2.1 Bacterial Strains and Inoculation

H. pylori strain J99 was incubated at 37°C under microaerobic conditions (5% O₂, 10% CO₂, 85% N₂). For agar plate culturing, strains were grown on Brucella (BD Biosciences, USA) agar plates supplemented with 10% horse serum (Gibco BRL, Life Technologies, USA) for 32 to 36 h. For liquid culturing, H. pylori was inoculated in Brucella broth containing 10% horse serum with a starting optical density at 600 nm of 0.2 optical density units (ODU) using a shaker at a speed of 150 rpm. H. pylori cells were collected when bacteria reached mid-log phase (1.5 ODU) or early stationary phase (4.0 ODU) [32]. When needed, 10 μg/mL chloramphenicol (Cm) or 10 μg/mL kanamycin (Km) was added to Brucella medium during inoculation. For E. coli inoculation, E. coli was grown on Luria-Bertani (LB) (BD Biosciences) agar plates or in broth media at 37°C. If required, 100 μg/mL ampicillin (Amp), Cm (25 μg/mL), or Km (25 μg/mL) were supplemented in LB media. The bacterial strains and plasmids used in this study are described in Table S1.

2.2 Construction of In-Frame Deletion ΔfliW2 Mutant of H. pylori

An in-frame deletion approach was employed to construct a ΔfliW2 knockout mutant strain of H. pylori J99 (genome accession number: NC_000921.1). We used the genomic DNA (gDNA) of J99 wild-type (WT) strain as a template, along with primers jhp1291(fliW2)-PCR-1-XbaI and jhp1291(fliW2)-PCR-2-BamHI (Table S2) to amplify a ~ 1.5 kb fragment containing fliW2 and its upstream and downstream flanking sequences. This amplicon was then ligated to XbaI/BamHI-cleaved pGEMTeasy to generate pGEMTeasy-fliW2 (Table S1). Next, we applied inverse PCR approach to generate an in-frame deletion of fliW2 with primers jhp1291(fliW2)-mut-3 and jhp1291(fliW2)-mut*-4, followed by ligation with HincII-cleaved Cm resistance cassette. The resultant pGEMTeasy::fliW2::Cm (pKO-fliW2) was delivered to H. pylori J99 WT strain using natural transformation. After homologous recombination occurred, the Cm-resistant transformants were inoculated for gDNA isolation to validate the deletion of fliW2 gene. The mutation of the ΔfliW2 #5 strain was confirmed by Southern blotting and Sanger sequencing analyses. In addition, we performed cDNA-qPCR analysis to rule out polar effect in this ΔfliW2 mutant strain (Figure S1).

2.3 RNA Isolation, cDNA Conversion, and cDNA-Quantitative PCR Analysis

Bacterial pellets were collected at 1.5 ODU by centrifugation and preserved in RNALater solution (Thermo Fisher Scientific, USA). Total RNAs were extracted using GENEzol TriRNA Bacteria Kit (Geneaid, Taiwan) according to the manufacturer's recommendations. The RNA concentration and its quality were determined using a Nanodrop spectrophotometer (NanoDrop ND-1000, USA) and agarose gel electrophoresis. Subsequently, a total of 0.5 μg RNA was converted to cDNAs using ReverTra Ace qPCR RT Master Mix with gDNA Remover (TOYOBO, Osaka, Japan) according to the manufacturer's protocols. Standard PCR was performed to ensure that there was no residual gDNA. For cDNA-qPCR analysis, cDNA samples were diluted ten-fold using an elution buffer (50 mM Tris–HCl, pH 8.0). The qPCR reaction was prepared containing diluted cDNA, 2× Fast SYBR Green Master mix (Thermo Fisher Scientific) and qPCR primers (Table S2), followed by execution in a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). The amplification cycling was set as follows: 95°C for 20 s, followed by 40 cycles of 95°C for 3 s and 60°C for 30 s. Data was collected during the extension step. The melting curve was included for qPCR primer quality control. Relative quantification of gene expression was determined and analyzed using the ABI software and 2^−ΔΔCt method [33], compared to gyrA Ct value as endogenous gene control.

