A Tripartite Efflux System Affects Flagellum Stability in Helicobacter pylori

Helicobacter pylori uses a cluster of polar, sheathed flagella for swimming motility. A search for homologs of H. pylori proteins that were conserved in Helicobacter species that possess flagellar sheaths but were underrepresented in Helicobacter species with unsheathed flagella identified several candidate proteins. Four of the identified proteins are predicted to form part of a tripartite efflux system that includes two transmembrane domains of an ABC transporter (HP1487 and HP1486), a periplasmic membrane fusion protein (HP1488), and a TolC-like outer membrane efflux protein (HP1489). Deleting hp1486/hp1487 and hp1489 homologs in H. pylori B128 resulted in reductions in motility and the number of flagella per cell. Cryo-electron tomography studies of intact motors of the Δhp1489 and Δhp1486/hp1487 mutants revealed many of the cells contained a potential flagellum disassembly product consisting of decorated L and P rings, which has been reported in other bacteria. Aberrant motors lacking specific components, including a cage-like structure that surrounds the motor, were also observed in the Δhp1489 mutant. These findings suggest a role for the H. pylori HP1486-HP1489 tripartite efflux system in flagellum stability. Three independent variants of the Δhp1486/hp1487 mutant with enhanced motility were isolated. All three motile variants had the same frameshift mutation in fliL, suggesting a role for FliL in flagellum disassembly.


