Identifying the suite of genes central to swimming in the biocontrol bacterium Pseudomonas protegens Pf-5

Abstract Swimming motility is a key bacterial trait, important to success in many niches. Biocontrol bacteria, such as Pseudomonas protegens Pf-5, are increasingly used in agriculture to control crop diseases, where motility is important for colonization of the plant rhizosphere. Swimming motility typically involves a suite of flagella and chemotaxis genes, but the specific gene set employed for both regulation and biogenesis can differ substantially between organisms. Here we used transposon-directed insertion site sequencing (TraDIS), a genome-wide approach, to identify 249 genes involved in P. protegens Pf-5 swimming motility. In addition to the expected flagella and chemotaxis, we also identified a suite of additional genes important for swimming, including genes related to peptidoglycan turnover, O-antigen biosynthesis, cell division, signal transduction, c-di-GMP turnover and phosphate transport, and 27 conserved hypothetical proteins. Gene knockout mutants and TraDIS data suggest that defects in the Pst phosphate transport system lead to enhanced swimming motility. Overall, this study expands our knowledge of pseudomonad motility and highlights the utility of a TraDIS-based approach for analysing the functions of thousands of genes. This work sets a foundation for understanding how swimming motility may be related to the inconsistency in biocontrol bacteria performance in the field.


InTRoduCTIon
The use of biocontrol bacteria as an agricultural tool to control crop diseases is growing, as they can provide many of the same benefits as fertilizers and pesticides without their accompanying damaging effects [1,2].Colonization of the plant rhizosphere is a crucial step in the provision of the beneficial effects of many biocontrol bacteria [3,4].One of the important factors affecting colonization success is motility of the beneficial bacteria.Successful colonization requires the movement of bacteria from the site of inoculation to the site of activity, and thus motility is an important factor in the field utility of biocontrol bacteria [5][6][7].
Bacteria use chemotactic systems combined with flagellated movement to detect and move towards more favourable environments and away from adverse conditions [7][8][9].Swimming is one of the main types of bacterial motility and is characterized by the movement of individual bacteria through an aqueous environment [9,10] powered by rotating flagella that can be triggered by environmental stimuli [11].Flagella are best studied in the peritrichous bacteria Escherichia coli and Salmonella enterica ser.Typhimurium, where more than 50 different proteins are involved [12][13][14], and flagellar synthesis and assembly are regulated via a three-tiered transcriptional cascade (reviewed in [15]).
Flagella structural genes are well conserved across Gram-negative bacteria, but there are substantial differences in the regulation of flagella synthesis and assembly between the enteric model systems and pseudomonads [15,16].Pseudomonad species and strains can have differing numbers of flagella, with well-studied strains possessing from one to seven flagella [17,18].The synthesis and assembly of the Pseudomonas aeruginosa PAO1 single polar flagellum is more complex than production of peritrichous flagella, involving ~60 genes and regulated by a four-tier hierarchy, consisting of different genes to those of E. coli [15,16].
Outside of well-known flagella and chemotaxis genes, there have been additional genes implicated in bacterial swimming motility [10].Using a P. aeruginosa PAO1 Mini-Tn5-lux mutant library, seven non-flagellar mutants were identified that had reduced swimming motility, including genes relating to nucleotide metabolism, RNA modification and central intermediary metabolism, and four hypothetical genes [19].A transposon insertion sequencing study of E. coli EC958 identified 14 non-flagellar genes and intergenic regions involved in enhanced swimming motility, showing that flagellar and chemotaxis genes are only part of the gene suite important for motility [2,20].Expanding motility research outside of flagella biosynthesis and chemotaxis genes will probably identify additional genes which influence swimming and therefore rhizosphere colonization and biocontrol activity.
The critical role of motility in rhizosphere colonization and biocontrol efficacy has been confirmed in multiple bacterial species [10].In Pseudomonas ogarae F113 (previously Pseudomonas fluorescens F113 [21]), hypermotile mutants are more competitive at rhizosphere colonization than the wild-type strain [4].P. fluorescens WCS365 mutants deficient in flagella-driven chemotaxis towards root exudates are poor colonizers of root tips in competition with other bacteria [6].In P. fluorescens NBC275, mutants showing a total loss of antifungal activity had reduced swimming motility [22].Swimming motility is vital for pseudomonads, and biocontrol bacteria more broadly, to be able to colonize the rhizosphere and effectively carry out biocontrol activities.
Pseudomonas protegens Pf-5 (hereafter referred to as Pf-5) is a plant growth-promoting bacterium which was originally isolated from the roots of cotton plants [23].Pf-5 produces a range of antibacterial and antifungal secondary metabolites [24] and can control pathogens of a range of crops [25][26][27][28].Work on Pf-5 has focused primarily on identifying genes involved in the regulation and production of secondary metabolites [1].Analysis of the Pf-5 genome has revealed many potential genes and molecular systems Pf-5 may utilize in its proliferation in soil and colonization of the rhizosphere [24,29].Swimming motility studies in Pf-5 have shown that mutations in the global regulators gacA and rpoS have no effect on swimming, a flagella biosynthesis mutant flhA has a deficiency in swimming, and knockouts of the polyurethanases pueA and pueB have no effect on swimming motility [30][31][32].
Here we used transposon-directed insertion site sequencing (TraDIS), a genome-wide approach, to identify genes involved in Pf-5 swimming motility.TraDIS is a powerful technique that combines high-density random transposon insertion mutagenesis and high-throughput sequencing to study gene fitness and link genotype and phenotype at a genome-wide scale [33,34].The data from our experiments provide insights into the suite of genes that are important for Pf-5 swimming motility beyond the classic set of swimming-related genes.

