Helicobacter pylori Biofilm Involves a Multigene Stress-Biased Response, Including a Structural Role for Flagella

Biofilms, communities of bacteria that are embedded in a hydrated matrix of extracellular polymeric substances, pose a substantial health risk and are key contributors to many chronic and recurrent infections. Chronicity and recalcitrant infections are also common features associated with the ulcer-causing human pathogen H. pylori. However, relatively little is known about the role of biofilms in H. pylori pathogenesis, as well as the biofilm structure itself and the genes associated with this mode of growth. In the present study, we found that H. pylori biofilm cells highly expressed genes related to cell envelope and stress response, as well as those encoding the flagellar apparatus. Flagellar filaments were seen in high abundance in the biofilm. Flagella are known to play a role in initial biofilm formation, but typically are downregulated after that state. H. pylori instead appears to have coopted these structures for nonmotility roles, including a role building a robust biofilm.

H elicobacter pylori has been coevolving with humans for tens of thousands of years (1). During this time, it has adapted to survive the hostile environment of the stomach and evade the immune system, allowing it to persist for the life of the host (2). H. pylori colonizes gastric epithelial surfaces and within the thin layer of mucus near the cells (3). More recently, H. pylori was found to colonize within gastric glands, repeated invaginations of the gastrointestinal tract, which may provide the bacteria a favorable niche (4,5). Even though most infections are asymptomatic, H. pylori persistence is considered a major risk factor for gastric and duodenal ulcers, gastric mucosaassociated lymphoid tissue (MALT) lymphoma, and gastric adenocarcinoma (6). H. pylori infections remain difficult to treat, and when left untreated (7), 1% to 2% progress to gastric cancer (8, 9).
H. pylori possesses several mechanisms to escape the challenging environment of the stomach, where the pH is around 2. These include urease production, flagellar motility, and chemotaxis, which are all required for the initial and sustained colonization of the gastric epithelial surface (10). Urease catalyzes the hydrolysis of urea, which is abundant in the stomach, into bicarbonate and ammonia and thus raises the pH to near neutral (10). pH elevation decreases the viscoelastic properties of mucus gel and improves the motility of H. pylori, which can then swim away from the lumen to reach safer niches, including those close to the gastric epithelial surface (11). H. pylori forms microcolonies at the cell surface in vitro (12,13) as well as within gastric glands (5). This microcolony mode of growth may be consistent with the bacteria being in a biofilm growth mode.
Biofilms are dense aggregates of microorganisms attached to a surface and embedded in an extracellular polymeric matrix (14). In contrast with the other mode of bacterial growth, free-floating or planktonic, biofilm cells tend to be more tolerant toward antimicrobials and host immune responses (14,15). Biofilms are also frequently associated with chronic disease, including pneumonia in cystic fibrosis patients, Lyme disease, and chronic otitis media (16)(17)(18). In those chronic diseases, biofilm growth is considered to be a survival strategy used by pathogens to escape antimicrobial therapies, to avoid clearance by the immune system, and to persist for the lifetime of the host.
Chronicity and recalcitrant infections are also common features associated with H. pylori (19). Yet, the role of biofilm growth in promoting H. pylori persistence is still not clear (20). The first suggestion of biofilm formation by H. pylori during colonization of the human gastric mucosa was found using biopsies and scanning electron microscopy (SEM) analysis (20)(21)(22). These studies demonstrated that gastric biopsy samples from H. pylori-positive patients showed dense layers of bacteria aggregated and attached to the mucosal surface. The bacteria were consistent in appearance with H. pylori, with cells in both the spiral and coccoid morphologies. The same bacterial-appearing structures were absent in H. pylori-negative patients; however, there has not yet been conclusive evidence showing that H. pylori forms a biofilm in vivo.
H. pylori has been well documented to form a biofilm in vitro. The first report of in vitro biofilm formation by H. pylori was described to occur in clinical, laboratory, and mouse-adapted strains and was observed at the air-liquid interface on glass coverslips when the bacteria were grown in brucella broth (BB) supplemented with slightly lower than normal fetal bovine serum (FBS) (23). The biofilms were mainly composed of coccoid bacteria, with a minority of spiral and rod-shaped cells (23). In subsequent reports, scientists analyzed the extracellular polymeric substance (EPS) of H. pylori biofilms and found proteomannans, lipopolysaccharide (LPS)-related structures, extracellular DNA (eDNA), proteins, and outer membrane vesicles (24,25).
Additionally, biofilm cells have been shown to exhibit high resistance in vitro to clarithromycin, which is one of the common antibiotics used to treat H. pylori infection (26). The MIC and the minimum bactericidal concentration (MBC) were increased by 16and 4-fold, respectively, in the biofilm cells compared to planktonic ones (26). However, despite the growing evidences of H. pylori biofilm formation both in vitro and in vivo (20)(21)(22)24), little is known about the genes involved in biofilm formation. We thus sought to characterize H. pylori biofilm and investigate global transcriptional changes during biofilm formation, with a particular focus on H. pylori strain SS1 because it is able to colonize mice and thus will be able to serve as a model for biofilm formation in vivo.

