Suppressor mutations restore motility in ΔfimX mutants

FimX is a cdGMP-binding regulator of pilus extension in P. aeruginosa, required for PilB polar localization and thought to promote pilus extension through an unknown mechanism [8,29–31]. Consistent with previous work, we confirmed that a PAO1 deletion mutant of fimX had significantly reduced twitching motility, 5–10% of the wild type (Fig 1B). However, this strain still twitched when compared to a non-motile major pilin (pilA::Tn5) mutant and remained susceptible to the pilus-targeting bacteriophage, PO4 (Fig 1C) [29]. FimX mutants did not produce detectable surface pili in sheared protein fractions (S1 Fig), further supporting a defect in pilus extension. This phenotype is distinct from that of mutants lacking another PilB regulator, the putative chaperone, PilZ [27,28], which do not twitch and are resistant to pilus-specific phages (Fig 1B and 1C).

We reasoned that we could leverage the ΔfimX pilus extension deficit to screen for suppressor mutants capable of restoring twitching motility, to inform on the function of FimX. After allowing fimX mutants to twitch at the sub-agar interface for an extended period (48–72 h), we isolated highly motile flares that emerged from the small primary twitching zone (Fig 1D). These cells were collected, isolated for single colonies, and after confirming their increased motility compared to the ΔfimX parent, were sequenced to identify candidate suppressor mutations. The mutations mapped to relevant proteins including the ATPase PilB, alignment subcomplex component PilO [37,38], members of the Pil-Chp network (PilJ and FimL) [7], and CpdA (additional mutants are listed in S1 Table). Each mutation increased motility to varying degrees (Fig 1E), to approximately 20–50% of the wild-type, but did not impact growth (S2 Fig) or phage susceptibility (S3 Fig). Complementing the suppressor mutants with fimX further increased twitching motility to levels similar to or greater than WT, except for PilO P168L, suggesting it has a fimX independent motility phenotype (S4A Fig). In contrast, complementation of the suppressors with a defective version of FimX – where its cdGMP-binding EVL motif was mutated to AAA – did not further increase twitching motility, consistent with previous reports that this non-functional variant is incapable of PilB binding [30]. Diminished twitching motility in the WT strain when FimX AAA was expressed in trans is likely due to formation of FimX-FimX AAA heterodimers, supported by their pairwise interactions in a bacterial 2-hybrid assay (S4B Fig).

ΔfimX suppressor mutations do not directly influence cdGMP metabolism

Previous efforts to isolate ΔfimX suppressors revealed several mutations which collectively upregulated intracellular cdGMP levels [31]. To test whether our mutants shared this phenotype, we assessed intracellular levels of cdGMP using a pcdrA:Lux cassette-based reporter system in liquid cultures [20]. Under these conditions, none of our mutants demonstrated higher cdGMP levels over time when compared to control cells expressing the well-characterized diguanylate cyclase (DGC), SadC (Fig 2A). To directly test the impact of cdGMP on ΔfimX suppression, we overexpressed DGCs from P. aeruginosa (sadC) or Escherichia coli (ydeH) in WT or ΔfimX backgrounds and tested twitching motility. In all cases, cells expressing the DGCs showed significantly reduced twitching motility compared to the vector control (Fig 2B). Together, these data reflect the observations from prior work [31] that suppression of the ΔfimX motility defect in PAO1 is not related to changes in cdGMP levels.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. FimX suppressor mutations do not impact intracellular cdGMP levels.

(A) Time-course curves of relative luminesence signal of a pcdrA:LuX cassette cdGMP reporter normalized to OD₆₀₀. Each timepoint is the mean of triplicate samples from three independent experiments + SD. EV-pMS402: empty pMS402 vector, sadC: sadC in pBADGr vector. (B) Quantification of sub-agar stab twitching motility zones for strains overexpressing diguanylate cyclases (DGCs). Representative crystal violet-stained twitching zones are shown to the left. DGC expression was induced with 0.1% arabinose. Bars represent the means of triplicate samples from three independent experiments ± SD. X-axis depicts a log₂ scale. Scale bar = 1 cm. EV: empty pBADGr vector, sadC: sadC in pBADGr vector, ydeH: ydeH in pBADGr vector. ***: 0.001 ≥ p (One-way ANOVA).

https://doi.org/10.1371/journal.pgen.1011802.g002

Elevated cAMP levels increase motility in the ΔfimX background

Since several of our motile suppressor mutations mapped to the Pil-Chp network or CpdA, we hypothesized that changes to cAMP homeostasis [5,7] increased twitching in the absence of fimX. We therefore assessed intracellular levels of cAMP in both liquid cultures and surface-adapted cellular populations using a PaQa:YFP reporter plasmid system [39]. WT cells had significantly higher levels of cAMP when the cells were grown on a surface versus in liquid, but elevated cAMP-dependent signal was lost in a pilA mutant (Fig 3A), consistent with T4P being a major input to cAMP generation in P. aeruginosa [40]. Deletion of CyaB or expression of a constitutively active CyaB R456L in trans further reduced or increased cAMP levels, respectively [15,41,42]. In these assay conditions, ΔfimX had lower than WT levels of cAMP and was unable to respond to surface contact. The ΔfimX suppressor mutants PilJ T570I, and CpdA V258G all had significantly elevated cAMP levels. In contrast, neither the PilB nor PilO P168L suppressor mutants had significant increases in cAMP levels compared to the ΔfimX parent. These data suggested that there were at least two routes to increase twitching in the fimX background, one of which was independent of cAMP levels.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. A subset of ΔfimX twitching suppressors elevate cAMP levels and upregulate associated phenotypes.