Introduction
Helicobacter pylori, a member of the phylum Campylobacterota, colonizes the human gastric mucosa where it can cause a variety of diseases, including chronic gastritis, peptic and duodenal ulcers, B cell MALT lymphoma and gastric adenocarcinoma [1]. H. pylori possesses a cluster of polar, sheathed flagella that it uses to migrate through the viscous mucus layer covering the stomach epithelium. Flagellar-mediated motility is required for host colonization, as non-motile H. pylori mutants are unable to colonize animal models [2,3].
The bacterial flagellum consists of three distinct parts referred to as the basal body, hook, and filament. The basal body includes a rotary motor and the flagellar protein export apparatus that transports axial components of the flagellum (e.g., rod, hook, and filament proteins) across the inner membrane [4,5]. Components of the archetypical flagellar motor of Escherichia coli and Salmonella enterica include the MS ring, C ring, rod, L ring, P ring, and a stator unit comprised of a complex of the MotA and MotB proteins. The stator unit is the torque generator for the motor and is powered by the proton motive force [6]. The MS and C rings form the rotor and transmit torque through the rod and hook to the filament, which acts as a propeller to push the cell forward [7]. The H. pylori flagellar motor accommodates up to 18 stators [8], making it the largest bacterial flagellar motor described to date. In sheath biosynthesis or function, we searched for homologs of H. pylori proteins that were prevalent in FS + Helicobacter species but underrepresented in FS − Helicobacter species. Using blastp, predicted proteins encoded in the genomes of 35 FS + Helicobacter species and 9 FS − Helicobacter species were examined for homologs of the 1458 predicted proteins encoded in the H. pylori G27 genome. The H. pylori G27 genome was chosen as a reference genome for the analysis since G27 is a model for studies of motility and flagellum biosynthesis in H. pylori. Helicobacter species and strains analyzed are listed in Table S1. The blastp comparison was performed with an E-value cutoff of 10 −20 . Under the conditions of the null hypothesis (that any protein is equally likely to have a homolog in the FS + species as in the FS − species), the number of homologs in FS + species is expected to follow the hypergeometric distribution. We therefore used the hypergeometric test (equivalent to one-tail Fisher's exact test) to obtain a p-value for each protein. False discovery rate (FDR) [30] and a more conservative Bonferroni correction were used for multiple testing correction. Results from the blastp analysis for all 1458 H. pylori G27 proteins are presented in Tables S2 and S3. Forty-two proteins that satisfy the Bonferroni-adjusted cutoff for statistical significance (p-value < 3.4 × 10 −5 ) in FS + Helicobacter species are listed in Table 1. An additional 35 proteins qualified as significantly overrepresented in FS + Helicobacter species when the less conservative FDR correction was applied (Table S2).  To test the predictions from the comparative genomics approach, we targeted some of the identified genes for deletion analysis. Homologs of four proteins, HP1486, HP1487, HP1488 and HP1489, were present in all the FS + Helicobacter species and absent in all the FS − Helicobacter species (Table S2), which made the genes encoding these genes good candidates for further investigation. Based on the functional annotations of the proteins and the linkage of the genes encoding them within an operon (Figure 1), these proteins are predicted to form a tripartite efflux system. Absent in the hp1489-hp1486 locus, however, is a gene encoding the ATPase for the ABC transporter, which presumably is present at a different locus in the genome. The H. pylori 26695 and G27 genomes encode 19 and 18 predicted proteins that contain an ATP-binding domain of an ABC transporter (pfam00005), respectively. Most of the H. pylori genes encoding ATP-binding domain proteins are associated with genes encoding other ABC transporter genes or encode an ABC transporter transmembrane domain, suggesting these genes do not encode the ATPase that functions with HP1486/HP1487 transmembrane domain proteins. A blastp analysis of the sequences of all the ATP-binding domain proteins from H. pylori 26695 versus the genomes of FS + and FS − Helicobacter species failed to identify a clear candidate for the ATPase of the HP1486/HP1487 ABC transporter that was supported by the protein functional annotation or yielded orthologous homologs. To determine if the HP1489-HP1486 efflux system has a role in flagellar biosynthesis or function in H. pylori, we deleted hp1489 and hp1487-hp1486 in H. pylori B128, and also deleted hp1489 in H. pylori G27, using a sucrose-based counter-selection method [47]. Motilities of all three mutants in soft agar medium were reduced 33 to 50% compared to the wild-type parental strains (Figure 2). Introducing a copy of hp1489 into the H. pylori G27 Δhp1489 mutant on a shuttle vector rescued motility ( Figure 2B), indicating loss of hp1489 was responsible for the motility defect in the mutant. Examination of the H. pylori B128 ∆hp1487-hp1486 and Δhp1489 mutants by transmission electron microscopy (TEM) revealed the strains possessed fewer flagella per cell than wild type ( Figure 3). Flagella of the Δhp1487-hp1486 and Δhp1489 mutants were sheathed, indicate sheath biosynthesis was not grossly impaired in the mutants. Taken together, these observations suggest the HP1489-HP1486 efflux system is required for normal flagellum biosynthesis and/or motility.  H. pylori 26695 operon that contains genes encoding the HP1486-HP1489 tripartite efflux system. Only the first 15 genes of the potential operon are shown. The larger arrows indicated the predicted open reading frames and are drawn approximately to scale. hp1489, hp1488, hp1487, and hp1486 are indicated in blue. The operon includes two flagellar genes, flgA (encodes chaperone for the P ring protein) and pflB (paralyzed flagella protein B). Other genes of known function are uvrD (encodes DNA helicase II), serS (encodes seryl-tRNA synthetase), xseB (encodes exodeoxyribonuclease VII small subunit), and ubiE (encodes demethylmenaquinone methyltransferase). The remaining genes encode proteins of unknown function, although hp1491 is predicted to encode a phosphate transporter. The smaller arrows indicate identified transcriptional start sites that were identified from the H. pylori 26695 transcriptome [46].
To determine if the HP1489-HP1486 efflux system has a role in flagellar biosynthesis or function in H. pylori, we deleted hp1489 and hp1487-hp1486 in H. pylori B128, and also deleted hp1489 in H. pylori G27, using a sucrose-based counter-selection method [47]. Motilities of all three mutants in soft agar medium were reduced 33 to 50% compared to the wild-type parental strains (Figure 2). Introducing a copy of hp1489 into the H. pylori G27 ∆hp1489 mutant on a shuttle vector rescued motility ( Figure 2B), indicating loss of hp1489 was responsible for the motility defect in the mutant. Examination of the H. pylori B128 ∆hp1487-hp1486 and ∆hp1489 mutants by transmission electron microscopy (TEM) revealed the strains possessed fewer flagella per cell than wild type ( Figure 3). Flagella of the ∆hp1487-hp1486 and ∆hp1489 mutants were sheathed, indicate sheath biosynthesis was not grossly impaired in the mutants. Taken together, these observations suggest the HP1489-HP1486 efflux system is required for normal flagellum biosynthesis and/or motility.  Other genes of known function are uvrD (encodes DNA helicase II), serS (encodes seryl-tRNA synthetase), xseB (encodes exodeoxyribonuclease VII small subunit), and ubiE (encodes demethylmenaquinone methyltransferase). The remaining genes encode proteins of unknown function, although hp1491 is predicted to encode a phosphate transporter. The smaller arrows indicate identified transcriptional start sites that were identified from the H. pylori 26695 transcriptome [46].
To determine if the HP1489-HP1486 efflux system has a role in flagellar biosynthesis or function in H. pylori, we deleted hp1489 and hp1487-hp1486 in H. pylori B128, and also deleted hp1489 in H. pylori G27, using a sucrose-based counter-selection method [47]. Motilities of all three mutants in soft agar medium were reduced 33 to 50% compared to the wild-type parental strains ( Figure 2). Introducing a copy of hp1489 into the H. pylori G27 Δhp1489 mutant on a shuttle vector rescued motility ( Figure 2B), indicating loss of hp1489 was responsible for the motility defect in the mutant. Examination of the H. pylori B128 ∆hp1487-hp1486 and Δhp1489 mutants by transmission electron microscopy (TEM) revealed the strains possessed fewer flagella per cell than wild type ( Figure 3). Flagella of the Δhp1487-hp1486 and Δhp1489 mutants were sheathed, indicate sheath biosynthesis was not grossly impaired in the mutants. Taken together, these observations suggest the HP1489-HP1486 efflux system is required for normal flagellum biosynthesis and/or motility.  wild type (WT), H. pylori B128 ∆hp1486/hp1487 mutant, and H. pylori B128 ∆hp1489 mutant was assessed by inoculating H. pylori strains into soft agar medium and allowing the cells to migrate from the point of inoculation. (B) Motility of H. pylori G27 wild type (WT) and H. pylori G27 ∆hp1489 mutant in soft agar medium. The two strains either had (+) or lacked (−) the shuttle vector pHel1489, which carried H. pylori G27 hp1489 under control of its native promoter. Bars indicate the mean of swim halo diameters measured 7 d post-inoculation for at least 9 replicates. Error bars correspond to one standard deviation. For both panels (A,B), the asterisk (*) indicates a significant difference between the mutant and wild type (p-value < 0.0001). Statistical significance was determined using a two-sample t-test. between the mutant and wild type (p-value < 0.0001). Statistical significance was determined using a two-sample t test.

