A large-scale genetic screen identifies genes essential for motility in Agrobacterium fabrum

The genetic and molecular basis of flagellar motility has been investigated for several decades, with innovative research strategies propelling advances at a steady pace. Furthermore, as the phenomenon is examined in diverse bacteria, new taxon-specific regulatory and structural features are being elucidated. Motility is also a straightforward bacterial phenotype that can allow undergraduate researchers to explore the palette of molecular genetic tools available to microbiologists. This study, driven primarily by undergraduate researchers, evaluated hundreds of flagellar motility mutants in the Gram-negative plant-associated bacterium Agrobacterium fabrum. The nearly saturating screen implicates a total of 37 genes in flagellar biosynthesis, including genes of previously unknown function.


Introduction
Flagellar motility is widespread in Gram-positive and Gram-negative bacteria, with motility systems in the Gram-negative enteric species Escherichia coli and Salmonella enterica being particularly well characterized. The gram-negative flagellum consists of an envelope-integrated basal body that drives rotary motion of an external hook-filament complex. The basal body includes a hollow rod connected to a series of rings embedded in each of the three envelope layers (inner membrane, peptidoglycan, and outer membrane). The rotary motion of the rod is driven by the flagellar motor, which consists of the rotor component along with stator modules that are mechanically stabilized by their association with the peptidoglycan wall. Flagellar motion is powered by the movement of protons through the stator modules and this movement is translated to locomotion by connection of the flagellar rod to a flexible extracellular hook that in turn connects to a propulsive filament. The spinning filament generates thrust by either clockwise or counterclockwise rotation (depending on the organism), and directionality of cell movement is ultimately controlled by a discontinuous pattern of "runs" (propulsive a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 motile starting strains BB01, YS01, BM01, IW01, CI01, and BL01. Donor E. coli strain DH5ɑ/ pAB181 (D223) was used to deliver a mini-Tn5 transposon for mutagenesis, with helper strain DH5ɑ/pRK600 (B001) [23]. Routine culture of A. fabrum and E. coli was carried out in Luria broth (LB) containing (per liter) 10 g tryptone, 5 g yeast extract, 5 g NaCl and 1 ml of 2N NaOH, with 12 g of agar added to solidify when appropriate. Cultures were incubated at 37˚C (E. coli) or 30˚C (A. fabrum). Motility agar contained (per liter) 5 g tryptone, 2.5 g yeast extract, 0.5 g CaCl 2 and 2 g of agar. For motility testing, cells were taken with wooden toothpicks, stabbed into motility agar, and allowed to incubate at 30˚C for 2-3 days. For library enrichment (described below), motility agar was made instead with 1.7 g/L agar. Where appropriate, antibiotics were used as follows: streptomycin (Sm), 200 μg/ml; ampicillin (Ap), 100 μg/ml; chloramphenicol (Cm), 30 μg/ml; kanamycin (Km), 30 μg/ml; and neomycin (Nm), 100 μg/ml.

Transposon mutagenesis and screening for motility defects
Bacterial conjugation by triparental mating was carried out by first growing the six initial Sm-resistant A. fabrum strains (the recipients), and donor and helper E. coli strains (for plasmids and strains details see S1 and S2 Tables) separately as patches on LB-agar with appropriate antibiotics. Cells were collected with toothpicks and suspended in liquid LB to equivalent levels of turbidity. Six matings (one for each recipient) were set up by combining 70 μl of each suspension, plating cell mixtures on LB-agar, and allowing overnight incubation. Resulting lawns were collected by suspending in LB containing 15% glycerol. Aliquots were stored at -80˚C. Transposants were selected by plating mating suspensions on LB-agar containing Sm and Nm. From each of the six matings, approximately 2 x 10 5 mutant colonies were selected. The selected libraries were collected in LB containing 15% glycerol and stored frozen as before.
For enrichment of non-motile A. fabrum mutants from each transposon library, 1 μl of suspension (containing approximately 1 x 10 7 cells) was stabbed into 0.17% motility agar and allowed to incubate for 72 hours. At this point a clean toothpick was inserted into the original stab site to pick up cells that had not migrated into the motility agar. These cells (approximately 1 x 10 6 per toothpick) were grown to saturation in LB-Sm/Nm, then combined with glycerol and stored as frozen aliquots for subsequent screening. For screening, the transposon libraries (enriched for non-motile mutants) were plated to single colonies on LB-Sm/Nm, and colonies were stabbed into motility agar one by one, with approximately 60 stabs per 100-mm plate. Non-motile clones were extracted and restreaked for further analysis to retest the phenotype and identify transposon insertion sites.