(A) Plasmid-based cAMP reporter assay quantifying cAMP-dependent YFP expression (PaQa) normalized to constitutive mKate expression (rpoD). Bars represent the means of duplicate samples from three independent experiments ± SD. (B) Quantification of sub-agar stab twitching motility zones for strains overexpressing the adenylate cyclase CyaB or a catalytically inactive mutant. Representative crystal violet-stained twitching zones are shown to the left. Expression was induced with 0.1% arabinose. Bars represent the means of triplicate samples from three independent experiments ± SD. Scale bar = 1 cm. EV: empty pHERD30T vector, His₆-CyaB (D234S): N-terminally hexa-histidine tagged CyaB or CyaB D234S in pHERD30T vector. For liquid-surface comparisons: *: 0.05 ≥ p ≥ 0.01; **: 0.01 ≥ p ≥ 0.001 (two-tailed Welch’s t-test). For surface only comparisons **: 0.01 ≥ p ≥ 0.001; ***: 0.001 ≥ p (one-way ANOVA). (C) Vfr-dependent expression of mRuby3 chromosomally inserted upstream of the native alignment complex pilMNOPQ operon. Bars represent the means of duplicate samples from three independent experiments ± SD. Cartoon schematic of the mRuby3 expression system is shown above the graph. **: 0.01 ≥ p ≥ 0.001; ***: 0.001 ≥ p (One-way ANOVA). (D) Representative PilO Western blot. Sample loading was normalized to a non-specific band shown below. Molecular weight is indicated to the left. Blot is representative of three independent experiments. Strains in the grey boxes indicate ΔfimX mutant background. CyaB R456L: CyaB R456L in pBADGr vector.

https://doi.org/10.1371/journal.pgen.1011802.g003

We next asked whether high cAMP levels were responsible for increased ΔfimX twitching. Expressing CyaB from a multicopy plasmid in the ΔfimX background was sufficient to significantly increase twitching motility compared to the empty vector control (Fig 3B), while expression of an inactive CyaB variant (S5A Fig) [43] was not. This increased twitching phenotype was also observed in a double mutant lacking the Pil-Chp response regulators PilG and PilH (S5B Fig), bypassing Pil-Chp-dependent CyaB activation [36]. Together, these data indicate that elevated cAMP levels due to Pil-Chp or CpdA mutations increase ΔfimX motility. The pilJ and cpdA suppressor strains with high intracellular cAMP in turn had increased Vfr-dependent expression [13] of an mRuby3 fluorescent protein reporter integrated upstream of the T4P pilMNOPQ operon (Fig 3C). Integration of the reporter did not prevent phage susceptibility of these strains (S6 Fig) indicating that pilus function is maintained. Consistent with this increased expression, levels of PilO were also elevated (Fig 3D) suggesting that more T4P machines are expressed in these backgrounds.

Notably, the PilB and PilO suppressors had no increase in Vfr-dependent transcription or abundance of T4P components (Fig 2C and 2D) compared to ΔfimX, consistent with low cAMP levels (Fig 3A). Another cAMP-dependent trait, type II secretion system-dependent secretion of proteases generating zones of clearing on skim-milk plates [44] was also unchanged (S7A and S7B Fig), collectively suggesting that the cAMP levels in those mutants were not significantly impacted [13,15], and that ΔfimX suppression occurs through non-cAMP related mechanisms in these backgrounds.

Determinants for pilB Δ1bp function

The suppressors in the non-cAMP related class predominantly mapped to the pilus extension ATPase, PilB. To probe this second set, we focused first on a mutant in which a single adenine was deleted in the last codon of the PilB open reading frame (pilB Δ1bp). This deletion produced a frameshift that extended the coding sequence by an additional 21 bases, or seven amino acids, to the next stop codon (Fig 4A). When modelled in Alphfold3, the extended segments were predicted to be solvent-exposed, protruding from the base of the hexamer and away from its putative interaction interfaces with additional T4P components (S8 Fig). To find the minimum length of this C-terminal extension that was sufficient to increase motility in ΔfimX, we generated single codon truncations from the 3′ end of the pilB Δ1bp sequence (Fig 4A). Each variant complemented motility in both a pilB::Tn5 background (S9A Fig), and in the WT (S9B Fig), confirming that they produced a functional gene product with no dominant negative effect through mixed oligomerization with WT PilB. In a strain lacking both pilB and fimX, all pilB Δ1bp truncations tested significantly increased twitching motility relative to the control complemented with WT PilB (Fig 4B). Notably, the longest extension rescued motility to a greater degree than the shortest one.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. PilB mutations increase twitching motility in the absence of fimX.

(A) Cartoon schematic of deleted base pair A1697 at the PilB C-terminus, shown in red, and resulting extended sequence. The pilB Δ1bp truncation sequences are shown below. (B) Quantification of twitching zone areas of ΔfimX pilB::Tn5 double mutants overexpressing pilB Δ1bp 3′ truncations. Representative crystal violet stained twitching zones are shown to the left. PilB expression was induced with 0.05% arabinose. Bars represent the means of duplicate samples from three independent experiments ± SD. (C) Quantification of twitching zone areas of ΔfimX PilB D566X point mutants chromosomal knock-ins. Representative twitching zones are shown to the left. Bars represent the means of triplicate samples from three independent experiments ± SD. (D) Full-length multiple sequence alignment of PilB homologues from various species highlighting the sequence surrounding PAO1 PilB T430, labelled in red. The Alphafold3 model of the predicted secondary structure in this region is shown to the left. (E) Quantification of twitching zone areas of fimX pilB double mutants overexpressing PilB T430X mutants. Representative twitching zones are shown to the left. PilB expression was induced with 0.05% arabinose. Bars represent the means of duplicate samples from three independent experiments ± SD. (F) Quantification of twitching zone areas of fimX pilB::Tn5 double mutants overexpressing PilB constructs with point mutations in positions adjacent to T430. Representative twitching zones are shown to the left. PilB expression was induced with 0.05% arabinose. Bars represent the means of triplicate samples from three independent experiments ± SD. Strains in the grey or blue boxes indicate the ΔfimX or ΔfimX pilB::Tn5 mutant backgrounds respectively. EV: empty pHERD30T vector, pilB Δ1bp-3 to 21: pilB Δ1bp-3 to 21 in pHERD30T vector, PilB L429/T430/R431X: indicated PilB substitution in pHERD30T vector. All scale bars = 1 cm. ns: p ≥ 0.05; *: 0.05 ≥ p ≥ 0.01; **: 0.01 ≥ p ≥ 0.001; ***: 0.001 ≥ p (One-way ANOVA).

