To confirm that V. cholerae minicells well represent cell pole microcompartments, we introduced hubP-yfp under native promoter in the ΔminD ΔparA1 background and purified minicells by sequential centrifugations for microscopic observation. As shown in Fig 1A, the purified minicells showed discrete fluorescent foci, suggesting that they indeed represent cell pole compartments and likely retain a native pole architecture apart from the disrupted interactions specific to the ParABS1 partitioning system (due to ΔparA1). Therefore, we proceeded to prepare minicells from HubP⁺ (ΔminD ΔparA1) and HubP^- (ΔminD ΔparA1 ΔhubP) strains for the quantitative identification of HubP-dependent changes in the minicell proteomes. We used the iTRAQ (isobaric Tags for Relative and Absolute Quantitation [30]) tandem mass spectrometry method to differentially label and simultaneously determine and compare the amounts of proteins derived from purified HubP⁺ and HubP^- V. cholerae minicells. From a total of ~ 800 identified proteins, ~ 50 were found to be enriched in the HubP⁺ background. Confirming the validity of the assay, they include expected polar proteins such as FlhG and ParC, as well as ~30 downstream chemotaxis proteins including various MCPs (Fig 1B and S1 Table). We then examined the subcellular localization of the 21 remaining proteins by generating plasmids for their ectopic expression as fusion proteins carrying C-terminal monomeric superfolder GFP (hereafter GFP) that were expressed in HubP⁺ and ΔhubP backgrounds. Among them, 12 protein fusions showed diffuse cellular fluorescence and three were found to form inclusion bodies as visible under phase contrast (S1A and S1B Fig). Six proteins–VC0632, VC1210, VC1380, VC1909, VC2232 and VCA0220 –exhibited polar fluorescent foci confirming their localized enrichment at the poles (Figs 1C, 2 and S1C). Importantly, the polar localization of these GFP fusions was severely disrupted in a ΔhubP background with VCA0220 forming additional non-polar foci, similar to ParC distribution in ΔhubP cells (see below), while other proteins exhibited diffuse fluorescence with or without stronger signal at the cell periphery (Fig 1C). Conversely, upon co-expression of the plasmid-based fusions with fluorescent HubP variants (plasmid-based HubP-PAmCherry or chromosomal HubP-YFP expression), all six proteins showed their fluorescence foci co-localizing with the HubP signal (Figs 2 and S1C). Together, these data not only corroborate that the six proteins are specifically targeted to the cell poles in V. cholerae but also that their polar localization is dependent on direct or indirect interactions with the polar landmark protein HubP.

Previous genome-wide assessments of essential genes in V. cholerae suggest that all six polar proteins identified as HubP partners here are dispensable for the viability and normal growth of the bacterium [31,32]. Indeed, we successfully constructed clean deletion mutants for each individual gene and none of them showed significant defects in culture growth (LB, S2A Fig) or irregular cell shape under microscopic observation (S2B Fig). Furthermore, plasmid-based expression of HubP-GFP led to the formation of polar fluorescent foci in all mutant backgrounds, suggesting that none of the six proteins act upstream of HubP in cell pole targeting (S2C Fig). Predicted protein folds, oligomeric states, topology and domain architectures for the four proteins are shown in S3 Fig.

VCA0220 has been previously annotated as hemolysin secretion protein HlyB due to indirect evidence of its involvement in secretion of the cytotoxic HlyA hemolysin encoded by an adjacent gene [33]. The protein is not to be confused, however, with homologs of the HlyB ATPase binding cassette (ABC) transporters of E. coli or Bordetella pertussis that are responsible for hemolysin secretion by the assembly of a HlyB-HlyD-TolC Type I secretion system ([34] for review). Lines of evidence, including sequence and fold prediction analyses, reveal VCA0220 as a classical chemotaxis transducer protein with a periplasmic Tar/Tsr ligand binding module and intracellular HAMP and methyl-accepting domains in tandem ([35,36]; S3A and S3B Fig), suggesting that any potential roles in hemolysin secretion are likely indirect. Indeed, the subcellular localization pattern of the protein in ΔhubP cells is very similar to that of ParC and downstream chemotaxis proteins (Fig 1C; [11]). In addition, we show here that HlyB is often displaced in a ΔparC background (S1D Fig), which further indicates that the protein belongs to the MCP superfamily of chemotaxis array-forming methyl-accepting signal transducers [36].