2.4 Motility Assay Using Soft Agar Plates

A tip-full of H. pylori cells were inoculated by being vertically touched to the surface of soft agar (0.3%) plates (pH 7 and 6; Brucella agar/10% horse serum). After a seven-day incubation, we measured the diameter of the migrated bacteria. The motility of each strain was calculated as mean ± SD from at least three independent experiments. The ΔflaA mutant strain serves as a non-motile control.

2.5 Flagellar Filament Examination Using Scanning Electron Microscopy

The grown H. pylori J99 WT, ΔflaA, and ΔfliW2 strains were diluted in Brucella broth containing 2% horse serum to have an initial OD₆₀₀ of 0.2. Next, we transferred 1.5 mL of diluted H. pylori cells to 24-well plates, which contained lysine-coated glass coverslips. After 30 to 60 h of inoculation under microaerophilic conditions at 37°C, the supernatants were discarded, and the cells were fixed with 2.5% glutaraldehyde in PBS buffer at 4°C for overnight. Fixed cells were washed twice in PBS buffer, and dehydrated by adding 1 mL of ethanol in water in increasing concentration of 70%, 85%, 95%, and 100% (v/v) at room temperature for 10 min. An additional two dehydrations were conducted using 100% ethanol for 15 min for each time, followed by a critical point drying at the Electron Microscopy Facility (National Yang Ming Chiao Tung University). The coverslips containing the adherent cells were immediately coated with Au/Pd on a sputter coater. The examination was performed using a field emission scanning electron microscope (Jeol JSM-7600F, Japan) at an acceleration voltage of 5 kV. The micrographs were taken from at least two different fields from two independent experiments.

2.6 Generation of Mouse Polyclonal Anti-FlaA and Anti-FliW2 Sera

We over-expressed and purified recombinant FlaA and FliW2 proteins from E. coli BL21 harboring pET29b-FlaA and pET22b-FliW2 (Table S1), respectively, according to the recommendations from the manufacturer (Sigma-Aldrich, USA). One hundred micrograms of purified recombinant FlaA and FliW2 proteins were mixed with complete Freund's adjuvant (1:1, v/v), and subcutaneously injected into female C57BL/6 mice (six to eight-week-old). Mice were boosted three times biweekly for 6 weeks. In the seventh week, the sera were collected from the immunized mice and stored at −20°C.

2.7 SDS-PAGE and Western Blotting Analysis

Total proteins of collected H. pylori cells were extracted and subjected to 10% SDS-PAGE. In brief, after transferring proteins to PVDF membranes and blocking in 5% skimmed milk, we probed the membranes in TBS buffer containing 0.1% Tween 20 at 4°C overnight with diluted (1:5000) mouse anti-FlaA, anti-FliW2 polyclonal antibodies (in-house), or mouse anti-GroEL (Hsp60) monoclonal antibody (Product number H3524, Sigma-Aldrich, USA). GroEL served as an internal control.

2.8 In Silico Predictions of FliW2, CsrA, and flaA RNA Secondary Structure

We used H. pylori J99 CsrA as a query to construct the phylogenetic tree analysis (Figure 1, left panel). The CsrA homologs (Table S3) from a range of representative taxa spanning proteobacteria were aligned using (standalone) MAFFT v7.526, the L-INS-i method, and default settings [34]. With the alignment as input, a distance matrix was generated using the WAG amino acid model implemented in phangorn v2.12.1 in R v4.4.1 [35]. A tree topology was generated via UPGMA (unweighted pair group method with an arithmetic mean) and also implemented in phangorn. Bootstrap analysis was carried out and the results were plotted using phangorn's plotBS() function. Protein homolog comparison (Figure 1, right panel) was performed using the H. pylori J99 CsrA (UniProt ID: Q9ZJH4), FliW1 (UniProt ID: Q9ZK60), and FliW2 (UniProt ID: Q9ZJL5) as queries for protein identity analysis. The alignments were executed using Clustal O (version 1.2.4) on the UniProt website with default settings. The UniProt accession numbers for the CsrA homologs and FliW homologs are listed in Tables S3 and S4, respectively.