https://doi.org/10.1371/journal.pgen.1011802.g004

Interestingly, although the WT PilB and pilB Δ1bp-21 sequences are identical in length, only the latter significantly increased twitching motility when introduced in the ΔfimX pilB::Tn5 background. (Fig 4B). These two proteins differ only in the C-terminal amino acid, with a valine substituted for an aspartate in the pilB Δ1bp-21 sequence. PilB D566 may therefore play an important role in modulating the function of the enzyme. The longer pilB Δ1bp sequences that support increased motility have additional positively charged residues (Fig 4A), suggesting that loss of negative charge at the C-terminus enhances PilB function. Since the above studies were done with multicopy plasmids, we next introduced single chromosomal point mutants at the PilB C-terminus in a ΔfimX background to assess recovery of twitching motility. A PilB D566V mutation significantly increased twitching motility compared to the ΔfimX parent (Fig 4C). We also observed a significant increase in twitching when the C-terminal amino acid was charge-swapped to a lysine, but not when we introduced glutamate (Fig 4C). The second-last amino acid in the native PilB sequence is a positively charged lysine, and we hypothesized that shortening the WT sequence by a single amino acid to expose the positive charge would also restore motility in a ΔfimX mutant. However, substituting D566 with a stop codon did not significantly increase motility relative to the control (Fig 4C). These data suggest that PilB C-terminal charge, in addition to C-terminal length, is an important regulator of PilB function in cells

Determinants for PilB T430P function

We next examined a second PilB suppressor (PilB T430P) which also mapped to the C-terminal domain, but at a site distal to the C-terminus (S8 Fig). A multiple protein sequence alignment of full-length PilB homologues from various species revealed that the region surrounding T430 was well conserved, although T430 itself was not (Fig 4D). We therefore reasoned that this amino acid was less important to overall PilB function or potentially a species-specific regulatory site. We made individual substitutions at T430, choosing amino acids found at this position in the other homologues. All mutant constructs complemented twitching motility in a pilB::Tn5 background and the WT (S10 Fig). In a pilB fimX double mutant, only the original T430P suppressor supported twitching motility (Fig 4E). Therefore, disruption of local PilB secondary structure, rather than specific side-chain chemistry, is likely responsible for increased motility.

An Alphafold3 model of a PilB monomer predicted that residue T430 is located in the second turn of a short α-helix (Figs 4D and S8). While T430 is not well conserved, other amino acids in this helical segment, such as an adjacent R431, are. Alignment of the model onto the crystal structure of hexameric PilB from Geobacter metallireducens bound to a non-hydrolysable ATP analogue shows that the equivalent N429 (T430) and R430 (R431) residues are positioned at the protomer-protomer interface or jutting into the lumen of the torus, depending on the phase of the ATP hydrolysis cycle (S8 Fig). We hypothesized that this secondary structural element could be involved in inter-subunit communication and that the introduction of a helix-breaking proline would alter these functional contacts. Supporting this idea, mutation of the adjacent residues to proline, L429P or R431P, significantly increased twitching motility in fimX pilB (Fig 4F). In contrast, R431A failed to increase motility in the double mutant, thus disruption of the α-helical segment rather than loss of this sidechain chemistry was likely responsible for altering PilB function. Interestingly, in the pilB::Tn5 or WT backgrounds where FimX was present (S11A and S11B Fig), twitching motility of PilB R431A and R431P variants was significantly decreased compared to control (S11A Fig). Therefore, the presence or absence of FimX seems to differentially modulate the function of these PilB variants.

Increased twitching of suppressors is due to enhanced PilB ATPase activity

We next investigated how the pilB Δ1bp and PilB T430P mutations restored pilus function in a ΔfimX background. Other PilB regulatory effectors, such as PilZ, are thought to serve as molecular chaperones and pilZ mutants lack detectable intracellular PilB on Western blots (Fig 5A). This is consistent with results from previous studies of the Xanthomonas citri PilZ homologue [27]. The pilZ mutants are also non-motile and phage resistant (Fig 1B and 1C), in keeping with their negligible levels of PilB. In contrast, PilB was still detectable in ΔfimX or FimX AAA backgrounds, although its levels were approximately half of WT (Fig 5A). Neither suppressor increased levels of PilB compared to those in ΔfimX or FimX AAA. Since FimX is important for PilB polar localization [8,27,30], the reduction in PilB levels in fimX mutants may be a consequence of fewer interactions between the ATPase and T4P machinery, making it more susceptible to degradation.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. PilB suppressor mutations increase ATP hydrolysis.

(A) Representative PilB Western blot. Sample loading was normalized to a non-specific band shown below. Molecular weight is indicated to the left. Blot is representative of three independent experiments. Quantification using BioMol green reagent of free phosphate released over a time course (B) or following overnight incubation (C) by ΔN1D-His₆-PilB-catalyzed ATP hydrolysis. PilB K332A, a non-catalytically active Walker mutant, was used as a negative control. Bars represent the means of duplicate samples from three independent experiments ± SD. **: 0.01 ≥ p ≥ 0.001; ***: 0.001 ≥ p (One-way ANOVA). (D) Quantification of free phosphate release by ΔN1D- His₆-PilB-catalyzed ATP hydrolysis in the presence of cdGMP containing reaction buffer (+Buffer) or with cdGMP and FimX (+FimX) following overnight incubation. Bars represent the means of duplicate samples from four independent experiments ± SD. For all PilB ATPase assays, activity was measured through the production of inorganic phosphate (P_i) at 37 °C using 100 nM of ΔN1D-PilB, and 1 mM of ATP. Where indicated, 200 nM of FimX and/or 1 µM of cdGMP was also added. Proteins in the green boxes indicate additional mutations to ΔN1D-PilB. (E) Quantification of twitching zone areas of ΔfimX or FimX AAA mutants with or without individual PilB suppressor mutation chromosomal knock-ins. Representative crystal violet stained twitching zones are shown to the left. Bars represent the means of triplicate samples from three independent experiments ± SD. Scale bar = 1 cm. ns: p ≥ 0.05; *: 0.05 ≥ p ≥ 0.01; ***: 0.001 ≥ p (Two-tailed Welch’s t-test).

