 4, 8) ( Table 1). The two genes which 280 we could not assay in our TraDIS screen were rpoN (PA14_57940) and pocB (PA14_25500) 281 due to a minimal insertion density in these genes in our starting base library. 282

Identification of novel components involved in twitching motility-mediated biofilm 283 expansion 284
From our TraDIS results we selected 39 genes that had not been previously implicated 285 in twitching motility (Table S2) to phenotypically characterize using single transposon 286 mutants from the non-redundant PA14 transposon mutant collection (47). We tested the 287 ability of each mutant to undergo twitching motility using a sub-surface stab assay. For those 288 target genes that had multiple transposon mutants available, we tested all mutants, bringing 289 the total number of assayed mutants to 52 (Table S2). From these assays, we detected 32 290 transposon mutants which had significantly altered levels of twitching motility compared to 291 wildtype ( Figure S2). 292 We selected 11 transposon mutants to further characterize based on biological interest 293 and especially dramatic changes in twitching ability ( Table 2). The genes containing these 294 transposon insertions appear to group into distinct functional classes including cellular 295 metabolism, signal transduction, cytokinesis and flagella-mediated motility (Table 2). 296 Biofilm and growth assays were conducted for these 11 selected mutants to determine 297 whether there was any effect on submerged biofilm formation ( Figure 1B) and also to 298 determine if the observed twitching motility defect ( Figure 1A) was due to a growth-related 1 3 effect ( Figure 1C). Of these only kinB was found to have decreased levels of biofilm 300 formation compared to wildtype ( Figure 1B), as reported previously (55). None of these 301 transposon mutants had an altered ability to grow in the minimal medium used for the biofilm 302 assay demonstrating that any alteration in biofilm formation was not a result of a growth-303 related effect ( Figure S1B). Of these 11 transposon mutants prlC, lon, kinB, fliF and fliG had 304 significant alterations in growth rate in LB media (the same media used for sub-surface 305 twitching motility assays) compared to wildtype ( Figure 1C). Specifically, fliF, fliG and prlC 306 had a shorter lag time than wildtype, and kinB and lon had a longer lag time than wildtype, 307 but all reached approximately the same final cell density ( Figure 1C). Based upon these 308 results a growth-related effect may account for some of the observed decrease in twitching 309 motility for kinB and lon. However, the observed alterations in twitching motility for prlC, 310  Figure 1A) had extra-long pili, which in some cases appeared to 327 intertwine with the observed flagella ( Figure 2E) and an overall reduction in polar T4P levels 328 compared to wildtype ( Figure 2F). While PA14 wildtype, PA14_66850 and prlC were found 329 to possess non-polar T4P in a few cases, there was no difference in the numbers of non-polar 330 T4P in the mutant strains compared to wildtype ( Figure 2G). 331 Investigating the role of PA14_66580 in T4P assembly and function 332 A clean deletion in PA14_66580 was generated in the orthologous gene (PAK_05353 333 with the gene product having 99.82 % amino acid identity to PA14_66580) in the P. 334 aeruginosa strain PAK. As was observed for the transposon mutant of PA14_66580 in PA14, 335 a reduction in twitching motility was also observed in the PAK deletion mutant PAK05353 336 ( Figure 3A). PA14_66580/PAK_05353 is encoded just upstream of the pilMNOP gene cluster 337 which encodes the components in the alignment subcomplex, and the outer membrane 338 associated secretin complex of PilP and PilQ which is involved in T4P outer membrane 339 extrusion (11-14). Additionally, PA14_66580/PAK_05353 is also annotated as a predicted 340 ExeA-like protein. ExeA is an ATPase which binds peptidoglycan and is involved in 341 transport and multimerization of ExeD into the outer membrane to form the functional 342 secretin of the Type II secretion system (56). Given this, we hypothesized that 343 PA14_66580/PAK_05353 may be involved in multimerization and/or localization of the PilQ 344 secretin complex. To investigate this we performed immunoblotting of whole cell lysates of 345 wildtype, PAK05353 and PAKpilQ strains harvested from agar plates for both the multimeric 346 and monomeric forms of PilQ ( Figure 3B). This revealed that PAK05353 was able to form 347 both multimers and monomers of PilQ to the same extent as wildtype indicating that 1 5

Functional Gene Enrichment Analysis 350