In-Situ Structures of Flagellar Motors in the Δhp1489 and Δhp1487/hp1486 Mutants Reveal Apparent Flagellum Disassembly Products
The impaired motility and reduced number of flagella in the Δhp1489 and Δhp1487/hp1486 mutants suggested a role for the HP1489-HP1486 efflux system in the assembly or stability of the flagellum. To address this hypothesis, we examined the structure of the flagellar motors of the H. pylori B128 Δhp1489 and Δhp1487/hp1486 mutants by cryo-ET. For the analaysis, we selected 77 subtomograms of motors from 48 reconstructions of Δhp1489 mutant cells and 137 subtomograms of motors from 70 reconstructions of Δhp1487/hp1486 mutant cells, which we analyzed by subtomogram averaging and classification. The resulting motor structures are presented in Figure 4.

In-Situ Structures of Flagellar Motors in the ∆hp1489 and ∆hp1487/hp1486 Mutants Reveal Apparent Flagellum Disassembly Products
The impaired motility and reduced number of flagella in the ∆hp1489 and ∆hp1487/hp1486 mutants suggested a role for the HP1489-HP1486 efflux system in the assembly or stability of the flagellum. To address this hypothesis, we examined the structure of the flagellar motors of the H. pylori B128 ∆hp1489 and ∆hp1487/hp1486 mutants by cryo-ET. For the analaysis, we selected 77 subtomograms of motors from 48 reconstructions of ∆hp1489 mutant cells and 137 subtomograms of motors from 70 reconstructions of ∆hp1487/hp1486 mutant cells, which we analyzed by subtomogram averaging and classification. The resulting motor structures are presented in Figure 4. between the mutant and wild type (p-value < 0.0001). Statistical significance was determined using a two-sample t test.