VC0632, hereafter DacB_Vc, is a conserved homolog of the E. coli DacB protein (DacB_Ec), also known as penicillin-binding protein 4 (PBP4) [37]. DacB homologs are bifunctional, homodimeric periplasmic proteins with known molecular architecture and peptidoglycan DD-endopeptidase and DD-carboxypeptidase activities (S3A and S3C Fig). Although DacB proteins are not essential for peptidoglycan biogenesis and cell growth neither in E. coli, nor V. cholerae, they have been proposed to cooperate with other cell wall modifying enzymes in envelope maturation and maintenance [37–39]. Indeed, the protein’s HubP-dependent targeting to the pole, as well as HubP’s own dependence on peptidoglycan binding for polar localization [11] support a model of cell pole-specific cell wall remodeling mechanisms.

VC1210 is annotated as a hypothetical, 31 kDa protein of unknown function and homologs of it are found only in a subset of Vibrio species. The subcellular localization of the VC1210-GFP fusion in a ΔhubP background suggests that it is expressed in the cytosol, which is consistent with the lack of a detectable N-terminal signal peptide or other motifs for protein secretion (Figs 1C and S3D). Fold prediction analyses suggested significant structural similarity to aminoglycoside nucleotidyl transferases that are involved in the development of resistance against aminoglycoside antibiotics, however, the Δvc1210 mutant did not show significant difference in sensitivity to kanamycin or gentamycin and the protein’s function at the pole remains unknown (S3E Fig). Hypothetical protein VC1380 is a short, 116 amino acid-long polypeptide which does not feature detectable conserved domains, transmembrane segments or secretion signal motifs (S3F Fig). Consistent with the localization pattern of the VC1380-GFP fusion in ΔhubP cells, the protein is indeed likely produced in the cytosol (Fig 1C), however its specific function remains to be further examined.

VC1909 is annotated as a hypothetical protein with unknown function and is predicted to have a single N-proximal transmembrane helix leaving the remaining ~ 230 C-terminal amino acids in the cytosol (S3A and S3G Fig). Interestingly, VC1909 is a homolog of the Aliivibrio (formerly Vibrio) fischeri protein VF1491, which was identified as a putative motility regulator in soft-agar motility assays on an A. fischeri transposon mutant collection [40]. In addition, a recent study reported that VC1909/VF1491 homologs co-localize and interact with HubP in Shewanella putrefaciens and V. parahaemolyticus and affect flagellar rotation reversals required for cell tumbling and changes in swimming direction [41]. Fold prediction tools detect similarities of the C-terminal VC1909 region to MIT (microtubule interacting and trafficking) and TPR (tetratricopeptide repeat) helical motifs that are often found in scaffolding proteins and can directly mediate protein-protein interactions in multi-component macrocomplexes [42]. Indeed, deletion of the cytoplasmic region of VC1909 resulted in diffuse localization of the GFP fusion, suggesting that interactions through the cytoplasmic domain are indeed essential for the protein’s polar localization (S3I Fig).

Finally, VC2232 is also annotated as a hypothetical protein with unknown function. It is predicted to be synthesized as a signal peptide-containing precursor which is exported in the periplasm and remains anchored to the inner membrane via a single C-proximal transmembrane helix and a short, positively charged cytosolic tail. Following cleavage of the signal peptide, the large periplasmic region is predicted to fold into three immunoglobulin-like domains. The latter are highly stable structural modules that are often found in multiple repeats and can mediate protein-protein interactions in periplasmic or cell surface molecules such as pili, non-fimbrial adhesins, periplasmic chaperones, usher proteins, ABC transporters, and a variety of enzymes [43] (S3A and S3H Fig). Truncated VC2232 mutants show that while the cytosolic tail is dispensable for the protein’s polar localization, its transmembrane region remains essential for proper cell pole targeting (S3J and S3K Fig). Indeed, the membrane-proximal periplasmic amino acids together with the inner membrane anchor are predicted to fold into a continuous coiled coil which likely stabilizes the functional quaternary structure of the protein.