Enrichment analyses of genes that had increased or decreased mutant populations in 351 the TraDIS output using the KEGG database (57) revealed that 3 key pathways were 352 significantly altered. These were: flagella biosynthesis, two-component systems (TCS) and 353 chemotaxis (Table S3 (increased population) and Table S4 (decreased population)). 354 We noticed that a number of flagella-associated structural and regulatory genes had 355 altered mutant abundances following selection for twitching motility-mediated biofilm 356 formation in our TraDIS assay, and some single mutants were confirmed to have significantly 357 altered levels of twitching motility compared to wildtype ( Figure 1A). Remarkably, this 358 revealed a strong correlation between gene products predicted to have a negative effect on 359 twitching motility (which corresponds to a positive log-fold change in our TraDIS output 360 (Table S1) (Table S1)  The chemotaxis pathway identified in our functional gene enrichment analysis 367 included mutants of swimming chemotaxis (che) genes which appeared to promote 368 (cheA/B/Z/Y) as well as inhibit (cheR/W) twitching motility-mediated biofilm formation. This 369 suggests a balance between bacterial chemotaxis and twitching motility, especially as the 370 chemotaxis pathway also controls flagella assembly. The TCS linked to twitching motility 371 were mostly known genes, for instance algZ/R involved in alginate biosynthesis, or the pil 372 genes in T4P production, but also included some unexpected genes related to osmotic 1 6 stability, such as cusS/R involved in copper efflux, or dctA/B/D/P for C4-dicarboxtrate 374 transport. 375

Discussion 376
In this study, we have successfully applied a physical separation-based TraDIS 377 approach to identify genes involved in twitching motility-mediated biofilm formation in P. 378 aeruginosa. Using this method, we could detect almost all genes currently known to be 379 involved in T4P assembly and twitching motility, in addition to a large number of genes 380 identified in our TraDIS output (Table S1) and a select group for further study (Table 2) not 381 previously known to be involved. 382 A functional enrichment analysis of all genes that have altered mutant abundances in 383 our assay identified 3 major groups of gene function that were affected during twitching 384 motility: flagella assembly, bacterial chemotaxis and TCS. Perhaps the most interesting from 385 these is the potential involvement of the flagella as suggested from the predicted (Table S1) 386 or determined ( Figure S2) differential effect of structural and regulatory flagella components 387 on twitching motility. Specifically, we observed a strong correlation between gene products 388 predicted to have a negative effect on twitching motility with proteins associated with the 389 outer part of the flagella body, and in contrast, proteins associated with the inner part of the 390 flagella body were predicted to have a positive effect on twitching motility (Figure 4). This is 391 intriguing as it suggests a differential effect on twitching motility by flagella components 392 based upon their cellular location and certainly warrants further investigation in future work. 393 For each of the 11 gene targets selected for further investigation (Table 2) the 394 twitching motility phenotype was confirmed in a sub-surface stab assay, with growth assays 395 performed to demonstrate that the observed twitching phenotype was not due to a growth 396 defect. Submerged biofilm formation was also assayed which revealed that only kinB had a 1 7 TraDIS assay did indeed selectively identify genes specific for twitching motility-mediated 399 biofilm expansion on a semi-solid surface. Overall these assays confirmed the twitching 400 motility phenotype observed for prlC, PA14_66580, pfpI, fliG and motY was not due simply 401 to a growth related defect. For these mutants TEM was used to determine whether the 402 twitching motility phenotype could be attributed to alterations in levels and/or localization of 403 surface assembled T4P. No pili were observed in a pfpI mutant ( Figure 2B), which explains 404 the observed lack of twitching motility ( Figure 1A). PfpI is an intracellular protease which 405 affects antibiotic resistance, swarming motility and biofilm formation in P. aeruginosa (58) 406 however, to our knowledge the current study is the first report to a role for PfpI in twitching 407 motility. Given the established role of intracellular proteases in controlling levels of a range 408 of chaperones and regulatory proteins it is likely that the protease activity of PfpI is required 409 for control of regulators or other proteins involved in T4P biogenesis and/or assembly. is an ATPase that binds peptidoglycan and is involved in transport and