Transposon mutant library experiments
The role of genes in swimming motility was investigated using a previously constructed P. protegens Pf-5 transposon mutant library, as described by Fabian et al. [36], which contains ~256 000 unique transposon insertion sites spread evenly throughout the genome, equating to an average of ~45 transposon insertion sites per non-essential protein coding gene.Two microlites of the Pf-5 transposon mutant library (containing 3.2×10 7 c.f.u.) was stab-inoculated into the centre of six swim agar plates (modified KMB with 0.3 % Bacto agar).After 24 h at 22 °C, 4-5 mm of the agar containing the most motile cells was removed from the edge of the swimming zone (cells from half of the circumference from all six plates were pooled to form one replicate; Fig. 1).This process was repeated with two additional sets of six plates providing three replicates (total of 18 plates).For each of the three replicates, a volume of KMB was added, equivalent to the agar volume collected, and the mixture was vortexed for 10-20 s to homogenize

Impact Statement
Biocontrol bacteria, such as Pseudomonas protegens Pf-5, are increasingly being used as an agricultural tool to control crop diseases, and motility is a key factor in their successful colonization of plant surfaces.Here we use a high-throughput approach to identify the suite of genes important for swimming motility in P. protegens Pf-5.These included flagella and chemotaxis genes, as well as a variety of cell surface, cell division and signalling genes.We also show that defects in the Pst phosphate transporter lead to enhanced swimming motility, a hitherto unreported link between phosphate transport and swimming motility.Understanding the genetic basis of swimming motility enhances our knowledge of key processes in biocontrol bacteria that are needed to ensure their competitive success.This will contribute to developing strategies to increase the utility of biocontrol bacteria in agricultural settings to prevent crop losses.Fig. 1.Schematic of the experimental methodology used to identify Pseudomonas protegens Pf-5 genes important for swimming motility, indicating processes performed per replicate (three replicates generated for control and experimental treatments).The reads obtained from TraDIS sequencing Pf-5 cells from the edge of the swimming zone (treatment) are compared with TraDIS sequencing reads from Pf-5 cells grown on media with 1.5 % agar (control).In the final panel the gene indicated in orange is important for swimming motility as the number of transposon insertion reads for this gene is significantly lower in the swimming assay compared to the control (visualized using Artemis software [100]).
the mixture.For each of the three replicates the mixture was pelleted at 16 000 g for 5 min and excess agar removed (termed the 'output pool').The same number of plates per replicate were used to generate three control samples.For the controls, 166 µl of the Pf-5 transposon mutant library (containing 1.7×10 8 c.f.u.) was spread on six modified KMB 1.5 % agar plates for each replicate.After 24 h at 22 °C, the lawn was scraped off the agar surface of all six plates and pooled in PBS (termed the 'control pool').