Since ΔhubP mutants in V. cholerae, as well as deletion mutants of the vc1909 homologs in S. putrefaciens or V. parahaemolyticus show defects in motility in soft agar plates [11,41], we examined the role of newly identified HubP partners in flagellar motility. As shown in Fig 3A, Δvc0632/dacB_vc, Δvc1210, Δvc1380 and Δvca0220/hlyB mutants are motile with wild-type (WT) levels of motility in soft agar plates. In contrast, the Δvc2232 mutant showed motility defects similar to that of the ΔhubP mutant, while the Δvc1909 mutant was not able to expand its growth outside the inoculated surface. These data indicate direct involvement of not only VC1909, but also VC2232 in motility regulation and hereafter we refer to the two genes as MotV and MotW, respectively. The rest of the study will mainly focus on these two proteins and their involvement in motility.

Motility defects can have different underlying causes such as lack of flagellum assembly or compromised flagellar integrity, expression of multiple appendages leading to lopho-, peri- or amphitrichous flagellation, or defects in the rotational or directional switching dynamics of the assembled polar appendage. Cell imaging by electron microscopy revealed that ΔmotV, ΔmotW and ΔmotV ΔmotW mutants feature single, seemingly intact flagella at the cell poles, morphologically undistinguishable from those in WT cells (S2D Fig). This result indicates that while ΔmotV and ΔmotW mutants are able to assemble flagellar structures, they are likely compromised in the rotational dynamics or directional switching of their motility organelles.

ΔhubP cells were shown to maintain their swimming capabilities but display a significant bias toward straight swimming with turning events observed 3–4 times less frequently and with lower velocity than WT cells (Fig 3C and 3D, [11]). To further examine how flagellar motility was affected in the ΔmotV and ΔmotW mutants and whether the motility defects correlate with those of the ΔhubP strain, we performed single-cell video-tracking experiments. Similar to the ΔhubP cells, the ΔmotV mutant exhibited defects in turning. Nevertheless, these defects were much more severe in the ΔmotV mutant, with the majority of cells swimming straight across the imaging field and never tumbling for the duration of the video-tracking experiments (Fig 3B and 3C). In addition, the average speed of swimming was somewhat higher in the ΔmotV mutant compared to WT (Fig 3D), which can be explained by the lack of tumbling and confirms defects in directional switching but not in rotational dynamics of the flagella.

The soft-agar growth phenotype and video-tracking experiments revealed seemingly conflicting results for the ΔmotV mutant. Indeed, the cells were found to swim with speeds comparable or superior to WT V. cholerae, however, the colonies were unable to expand from the inoculated soft-agar area. To reconcile these results, we carried video-tracking experiments on cells embedded in semisolid agarose between the microscope slide and cover slip. In these conditions, mimicking the motility assay in soft agar on plates, the ΔmotV cells were often found stuck against the agarose matrix and exhibited very little displacement ability compared to WT V. cholerae (Fig 3B and 3E). Together, these results confirm that both the swimming by flagella rotation and directional switching are necessary for colony expansion in non-liquid media, with cell turning likely required to overcome the physical obstacles presented by the media matrix.

In contrast to the ΔmotV cells, the ΔmotW mutant showed defects in both the tumbling frequency and swimming speed, indicating a distinct role in motility regulation (Fig 3B, 3C and 3D). Furthermore, we examined the motility of the ΔmotV ΔmotW double mutant. Although the double mutant assembled monotrichous flagella similar to WT V. cholerae, the strain was totally defective in soft agar expansion, similar to the ΔmotV single mutant. However, more severe motility defects were observed in video-tracking experiments in liquid: the vast majority (85%) of cells were immotile throughout the entire length of recording (~13 seconds), indicating that MotV and MotW work separately to control flagellar motility.