Arbitrarily primed PCR and Sanger sequencing
In the first round of Arbitrary (Arb) PCR, 1.5 μl of lysed, boiled cells from each strain was added to 15.4 μl of water, 2 μl of Taq buffer, 0.5 μl 10 mM dNTP, 0.15 μl Taq, 0.15 μl of 100 μM of forward primer (2100), and 0.3 μl of 100 μM reverse primer (2102 or 2103). PCR was carried out under the following conditions: after initial denaturation (94˚C for 1 min), cycling 6 times (94˚C for 15 sec, 33˚C for 45 sec, 70˚C for 45 sec), and cycling 30 times (94˚C for 15 sec, 43˚C for 30 sec, 70˚C for 45 sec). In the second round of Arb-PCR, 0.9 μl of the amplified DNA from the first round of PCR was added to 16.1 μl of water, 2 μl of Taq buffer, 0.5 μl of 10 mM dNTP, 0.15 μl Taq, 0.15 μl of 100 μM forward primer (2101) and 0.15 μl of 100 μM reverse primer 2104. The second round of PCR was carried out under the following conditions: after initial denaturation (94˚C for 1 min), cycling 30 times (94˚C for 15 sec, 55˚C for 15 sec, 70˚C for 45 sec). Sanger Sequencing was carried out on the amplified PCR products to identify transposon insertion sites. The first-round arbitrary primer 2102 was normally used, but in cases of unacceptable product or low-quality sequence for a given mutant, the alternative primer 2103 was used (see S3 Table for primer  sequences).
From the data obtained from Sanger Sequencing, the . . .GAGACAG sequence at the end of the mini-transposon was located and the 30 nucleotides following this were used to find the position in the A. fabrum genome, using BLASTN (in GenBank accession AE007869.2) [24]. In each case, the transposon insertion location, directionality, and identity of the disrupted gene was noted (given in S1 and S2 Files). Where appropriate, the corresponding protein sequences were analyzed bioinformatically using BLASTP, signalP, and Pfam [25,26]. A manual annotation of the A. fabrum motility gene cluster was also carried out using BLASTP (given in S3 File) [24].

Construction of strains with in-frame deletions
Plasmids that contained homology regions for each target gene were created. Homology regions were designed to contain 300bp of DNA on each side of the gene of interest, maintaining the first and last several codons for that gene. Sequences for each target gene were retrieved from GenBank accession number AE007869.2. Parent plasmid pJG1108 containing XbaI-SalI-gus-sacB-kanR was digested with XbaI and SalI. Inserts were amplified from BB01 DNA using primers listed in S3 Table and prepared for 3-way ligations in which the joint between right and left homology regions is the 6-base sequence CCCGGG (XmaI, encoding Pro-Gly). Ligation products were transformed into E. coli DH5α and sequence verified. Deletion plasmids were transferred into A. fabrum strain BB01 by triparental mating, as described above for transposon mutagenesis and transconjugant clones were selected on LB-SmNm. Four individual colonies from each mating were carried over to selection on LB containing sucrose and X-Gluc (5-Bromo-4-chloro-3-indoxyl-beta-D-glucuronide cyclohexylammonium salt) (100 μg/mL). After 72 hours, cells from white colonies were suspended in PCR lysis buffer (5mM Tris pH 8.0, 2mM EDTA, 0.5% Triton X-100) and heated to 95˚C for 5 minutes. Confirmatory PCR was carried out with primers listed in S3 Table. Successful deletion was verified by band down-shift on an agarose gel. Motility tests for deletion strains were carried out as described above.