TradIS sequencing and bioinformatic analysis
DNA was extracted from the agar/cell mixture from the output and control pools using a Wizard Genomic DNA Purification Kit (Promega).The manufacturer's protocol was followed with two modifications: centrifugation of the precipitated DNA was increased to 18 000 g for 5 min and to 14 000 r.p.m. for 10 min for the ethanol-washed DNA.Agar was removed from the DNA using the Wizard SV Gel and PCR Clean-Up System (Promega).Two of the three replicates were selected at random for TraDIS sequencing at the Ramaciotti Centre for Genomics (UNSW, Sydney, Australia), following standard practice for TraDIS sequencing as previously described in Barquist et al. [33] with an Illumina MiSeq platform to obtain 52 bp single-end genomic DNA reads.Reads were quality checked using FastQC (v0.11.5; [37]) and transposon insertion sites were mapped to the Pf-5 genome and analysed using the Bio-Tradis pipeline [34] as previously described [36].Briefly, this included allowing a 1 bp mismatch in the transposon tag, mapping reads with more than one mapping location to a random matching location, and excluding transposon insertions in the 3′ end of each gene.After matching the transposon tag, an average of 1.56 million reads per replicate were mapped to the Pf-5 genome (Table S1, available in the online version of this article).A linear regression of the gene insertion indexes of the replicates was completed in R [38].Correlation coefficients between the insertion indexes for all pairs of replicates were >0.92 (P<0.01;Fig. S1) which validates the reproducibility of our replicates and is consistent with the reproducibility of transposon insertion sequencing replicates in other studies [39].The transposon insertion sites in the output pools were compared with those of the control pool with the tradis_comparison.R script with default parameters.Only genes with >10 reads in both replicates of either the control or output pools were included to avoid genes being falsely classified as important for fitness.Genes with a log 2 -fold change of ≥2 in the number of transposon insertion reads in the output pool compared to the control pool and a q-value <0.01 were used for further analysis [39][40][41].
Clusters of Orthologous Groups (COG) assignments [42] for each Pf-5 gene were compiled using eggNOG-mapper [43] with 87.3 % of Pf-5 coding genes assigned a COG code [36].The sum of all categories does not equal the total number of genes of interest, as some genes are assigned multiple COG codes, and some no code.Orthologues of Pf-5 genes were identified using the Pseudomonas Genome Database available at https://pseudomonas.com [44].

Construction of gene knockout mutants
In-frame chromosomal gene deletion mutants were generated for Pf-5 genes pstS (PFL_6119), phoB (PFL_6108) and phoR (PFL_6109) via an overlap-extension PCR method, followed by allelic exchange with the suicide vector pEX18Tc [45], using a protocol adapted from Kidarsa et al. [46].To create the mutant allele, fragments of 500-1100 bp flanking upstream and downstream of each gene of interest were first PCR amplified using the upstream (UpF/UpR) or downstream (DnF/DnR) primer pair.All PCRs were performed using KOD Hot Start DNA polymerase (Novagen) according to Kidarsa et al. [47], using the primers listed in Table S2.The upstream forward primer and downstream reverse primer each had a 5′ extension adding an XbaI restriction site, while the upstream reverse and downstream forward primers were designed to be in-frame with the gene of interest and had a 5′ linker of 12 bp complementary to each other to allow overlapped annealing during the secondary PCR.The amplicons of the upstream and downstream primary PCRs were gel-purified, mixed 1 : 1 (50 ng each) and used as the template for the secondary PCR using the UpF and DnR primers.The resultant full-length product was gel-purified, digested with XbaI, treated with Calf Intestinal Alkaline Phosphatase (New England Biolabs), and then cloned into the pEX18Tc vector (linearized with XbaI) using T4 ligase (New England Biolabs).
The recombinant vectors were transformed by electroporation into ElectroMAX DH5α-E competent E. coli cells (ThermoFisher Scientific) and mutant alleles were verified using pEx18Tc sequencing primers (Table S3).All vectors were subsequently electrotransformed into mobilizing strain E. coli S17-1 competent cells [48].Biparental matings were performed between the vectorcontaining E. coli S17-1 and parental Pf-5 strain for conjugative transfer of each vector into Pf-5, as described in Lim et al. [49], but using nutrient agar containing 1.5 % (v/v) glycerol.Transconjugant Pf-5 colonies were selected on KMB agar containing 200 µg ml −1 tetracycline (vector conferred resistance) and 100 µg ml −1 streptomycin (innate resistance of Pf-5).Surviving colonies were grown without selection in LB broth for 3 h with shaking at 200 r.p.m. and plated on 10 % sucrose LB agar to resolve merodiploids (counter-selection against sacB-carrying cells).Sucrose-resistant colonies were patched in parallel onto LB agar containing 10 % sucrose and KMB agar with 200 µg ml −1 tetracycline to further confirm the absence of the pEX18Tc vector backbone.Tetracyclinesensitive colonies were screened by PCR with primers annealing to chromosomal regions external to the target gene (Table S3) to detect mutants with truncated amplicon sizes compared to a parental strain control.The deletion of each gene was confirmed by PCR and sequencing of genomic DNA from each Pf-5 mutant colony before being stored in 25 % glycerol at −80 °C until required.
Growth curves were conducted to check for general growth defects in knockout mutants.Overnight cultures of wild-type Pf-5 and the mutant strains ∆pstS, ∆phoB and ∆phoR were each grown at 27 °C with shaking at 200 r.p.m. for 16 h in modified KMB.The cultures were sub-cultured 1 : 25 into fresh modified KMB and incubated with shaking at 200 r.p.m. until reaching an OD 600 of 0.6.Cultures were kept on ice while serial dilutions were performed in modified KMB to reach a final density of 1.2×10 4 c.f.u.ml −1 in 150 µl modified KMB in a 96-well plate (four replicates per strain).Plates were incubated in a Pherastar plate reader at 27 °C with shaking at 200 r.p.m. and OD 600 readings were taken at 6 min intervals for 24 h.