Whereas MotV homologs have been recently shown to interact with HubP in bacterial two hybrid assays [41], there are no reports of HubP-dependent cell pole recruitment of MotW homologues in the literature. Microscopy of truncated mutants suggested that MotW’s transmembrane region and periplasmic domains are required for its HubP-dependent targeting to the pole, whereas its short cytosolic tail appears dispensable for proper localization (S3J and S3K Fig). To corroborate interactions between HubP and MotW, we generated several strains encoding FLAGx3- or HA-tagged HubP, MotW and MotV proteins from their endogenous chromosomal loci. Strains expressing FLAGx3 or HA-tagged HubP or MotW exhibited unperturbed morphology, cell growth and motility phenotypes (S4B Fig) and western blot analyses of the cell lysates exhibited specific signals for the respective epitope tags on the two proteins: the SDS-PAGE profile of tagged MotW is consistent with its theoretical monomeric molecular weight in the ~50 kDa range, whereas HubP’s running behavior is consistent with that of a detergent-resistant dimer in the high molecular weight region (S4C Fig). Eluted samples following stringent affinity pull-downs using anti-FLAG or anti-HA affinity resins were found to indeed contain both proteins further indicating direct or indirect interaction between the two polar markers (S4D Fig). Finally, bacterial two-hybrid assays based on the functional complementation between adenylate cyclase fragments fused to HubP’s and MotV’s periplasmic domains further support interaction between the two proteins, which is likely stabilized by additional factors in their native periplasmic context in vivo. Similar attempts to detect MotV-HubP interactions were not successful, as the insertion of epitope-encoding sequences at the 5’ or 3’ ends of the motV gene invariably led to strains featuring non-motile phenotypes and no MotV-specific bands were detected upon analyses of the corresponding cell lysates (e.g. S4C Fig).

While MotV-GFP and MotW-GFP exhibited similar bipolar foci in hubP⁺ cells when expressed from plasmids (Figs 1C and 2), it is known that protein expression levels can affect uni- and bi-polarity of polar landmark proteins and their interaction partners [12,44]. To avoid possible overexpression effects resulting from the inducible plasmid-based expressions described above, we generated V. cholerae strains expressing fluorescent fusion proteins at the native motV and motW loci and under control of the respective endogenous promoters (for expression of MotV-CFP or MotW-CFP, respectively). Although the overall co-localization of the two proteins with HubP remained consistent with the plasmid-based co-localization data, we observed increased proportion of cells carrying unipolar MotW loci in cells showing bipolar HubP (HubP-YFP) fluorescence (Fig 2). In addition, whereas the plasmid-based localization experiments indicated that polar localization of both MotV and MotW is HubP-dependent, the cell pole-targeting of each protein was independent of that of the other (S4A Fig). Together, these results further indicate distinct involvement of MotV and MotW in motility. In addition, different timing for the expression and/or local recruitment of HubP and its downstream interaction partners suggests complex maturation of the cell poles with long-axis asymmetry.

To further explore the function of MotV and MotW in motility and to understand in which pathways MotV and MotW are involved, we carried out suppressor development. Spontaneous mutants of the ΔmotV and ΔmotW strains that showed improved motility could be detected and selected in soft agar plates without treatment with mutagenic agents. (Figs 3A and S5). Whole-genome sequencing of the motility reversal isolates showed that four proteins, VieS (VC1653), FliM (VC2126), FliG (VC2132) and CheV4 (VCA0954), were often found to be mutated in suppressor strains (Table 1). FliG mutations were found in both ΔmotV and ΔmotW suppressors, while FliM mutations were only found in ΔmotV suppressors, together or not with mutations in FliG (Table 1). Interestingly, FliG and FliM interact directly and together with multiple copies of FliN build the flagellar C-ring complex, a conserved cytosolic component responsible for both torque generation and directional switching and therefore essential for flagellar motility. In accordance with this, ΔfliG and ΔfliM mutants are fully deficient in motility in soft agar plates (Fig 3A).

As opposed to the ubiquitous C-ring components, only a limited subset of flagellated bacteria encode MotV and MotW homologs (Fig 4A). Given the emergence of multiple suppressor strains with mutations in the fliG and fliM genes, we hypothesized that motV⁺/motW⁺ bacteria could possess distinct amino acid residues or motifs in their FliG and/or FliM components that are responsible for MotV- and/or MotW-specific regulatory inputs through the C-ring. Indeed, phylogenetic analyses of FliG and FliM identified amino acids conserved only in motV⁺/motW⁺ species. Remarkably, FliG^G119 and FliM^D180 were among these residues and both were found to be mutated in the ΔmotV suppressor mutant A2 (Table 1 and Fig 4B). FliG^G119 maps at the linker connecting the N-terminal and middle domains of the FliG C-ring unit, which has been proposed to undergo significant conformational changes during directional switching of the flagellum. FliM^D180, on the other hand, maps on the inner surface of FliM’s middle domain, close to the interface with the protein’s C-terminal domain (Fig 5A, 5B and 5C). Importantly, the two residues are physically distant and do not partake in pairwise interactions (Fig 5B and 5C).