https://doi.org/10.1371/journal.pgen.1011802.g005

Since PilB is an ATPase used to power pilus extension, we reasoned that the mutations might increase enzymatic activity. PilB has two N-terminal domains (N1D and N2D) connected by a flexible linker to the C-terminal domain (S8 Fig) [45]. The N1D can be removed without disrupting hexamer formation to improve stability and solubility for in vitro studies [27,46–50]. We purified ΔN1D-PilB variants to assess ATP hydrolysis efficiency in vitro (S12 Fig). Both PilB suppressors hydrolyzed more ATP than the WT and catalytically inactive Walker A mutant (PilB K332A) [1] controls over time (Fig 5B), as well as following overnight incubation (Fig 5C). Despite mapping to disparate sites in the protein, both PilB suppressors upregulated enzymatic activity when compared under these conditions. Consistent with previous work [30], the addition of FimX and cdGMP did not stimulate the ATPase activity of WT PilB. Interestingly, when we added FimX and cdGMP to the purified PilB variants, we saw a decrease in ATP hydrolysis for pilB Δ1bp but an increase for PilB T430P (Fig 5D). While this result is difficult to interpret, it suggests that FimX may modulate PilB enzymatic activity. In a genetic context, pilB Δ1bp and PilB T430P mutations integrated into the chromosome restored twitching motility to at least WT levels in the presence of FimX (Fig 5E), but not FimX AAA. Taken together, this indicates that these PilB suppressor mutants can continue to respond to cdGMP-bound FimX input and regulation.

FimX promotes pilus extension to produce longer pili

To test the impact of increased ATP hydrolysis by PilB on pilus assembly, we imaged pilus dynamics using fluorescence microscopy (Fig 6A). The major pilin subunit, PilA, was first labelled with a fluorescent maleimide dye at A86C as previously described [4,51]. All PilA A86C mutants remained motile with slightly reduced twitching compared to the wild type (S13 Fig). WT cells produced T4P at the cell poles as expected, with some cells producing pili at both poles (Fig 6A). Notably, the ΔfimX mutant produced very short pili that were difficult to detect, often presenting only as bright polar puncta which we scored as a polar accumulation of labelled PilA or very short T4P, below the limit of detection. In addition, fewer cells were piliated compared to WT (Fig 6B and 6C). We chose to quantify pili using the ΔfimX pilB Δ1bp mutant as this strain showed the most robust twitching phenotype of our suppressors. Furthermore, in the presence of FimX, the pilB Δ1bp mutation alone had greater than WT levels of twitching, providing an optimal background to study FimX’s functional contribution. In the ΔfimX pilB Δ1bp background, piliation was restored on many cells (Fig 6B and 6C), with longer pili compared to the ΔfimX mutant (Fig 6A). The percentage of piliated cells in the ΔfimX pilB Δ1bp mutant was similar to WT (Fig 6B and 6C), but the average maximum pilus length was significantly shorter (Fig 6D), likely explaining its comparatively reduced motility (Fig 5E). Maximum pilus length was significantly increased compared to the ΔfimX parent, indicating that the pilB Δ1bp mutation caused cells to produce longer pili, perhaps through enhanced ATP hydrolysis. The pilB Δ1bp alone background produced the longest pili (Fig 6A and 6D) and had the highest percentage of piliated cells (Fig 6B and 6C), consistent with this strain being the most motile (Fig 5E). Longer and long-lived pili in PAO1 (~2–3 µm) comprise a rare subpopulation [4]. We therefore anticipate that more ΔfimX cells are likely piliated, but since the maximum pilus length in these mutants is ~ 0.75 µm, most pili may not be detectable.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. FimX mutants produce short pili due to slow extension rates.

(A) Representative images of PilA A86C type IV pili labelled with maleimide-Alexafluor 488 with background fluorescence subtracted. White arrows indicate visible pili. All scale bars = 2 µm. Number of cells producing visible PilA A86C type IV pili expressed as a percentage of total cells (B) or normalized relative to the WT arbitrarily set to 1 (C). Bars represent the means ± SD of three independent experiments (WT n = 193, ΔfimX n = 181, ΔfimX pilB Δ1bp n = 210, pilB Δ1bp n = 197 total cells). Technical replicates are indicated by colour. *: 0.05 ≥ p ≥ 0.01; ***: 0.001 ≥ p (One-way ANOVA). (D) Maximum PilA A86C pilus length. Each sample depicts the means ± SD of three independent experiments, each point represents a single PilA A86C pilus length (WT, ΔfimX pilB Δ1bp, and pilB Δ1bp n = 60, ΔfimX n = 45 total pili). Technical replicates are indicated by colour. (E) Rate of PilA A86C pilus extension or retraction. Each sample depicts the means ± SD of three independent experiments, each point represents the rate of extension and/or retraction of a single PilA A86C pilus (ΔfimX pilB Δ1bp n = 46, pilB Δ1bp n = 44 total extension and ΔfimX pilB Δ1bp n = 40, pilB Δ1bp n = 41 total retraction events). ns: p ≥ 0.05; **: 0.01 ≥ p ≥ 0.001 (Two-tailed Welch’s t-test).

https://doi.org/10.1371/journal.pgen.1011802.g006

The pilB Δ1bp mutants afforded an opportunity to quantify pilus dynamics in the presence or absence of FimX. We quantified the rates of pilus extension and retraction over time in the ΔfimX pilB Δ1bp versus pilB Δ1bp to assess FimX’s contribution to pilus dynamics. The FimX-expressing strain had a significantly increased average rate of pilus extension (Fig 6E) while the average rate of retraction was not significantly different, supporting FimX’s role as a modulator of PilB-dependent pilus extension. These data suggest that FimX increases pilus extension efficiency, potentially by acting on PilB.