Phenotypic assays with knockout mutants
Swimming assays were performed on a modified BM2 minimal medium [0.5 % (w/v) casamino acids, 2 mM MgSO 4 , 10 µM FeSO 4 , 0.4 % (w/v) glucose] supplemented with potassium phosphate buffer (0.05 or 6.6 mM; pH 7) [19] and 0.3 % Bacto agar.To prepare the inoculum, cells of the Pf-5 parental strain and each mutant were scraped from a culture grown on KMB agar for 48 h, then resuspended in water and diluted to an OD 600 of 0.2.Two microlitres of the cell suspension was stab-inoculated into the centre of swim agar plates (n=3-4).All cultures were incubated at 22 °C for 20 h and the diameter of the motile zones were measured.We analysed the data using an unbalanced two-way ANOVA (type III) with Tukey's HSD post-hoc analysis using the R packages car [49] and agricolae [50].
A droplet collapse assay was used to test for surfactant activity of the Pf-5 knockout mutants [51].Cultures were grown for 24 h at 27 °C on modified BM2 agar plates (as above).The cultures were scraped from the plates and suspended in water to a final density of 1×10 10 c.f.u.ml −1 and 10 µl droplets were spotted on parafilm in triplicate.A flat droplet indicated the cells produced the surfactant orfamide A, while domed droplets indicated the cells did not produce orfamide A [52].

data availability statement
Sequence data from TraDIS sequencing are available from the European Nucleotide Archive.Sequence files for the project are available under the project accession number PRJEB56281, with the control sequencing reads under sample accession numbers ERS13490282 and ERS13490283 and swimming sequencing reads under sample accession numbers ERS13490284 and ERS13490285.

Identifying genes contributing to Pf-5 swimming fitness
We grew a P. protegens Pf-5 saturated transposon mutant library in swimming agar (0.3 %) for 24 h at 22 °C and collected the cells at the edge of the motile zone (output pool).DNA from these cells was subjected to TraDIS sequencing and the read frequency for each gene was compared with that of cells from the Pf-5 transposon mutant library grown on 1.5 % agar (control pool).These swimming assays identified a set of 249 genes for which loss of function affected Pf-5 swimming fitness (4.40 % of non-essential genes [36]; Figs 2 and S2).Loss of function of 189 of these genes was strongly detrimental for swimming fitness, based on cells with mutations in these genes being present at significantly lower levels in the output pool compared to the control (log 2 -fold change <−2).Loss of function of the remaining 60 genes was beneficial for swimming fitness, with cells carrying mutations in these genes present at significantly higher levels in the output pool compared to the control (log 2 -fold change >2).The fold change values for the full set of Pf-5 genes are available in Dataset S1.

Functional analysis of genes contributing to swimming fitness
A functional overview of the Pf-5 genes that affected motility fitness was obtained by classifying the genes using COG categories [42].As expected, genes annotated with the functional categories of cell motility (N) and signal transduction (T) made up a large proportion of the genes that detrimentally affected swimming fitness when their function was lost (30.7 %; Fig. 3).The functional loss of a large proportion of the genes in the following categories also detrimentally affected swimming motility: cell wall/membrane/envelope biogenesis (M); post-translational modification, protein turnover and chaperones (O); energy production and conversion (C); replication, recombination and repair (L); and unknown function (S; Fig. 3).From the set of genes that enhanced swimming fitness when their function was lost, the highest proportion are involved in signal transduction (T), followed by translation, ribosomal structure and biogenesis (J), inorganic ion transport and metabolism (P), and function unknown (S; Fig. 3).Just over 20 % of this gene set have no COG code assigned, including pseudogenes, tRNAs, rRNAs and conserved hypothetical proteins.