To further examine the functional significance of these co-evolved residues and their role in ΔmotV suppression, we tested the effects of FliG^G119D and FliM^D180N mutations by introducing them into WT and ΔmotV background, independently or in combination. In the motV⁺ background, the fliG^G119D mutation abolished the motility in soft agar, similar to fliG deletion, indicating functional incompatibility. In contrast, the fliM^D180N motV⁺ mutant retained ~80% of motility relative to WT (Fig 3A), but introducing the same fliM^D180N mutation in a ΔmotV background did not rescue the immotile phenotype. Strikingly, introducing the fliM^D180N mutation in the fliG^G119D motV⁺ background overrode the motility defect of the fliG^G119D mutation (Fig 3A). Furthermore, a fliG^G119D fliM^D180N ΔmotV triple mutant retained motility in soft agar, at ~50% relative to WT V. cholerae and comparable to that of suppressor mutant A2, which carries some additional mutations (Table 1 and Fig 3A). In addition to motility in soft agar, video-tracking experiments indicated that the fliG^G119D fliM^D180N ΔmotV triple mutant recovered the ability to tumble (Fig 3B). Furthermore, this mutant swam as fast as the ΔmotV mutant in liquid and was not trapped in the 0.25% agarose matrix (Fig 3B, 3D and 3E). In contrast, FliG^G119D and FliM^D180N mutations did not suppress motility defects by ΔhubP or ΔmotW (Fig 3A). Together, these results suggest that the combination of FliG^G119D and FliM^D180N mutations were sufficient to suppress the effect of ΔmotV, probably transforming the C-ring to be function-independent of the MotV protein.

ΔmotV suppressor A1 was also found to harbor a mutation in fliG, FliG^F199L, which maps at the second FliG interdomain linker (MC) known to undergo significant conformational changes upon directional switching (Fig 5A). Introducing the FliG^F199L mutation in ΔmotV background slightly enhanced motility in soft agar but the rescue phenotype was not comparable to that of suppressor A1. Interestingly, A1 also features a 14-base-pair deletion in cheV4, leading to a frameshift from the protein’s 161th amino acid and emergence of a stop codon 2 amino acids later (thus the mutant will be referred here as CheV4^1-161). Remarkably, introducing cheV4^1-161 to the ΔmotV mutant was sufficient to rescue motility in soft agar and the resulting ΔmotV cheV4^1-161 cells also showed restored ability to swim and tumble in video-tracking analyses (Fig 3B and 3C). CheV proteins are CheW-CheY hybrids (Fig 5D), where the CheW module is proposed to bridge a CheA histidine kinase to its cognate MCP and allow phosphorylation of the CheY module, which in turn would bind the N-terminal region and middle domain of FliM to favor flagellar directional switching and cell tumbling. The observed suppressor mutation cheV4^1-161 is equivalent to removal of the CheY module from the CheV4 protein, possibly by allowing the CheW module to transmit sensory inputs in a MotV-independent manner through interactions with canonical CheY proteins. CheV4 and CheV4^1-161 localize at the cell pole (Fig 5F) and feature mainly unipolar localization similar to that of the better-studied CheW1 protein [22,23]. Unlike CheW1, however, CheV4 and CheV4^1-161 remained polar and without additional foci in a ΔparC background. Interestingly, CheV4 was found to be enriched in the HubP⁺ minicells (S1 Table). Indeed, in a ΔhubP background CheV4 distribution shows the emergence of additional foci, suggesting that polar localization of the CheV4 protein could be at least partly HubP- and/or MotV-dependent (Fig 5F).

VieS mutations were only found in ΔmotW suppressors and not in ΔmotV suppressors. VieS is a hybrid sensor kinase which has been shown to positively regulate motility by activating the phosphodiesterase (PDE) VieA and altering the cellular levels of the second messenger c-di-GMP [46]. Interestingly, observed VieS suppressor mutations (VieS^A669V, VieS^H1002Y and VieS^T1030I) map at or near residues involved in canonical phosphotransfer through the protein’s intrinsic receiver domain, raising the possibility for direct phosphorylation of the cognate phosphodiesterase instead (Fig 5G). Although а ΔvieS mutant did not show motility defects in our hands, VieS mutants with any of the above single amino acid substitutions exhibited hypermotile phenotypes in soft agar plates (S5 Fig), suggesting that motility defects of the ΔmotW mutant could be compensated by VieS/VieA-induced hypermotility in the suppressors.