To compare the structure and conservation of H. pylori J99 FliW1 and FliW2 (Figure S2), we used MMseqs2 (version 71dd32ec43e3ac4dabf111bbc4b124f1c66a85f1) for a single-against-many search for homologous sequences using FliW1 (UniProt ID: Q9ZK60) and FliW2 (UniProt ID: Q9ZJL5) as query sequences. This strategy was implemented in ColabFold v1.5.5 [36], a structural prediction tool containing AlphaFold2 and hosted as a Jupyter Notebook (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb). ColabFold uses three databases: UniRef30 (a clustered version of UniRef100), PDB70 and a bespoke environmental dataset hosted by ColabFold (https://colabfold.mmseqs.com) in the homology search. The procedure employed limits the number of sequences in each multiple sequence alignment (MSA), reducing redundancy therein, while collecting sequences within (and retaining similar sequences between) buckets with different degrees of similarity to the query sequence, that is, it samples diverse sequences inferred by the query while reducing the overall alignment size. ColabFold was run with default parameters via the online notebook. For the MSA these parameters were “msa_mode = mmseq2_uniref_env and pair_mode = unpaired_paired.”

Protein sequence alignment of the H. pylori J99 CsrA (UniProt ID: Q9ZJH4), FliW1 (UniProt ID: Q9ZK60), and FliW2 (UniProt ID: Q9ZJL5) homologs were carried out using CLC Genomic Workbench (version 24.0.1). The 3D structure of FliW2-CsrA interacting regions was predicted using ColabFold based on DeepMind AlphaFold2 [36] and reconstructed using PyMoL. The RNA secondary structure of the 5'UTR of flaA mRNA (50 bases) was executed in the RNA Folding Form V2.3 at the UNAFold Web Server (http://www.unafold.org/) with a default setting.

2.9 In Vitro Pull-Down Assays

2.10 Bacterial Two-Hybrid Analysis

We chose an in vivo approach, the Bacterial Adenylate Cyclase-based Two-Hybrid (BACTH) system (Euromedex, France), to examine the interaction between CsrA and FliW2. In brief, pKT25 (N-terminal T25 segment fusion) serves as a vector for constructing a bait protein, while pUT18 (C-terminal T18 segment fusion) vector is used for a target protein. The fliW2 gene and csrA gene carrying various deletions were individually PCR-amplified, BamHI/Acc65I-digested, and cloned into the BamHI/Acc65I-cleaved pKT25 and pUT18 vectors (Table S1). After Sanger sequencing validation on the constructed clones, pKT25- and pUT18-derivative clones were co-transformed into an E. coli reporter strain DHM1. The transformants were grown in LB broth with ampicillin, kanamycin, and IPTG for 16 h, then the cells were lysed and extracted for β-galactosidase activity measurement. An increase of β-galactosidase activity in Miller units is an indication of positive protein–protein interactions. E. coli DHM1 cells harboring vectors pKT25 and pUT18 served as a negative control, whereas the cells harboring pKT25-FeoB and pUT18-FeoC were a positive control [37]. We also included background signal controls to distinguish false-positive protein interactions.

2.11 Ethics Statement

Animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of National Cheng Kung University, Taiwan (approval number: 109127). Animal well-being, sedation, and analgesia were monitored and administered as indicated to minimize stress and pain associated with any veterinary procedures.

2.12 Statistical Analysis

Statistical analysis was assessed using an unpaired t test with Welch's correction, one-way or two-way analysis of variance (ANOVA) (GraphPad Prism 8). Statistical significance was represented as *, p < 0.05; **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001 unless indicated.