Swimming motility requires a large number of flagellar and chemotaxis genes
Of 48 genes related to flagella structure, biosynthesis, assembly and regulation, which are mainly located in two regions of the Pf-5 genome, our TraDIS analysis identified 41 as being significantly important for swimming motility (Fig. 4).In addition, loss of eight chemotaxis genes affected Pf-5 swimming fitness (Fig. 4).The direction of the fold changes for the flagella genes and most of the chemotaxis genes are consistent with the known roles of orthologous proteins in P. aeruginosa.In most cases gene loss is detrimental to swimming fitness, but one notable exception is the regulator gene PFL_4484 that encodes the anti-sigma-28 factor FlgM. Loss of function mutations in this gene strongly enhanced Pf-5 swimming motility.During flagellar synthesis in P. aeruginosa, FlgM negatively regulates flagella gene expression by binding sigma-28 factor FliA [15,53].When FlgM activity is lost following anti-anti-sigma factor HsbA binding, this allows FliA-regulated genes to be transcribed and motility improves [54].These past observations in P. aeruginosa are consistent with the enhanced Pf-5 swimming fitness seen in this study when FlgM function is lost.
The majority of chemotaxis genes also showed phenotypes consistent with previously reported gene functions.The chemosensory genes of Pf-5 are organized in a similar way to the five P. aeruginosa PAO1 gene clusters that encode chemosensory signalling proteins, except there are no cluster II orthologues, only some orthologues of cluster IV genes are present and chpA is truncated [55,56].Loss of most of the Pf-5 genes orthologous to those in P. aeruginosa clusters I and V (PFL_1668-PFL_1677 and PFL_4481-PFL_4482, respectively) detrimentally affected swimming (Fig. 4).In P. aeruginosa the Che pathway, comprising cluster I and V genes, is essential for chemotaxis, which is consistent with the results of this study [57].
While most genes showed expected phenotypes, TraDIS data indicated genes cheZ and PFL_4482 had unexpected effects on swimming fitness.For cheZ (cluster I), loss of function was beneficial for Pf-5 swimming.By contrast, in E. coli, loss of cheZ function results in a state of tumbling and random movement [58,59].Loss of function of PFL_4482 (orthologue of cheV in cluster V) was also beneficial for swimming.The exact role of CheV in bacterial cells is not well understood, but it possibly acts as a phosphate sink in enterobacteria by competing with CheY for phosphorylation by CheA to offset chemoreceptor overstimulation [60].
In addition, some of the methyl-accepting chemotaxis (MCP) encoding genes also had phenotypes of interest.In Pf-5 there are 42 genes that encode known chemosensory receptors, comprising the named genes pilJ, pctC, pctA_1, bdlA, aer_1, aer_2, aer_3 and pctA_2, and a further 34 which have no specific assigned function and are annotated only as MCPs.MCPs are critical components of chemosensory signalling systems, but the majority are not located in a cluster or pathway [55].Three of the Pf-5 genes encoding MCPs influenced swimming fitness when disrupted: loss of function of PFL_0778 and pctA_2 was detrimental to swimming fitness, whereas loss of PFL_5046 function was beneficial for swimming fitness.The chemoattractants of these three genes remain unknown.Loss of function of aer_2, which encodes an aerotaxis receptor, detrimentally affected Pf-5 swimming when disrupted.This gene is homologous to the aer gene in P. aeruginosa PAO1 that encodes the most prevalent aerotaxis receptor in pseudomonads [61].Consistent with our results, loss of the gene encoding the main aerotaxis receptor in the biocontrol bacteria Pseudomonas chlororaphis PCL1606 detrimentally affected swimming motility [62].