Unless otherwise noted, the whole-genome sequenced N16961 strain [60] was used as WT V. cholerae. Plasmids were constructed by either conventional digestion-ligation methods or by isothermal assembly [61]. List of bacterial strains, plasmids, and oligonucleotides are indicated in Supporting information S2, S3 and S4 Tables, respectively. All expression constructs were verified by Sanger DNA sequencing (GATC Biotech/Eurofins Genomics, Ebersberg, Germany). Mutants were generated by allelic exchange [62]. Plasmids were introduced in V. cholerae by electroporation. Unless otherwise specified, bacterial cells were grown in lysogeny broth (LB) or LB agar (1.5%) and, as appropriate, antibiotics were used at the following concentrations: ampicillin 100 μg/mL, chloramphenicol 25 μg/mL (for E. coli) or 5 μg/mL (for V. cholerae), kanamycin 25 μg/mL, streptomycin 100 μg/mL. Cell growth curves and growth rate measurements were done by microplate reader (Tecan Infinite M200 Pro, Tecan, Switzerland) [63].

In order to avoid false positives resulting from spontaneous reversal of the BTH101 cyaA mutant strain [64], the bEYY2122 host strain was first from constructed from the latter by deletion of the adenylate cyclase cyaA gene altogether via transduction of ΔcyaA::kan followed by flipping out of the kan resistance cassette. Two test plasmids encoding complementary adenylate cyclase fragments fused to the protein domains of interest, the Gcn4 leucine zipper motif (positive control) or without additional protein sequences (negative control) were subsequently introduced and liquid cultures of cotransformant colonies were spotted on M9 maltose agar plate containing 50 μg/mL of 5-Bromo-4-Chloro-3-Indolyl β-D-Galactopyranoside (X-gal), 500 μM of Isopropyl β-D-1-thiogalactopyranoside (IPTG) and appropriate antibiotics. The plates were then incubated for 2 days at 30°C and protein interactions were evaluated by both colony growth and development of blue coloration resulting from functional complementation of the split adenylate cyclase protein.

Overnight cultures of bEYY1208 (ΔminD ΔparA1) and bEYY1319 (ΔminD ΔparA1 ΔhubP) were prepared in LB broth at 37°C with agitation (170 rpm). 3 mL of bEYY1208 and 7 mL of bEYY1319 overnight culture were inoculated in 3 L and 7 L of fresh LB media, respectively, and the cell cultures were grown until OD_{600 nm} = 0.8. Cells were harvested as follows: rod cells were sedimented by low speed centrifugation (3000 g at 4°C) for 7 min, then the supernatant was transferred into a new flask and centrifuged at higher speed (9000 g at 4°C) for 20 min. The pellet, enriched in minicells, was resuspended in LB and the series of centrifugations was repeated with an extended low-speed step of 40 min. Purity of the minicell preparations was examined by microscopy and viability on LB plate. Total protein amounts in the minicell preparations were quantified by Coomassie (Bradford) Protein Assay Kit (Pierce Biotechnology, Rockford, IL). Purified pelleted minicells with > 11 μg of total protein content were flash-frozen in liquid nitrogen. Two replicates of each mutant (hubP+ and hubP-) were subjected to iTRAQ by ITSI-BIOSCIENCES (Johnstown, PA).

Cells were grown in M9 minimal media supplemented with glucose (0.2%), casamino acids (0.1%) and thiamine (1 μg/mL) at 37°C with agitation (180 rpm). When necessary, fluorescent protein fusions encoded on overexpression plasmids were induced with 0.02% arabinose for 1 h prior to analysis. Cells in exponential phase were spotted on agarose pads (1% agarose in 1x M9 media) that was mounted on the glass slide. Cell images were acquired using DM6000-B (Leica, Wetzlar, Germany), Zeiss Observer.Z1 (Carl Zeiss, Oberkochem, Germany) and Eclipse Ti (Nikon, Tokyo, Japan) microscopes with the Metamorph software (Molecular Devices, San Jose, CA). Image J, Adobe Photoshop and an ImageJ plug-in MicrobeJ [65] were used for image processing and analyses.

For the soft agar plate-based assays, cells were grown overnight in LB at 37°C with agitation. 1 μL of culture was pricked into semisolid LB agar (0.3%) plates. The WT strain was always included in each test plate as standard. Plates were incubated overnight and the swimming diameter of each strain was measured and compared to the control.