In-Situ Structures of Flagellar Motors in the Δhp1489 and Δhp1487/hp1486 Mutants Reveal Apparent Flagellum Disassembly Products
The impaired motility and reduced number of flagella in the Δhp1489 and Δhp1487/hp1486 mutants suggested a role for the HP1489-HP1486 efflux system in the assembly or stability of the flagellum. To address this hypothesis, we examined the structure of the flagellar motors of the H. pylori B128 Δhp1489 and Δhp1487/hp1486 mutants by cryo-ET. For the analaysis, we selected 77 subtomograms of motors from 48 reconstructions of Δhp1489 mutant cells and 137 subtomograms of motors from 70 reconstructions of Δhp1487/hp1486 mutant cells, which we analyzed by subtomogram averaging and classification. The resulting motor structures are presented in Figure 4.  The color scheme is: orange, cage; brown, PflA/PflB, P-disk and basal disk; gray, L ring and P ring; cyan, C ring, MS ring, export apparatus, rod, and hook; green, outer membrane/flagellar sheath; light green, inner membrane.
Many of the motors of the ∆hp1489 and ∆hp1487/hp1486 mutants were normal in appearance and attached to a flagellar filament ( Figure 4A). These flagella were presumably functional and responsible for the motility observed with the mutants. Several of the motors of the mutants, however, appeared to be incomplete, and often lacked an associated filament ( Figure 4B-D). In contrast, such incomplete motors were rarely observed for the H. pylori B128 wild-type strain. To gain a better understanding of the aberrant motors of the ∆hp1489 and ∆hp1487/hp1486 mutants, we used a multivariate statistical analysis to resolve distinct motor structures that corresponded to the full motor and three specific class averages for the aberrant motors ( Figure 4E-H). The number of subtomograms for each class of motor are indicated in Table 2. In the class 1 motor or full motor, associated hook and flagellar sheath structures were observed ( Figure 4A), and the motor contained the core and accessories features reported in previous cryo-ET reconstructions of the intact H. pylori motor ( Figure 4E,I) [8]. Similar to the complete motor, the class 2 motor had the associated hook and sheath stuctures ( Figure 4B,F,J). In addition, core components of the motor (MS ring, C ring, LP ring, rod and flagellar protein export apparatus), the basal disk, and the outer disk were evident in this class of motor ( Figure 4F,J). The cage and other motor components normally found near the inner membrane though were absent in this class of motor ( Figure 4F,J).
The class 3 motor lacked the associated hook, filament and flagellar sheath, although the outer membrane above the rod was deformed slightly with a convex bulge ( Figure 4C,G,K). Core components of the motor, as well as the basal disk and the outer disk, were visible in the second class of motor, but the cage and motor components normally found near the inner membrane were missing ( Figure 4G,K).
The class 4 motor also lacked the associated hook, filament and flagellar sheath ( Figure 4D). In contrast to the other classes of motors, the class 4 motor lacked most of the core components of the motor and the only visible structures were the LP ring, the basal disk and the outer disk ( Figure 4H,L). As with the class 3 motor, the outer membrane above the rod was deformed slightly with a convex bulge ( Figure 4H). The flagellar substructure resembled the outer-membrane-associated relic complexes that result from flagellum disassembly and have been observed in a variety of bacterial species [27,28]. These flagellum disassembly products are referred to as PL-subcomplexes, and so we will refer to this class of motor substructure as the PL-subcomplex.
In bacterial species that lack a flagellar sheath, the PL-subcomplex is associated with a crater-like structure in the outer membrane that is sealed at the bottom by the P ring, L ring proteins or some other undefined molecule [27,28]. The H. pylori PL-subcomplex is not associated with a crater-like structure in the outer membrane, but instead the outer membrane over the PL-subcomplex is convex ( Figure 4H) and resembles the PL-subcomplexes observed in Vibrio species [28], which like H. pylori have sheathed flagella.

Isolation of Mutants That Suppress the Motility Defect in the ∆hp1487/hp1486 Mutant
To identify mutations that suppress the motility defect associated with loss of the HP1489-HP1486 efflux system, we enriched for variants of the H. pylori B128 ∆hp1487/hp1486 mutant that had enhanced motility in motility agar medium. Three independent motile variants were isolated (designated ESM_MV1 through ESM_MV3) that had motilities in soft agar medium that approximated that of wild type H. pylori B128.
Whole genome sequencing of the ∆hp1487/hp1486 motile variants and the ∆hp1487/hp1486 parental strain identified several phase variations resulting from the deletions or insertions in homopolymeric runs, as well as single nucleotide polymorphisms (SNPs) present in the motile variants that were not in the parental strain (Table 3). All three ∆hp1487/hp1486 motile variants contained the identical frameshift mutation in fliL (Table 3). While some of the alleles appeared in more than one motile variant (e.g., a SNP in murJ), the mutation in fliL was the only allele that occurred in all three motile variants, and moreover, was the only gene that carried a mutation in all three motile variants ( Table 3). The genome sequencing data suggested strongly that the mutation in fliL was responsible for the enhance mobility of the ∆hp1487/hp1486 motile variants in soft agar medium.  The mutation in fliL in the ∆hp1487/hp1486 motile variants was unexpected since fliL is reported to be required for motility in H. pylori [29]. The fliL mutation was a deletion of an A in a homopolymeric run of 9 A residues near the start of the open reading frame of the gene ( Figure 5). The frameshift mutation, which was in codon 16, altered the amino acid sequence at the point of the mutation and resulted in a stop codon shortly after the deletion. A couple of nucleotides downstream of the introduced stop codon, however, is a potential start codon in the correct reading frame that would allow for the expression of a truncated FliL protein ( Figure 5). Moreover, the ATCATG motif upstream of this potential start codon is a reasonable match for a Shine-Dalgarno sequence and is appropriately positioned relative to the start codon. H. pylori B128 FliL is 183 amino acids in length, and the truncated protein resulting from the potential start codon indicated in Figure 5B is 154 amino acid residues in length. a truncated FliL protein ( Figure 5). Moreover, the ATCATG motif upstream of this potential start codon is a reasonable match for a Shine-Dalgarno sequence and is appropriately positioned relative to the start codon. H. pylori B128 FliL is 183 amino acids in length, and the truncated protein resulting from the potential start codon indicated in Figure 5B is 154 amino acid residues in length. The altered amino acid sequence following the frameshift mutation is indicated in red type. A potential start codon that brings the nucleotide sequence into the correct reading frame is indicated in green type.