defects in cell envelope and cell division genes are detrimental to swimming motility
There are 43 cell envelope and cell division genes that detrimentally affected Pf-5 swimming fitness when their function was lost (Fig. 5a, b).Mutations in these genes are likely to affect cell shape or cell surface composition [63][64][65].Swimming motility is based on cells sensing environmental cues and switching between 'runs' and tumbling movement to move up or down a chemical gradient [6].Changes in cell morphology have the potential to impact on the drag and other forces experienced by the cell and therefore their swimming fitness [66].For example, in E. coli AW405 changes in cell length are linked to changes in swimming patterns, and E. coli KR0401 rodZ mutants defective in peptidoglycan synthesis are spherical (instead of the normal rod shape) and non-motile [67][68][69].
Nineteen of the 43 cell envelope genes have functions in lipopolysaccharide (LPS) biosynthesis, including genes associated with LPS core biosynthesis (PFL_0518, PFL_0519, PFL_0521, wapB, galU, wzt) as well as those related to O-antigen biosynthesis (wbpL, rfbG, rfbH).Mutations causing defects in LPS are likely to affect cell surface structures and have previously been shown to affect motility in multiple bacterial species, including P. aeruginosa PAO1 and Pseudomonas syringae pathovar glycinea [70,71].Impacts of LPS defects on swimming motility have been shown to be due to changes in flagella and/or changes in cell adhesive properties.In E. coli and S. enterica serovar Typhimurium the expression of flagellar genes is impaired by cell envelope stress, specifically the truncation of LPS [72].In P. aeruginosa PAO1 mutants with non-wild-type LPS have reduced swimming motility due to changes in cell-cell and cell-substrate adhesion forces [73].
Ten of the 43 cell envelope-related genes are involved in peptidoglycan turnover (Fig. 5a).This includes genes from each stage of peptidoglycan turnover, from synthesis of muropeptides (ddlB and murF), to incorporating new muropeptides into existing peptidoglycan (mrcA, PFL_3300, PFL_5449 and PFL_5608) and modifying/degrading the peptidoglycan (dacA, mltF, PFL_0563 and PFL_4436).The loss of function of four of these genes had very strong negative effects on swimming fitness (log 2 -fold change <−4), indicating that the correct assembly of the precursor molecule d-alanyl-d-alanine by DdlB and MurF, and insertion of new muropeptides into the cell wall are critical for swimming fitness.In P. aeruginosa PAO1 and E. coli, the proteins encoded by dacA and PFL_5449 (RlpA family lipoprotein) have been shown to be important for maintaining cell shape [74,75].It is therefore likely that the effect of the loss of peptidoglycan turnover-related genes on Pf-5 swimming fitness is through alterations to typical cell shape.
Thirteen cell division-related genes detrimentally affected swimming fitness, with 10 having a strong negative effect (log 2 -fold change <−4; Fig. 5b).These genes include those involved in localizing the cell division machinery to the centre of the cell (minC and minD) and breaking down peptidoglycan for septal splitting (nlpD) along with genes involved in segregation of origin of replication domains (smc, scpA), chromosome or plasmid molecules (parA, parB), and the chromosome terminus (ftsK).Loss of function of cell division genes probably results in abnormal cell morphology.For example, minC and minD mutants in the plant pathogen Xanthomonas oryzae pv.oryzae form short filaments and minicells and a minC knockout was not able to swim [76].P. aeruginosa PAO1161 parA and parB mutants also have swimming defects [77,78].

Swimming motility is affected by defects in signal transduction genes
Loss of function of 14 signal transduction genes affected swimming fitness, including the sensor kinase gacS which is part of the global regulatory GacSA signal transduction system in pseudomonads (Fig. 5c; [30]).In this study loss of gacS function was detrimental to Pf-5 swimming motility.In other pseudomonads the effect of gacS mutation on swimming motility has been shown to be varied.For example, gacS mutants of P. aeruginosa PA14, P. chlororaphis O6 and P. ogarae F113 showed enhanced swimming compared with the wild-type [79][80][81][82].In contrast, a gacS mutant of P. fluorescens NBC275 had decreased motility compared with the wild-type strain [22].
The loss of function of five genes involved in c-di-GMP turnover affected Pf-5 swimming fitness.Turnover of the second messenger c-di-GMP controls a range of cellular processes, including motility, and a low concentration of intracellular c-di-GMP is associated with motile cells [83].The loss of PFL_4827, an orthologue of bifA that encodes a phosphodiesterase in P. aeruginosa PAO1, was detrimental to Pf-5 swimming.Phosphodiesterases are part of the c-di-GMP regulation system in pseudomonads and BifA has been shown to degrade c-di-GMP in P. aeruginosa PA14 [84].In P. ogarae F113 and Pseudomonas putida KT2440, bifA knockouts had reduced c-di-GMP breakdown, causing higher c-di-GMP levels and a reduction in swimming motility [85,86].The loss of function of PFL_0190 and PFL_0507 also had detrimental effects on Pf-5 swimming motility.These genes are annotated as diguanylate cyclase proteins, but in P. aeruginosa PAO1 the orthologue of PFL_0507, dipA, functions as a phosphodiesterase and knocking it out resulted in reduced motility [87].PFL_0190 and PFL_0507 contain EAL domains, further evidence that the proteins they encode probably function as phosphodiesterases [88].Enhanced Pf-5 swimming fitness is seen with the loss of function of PFL_0675 and PFL_5686 which encode diguanylate cyclase proteins that synthesize c-di-GMP.PFL_0675 is an orthologue of P. aeruginosa PAO1 gcbA which encodes a diguanylate cyclase.In a separate Pf-5 study, a PFL_5686 mutant showed wild-type swimming motility [89].In P. aeruginosa PAO1 and P. fluorescens Pf0-1, gcbA knockouts had increased swimming motility compared to the wild-type strain [90,91].
Loss of function of two signal transduction genes enhanced Pf-5 swimming fitness (PFL_0554 and PFL_5871).The proteins encoded by these genes are both annotated as HD-related output domain (HDOD)-containing proteins.The HDOD is a domain of unknown function, but it is suggested to have a role in regulation and signalling [92].In the plant pathogen Xanthomomonas campestris pv.Campestris, loss of the HDOD-containing protein GsmR results in reduced swimming motility via changes to flagellum synthesis [92].Similarly, mutants of Campylobacter jejuni gene CJ0248, which contains an HDOD, showed an altered motility phenotype with less migration of cells [93].In contrast, a mutant of the HDOD-containing protein sadB in P. ogarae F113 had enhanced swimming motility [82], consistent with the enhancement of Pf-5 swimming when the functions of PFL_0554 and PFL_5871 were lost.