Construction of plasmids for complementing knock-out strains
For complementing mutants ΔATU0568, ΔATU0583, ΔflgN, and ΔmotF in A. fabrum, plasmid pKJ056 was used (S1A Fig). This plasmid allows expression of downstream genes by readthrough transcription of the kanamycin-resistance (kanR) gene. All of the genes to be complemented were amplified from the A. fabrum C58 genome with their respective primers (S3 Table). pKJ056 was digested with EcoRI and BamHI as were all the amplified products and then ligated together. Ligated plasmids were transformed into E. coli DH5α and sequence verified. For visNR complementation experiments, three replicative plasmids were created that contain visN, visR, or visNR under the control of the P vis promoter. Parent plasmid pPG012 containing BamHI-XbaI-kanR was digested with BamHI and XbaI (S1B Fig). Inserts were amplified from strain C237 with primers listed in S3 Table. Ligation products were transformed into DH5α and sequence verified. All complementation plasmids were transferred into their respective knock-out strains by triparental mating as described above for transposon mutagenesis. Motility tests were carried out as described above. Complete DNA sequences of parent plasmids are provided in S4 File.

An enrichment-aided screen for motility mutants in A. fabrum
A screen for motility mutants in A. fabrum was carried out by first mutagenizing Sm-resistant derivatives of the plasmid-cured strain UBAPF2 [22]. This strain imposes reduced environmental hazard for use in an undergraduate lab, as it is unable to infect plants, and its genome is somewhat reduced without loss of motility, increasing the probability of finding non-motile mutants in a forward genetic screen. Mutagenesis by triparental mating with the donor strain DH5ɑ/pAB181 (S2A and S2B Fig) resulted in over 10 6 Sm/Nm-resistant transposants as a starting population for screening. Given the proportion of the genome expected to control flagellar motility (around 0.8%, or 40-50 genes), we anticipated the need to assay approximately 50,000 transposants in order to approach saturation. Rather than use this approach, we preenriched the mutant population for non-motile cells S2C Fig. This was done by first injecting motility agar with the mutant population and then allowing cells to swim away from the site of inoculation for several days. The sub-population remaining at the site of injection was then recovered in a manner that avoided "bottlenecking" effects (see Methods). This enrichment procedure was carried out across multiple plates from six independent transposon libraries to ensure the maintenance of genetic diversity in the enriched libraries.
Libraries enriched for non-motile mutants were plated to single colonies, and these were screened colony-by-colony for motility defects. Only strongly non-motile mutants were carried forward in our analysis. With the enrichment strategy described above, over 25% of colonies assayed were strongly defective in motility (S2 Fig). Around 500 such mutants were streaked to isolation and retested for motility. Of these, 360 confirmed mutants were used to map transposon insertion sites by arbitrarily primed PCR. This analysis revealed 314 unique insertions possibly associated with the motility defect. Most of these (301 insertions) were confined to a known cluster of flagellar motility genes, comprising nucleotides (512,099-569,457) in the A. fabrum genome (GenBank Accession AE007869.2, and S1 File). The remaining 13 insertions were distributed across the genome (S2 File). Remarkably, genes disrupted outside the motility cluster were always represented by only a single insertion, while genes disrupted inside the motility cluster (Fig 1) were always represented by at least two independent insertions (an average of 8 insertions per gene). Therefore, we have focused our follow-up analysis on genes within the cluster.