The HP1489-HP1486 Tripartite Efflux System Has an Apparent Role in Stabilizing the H. pylori Flagellar Motor
Using a comparative genomics approach, we found homologs of the HP1489-HP1486 efflux system components were present in all 35 FS + Helicobacter species but absent in the 9 FS -Helicobacter species that we examined. Deletion of hp1489 or hp1487/hp1486 in H. pylori resulted in reduced motility as assessed by the ability of the mutant strains to migrate from the point of inoculation in soft agar medium ( Figure 2). In the case of the H. pylori G27 Δhp1489 mutant, motility was restored by introducing a copy of hp1489 into the mutant on the shuttle vector pHel3 (Figure 2), providing strong evidence that the loss of HP1489 was responsible for the motility defect in the original Δhp1489 mutant.
Classification of the motor structures from the Δhp1489 and Δhp1487/hp1486 mutants revealed many of them were missing the cage and other components associated with the inner membrane ( Figure 4F,G). At least three distinct classes of aberrant motors were apparent, one of which resembled the PL-subcomplex observed from flagellum disassembly in various bacterial species [27,28]. The relationship between the H. pylori PL-subcomplex The altered amino acid sequence following the frameshift mutation is indicated in red type. A potential start codon that brings the nucleotide sequence into the correct reading frame is indicated in green type.