Swimming motility also requires conserved genes of unknown function
Loss of function of 24 genes encoding conserved hypothetical proteins affected Pf-5 swimming fitness (Dataset S1), with six of these genes enhancing swimming fitness, and the other 18 genes negatively affecting swimming fitness.Loss of function of 13 genes encoding conserved hypothetical proteins had a very strong effect on swimming fitness (log 2 -fold change >4 or <−4) so characterization of these genes is of particular interest and may provide further insights into swimming motility in this plantassociated bacterium.
Genomic context and orthologue function analysis suggested possible functions for five hypothetical genes.PFL_1677 is in an operon with two CheW-like genes (PFL_1675-1677), suggesting a possible role in chemotaxis, and loss of function of all three ).An unbalanced two-way ANOVA with Tukey's HSD post-hoc analysis was conducted in R [38].*Significant difference at α=0.05. Figure generated using the R packages ggplot2 [101] and ggsignif [102].
genes in the operon detrimentally affects swimming fitness.A further two are probably cell envelope genes (PFL_5098, PFL_5485).PFL_5098 is part of a cluster of 11 O-antigen biosynthesis genes (PFL_5093-5103) that all have detrimental effects on swimming fitness when lost.PFL_5485 is in an operon (PFL_5485-5488) which also contains genes encoding a GtrA family protein and a glycosytransferase.In Pseudomonas donghuensis HYS GtrA is involved in the glucosylation of LPS O-antigen [94].There is also evidence that PFL_1916 may be associated with energy transduction as it is in a two-gene operon with ccoG_1 which encodes a cytochrome oxidase assembly factor.
Conserved hypothetical proteins comprise the full operon PFL_2736-2739 and the majority of operon PFL_5239-5241.In each of these operons the loss of each of the members were detrimental to Pf-5 swimming fitness, but genomic context and orthologue function analysis have not shed any light on their potential functions.Now that we have connected these genes with swimming motility, this may assist in future functional characterization.