In P. aeruginosa and other bacterial species, fluorescently labelled FimX fusions display defects in polar localization when PilB interaction is impaired [27,30,33]. Consistent with these other studies, most ΔfimX mutant cells complemented with an N-terminal fusion of mNeonGreen (mNGr) to FimX showed bright puncta localized to a single cell pole (Fig 7A), with a smaller population of cells displaying bipolarly-distributed puncta (Fig 7B). The WT FimX fusion also complemented twitching motility (S14A Fig), while cells expressing an empty vector control had no discernable puncta (S14B Fig). The number of cells with identifiable polar puncta was decreased by ~20–50% when pilB was deleted or if a mNGr-FimX AAA fusion was used (Fig 7B). Importantly, these data indicate that FimX localization patterns in our strains are consistent with previous reports [8,27,30] and are contingent on interaction with PilB.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. FimX polar localization is disrupted in the absence of PilB binding.

(A) Representative images of cells expressing an mNeonGreen N-terminal FimX or FimX AAA fusion with background fluorescence subtracted. All scale bars = 2 µm. (B) Quantification of cells expressing an mNeonGreen-N-terminal FimX or FimX AAA fusion scored for bipolar, unipolar, or no discernable puncta and expressed as a percentage of total cells. Bars represent the means ± SD of three independent experiments (ΔfimX + mNGr-FimX n = 216, ΔfimX pilB::Tn5 + mNGr-FimX n = 213, ΔfimX + mNGr-FimX AAA n = 235, ΔfimX pilB::Tn5 + mNGr-FimX AAA n = 206, total cells). Strains in the blue or grey boxes indicate expression of the mNeonGreen N-terminal FimX or FimX AAA fusions respectively. mNGr-FimX or mNGr-FimX AAA: mNeonGreen N-terminal FimX or FimX AAA fusion in pHERD30T vector. Expression was not induced with arabinose.

https://doi.org/10.1371/journal.pgen.1011802.g007

PilB enhancing mutations are uncommon

Given the relative ease of selection for fimX suppressor mutations that mapped to PilB and improved twitching motility, we wondered if similar PilB mutations were common in sequenced strains of P. aeruginosa. We compared 417 PilB orthologue sequences from the Pseudomonas.com database to identify comparable mutations. Of these, 297 were the same length as PAO1 PilB at 566 residues, and 82 were longer. Of those sequences, 72 were longer by one residue, 8 by two residues, 1 by four residues, and 1 by 15 residues (S15A Fig). A full length sequence alignment of representative orthologues revealed that these sequences did not comprise a C-terminal extension. Instead, the additional length was due to insertions elsewhere in the protein, such as at the N-terminus (+15 residue sequence) or in expanded loop regions (S15B Fig). Strikingly, the C-terminal residue of all sequences was an aspartic acid. We also assessed the local sequence surrounding PilB T430 for helix-disrupting residues which may confer a similar effect to T430P. We examined the region + /- three amino acids from the residue aligned with PAO1 PilB T430 in a full-length protein sequence alignment, to capture the most likely positions to influence stability of a 3-turn α-helix. Of the 417 sequences, 4 had a proline residue and 8 had a glycine within this window, accounting for less than 1% and 2% of the total sequences respectively (S15C Fig). This analysis of P. aeruginosa sequences suggests that PilB mutations similar to the ones identified in this screen are uncommon.

Bacterial strains and growth conditions

All strains and plasmids used in this study are outlined in S2 Table. Bacterial strains were grown aerobically at 37 °C in Lysogeny Broth (LB) media (10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl), supplemented with the appropriate concentration of antibiotics where necessary unless otherwise specified. Antibiotics were used at the following concentrations: 15 µg/mL or 30 µg/mL of gentamicin for E. coli or P. aeruginosa respectively unless otherwise indicated. 50 µg/mL or 200 µg/mL kanamycin for E. coli or P. aeruginosa respectively. 100 µg/mL of ampicillin or 200 µg/mL of carbenicillin for E. coli or P. aeruginosa respectively. Plasmid constructs were generated using E. coli DH5α before introduction into the appropriate cellular background by heat shock (E. coli) or electroporation (P. aeruginosa or E. coli) where appropriate [66,67].

Plasmid and strain construction

Fragments of interest were PCR amplified from isolated mPAO1 genomic DNA. All PCR primers are listed in S3 Table. For point mutations, two separate PCR amplified DNA fragments were stitched together through a second single overlap extension PCR, with each DNA overhang including the mutation of interest introduced during primer synthesis. For plasmid construction, the in-frame genetic insert and corresponding vector were digested with the appropriate FastDigest restriction enzymes (ThermoFisher Scientific) as per manufacturer instruction and ligated with T4 DNA ligase. Ligated products were transformed via heat shock into E. coli DH5α and selected on 1.5% agar LB media plates supplemented with the appropriate antibiotic. Following selection, single colonies were harvested, and the plasmid was extracted using a GeneJet plasmid prep kit (ThermoFisher Scientific) for sequence confirmation (Plasmidsaurus).