3.1 Evolutionary Characteristics of the CsrA and FliW2 Homologs in H. pylori and Proteobacteria

We identified two fliW homologs (jhp1081-encoded FliW1 and jhp1291-encoded FliW2) from the H. pylori J99 genome using the BLAST analysis. Subsequently, we constructed multiple sequence alignments (MSAs) using MMseqs2 for comparison against diverse sequences (Figure S2). Our result showed a more consistent sequence coverage distribution across the length of FliW2 than across the length of FliW1, particularly among the bulk of the sequences that have intermediate similarities to the query. It appears that the N-terminal region of FliW1 is less well-represented in MSAs than that of FliW2. This analysis implied that the H. pylori FliW2 is more likely to be conserved than the FliW1. Our proteome analysis of the H. pylori J99 wild-type strain also showed that FliW2 production was ten-fold more than that of FliW1 in neutral or acidic media (Table S5). Further phylogenetic tree analysis was carried out to investigate whether H. pylori CsrA has an association with the FliW2 protein over macroevolutionary time scales. We identified the presence of csrA and fliW genes and extracted their protein sequences from the genomes of the representative members of Proteobacteria. Interestingly, H. pylori's CsrA sequence is approximately 15 residues longer at the carboxyl terminus than that of Gammaproteobacteria (Figure 1, right panel). Based on evidence presented elsewhere in this paper, H. pylori CsrA likely interacts with FliW2. As well, we found that most Gammaproteobacteria do not contain FliW proteins (Figure 1, right panel). Therefore, the genomic characteristics of these CsrA and FliW2 show that H. pylori differs from other Proteobacteria (Figure 1, left panel). We also predicted the RNA secondary structure of the 5′ un-translational region (UTR) of the major flagellin flaA transcript, revealing two hairpin structures containing the CsrA binding motif NGGA (Figure S3). This finding was in agreement with the hypothesis that CsrA dimers bind to the flaA mRNAs in C. jejuni [14]. We therefore hypothesized that FliW2 modulates CsrA by allosteric obstruction of target transcripts (i.e., the major flagellin flaA), thus de-repressing the motility of H. pylori.

3.2 Disruption of fliW2 Impairs the Flagellar Motility of H. pylori

To discover the function of FliW2, we constructed an in-frame deletion mutant of fliW2 using a gene replacement approach. The resultant ∆fliW2 mutant was validated using Sanger sequencing and cDNA-qPCR analysis, confirming that the ∆fliW2 mutant had no polar effects (Figure S1). We examined the motility of the ∆fliW2 mutant in the soft agar plates. Our results showed that, unlike the WT cells, the ∆fliW2 mutant cells lost motility in both the neutral and acidic media (pH 7 and 6), similar to the non-motile ∆flaA cells (Figure 2A). We also studied the flagellar structure of the ∆fliW2 mutant cells using scanning electron microscopy. The ∆fliW2 mutant cells were primarily aflagellate or had only short-protruding flagella, differing from the long tangled flagellar filaments observed in the WT cells (Figure 2B). To further investigate the non-motile ∆fliW2 mutant cells, we analyzed the expression of the major flagellin FlaA by Western blotting. The FlaA expression was substantially reduced in the ∆fliW2 mutant cells (Figure 2C), indicating that FliW2 modulates FlaA expression through direct or indirect regulations. In order to test our hypothesis, we investigated whether or not FliW2 interacts with CsrA.