The HP1489-HP1486 Tripartite Efflux System Has an Apparent Role in Stabilizing the H. pylori Flagellar Motor
Using a comparative genomics approach, we found homologs of the HP1489-HP1486 efflux system components were present in all 35 FS + Helicobacter species but absent in the 9 FS − Helicobacter species that we examined. Deletion of hp1489 or hp1487/hp1486 in H. pylori resulted in reduced motility as assessed by the ability of the mutant strains to migrate from the point of inoculation in soft agar medium (Figure 2). In the case of the H. pylori G27 ∆hp1489 mutant, motility was restored by introducing a copy of hp1489 into the mutant on the shuttle vector pHel3 (Figure 2), providing strong evidence that the loss of HP1489 was responsible for the motility defect in the original ∆hp1489 mutant.
Classification of the motor structures from the ∆hp1489 and ∆hp1487/hp1486 mutants revealed many of them were missing the cage and other components associated with the inner membrane ( Figure 4F,G). At least three distinct classes of aberrant motors were apparent, one of which resembled the PL-subcomplex observed from flagellum disassembly in various bacterial species [27,28]. The relationship between the H. pylori PL-subcomplex and the other two classes of aberrant motors is unclear. The different classes of motors may represent intermediates in the flagellum disassembly pathway. While it is possible the aberrant motors are assembly intermediates, this seems less likely. Qin and co-workers examined the ultrastructure of flagellum assembly intermediates in H. pylori using cryo-ET to visualize over 300 flagellar motors. The earliest flagellum assembly intermediate Qin and co-workers observed had progressed through rod assembly, and this intermediate contained the cage and the other motor components present in the mature H. pylori flagellum [8]. This argues against the aberrant motors in the ∆hp1489 mutant being assembly intermediates since they lack the cage. Two significant questions concerning the HP1489-HP1486 efflux system are: (i) what is the substrate transported by the efflux system; and (ii) how does the efflux system stabilize the flagellum? Regarding the first question, tripartite efflux systems transport a diverse range of proteins, xenobiotics, antibiotics, metal ions and other small molecules [25,26]. While the substrate for the HP1489-HP1486 efflux system is not known, van Amsterdam and co-workers reported that disrupting hp1489 in H. pylori 1061 resulted in increased sensitivity to ethidium bromide [48]. It is not clear from the results of van Amsterdam and co-workers whether the HP1489-HP1486 efflux system transports ethidium bromide, or if loss of the efflux system increases the sensitivity of the hp1489 mutant to ethidium bromide by altering the permeability of the outer membrane. Regarding the second question, the HP1489-HP1486 efflux system may stabilize the flagellum by transporting a molecule that would otherwise accumulate in the cytoplasm of inner membrane and destabilize the flagellum. Alternatively, the transported molecule may be targeted to the outer membrane where it is needed to stabilize the flagellum. It is possible the structure of the HP1489-HP1486 efflux system itself has a role in stabilizing the flagellum. For example, the efflux system may interact with components of the flagellar motor to stabilize the flagellum. Identifying the ATPase that functions with HP1487 and HP1486 will likely prove helpful for examining the validity of the above models for how the efflux system stabilizes the flagellum. The HP1489-HP1486 homologs in Helicobacter anseris MIT 04-9362 are encoded in an operon that contains a gene encoding a predicted ATPase for an ABC transport system. HP0853 is an orphan ATPase (i.e., the gene encoding the protein is not associated with genes encoding transmembrane components of an ABC transport system) and an ortholog of the H. anseris ATPase. We are currently investigating whether HP0853 is the ATPase that functions with HP1487 and HP1486.
The occurrence of the identical frameshift mutation in fliL in all three ∆hp1487/hp1486 motile variants strongly suggests the fliL mutation is responsible for suppressing the motility defect in the parental ∆hp1487/hp1486 mutant ( Table 2). This result was unexpected since fliL was reported to be required for motility in H. pylori [29]. As indicated in Figure 5, the frameshift mutation in fliL is predicted to result in the expression of a short polypeptide corresponding to the first 16 amino acids of the FliL N-terminus plus an additional 12 amino acid residues. The frameshift mutation could also result in the expression of an N-terminal truncation of FliL that is missing the first 29 amino acid residues ( Figure 5). It is possible that both truncated FliL proteins are expressed and needed to suppress the motility defect of the ∆hp1487/hp1486 mutant. We are currently working to determine if one or both of the truncated FliL proteins are required for suppressing the motility defect of the ∆hp1487/hp1486 mutant. We hypothesize that the mutant fliL allele suppresses the motility defect in the ∆hp1487/hp1486 mutant by preventing or limiting flagellum disassembly, and studies are planned to examine the validity of this hypothesis. If our hypothesis tests valid, this would implicate FliL in playing a role in flagellum disassembly in H. pylori and perhaps other bacteria. Such a finding would be very exciting as little is known regarding the proteins and factors that control flagellum disassembly.

Identification of Genes Potentially Involved in Sheath Formation
Although many bacteria possess flagellar sheaths, little is known about the proteins and other factors that contribute to sheath biosynthesis in any bacterial species. To address this gap in our knowledge, we searched for H. pylori proteins that are conserved in FS + Helicobacter species but are absent or underrepresented in FS − Helicobacter species with the idea that some of these proteins have roles in sheath biosynthesis or function. A caveat of the approach is the FS + and FS − Helicobacter species segregate into distinct phylogenic groups [49], which means some of the proteins identified as specifically occurring in FS + Helicobacter species may be shared among the species of that group even though they have roles unrelated to the sheath, flagellum or motility.
Despite these limitations, the comparative genomics approach enabled the identification of candidates for proteins that have roles in sheath biosynthesis. One such candidate is the cardiolipin synthase ClsC [43]. Cardiolipin is a phospholipid that accumulates at the cell pole and septal regions in rod-shaped bacteria due to its ability to form microdomains that exhibit a high intrinsic curvature and thus have a lower energy upon localization to negatively curved regions of the membrane [50,51]. Given that the flagellar sheath is a long narrow tube with a high degree of negative curvature, it is reasonable to postulate that cardiolipin accumulates in the sheath. Consistent with this hypothesis, preparations of sheathed flagella from H. pylori G27 contain significant amounts of cardiolipin [52]. Thus, FS + Helicobacter species may need ClsC to satisfy an increased demand for cardiolipin in sheath biosynthesis.
Some of the proteins found preferentially in FS + Helicobacter species may have roles in mitigating the action of the sheathed flagellum. Rotation of the sheathed flagellum in Vibrio species is a major source of OMVs [21,22], and the generation of these OMVs is important in communication between V. fischeri and its host and symbiotic partner E. scolopes [22]. In contrast, H. pylori has presumably evolved to avoid surveillance by the host innate immune system. Several of the proteins found preferentially in FS + Helicobacter species enzymes involved in LPS biosynthesis or modification (Table 1), which contribute to the low endotoxin and immunological activity of H. pylori LPS [53][54][55]. These LPS modification enzymes may be a strategy that FS + Helicobacter species use to avoid activating the host innate immune response by OMVs released from the bacterium during the rotation of the sheathed flagellum.
Several of the proteins conserved in FS + Helicobacter species belong to the SLR family ( Table 1), members of which contain multiple copies of a degenerate amino acid repeat motif that is characteristic of eukaryotic Sel1 proteins [56]. Although the physiological roles of the H. pylori SLR proteins are unclear, some of them bind β-lactam compounds [37,38] and have been suggested to interact with immunomodulatory peptidoglycan fragments to affect the innate immune response [57]. Rotation of the H. pylori sheathed flagellum may result in the release of peptidoglycan fragments from the periplasmic space, and FS + Helicobacter species may use the peptidoglycan-binding SLR proteins to avoid eliciting a host immune response.