A 'parts list' for A. fabrum flagellar motility that includes previously undescribed genes
Genes hit in our screen represent 37 proteins: 33 with known or suspected functions. At the time this screen was carried out, four of these proteins were of unknown function. Most of these proteins map to specific components of a model Gram-negative bacterial flagellum, as depicted in Fig 2. These components include the flagellar secretion apparatus (FliP, FliI, FlhB, FliQ, FlhA, FliR), the cytoplasmically localized C ring (FliM, FliN, FliG), the inner membrane-localized MS ring (FliF), the proximal rod junction (FliE), the proximal rod (FlgB, FlgC, FlgF), the P ring (FlgI), the L ring (FlgH), the distal rod (FlgG), core stator proteins (MotA, MotB), stator-associated proteins (MotC, MotE, FliL), hook (FlgE), hook-filament junction (FlgK, FlgL), and filament (FlaA). Other genes disrupted in the screen encode proteins that may play transient roles in directing flagellar assembly, including the rod capping protein (FlgJ), the hook capping protein (FlgD), the hook length regulator (FliK), and the P ring assembly chaperone (FlgA). Disrupted genes with transcriptional or translational regulatory functions include the Class IB transcriptional regulator Rem, and the functionally coupled translational regulators FlbT and FlaF.
At the time this screen and analysis were carried out, four genes were of unknown function, and they were subjected to further analysis. These genes were designated ATU0568, ATU0583, ATU0585, and ATU8132 (according to the naming system in GenBank accession AE007869.2, and see Fig 1). A BLASTP search revealed that ATU0568 contains a DUF4231 domain from the SLATT superfamily, which contains a pair of N-terminal transmembrane helices and a helical C-terminal cytoplasmic region [27]. A Pfam search of ATU0568 did not yield any homology or domain similarities to previously characterized proteins [26]. Bioinformatic analysis for ATU0583 suggested that this protein does not contain a signal peptide, nor does it contain any domain similarities to known proteins. A BLASTP of ATU0585 did not yield any regions of local similarities to previously characterized proteins. ATU0585 is not predicted to

PLOS ONE
have a signal peptide according to signalP, and Pfam analysis of this protein did not yield any domain similarities to known proteins [25,26]. A BLASTP search of ATU8132 indicated that this protein shares homology with a FliL superfamily domain in Agrobacterium fabrum. Sig-nalP predicted a potential signal peptide in ATU8132 suggesting that this protein is not localized in the cytoplasm, but it could be membrane localized or secreted. Pfam analysis revealed that the N-terminal residues of ATU8132 (aa 4-12) are predicted to be a hydrophobic region of a signal peptide, while aa 18-177 contain a membrane-bound region and reside on the outside of the membrane (in the periplasm or extracellular region).
A recent study by Sobe et al. [28] revealed a four-gene cluster that is required for motility in Sinorhizobium meliloti (S. meliloti) strain RU11/001. This cluster in S. meliloti (SMc03056-SMc03071-SMc03072-SMc03057) shows high synteny to an analogous region in A. fabrum shown in this study (ATU0583-flgJ-ATU0585-ATU8132) (Fig 3). Their study reveals that absence of the S. meliloti orthologs of ATU0583, flgJ, and ATU0585 result in the absence of FlaA flagellin production. This is consistent with these three genes being involved at an earlier stage of flagellar biosynthesis [9]. By homology, this suggests that ATU0583, FlgJ (A. fabrum), and ATU0585 also function prior to flagellin secretion. A multiple sequence alignment of ATU0585, Smc03072, and FlgN (Salmonella enterica) reveals several identical amino acids justifying renaming ATU0585 as FlgN. The S. meliloti study also revealed that mutants lacking the S. meliloti ortholog of ATU8132 were able to produce FlaA at levels comparable to wildtype, consistent with a role in a late stage of flagellar assembly or motor function and renamed it motF. In this study, ATU8132 has also been renamed motF.
The four focal genes discussed above (ATU0568, ATU0583, flgN, motF) were deleted in a manner intended to eliminate possible polar effects. This was done by removing most of the gene but retaining the first and last several codons of each coding sequence. We refer to these as "in-frame deletion" strains. All four of these strains exhibited motility defects similar to those of the rest of the motility mutants identified in the transposon screen. These deletion strains could be restored to normal motility by complementation plasmids constitutively expressing the corresponding genes (Fig 4).