Genetic manipulation of mPAO1 chromosomal DNA was performed by double homologous recombination as previously described [68]. In brief, for genetic deletions, ~ 500 bp upstream and downstream of the desired target sequence were PCR amplified and purified from isolated PAO1 genomic DNA. Upstream and downstream sequences were engineered to contain ~50–100 bases of the 5′ or 3′ ends of the gene deletion target to ensure minimal disruption of adjacent genes. Upstream and downstream fragments were digested with the appropriate restriction enzymes along with the suicide vector backbone pEX18Gm. Digested fragments were ligated using T4 ligase and transformed into via heat shock into E. coli DH5α and selected for on 1.5% agar LB media plates supplemented with gentamicin. Purified and sequenced plasmids were then transformed into E. coli SM10 (λpir). A single colony of SM10 harbouring the plasmid was grown overnight in 3 mL of LB media supplemented with gentamicin. The following day, the SM10 culture was mixed with the appropriate strain of mPAO1 overnight culture in a 1:1 ratio (by volume), pelleted by centrifugation, the resulting cell pellet was resuspended in 50 µL of LB and then spotted on 1.5% agar LB plates dried in a biological safety cabinet for 20–30 min. The spot was allowed to dry at room temperature before being incubated at 37 °C overnight. The following day, the cell spot was scraped into fresh LB media and plated onto premixed Pseudomonas isolation agar (Difco) supplemented with 100 µg/mL of gentamicin. The plates were incubated again at 37 °C overnight. Single colonies were then re-streaked again on Pseudomonas isolation agar plates supplemented with 100 µg/mL of gentamicin and incubated overnight at 37 °C. The next day, cells were streaked in a zig-zag pattern on 10% LB (1 g/L tryptone, 0.5 g/L yeast extract) with no added NaCl, 1.5% agar, and 5% sucrose media plates and incubated overnight at 37 °C. Single colonies from the sucrose media selection plate were patched again onto a fresh sucrose media plate as well as an LB media plate supplemented with gentamicin. Patches which grew on sucrose media but not LB with gentamicin were assessed for chromosomal deletion via colony PCR.

For chromosomal knock-in of a point mutation, the same protocol as outlined above was performed only differing at initial construct design. Knock-in constructs were instead generated ~500 bp upstream or downstream of the mutation site through single overlap extension PCR. Overlap extension primers were also designed to silently introduce or remove a restriction cut site in the genome where applicable to facilitate downstream quality assurance. The remaining procedure was then carried out as described.

Twitching motility assays

Bacterial strains were streaked onto 1.5% agar LB media plates supplemented with gentamicin where applicable and grown overnight at 37 °C. The following day, cells were picked using a sterile pipette tip and stabbed down to the bottom layer plastic of 1% agar LB media tissue culture treated plates, supplemented with antibiotic or arabinose where applicable, dried for 30 min in a biological safety cabinet. Plates were incubated overnight at 37 °C. The following day, the agar was carefully removed with a pipette tip and discarded. The twitching zones adhered to the plate were stained with 0.1% crystal violet dye for a minimum of 10 min, washed twice with DI water, and left to air dry overnight at room temperature before imaging with a scanner. Twitching zone areas were then digitally traced using ImageJ, and either duplicates or triplicates were averaged.

Phage-susceptibility assays

Bacterial strains were inoculated into 3 mL of LB media liquid cultures and grown overnight at 37 °C. The next morning, 5 µL of overnight culture was sub-cultured into 3 mL of fresh LB media and grown at 37 °C to mid-logarithmic phase (OD₆₀₀~ 0.5-1.0). Cells were normalized to an OD₆₀₀ of 0.1 in LB. For plate-based assays, 10 µL of bacterial suspension was mixed with 5 µL of serially diluted PO4 phage stock (3e5 PFU/mL original titer). 10 µL of cell-serial phage dilution was spotted onto 1.5% agar LB media dried for 30 min in a biological safety cabinet. The plate was incubated overnight at room temperature. The following day, the plate was incubated again at 37 °C for 2–3 hours before imaging with a scanner.

For liquid-based assays, 2 µL of OD₆₀₀ normalized cell suspension was added to 200 µL of LB media total in a flat bottom 96 well plate. To each well, 1 µL of 3e5 PFU/mL PO4 phage stock was also added. The plate lid was sealed with masking tape and then incubated overnight in a 37 °C shaking plate reader, with OD₆₀₀ readings taken every 30 min for 16–18 hours. Curves were normalized to LB media blank wells in the same plate.

Sheared surface protein preparation

Surface pilin and flagellin protein levels were assessed as previously described [69]. Briefly, strains were streaked in grid-like pattern on a 1.5% agar LB media plate dried for 30 minutes in a biological safety cabinet. The plate was incubated at 37 °C overnight. The next morning, cells were gently scrapped off the surface using a combination of a glass coverslip and sterile inoculation loop into 1.2 mL of sterile 1X PBS. Surface appendages were sheared by vortexing the suspension for 30 s. The homogenized suspension was transferred to a 1.5 mL Eppendorf tube and pelleted by centrifugation at maximum speed for 30 min. The supernatant was decanted into a fresh tube and 0.1 volume of 5M NaCl and 30% (w/v) polyethylene glycol (MW range 8k) was added. The tubes were then incubated on ice for 90 min, inverting the tubes periodically. The precipitated protein was collected by centrifugation again for 30 min at maximum speed. The supernatant was decanted, and the precipitated protein resuspended in 50 µL of 1X SDS loading dye (62.5 mM Tris-HCl pH 6.8, 0.5% (w/v) SDS, 5% (v/v) β-mercaptoethanol, 10% (v/v) glycerol, 0.25% (w/v) bromophenol blue), and boiled for 10 min. The tubes were centrifuged for another 5 min at maximum speed and separated on a 15% SDS-polyacrylamide gel. Proteins were visualized with Coomassie blue staining.

Bacterial 2-hybrid assay

Single colonies of E. coli BTH101 co-transformed with both T18/T25 tagged target proteins were inoculated into 3 mL LB media overnight cultures supplemented with kanamycin, ampicillin, and 0.5 mM IPTG. The following morning, the cultures were normalized to an OD₆₀₀ of 0.1 in fresh LB and 4 µL was spotted onto MacConkey agar plates (40 g/L, Difco) supplemented with kanamycin, ampicillin, 0.5 mM IPTG, and 1% maltose pre-dried for 40 min in a biological safety cabinet. The spots were allowed to dry at room temperature before incubation at 30 °C in a standing incubator. The plates were imaged after 24 hours.