3.3 Interaction Between FliW2 and CsrA Proteins

To demonstrate protein interactions between FliW2 and CsrA, we carried out pull-down assays. We generated nickel ion magnetic beads bound with recombinant histidine-tagged CsrA protein (CsrA_His). By applying the total lysate of the H. pylori WT strain to the beads, it allowed us to detect the proteins that interact with CsrA_His under native conditions. After washing, the FliW2 signal was revealed in the eluted fraction by Western blotting (Figure 3A, Lane 7). To determine whether or not FliW2 binds to CsrA directly, we overexpressed the recombinant CsrA_His (pET22b-CsrA) (Figure 3B, Lane 3) and the recombinant tag-free FliW2 (pET22b-FliW2-noHis) (Figure 3B, Lane 5), respectively, in E. coli BL21 (DE3) cells. Then we pooled the E. coli lysate containing FliW2 proteins with the other E. coli lysate, which contained CsrA_His proteins. As we hypothesized, the tag-free FliW2 was co-eluted with the CsrA_His through nickel column separation, showing that the FliW2 directly interacted with the CsrA (Figure 3B, Lane 7). The overexpressed protein band (~ 15 kDa) was then excised, purified, and analyzed by liquid chromatography with tandem mass spectrometry. The matched peptide fragments of H. pylori FliW2 were identified (Figure 3C). The combined results of these investigations demonstrate that H. pylori FliW2 interacts directly with CsrA.

3.4 AlphaFold2 Prediction of FliW2-CsrA Interacting Regions

To identify the FliW2-CsrA interacting regions, we took advantage of the deep learning algorithm AlphaFold2 (AF2) to predict the pattern of protein interactions. As shown in Figure 4A (Left panel), the β-barrel region of FliW2 formed a cleft that is predicted to interact with CsrA. The structure of H. pylori CsrA contains the N-terminal β-strands region and a loop–helix–loop–helix region, where we predicted that the C-terminal extension helix–loop–helix (residues 55–76) would interact with FliW2 (Figure 4A, right panel). The findings of earlier loss-of-function studies of the FliW and CsrA in B. subtilis and C. jejuni [38, 39] allowed us to postulate the crucial residues of FliW2 (F27, Q106, and V108) and CsrA's core proximal cluster (N55) for protein binding (Figure 4B,C). Subsequently, we designed successive truncations of CsrA to determine the region that is essential for the CsrA-FliW2 binding based on our predictions. The deletion of the C-terminal extended helix–loop–helix structure of CsrA occurs in CsrA_1-72 (4-residue deletion), CsrA_1-63 (13-residue deletion), CsrA_1-58 (18-residue deletion), and CsrA_1-54 (22-residue deletion) (Figure 4A, right panel).

3.5 Determination of the Critical Regions of CsrA for FliW2 Interaction In Vivo

Following our AF2 prediction, we performed in vivo bacterial two-hybrid analysis to determine FliW2-CsrA interactions and the regions required for the FliW2-CsrA binding. We chose the Bacterial Adenylate Cyclase-based Two-Hybrid (BACTH) system to generate T25-fused bait proteins using pKT25 (N-terminal T25 fusion) and T18-fused target proteins using pUT18 (C-terminal T18 fusion). In our first set of experiments, we detected the activity of β-galactosidase by analyzing T25-fused FliW2 and T18-fused CsrA, and T18-fused CsrA truncating proteins (Figure 5, upper panel). The increase in β-galactosidase activity observed in samples 3, 5, and 7 showed that FliW2 was bound to full length CsrA, CsrA_1-72, and CsrA_1-63. With the elimination of the C-terminus end 18 and 22 residues of CsrA, FliW2 could not bind to CsrA_1-58 or CsrA_1-54 and the β-galactosidase activity was not significantly elevated compared to negative controls (Samples 9 and 11). To verify whether the C-terminal extension of CsrA for FliW2 binding was not an artificial defect, we created T25-fused CsrA proteins (full-length or with truncations) and T18-fused FliW2 in our second set of experiments (Figure 5, lower panel). Consistently, the binding between CsrA and FliW2 was absent when the C-terminal 18 and 22 residues of CsrA were deleted (Samples 20 and 22). Taken together, these results showed that the C-terminal region of CsrA is crucial for interaction with FliW2.