Intergenic insertions associated with motility defects
Several insertions associated with motility defects occurred in intergenic regions within the motility cluster (see Fig 1). These are generally interpreted to result in mis-expression of a nearby gene (by disrupting a promoter, overexpression, or antisense RNA expression). It was less straightforward to explain the three intergenic insertions occurring between the genes flaD and ATU0568. As discussed below, flaD is not required for motility, though ATU0568 is. These "flaD-ATU0568" intergenic insertions are intriguing because they occur over a region of 161 bp, with the two insertions farthest from ATU0568 (called IG1 and IG2) transcribing toward flaD (see Fig 1). There is an alternative start codon located 69 bp upstream of the annotated start codon of ATU0568. However, this is not predicted to be disrupted by either insertion (IG1 and IG2) since the closest insertion (IG2) is located 105 bp upstream of this alternative start codon. Sequence specific details of this intergenic region are provided in S5B Fig. Within this region, there is also a predicted open reading frame in the reverse orientation that slightly overlaps with ATU0568 that could explain this motility phenotype. To investigate whether an unannotated gene exists in this region, the sequence disrupted in IG1 and IG2 was manipulated by deleting 10 bp and replacing it with a 6-bp XmaI sequence. This changes the frame and sequence in this region. These two deletion strains exhibited normal motility (S6

Suppressibility of the ΔmotF motility phenotype
To begin exploring the functions of the four genes analyzed in this study (ATU0568 ,  ATU0583, flgN, and motF), we used the in-frame deletion strains to test whether the mutant phenotype could be reversed by intergenic suppression. Genetic suppression was only observed for the ΔmotF strain. This suppression phenomenon is shown in Fig 5, where extended cultivation of ΔmotF in motility agar results in a blebbing pattern not seen in a nonsuppressible mutant such as ΔflaF. Clones extracted from these zones of resumed motility exhibit near-wild-type motility upon retesting. Unlike wild-type cells, however, these suppressor strains seem to generate hypermotile derivatives, as evidenced by a blebbing pattern around the normal ring of motile cells. Future investigation to determine the molecular basis of ΔmotF suppressibility will allow us to make connections between this new gene and previously studied motility functions. For instance, the study by Sobe et al. [28] reported a suppressor phenotype in strains lacking motF, in which suppressors regained partial motility (25%). They performed whole-genome sequence and found that all five suppressor strains contained mutations that mapped to the coding region of motA (G136S and Y248H).

Analysis of master regulators visN and visR
A readily noticeable discrepancy between our screening data (Fig 1) and known flagellar motility pathways in Rhizobiaceae is the absence of any mutants in the master regulator genes visN and visR. These two genes are adjacent to one another and encode a likely dimeric LuxR family transcriptional regulator required for motility in several Rhizobiaceae species including S. meliloti and A. fabrum [9]. The absence of insertions in the visNR operon is particularly striking given that the fliF gene upstream (which is roughly the same size as visNR) was disrupted by 18 different insertions. We have considered several explanations for this discrepancy, including the possibility that visN and visR have partially redundant functions in A. fabrum. To test this hypothesis, the visNR locus was removed from the motile strain BB01 and then modified with four complementation plasmid derivatives: pPG012 (vector-only control), pKJ121 (visN expression plasmid), pKJ122 (visR expression plasmid) and pKJ120 (visNR expression plasmid). This complementation analysis indicates that visN and visR are non-redundant: neither alone can restore motility to the ΔvisNR strain; but co-expression of visNR completely restores motility (Fig 6A). This is consistent with results previously reported by Xu et al. [29]. The second hypothesis posits that visN and visR mutants did not arise in the screen because these mutants are hyper-adherent to neighboring cells. This property has been documented previously [29]. To test this, the ΔvisNR mutant was mixed with BB01 that had been modified to constitutively express lacZ. Strong cell-cell adherence occurring in the mixed culture would result in sectored colonies upon plating on medium with the beta-galactosidase indicator Xgal (5-Bromo-4-Chloro-3-Indolyl β-D-Galactopyranoside). Sectored colonies were not observed in this analysis (S8 Fig). However, we observed that the visNR+ colonies are substantially larger than ΔvisNR colonies (Fig 6B), presumably due to visNR being important for secretion of exopolysaccharides that are normally produced in abundance by wild-type cells [29]. Fig 6B shows colonies that had been incubated for 3 days, whereas during the screening process colonies were incubated for 2 days. From this, we presume that students may have systematically been biased against visN or visR mutants due to their unusually small colony morphology.