Selecting ΔfimX twitching suppressor mutants

Motility suppressors were isolated through the standard twitching assay described with some alterations. ΔfimX bacteria were first streaked for single colonies on an LB agar plate. Using a sterile pipette tip, a single colony was stab-inoculated into a standard twitching plate ~4–5 times in a row. The tip was discarded, and a new colony was then stab inoculated in the same manner. This was repeated for a total of 35–40 individual twitching zones on a single assay plate. The plate was sealed with parafilm and incubated at 37 °C for 48–72 hours. Following incubation, the top agar layer was carefully removed and discarded. Flares emerging from small twitching zones were collected with a sterile cotton swab and streaked again for single colonies. Single mutant colonies were reassessed for twitching motility. Those with increased twitching relative to the ΔfimX strain were collected and their genomic DNA isolated using a Wizard genomic DNA isolation kit (Promega) as per manufacturer’s instructions. Isolated chromosomal DNA was sent for Illumina Sequencing (SeqCenter) and mutations were identified using BreSeq [70]. Where appropriate, candidate mutations were reintroduced into the chromosome as described above.

cdGMP reporter assay

The cdGMP reporter assay was conducted as previously described [20] with some alterations. Briefly, cells transformed with the reporter construct (pcdrA:LuX-PMS402) were grown in LB media liquid cultures supplemented with kanamycin (as well as gentamicin for the high control carrying sadC-pBADGr) overnight at 37 °C. Overnight cultures were diluted (1:25) in 3 mL of fresh LB media with antibiotics and grown at 37 °C to mid-logarithmic phase. Cultures were then normalized to an OD₆₀₀ of 0.1 and 2 µL of suspension was added to 200 µL of LB media total supplemented with antibiotics in a black flat clear-bottom 96 well plate. The plate was sealed with masking tape and then incubated overnight in a 37 °C shaking plate reader, with chemiluminescence and OD₆₀₀ readings taken every 15 min for 16–18 hours. Relative luminescence signal was normalized to OD₆₀₀ reading at the same timepoint.

cAMP reporter assay

The cAMP reporter assay was conducted as previously described [39] with some alterations. Briefly, cells transformed with the reporter construct (rpoD:mKate/PaQa:YFP – O’Toole Lab) were grown in LB media liquid cultures supplemented with carbenicillin (as well as gentamicin for the high cAMP control carrying CyaB R456L-pBADGr) overnight at 37 °C. Overnights were sub-cultured 1:25 in 5 mL of M8 minimal media containing 0.2% glucose, 0.5% casamino acids, and 1 mM MgSO₄ supplemented with carbenicillin to an OD₆₀₀ of 0.2-0.4. At this point, 200 µL of liquid culture was spread onto 1% agar M8 media plates dried for 20 min in a biological safety cabinet. Liquid cultures and plates were then incubated for an additional 5 hours at 37 °C. After the incubation, 1 mL of sterile PBS was added to each plate and cells were scraped up with a sterile inoculation loop. 100 µL of cell suspension was taken from each plate and normalized to an OD₆₀₀ of ~0.2 in sterile 1X PBS. The liquid cultures were also normalized to an OD₆₀₀ of ~0.2 in sterile 1X PBS. 200 µL of normalized cell suspensions were then added to a black flat clear bottom 96 well plate (Corning). mKate (Ex: 588/Em: 635) and YFP (Ex: 500/Em: 541) fluorescence were measured, and the signal was normalized by dividing YFP/mKate for both surface-adapted and liquid samples.

Vfr reporter assay

For vfr promoter activity detection, cells harbouring a gene encoding mRuby3 chromosomally integrated upstream of pilM and under control of the pilMNOPQ promoter were streaked onto 1.5% agar LB media plates and incubated overnight at 37 °C. The next morning, cells were harvested into sterile 1X PBS and normalized to an OD₆₀₀ of ~0.1. 200 µL of cell suspension was then added to a black flat clear bottom 96 well plate. mRuby3 (Ex: 558/Em: 600) fluorescence and exact OD₆₀₀ were then measured, and the signal was normalized by dividing mRuby3/OD₆₀₀.

Western blot analysis

Whole cell lysates were first separated on a 15% or 12% SDS-polyacrylamide gels for PilA and PilB respectively. Samples were transferred to a nitrocellulose membrane at 225 mA for 1 hour. The membranes were blocked with 5% (w/v) skim milk (Bioshop) dissolved in 1X Tris-buffered saline (TBS: 20 mM Tris-HCl pH 7.6, 150 mM NaCl) overnight or for 2 hours for PilA or PilB respectively. Primary incubation was performed as previously described using rabbit antisera [71]. The membranes were washed 4 times for 5 min with 1X TBS before incubation using alkaline phosphatase-conjugated goat-anti rabbit IgG secondary antibody (BioRad) in a 1:3500 dilution. The membranes were washed 4 times with 1X TBS and developed using alkaline phosphatase developing reagent according to manufacturer’s protocol, before imaging the membranes in an Azure Biosystems 400 transilluminator.

Skim milk T2SS secretion assay

Liquid cultures were grown overnight in LB media at 37 °C. The next morning, each culture was normalized to an OD₆₀₀ of 0.2 and spotted onto tryptic soy agar plates supplemented with 1.5% skim milk powder (Bioshop) dried for 30 min in a biological safety cabinet. The spots were allowed to dry at room temperature before incubation overnight at 37 °C. The resulting zones of clearance were then imaged with a scanner, and the zone areas were quantified digitally using ImageJ.

Protein modelling and sequence alignments

Alphafold3 was used to predict protein structures and generate predicted aligned error plots using the unmodified default settings [72]. All models and plots were visualized in UCSF ChimeraX [73]. Full length PilB sequences were aligned using Geneious Prime and visualized using ESPript3 [74].