Strain Construction
Primers used for PCR are listed in Table S4 and strains used in this study are listed in Table 4. Genomic DNA (gDNA) from H. pylori G27 was purified using the Wizard genomic DNA purification kit (Promega, Madison, WI, USA) and used as the template to construct all deletion mutants. Target sequences from gDNA were amplified using Phusion polymerase (New England Biolabs, Ipswich, MA, USA). To facilitate cloning into the pGEM-T Easy vector (Promega) amplicons were incubated with Taq polymerase (Promega) at 72 • C for 10 min to add 3 -A overhangs. H. pylori B128 hpb128_199g42 (hp1489 homolog) and hpb128_199g40 along with hpb128_199g39 (hp1487 and hp1486 homologs) were deleted as follows. Regions flanking the target genes were amplified from gDNA using primer pairs P117/P118 plus P119/P120 for hp1489, and P177/P178 plus P179/P180 for hp1487/hp1486. The primer pairs introduced complementary sequence to the ends of the amplicons that were used for overlapping PCR. The complementary sequences also included NheI and XhoI restriction sites for a subsequent cloning step. After joining the flanking regions for each target gene by PCR SOEing, the resulting amplicons were cloned into pGEM-T Easy to generate the plasmids pJC049 (carried flanking sequences for hp1489), and pJC076 (carried flanking sequences for hp1487/hp1486). Plasmid pJC038 carries a kan R -sacB cassette in which sacB was under control of the H. pylori ureA promoter [43]. The kan R -sacB cassette from pJC038 was introduced into unique NheI and XhoI sites in plasmids pJC049 and pJC076 to generate the suicide plasmids pJC051 and pJC080, respectively. The suicide plasmids were introduced by natural transformation into H. pylori B128 or H. pylori G27 and transformants were screened for kanamycin resistance. The kan R -sacB cassette in the target genes was removed by transforming the kanamycin-resistant isolates with plasmids pJC049 or pJC076, and transformants in which the kan R -sacB cassette was replaced with the unmarked deletion were isolated following a sucrose-based counter-selection [47]. Regions flanking the targeted genes were amplified by PCR to identify transformants in which the target gene had been deleted, and the resulting amplicons were sequenced (Eton Biosciences) to confirm the target gene had been deleted.

Motility Assay
Motility was evaluated using a semisolid medium containing MHB-HS, 20 mM MES (2-(4-morpholino)-ethane sulfonic acid) (pH 6.0), 5 µM FeSO 4 and 0.4% Noble agar. A minimum of three biological samples and three technical replicates were used to assess the motility of each strain. H. pylori strains grown on TSA supplemented with 10% heatinactivated horse serum for 4 days were stab-inoculated into the motility agar and incubated at 37 • C under an atmospheric condition consisting of 10% CO 2 , 4% O 2 and 86% N 2 . The diameters of the resulting swim halos were measured 7 d post-inoculation. The two-sample t test was used to determine statistical significance for the results.

Complementation of ∆hp1489 Mutation
Primers 129 and 130 were used to amplify a~500 bp region immediately upstream of the start codon for hp1491 from H. pylori G27 gDNA, which included the promoter for the gene. Primers 131 and 132 were used to amplify hp1489 from H. pylori G27 gDNA. The primer pairs introduced complementary sequence to the ends of the amplicons that were used for overlapping PCR. The complementary sequences also included BamHI and XhoI restriction sites for a subsequent cloning step. After joining the flanking regions for each target gene by PCR SOEing, the resulting amplicons were cloned into pGEM-T Easy to generate plasmid pJC083. A BamHI-XhoI fragment bearing the hp1491 promoter fused to hp1489 (P hp1491 -hp1489) was excised from pJC083 and cloned into the H. pylori shuttle vector pHel3 [60]. The resulting plasmid, pHel1489, was introduced into HP122 (H. pylori G27 ∆hp1489) by natural transformation to generate strain HP128.