Discussion
In this study, we carried out a comprehensive forward genetic screen for motility mutants in Agrobacterium fabrum in which 37 genes were identified as being required for motility based on strong loss-of-motility phenotypes. Based on the suspected functions of proteins encoded by these genes, nearly every molecular component generally required for flagellar assembly in Gram-negative bacteria was identified, in addition to several Alphaproteobacteria-specific functions and four proteins less well characterized. These four proteins (ATU0568, ATU0583, FlgN, and MotF) are all found within a 16-kb region on the right side of the motility gene cluster [30]. Based on database searches, these proteins do not appear in organisms outside of Alphaproteobacteria, suggesting they are specialized features within this class, and their presence almost exclusively in the family Rhizobiaceae suggests a particularly specialized role in these largely plant-associated bacteria. The ΔmotF strain can spontaneously mutate to generate suppressor strains with restored motility (the other three mutants do not have this tendency).
Of the four flagellar filament genes in A. fabrum (flaA, flaB, flaC, and flaD), only flaA was identified in this screen, which was expected based on previous work showing that this is the only required flagellin, with the others serving subsidiary functions. Earlier studies have shown that ΔflaA mutants form straight flagellar filaments that result in very slow tumbling motion [18]. It appears that FlaA incorporates a functionally crucial helical attribute into the flagellar filament. A key residue in FlaA distinguishing it from the subsidiary flagellins is an Asn residue at position 129 that plays a role in establishing this helical property. It has been shown, however, that FlaA must function with at least one of the three subsidiary flagellins [16][17][18].
Three of our transposon mutants had insertions in three unique locations between the genes flaD and ATU0568. Within this intergenic region of 279 bp, they were located at position 84, 105 and 245. The transcription from the transposon read to the left for insertions at 84 and 105 and right for insertion 245. With these insertions spread so broadly across this region and the upstream flanking gene (flaD) not required for motility, we can only speculate how these transposon insertions bring about loss of motility. For two of these (IG1 and IG2), disruption of a 10-bp segment corresponding to the wild-type sequence, did not noticeably affect motility. We suspect that these intergenic insertions may disrupt an unusually large regulatory region upstream of ATU0568 or may disrupt or mis-regulate some independent and unannotated feature required for motility.
Mutagenesis by transposon insertion can have polar effects on polycistronic operons. In this screen, with modest transcription emanating from the kanR gene of the transposon into the genome, we did not expect polar effects when the transposon was transcribed inserted in the same direction of the disrupted gene; but we expected possible polar effects when the transposon was transcribed in the opposite direction of the gene. We also observed that there was a strong directionality bias for transposon insertion in some genes such as flaA and ATU0568, but for most genes implicated in the study, insertions in both directions could be found.
The regulatory genes rem, flbT, and flaF were all hit in this screen. The Rem protein is a Class I transcriptional regulator that activates the expression of Class II structural and regulatory genes [8]. Two of these regulatory genes (flbT and flaF) are highly conserved among Alphaproteobacteria. In Brucella melitensis, FlbT acts as a translational activator for the synthesis of flagellin [31]. Like flbT, flaF is also conserved among Alphaproteobacteria, and is generally located upstream and is cotranscribed with flbT [15,32,33]. The visN and visR genes were not hit in this screen despite being essential for motility as top-level master transcriptional regulators [9,10,29]. As previous studies have shown, VisN and VisR not only positively regulate flagellar synthesis, but are also negative regulators of unipolar polysaccharide (UPP) synthesis and positive regulators of exo genes that control succinoglycan biosynthesis [29,34,35]. ΔvisNR mutants are substantially smaller than wild type presumably due to decreased succinoglycan synthesis and more dry than wild type due to increased cellulose and UPP production [29]. We hypothesize that ΔvisNR mutant colonies may not have been selected by the student researchers carrying out this screen due to their unusually small colony morphology.
The Agrobacterium fabrum C58 genome encodes roughly 20 methyl-accepting chemotaxis protein (MCP) homologs, and 9 che genes for chemotactic control of flagellar activity [30,36,37]. No che or mcp genes were hit in this screen, likely because we focused on mutants with strong non-swimmer phenotypes; mutants lacking che/mcp genes show significant but reduced motility [38]. A screen for mutants with partial loss of motility would surely point to chemotaxis functions, as well as other pathways contributing in more subtle ways to flagellar assembly and control.