Protein purification

N-terminally truncated PilB (ΔN1D-PilB: M1-D180) and full length FimX were subcloned into pET28b with an N-terminal His₆-tag for inducible expression. PilB was purified as previously described with some alterations. Briefly, ΔN1D-PilB expressing overnight cultures of BL21-CodonPlus (DE3)-RIL were sub-cultured (1:100) in 1 L of LB supplemented with kanamycin and incubated at 37 °C with shaking until mid-logarithmic phase (~0.4-0.8 OD₆₀₀). Cells were induced with 0.5 mM IPTG overnight at 18 °C with shaking. The next morning, cells were harvested by centrifugation at 3000 xg for 20 min. Cell pellets were resuspended in lysis buffer (50 mM HEPES pH 7.3, 250 mM NaCl, 10 mM imidazole) and incubated on ice with 20mg of lysozyme for 15 min with periodic inversion. Cells were then lysed using sonication and the supernatant clarified through centrifugation at 31000 xg for 30 min. Supernatant was then incubated with ~5 mL of Ni²⁺-NTA resin (Millipore) pre-equilibrated with lysis-buffer for 1 hour on ice with gentle shaking. The resin was washed with 3 separate 30 mM imidazole washes (15 mL) before elution with 250 mM imidazole. The elution fraction was dialyzed overnight at 4 °C into 7.3 mM HEPES pH 7.3, 200 mM NaCl and concentrated using an Amicon concentrator (Millipore-Sigma). Protein was then aliquoted, flash frozen in liquid nitrogen, and stored at -80 °C. FimX was expressed as described above and purified as previously described [30], except for the final purification step involving separation by size-exclusion chromatography (Superdex 200, Cytiva) equilibrated using 50 mM Tris-HCl pH 8.0, 250 mM NaCl, and 5 mM β-ME. Protein-containing fractions were determined using SDS-PAGE and then pooled for concentration using an Amicon concentrator.

ATPase activity assay

All ATPase assays contained equimolar amount of ΔN1D-PilB or mutants (100 nM) to ensure valid comparison. PilB reaction buffer was prepared containing 50 mM HEPES pH 7.3, 200 mM NaCl, 5 mM MgCl₂, 1 µM ZnSO₄, and 1 mM ATP. Where applicable, the reaction buffer was supplemented with 1 µM cdGMP, and/or 200 nM of purified FimX. Aliquots of ΔN1D-PilB were thawed and the tube was centrifuged at maximum speed for 10 min at 4 °C to pellet insoluble precipitate. The remaining supernatant was transferred to a fresh tube and kept on ice. ATPase reactions were performed in a clear flat-bottom 96 well plate (Corning) in a total volume of 100 µL at 37 °C in a standing incubator. Following incubation and at the appropriate time point, 50 µL of reaction was mixed with 100 µL of BioMol green reagent (Enzo Lifesciences) incubated at room temperature for 15 min at which point absorbance at 620 nm was read with a spectrophotometer. Each ΔN1D-PilB reaction series was normalized to the timepoint at 0 min to ensure consistent background signal reduction and final values were converted nmoles of P_i using an inorganic phosphate standard curve. For overnight incubations, the 96-well plate lid was sealed with masking tape, and the reaction was performed as described with samples being normalized to an equivalent reaction well lacking PilB.

Fluorescence microscopy experiments

T4P labeling experiments were performed using a cysteine knock-in method described previously, with some species-specific changes [4,75]. 100 µl of overnight cultures of P. aeruginosa were spotted onto prewarmed LB agar plates and grown upright at 37 °C for 4–5 hours. Plate-grown cells were then scraped off of plates using a P200 pipette tip and gently resuspended into 100 µl of fresh LB. AlexaFluor488-maleimide (Sigma) was added to the cell suspension for a final concentration of 50 ng/µl and incubated for 45–60 min in the dark at room temperature. Cells were washed once with 200 µl fresh LB using centrifugation settings of 18,000 x g for 1 min. Cell were resuspended into 200 µl of fresh LB. mNeonGreen-FimX strains were prepared exactly the same way except with gentamycin supplemented in the medium and with no AlexaFluor488-maleimide labeling steps. All imaging was performed under 1% agarose pads made with PBS solution. Cell bodies were imaged using phase-contrast microscopy on a Nikon Ti2-E microscope using a Plan Apo 100X oil immersion objective, a GFP/FITC/Cy2 filter set for pili and mNeonGreen-FimX, a Hamamatsu ORCA-Fusion Gen-III cCMOS camera, and Nikon NIS Elements Imaging Software. To quantify pilus length, number of pili per cell, and extension and retraction rates, cells grown on LB agar plates were imaged by time-lapse microscopy every 10 or 1 s for 1 min or 30 s respectively. Length measurements of pili were quantified manually using the segment function in Fiji from 10 s timelapse images. The maximal length observed in extending T4P that exhibited a subsequent retraction event was quantified by measuring the pilus tip to the cell body. Interrupted or paused extension events were not quantified. Pilus extension and retraction rates were quantified from 1 s interval timelapses. Pili exceeding a length of 0.5 µm that exhibited continuous extension or retraction for at least three consecutive 1 s intervals were selected for rate measurements. A pilus that met these criteria was first measured from the pilus tip to the cell body in a single frame. The same pilus was then measured again from the pilus tip to the cell body in a subsequent frame to determine the starting and final filament lengths following a single extension/retraction event. Extension or retraction rates were calculated as the change in the manually measured length of a single pilus (µm) divided by elapsed time (s) See the dataset at https://doi.org/10.5281/zenodo.16995712 for the raw values used to calculate rates. Notably, populations of P. aeruginosa containing labeled T4P exhibit variability in labeling efficiency for unknown reasons. Therefore, T4P fluorescence signal contrast was adjusted so that T4P were clearly visible for each strain.

Statistical analyses

Microsoft Excel was used to determine averages of technical replicates and GraphPad Prism 8 was used to plot averages of biological replicates and calculate standard deviations. Where indicated, either an unpaired t-test with Welch’s correction or a one-way analysis of variance (ANOVA) with Dunnett’s test for multiple comparison to a control strain were performed (ns: p ≥ 0.05; *: 0.05 ≥ p ≥ 0.01; **: 0.01 ≥ p ≥ 0.001; ***: 0.001 ≥ p). Significance metrics are indicated in the corresponding figure captions.