For video-tracking, cells were grown in EZ Rich Defined Media (Teknova Inc, Hollister, CA) to exponential phase. For liquid environments, 2 μL of each culture were directly spotted onto the microscope slide. For semisolid environments, each cell culture was first mixed with agarose (final concentration 0.25% w/v) in EZ media and then 2 μL of the mixture was placed on a microscope slide. After placing a coverslip on top, images were recorded with the Leica microscope (see above). For liquid environments, we used a streaming acquisition function with 37 msec frequency of image capture. For semisolid environments, images were taken every 20 sec for 30 min. At least two individual experiments with three replicates were performed. Trajectories of individual cells were tracked using the ImageJ plug-in MTrackJ [66]. The swimming speed of each cell was calculated as the total travel distance divided by the total time spent in the viewing field, thus cells with more tumbling could be assigned smaller swimming speed values. The number of tumbling events was counted, and the tumbling frequency was calculated as the sum of the travel distance divided by the total number of tumbling events from each sample. For the semisolid environments, we implemented mean square displacement (MSD) analysis using a Matlab script as described previously [67].

2-mL WT and mutant V. cholerae cultures were grown to exponential phase in antibiotics-free LB medium and under mild agitation to prevent surface shearing of the flagella. Cells were gently sedimented, washed with 1x Tris-buffered saline (TBS) and resuspended in the same buffer. 5 μL of the cell suspension were spotted on glow-discharged carbon-coated copper grids (Agar Scientific, Stansted, United Kingdom). After 1 min incubation, the extra liquid was blotted off and the grids were passed sequentially through three drops of 2% w/v uranyl acetate solution, with 15–30 sec incubation in the last drop, before blotting and air-drying. Images were taken on a 200 kV Tecnai F20 Thermo Fischer transmission electron microscope equipped with an Eagle 4k x 4k CCD camera at nominal magnification of 5000 x for the representative micrographs in S2D Fig. and a range of lower magnifications to allow throughput for flagella vs. cell number quantification.

Spontaneous suppressor mutants of ΔmotV and ΔmotW strains that show restored motility were isolated as follows. 1 μL (~10⁶ cells) of overnight culture of the test strain was inoculated in the motility plate and incubated for overnight at 37°C. Cells that traveled farthest were recovered for the second round of screening. After 4 (for ΔmotV) or 5 (for ΔmotW) rounds of screening in the motility plate, single colonies were isolated and confirmed for the suppressor phenotype. For each ΔmotV and ΔmotW strain, 8 independent experiments were carried out and 6 mutants were arbitrarily selected for whole-genome sequencing (WGS).

5 μg of genomic DNA in 130 μL H₂O was sheared with a Covaris S220 focused ultrasonicator (Covaris, Wobum, MA) with target peak at 500 bp. After blunting the DNA ends with T4 DNA polymerase (New England Biolabs, Ipswich, MA) and Klenow Enzyme (Roche, Basel, Switzerland) followed by A-tailing with Klenow Fragment (New England Biolabs), adaptor DNA [68] was ligated. DNA fragments ranging 400~900 bp in size were purified using Pippin Prep (SAGE Science, Beverly, MA), then amplified with Illumina primers PE1.0 and PE2.0. PCR products were purified using QIAGEN MinElute columns (QIAGEN, Hilden, Germany) then size-selected using AMPure XP beads (Beckman Coulter, Brea, CA).

Illumina sequencing (Single-end 150 bp or Paired-end 75 bp) results were accordingly trimmed and mapped to the bEYY1013 sequence using Bowtie I [69] or Burrows-Wheeler Aligner [70]. Bam files were converted to mpileup format then SNPs were identified by a script VarScan [71], as well as visual screening with Integrative Genomics Viewer [72]. Final confirmation of SNPs and small deletions were made by PCR followed by Sanger sequencing. WGS data is available from the ArrayExpress database (accession number # E-MTAB-10691).