Isolation of Variants of ∆hp1487/hp1486 Mutant with Enhanced Motilities
Enrichment of variants of the ∆hp1487/hp1486 mutant with enhanced motility in soft agar medium was accomplished by six rounds of inoculating the mutant in the motility agar medium, allowing the bacteria to migrate from the point of inoculation, picking cells from the edge of the swim halo, and inoculating fresh motility agar medium. Following the last iteration, cells from the edge of the swim halo were streaked onto TSA-HS to obtain single colonies. Three independent enrichments were done and a single motile isolate from each enrichment was saved and characterized.

Genome Sequencing and Analysis
For whole genome resequencing, H. pylori gDNA was sonicated to approximately 500-bp fragments in a total of 500 ng. Genomic libraries were prepared using the NEBNext Ultra II DNA library prep kit for Illumina (New England Biolabs), including end-repair, adaptor ligation, and addition of an individual i5/i7 index primer. Illumina sequencing of the genomic libraries was done by Genewiz from Azenta Life Sciences. Sequence analysis was performed using the breseq computational pipeline [61]. A reference genome was constructed by mapping reads on an NCBI genome for H. pylori B128 (Accession no.: GCA_007844075.1). Reads were mapped to the reference genome, and a minimum variant frequency of 0.8 was used to call variants and identify SNPs. Variants and SNPs in the strains were compared manually to identify ones that were present only in the suppressor strains.

Transmission Electron Microscopy
H. pylori strains were grown to late-log phase (OD 600~1 .0) in MHB-HS. Cells from 1 mL of culture were pelleted by centrifugation (550× g) then resuspended in 125 µL of 0.1 M phosphate-buffered saline. Cells were fixed by adding 50 µL of 16% EM grade formaldehyde and 25 µL of 8% EM grade glutaraldehyde to the cell resuspension. Following incubation at room temperature for 5 min, 10 µL of the cell suspension were applied to a 300 mesh, formvar-coated copper grid and incubated at room temperature for 5 min. The cell suspension was wicked off the grids using a filter paper, and the grids were washed 3 times with 10 µL of water. Cells were stained by applying 10 µL of 1% uranyl acetate to the grids for 30 s. After removing the stain with filter paper, the grids were washed three times with 10 µL of water and then air-dried. Cells were visualized using a JEOL JEM 1011 transmission electron microscope. The number of flagella per cell (n = 50) were determined for each strain. A Mann-Whitney U test was used to identify statistically significant differences in number of flagella per cell for the various H. pylori strains.

Sample Preparation for Cryo-EM Observation
H. pylori strains were cultured on plates for 4 d. The cells on the agar surface were collected then washed with PBS buffer. Colloidal gold solution (10 nm diameter) was added to the diluted H. pylori samples and then deposited on a freshly glow-discharged EM grid for 30 s. The grid was blotted with filter paper and rapidly plunge-frozen in liquid ethane in a homemade plunger apparatus as described [62].

Cryo-ET Data Collection and Image Processing
The frozen-hydrated specimens of H. pylori strains were transferred to Titan Krios electron microscope (Thermo Fisher) equipped with a 300 kV field emission gun and a Direct Electron Detector. The data was acquired automatically with Tomography from Thermo Fisher. A total dose of 50 e − /Å 2 is distributed among 35 tilt images covering angles from −51 • to +51 • at tilt steps of 3 • . The tilt series were aligned using gold fiducial markers and volumes reconstructed by the weighted back-projection method using IMOD and Tomo3d software to generate tomograms [63,64]. In total, 207 tomograms from H. pylori B128 WT cells, 48 tomograms from H. pylori B128 ∆hp1489 cells and 70 tomograms from the H. pylori B128 ∆hp1487/hp1486 double mutant cells were generated.

Sub-Tomogram Analysis with i3 Packages
Bacterial flagellar motors were detected manually using the i3 program [65,66]. The orientation and geographic coordinates of selected particles were then estimated. In total, 511, sub-tomograms of H. pylori B128 WT, 77 sub-tomograms of H. pylori B128 ∆hp1489 mutant and 137 sub-tomograms of the H. pylori B128 ∆hp1487/hp1486 double mutant flagellar motors were used to sub-tomogram averaging analysis. The i3 tomographic package was used to generate class averages of the motor as described [62].