multimerization of 416 ExeD into the outer membrane to form the secretin of the Type II secretion system (56). Our 417 TEM data revealed that PA14_66580 had reduced numbers of pili compared to wildtype 418 ( Figure 2C, F), which correlates with the observed reduction in twitching motility in both 419 PA14 and PAK strain backgrounds ( Figure 1A and Figure 3A). Given that no difference in 420 the expression of monomeric or multimeric PilQ was observed in a mutant of PA14_566580 421 in PAK (PAK05353) (Figure 3B), we suggest that the reduction in surface T4P and twitching 422 motility levels is not due to a lack of secretin formation. PA14_66580 could instead be 1 8 involved in stabilization of the secretin pore and/or formation of the assembly and motor 424 subcomplexes in order to allow full functionality of the T4P. 425 A prlC mutant was found to have increased levels of twitching motility compared to 426 wildtype ( Figure 1A), a reduction in polar surface assembled T4P ( Figure 2F) and to have a 427 putative interaction between the surface-assembled flagella and T4P ( Figure 2D Alternatively, given the putative interaction of the flagella and T4P observed ( Figure 2D), 442 PrlC may be involved in processing flagella-associated proteins to ultimately affect the 443 putative interaction between these two motility machines and thus the function of the T4P (as 444 suggested from Figure 4). 445 We observed that a motY mutant had increased twitching motility levels ( Figure 1A) 446 but reduced levels of T4P (Figure 2A, F). MotY is a peptidoglycan binding protein which is 447 required for MotAB-mediated flagella motor rotation and is associated with the outer-  *A positive log-fold change indicates that the gene product has a negative effect on twitching motility, while a negative log-fold change has a positive effect on twitching motility; # result is not significant (Q-value is >0.05) but confirmed by visual inspection to have differential insertion levels during twitching motility-mediated biofilm expansion; NA -not assayed due to a minimal insertion density in these genes in our starting base library.  T4P compared to wildtype and in some cases these pili appeared to interact with the flagella 697 (potential interactions marked with white arrows) (E). In each image the pili are arrowed in Supplementary Table Legends  725   Table S1. Full list of all gene hits obtained in TraDIS screen. Table contains PA14 gene  726 locus tag, gene name, annotated gene function, the logFC (log2 fold change) in transposon 727 insertions from mutants obtained from the inner (non-twitching) zone vs mutants obtained 728 from the outer (active-twitching) motility zone, and the statistical q value for all replicates. 729 Table S2. Full list of PA14 transposon mutants used in the current study. PA14 mutants 730 from the PA14 transposon mutant collection (47) with gene name, annotated gene function 731 and the plate and well where the mutant was taken from in the collection. 732 Table S3. Enrichment analyses of genes that had an increased mutant population in the 733 TraDIS output. KEGG database (57) was used to identify pathways enriched in mutants 734 with increased mutant populations in the TraDIS output (Table S1). 735 Table S4. Enrichment analyses of genes that had a decreased mutant population in the 736 TraDIS output. KEGG database (57) was used to identify pathways enriched in mutants 737 with decreased mutant populations in the TraDIS output (Table S1)  undergo twitching motility-mediated biofilm expansion away from the inoculation site. After 743 72 hrs mutants were harvested from the inner non-motile region (outlined in white) and the 744 outer active twitching edge (outlined in black). Genomic DNA was extracted from each pool 745 of mutants and sequenced using a mass parallel approach. (B) Growth rates in minimal 746 media for 11 selected transposon mutants assayed for a submerged biofilm defect in Figure  747 3 3 M63 minimal media (same media used for submerged biofilm assays). There was no 749 significant different between growth rates of transposon mutants compared to wildtype 750 predicted by a one-way ANOVA with Dunnett's multiple comparison test. Data are 751 represented as the mean ± standard deviation for 2 independent replicates performed in 752 triplicate. 753 754 Figure S2. Twitching motility of selected transposon mutant targets identified using 755 TraDIS. Sub-surface twitching motility-mediated interstitial biofilm expansion at 756 agar/plastic interface after 24 h incubation at 37 °C is presented as the average surface area in 757 mm 2 ± standard deviation normalized against wildtype as obtained from 2 independent 758 experiments performed in triplicate. 759