Prediction of protein topology and modeling of protein folds was carried out with Phobius [73], GenomeNet (https://www.genome.jp/en/release.html), Robetta (https://robetta.bakerlab.org), Phyre2 (http://www.sbg.bio.ic.ac.uk/~phyre2), CCfold (http://pharm.kuleuven.be/Biocrystallography/cc) and AlphaFold [74]. Protein fold visualizations were carried out in Pymol (Schrödinger, New York, NY) and Chimera UCSF (https://www.rbvi.ucsf.edu/chimera). Gene co-occurrence was analyzed with STRING (https://string-db.org/) and phosphorylation residue conservation was visualized with WebLogo (https://weblogo.berkeley.edu). For phylogenetic analyses, we used gammaproteobacterial species available in the UniPlot database (https://www.uniplot.org), excluding ones with very small genome (less than 3000 genes) and redundant species. FliG and FliM homologs were retrieved using their pfam numbers (PF01706 and PF02154, respectively). Alignments were done by ClustalOmega [75] followed by visualization in Unipro UGENE [76].

For epitope tagged protein expression and identification of HubP- and MotW-specific bands, V. cholerae strains expressing differentially tagged HubP and MotW proteins were first grown in LB and whole cell lysates were subjected to SDS-PAGE and western blot analyses (see below). Multiple strains designed to express epitope-tagged MotV variants were also tested, however, no signal corresponding to a tagged MotV protein was detected.

The V. cholerae strain expressing HubP-3xFLAG and MotW-HA fusions from their native loci was grown in 2 L of LB at 37°C until OD_{600 nm} = 0.6. Cells were then centrifuged and resuspended in lysis buffer containing 20 mM HEPES pH 8.0, 100 mM NaCl, and cOmplete, EDTA-free protease inhibitors (Roche) and flash-frozen in liquid nitrogen. For membrane protein extraction, cells were thawed, lysed using an Emulsiflex C5 cell homogenizer (Avestin, Ottawa, Canada) and centrifuged at 12 000 x g for 15 min to remove cell debris. The supernatant was then subjected to ultracentrifugation using an SW28 Ti rotor at 26 500 rpm and 4°C for 1 h to pellet the membrane fraction. The latter was resuspended using a Potter-Elvehjem homogenizer in membrane resuspension buffer containing 20 mM HEPES pH 8.0, 120 mM NaCl, 10% glycerol, 5 mM MgCl₂, 10 μM AppCp (Jena Bioscience, Jena, Germany), 2 μM c-di-GMP (MilliporeSigma, Burlington, MA) and 1 tablet per 50 mL cOmplete protease inhibitors. A mix of detergents was then added for membrane solubilization as follows: 0.4% w/v digitonin (MilliporeSigma), 0.4% w/v n-dodecyl-β-D-maltopyranoside (anagrade β-DDM, Anatrace, Maumee, OH), 0.4% w/v decyl maltose neopentyl glycol (DM-NPG, Anatrace), and 0.2% w/v lauryl maltose neopentyl glycol (LM-NPG, Anatrace). Following 1 h incubation at 18°C and under mild agitation, the solubilized membrane fraction was cleared by a second high-speed centrifugation as above. The supernatant was then incubated with anti-FLAG M2 affinity gel (100 μL resin/L of culture, MilliporeSigma), under mild agitation at 4°C for 1 h. After gravity elution of the non-bound fraction, the resin was washed extensively with affinity buffer containing all membrane resuspension buffer components and 0.008% w/v LM-NPG (>30 column bed volumes allowed to flow through the resin sequentially). The bound complexes were then eluted using four column bed volumes of elution buffer (affinity buffer supplemented with 3xFLAG peptide at 100 μg/mL), concentrated on a 100 kDa cut-off Amicon Ultra (MilliporeSigma) centrifugal filter and loaded on a 4–20% SDS-PAGE gradient gels for western blot analysis. Following gel electrophoresis, the migrated proteins were transferred onto 0.2 μm PVDF membranes (Amersham Hybond P, GE Healthcare, Chicago, IL) using a Trans-blot Turbo Transfer system (Bio-Rad, Hercules, CA). After a blocking step with 5% skim milk in TPBS (0.05% Tween-20 in 1× PBS), immunoblots were probed with mouse primary antibodies (anti-FLAG M2 antibody (dilution 1:1000, Sigma-Aldrich) or anti-HA antibody (dilution 1:1000, Thermo Fisher Scientific, Waltham, MA). Alexa-Fluor 680-conjugated donkey anti-mouse antibody (dilution 1:10,000, Abcam, Cambridge, United Kingdom) was used as secondary antibody. The signal was detected using a Li-Cor Odyssey Fc system (LI-COR Biosciences, Lincoln, NE) in the 700 nm channel.