A role for phosphate transport in Pf-5 swimming fitness
A range of genes involved in phosphate utilization showed differential effects on swimming fitness.Loss of function of pitA_1, encoding a low-affinity inorganic phosphate transporter, and ppx, encoding exopolyphosphatase, were detrimental to swimming fitness (Fig. 5d).In P. aeruginosa PAO1, ppx encodes an exopolyphosphatase which cleaves P i residues from accumulated polyP chains.In P. aeruginosa PAO1, a ppx mutant had reduced swimming motility, consistent with the Pf-5 TraDIS data [95].
Interestingly, swimming fitness increased when the genes encoding the Pst high-affinity inorganic phosphate transporter (pstBACS) were lost, with pstS showing the second greatest fold increase across the genome (Fig. 5d).In P. aeruginosa this transporter is active in low-phosphate conditions and is part of a wider system of sensing and responding to environmental inorganic phosphate conducted by the Pho regulon [96], but has not previously been connected to swimming.We made gene knockouts in the transporter gene pstS, and the regulatory genes phoB and phoR (not significantly affected in the TraDIS swimming dataset) to investigate the role of the Pho regulon in swimming.We conducted swimming assays with the three mutants and the parental strain under phosphate-replete (6.6 mM) and phosphate-depleted conditions (0.05 mM).Under both phosphate levels ΔpstS had a significantly larger swimming zone diameter than the parental strain whereas ΔphoB and ΔphoR showed the same swimming phenotype as the parental strain (F (9,39) = 2.32, P=0.03; Figs 6 and S3) consistent with our TraDIS results.To confirm increased swimming motility was not due to overall growth improvements in these mutants, we conducted growth assays.Mutants of phoB and phoR showed identical growth curves to the parental strain, but the pstS mutant had slower growth in liquid culture compared to the parental strain, indicating swimming improvements were not due to an enhanced growth rate (Fig. S4).A droplet collapse assay was conducted and showed that production of the biosurfactant orfamide A was identical between the mutants and parental strain [51].The biosurfactant orfamide A has previously been shown to be involved in swarming motility in Pf-5 [97].Overall, both the gene knockouts and TraDIS data are consistent with PstS playing an important role in swimming motility in Pf-5, whereas the regulators PhoB and PhoR do not affect swimming motility.Expression of pstS is highly induced under phosphate-depleted conditions in P. aeruginosa PAO1 [98] so it is interesting that in Pf-5 PstS appears similarly important for swimming motility in both phosphate-replete and phosphate-depleted conditions.In P. aeruginosa there is a link between phosphate, the Pho regulon and swarming motility [98,99], but there are no reports of links with swimming.

ConCLuSIonS
In this study TraDIS successfully identified 249 genes involved in swimming motility of P. protegens Pf-5.This method identified both genes that are known to contribute to swimming motility as well as genes not previously associated with this phenotype.Successful swimming motility required the function of chemotaxis genes, genes involved in flagella structure, assembly and regulation, and signal transduction genes involved in c-di-GMP turnover.We also identified additional genes for which the loss of function negatively affected swimming motility, including cell division, peptidoglycan turnover and LPS biosynthesis genes, with effects on swimming probably due to aberrations in cell morphology.Of particular interest is our finding that phosphate transport plays a role in swimming motility.Both TraDIS and specific gene knockout results link phosphate transport by the Pit and Pst transporters with swimming motility in Pf-5, a connection not previously reported.Phosphate is an essential nutrient for cell function, and as phosphate is often limited in agricultural soils the link between phosphate transport and motility in biocontrol bacteria could be crucial for enabling biocontrol activity and preventing crop disease.
Overall, this study expands our knowledge of pseudomonad motility and highlights the utility of a TraDIS-based approach for systematically analysing thousands of genes and identifying the extended suite of genes affecting the fitness of the plant growthpromoting rhizobacteria Pf-5 during swimming.Understanding the genetic basis of swimming motility enhances our knowledge of key processes in biocontrol bacteria that are needed to ensure their competitive success.This knowledge will help address the issue of inconsistency in biocontrol bacteria effectiveness and reliability in the field and contribute to developing strategies to increase use of biocontrol bacteria in agricultural settings to prevent crop losses.

Fig. 2 .
Fig. 2. Genome position of P. protegens Pf-5 genes identified by TraDIS with log 2 -fold change >2 or <−2 from swimming assays when compared with the control.Genes with a log 2 -fold change of <−2 indicate that loss of their function was associated with reduced swimming fitness and >2 indicate that loss of their function was associated with enhanced swimming fitness.Scatterplot generated using the R package ggplot2 [101].

Fig. 4 .
Fig.4.Log 2 -fold change of P. protegens Pf-5 flagella and chemotaxis cluster I and V genes during swimming (not including genes that encode methylaccepting chemotaxis proteins).Genes are clustered in two main regions of the genome, indicated as regions 1 and 2. Dotted vertical lines are positioned at log 2 -fold changes of −2 and 2. Gene names from P. aeruginosa orthologues are in parentheses.*Genes with non-significant log 2 -fold change.Data visualized using the R package ggplot2[101].

Fig. 5 .
Fig. 5. Log 2 -fold change of P. protegens Pf-5 genes related to (a) cell envelope, (b) cell division, (c) signal transduction and (d) phosphate turnover during swimming.Dotted vertical lines are positioned at log 2 -fold changes of −2 and 2. All genes have significant log 2 -fold change.Data visualized using the R package ggplot2 [101].