RESULTS
Biofilm formation and growth condition. H. pylori strain SS1 has been extensively used as a murine model of H. pylori infection. H. pylori SS1 biofilms, however, are difficult to detect when the bacteria are grown in the standard nutrient-rich media routinely used for H. pylori culture. A previous study reported that H. pylori biofilm formation was significantly dependent on the growth media used (27). We thus, evaluated the ability of the H. pylori SS1 strain to form a biofilm using the crystal violet biofilm assay, as well as bacteria grown under various growth conditions that included different growth media, incubation times, and concentrations of serum. H. pylori SS1 and the other H. pylori strains used in this study are listed in Table 1.
Using brucella broth (BB) medium supplemented with 10% FBS (BB10), the condition usually used for H. pylori liquid growth, no biofilm was detected. We therefore explored smaller amounts of serum, as these have been reported elsewhere to promote adhesion of H. pylori strain 26695 (28) and biofilm formation of H. pylori reference strain ATCC 43629 and H. pylori clinical strains 9/10 (27). While only a slight biofilm was observed when H. pylori SS1 strain was grown with BB supplemented with 6% FBS, a pronounced biofilm (P Ͻ 0.01) was detected in BB supplemented with 2% FBS (BB2) (Fig. 1A). The H. pylori growth rate was slightly reduced in BB2 compared with BB10, which suggests that the increase of biofilm formation was not due to increased growth (data not shown). Ham's F-12 similarly only supported biofilm formation with low percentages of FBS (Fig. 1B). Further experiments identified that 3 days of growth in BB2 led to the greatest amount of biofilm (Fig. 1C). These results thus, suggest that BB medium supplemented with 2% serum and growth for 3 days is an optimal condition for studying H. pylori SS1 biofilm formation.
Biofilm characterization. To confirm and extend the results obtained with the crystal violet biofilm assay, biofilms of H. pylori SS1 were visualized by confocal laser scanning microscopy (CLSM) and staining with FilmTracer FM 1-43, a dye that fluoresces once inserted into the cell membrane. After 3 days of growth in BB2, we observed a thick bacterial biomass that nonhomogeneously covered the surface, consistent with a well-developed biofilm (Fig. 2). Using z-stack images, the thickness of the H. pylori SS1 biofilm was determined to be 11.64 Ϯ 2.63 m 3 /m (see Movie S1 in the supplemental material). As expected, H. pylori SS1 grown in BB10 did not form a biofilm that could be visualized by CLSM (data not shown).
To further characterize the EPS that composed the SS1 biofilm matrix, BOBO-3 and FilmTracer SYPRO Ruby biofilm matrix stains were used to stain extracellular DNA (eDNA) and extracellular proteins, respectively, as described previously (29,30). Both of these molecules extensively stained the biofilm EPS, consistent with the idea that the H. pylori SS1 biofilm matrix contains significant amounts of eDNA and extracellular proteins ( Fig. 2B and C). Because the same molecules have been detected in other H. pylori strains, these results suggest that the H. pylori EPS is typically composed of eDNA and proteins (24,31).
We also performed live-dead staining with the FilmTracer LIVE/DEAD biofilm viability kit, to define whether the biofilm cells were alive or dead. This approach revealed a To determine the importance of extracellular proteins and eDNA in the biofilm matrix of H. pylori SS1, we employed enzymatic treatment using DNase I and proteinase K. Proteinase K treatment significantly dispersed preformed biofilms (P Ͻ 0.01) (Fig. 3). H. pylori preformed biofilms were, however, resistant to DNase treatments. These data suggest that DNA may play only a minor role in the biofilm matrix; however, extracellular proteins likely play an important role in the biofilm architecture of H. pylori, as has been reported in other H. pylori strains (24,31). These results suggest that many H. pylori strains, including SS1, use a protein-based biofilm matrix.
Transcriptomic profiling of biofilm versus planktonic cells. To gain insight into the genes involved in H. pylori biofilm growth, we performed a transcript profiling experiment using transcriptome sequencing (RNA-seq). For this experiment, we grew H. pylori SS1 in BB2 in six-well plates for 3 days and collected the free-floating planktonic cells and the bottom-attached biofilm ones from the same wells. We collected RNA pylori SS1 was grown in BB medium supplemented with 2% FBS, and biofilm formation was evaluated at different time points. Experiments were performed three independent times with at least 6 technical replicates for each. Statistical analysis was performed using ANOVA (*, P Ͻ 0.05; and **, P Ͻ 0.01).

Hathroubi et al.
® from three biological replicates grown on 2 separate days. A total of 10 to 20 million reads per sample were generated by RNA-seq. These reads were then mapped to the H. pylori SS1 complete reference genome (32) and revealed a clear clustering of the biofilm-grown cells in a distinct population compared to the planktonic ones (Fig. 4A). This transcriptomic analysis showed that 122 of 1,491 genes (8.18%) were significantly differentially expressed (P Ͻ 0.01 and log 2 fold change of Ͼ1 or ϽϪ1) between H. pylori biofilm and planktonic populations ( Fig. 4B and Fig. 5). Sixty-one genes were significantly upregulated in biofilm cells compared to their planktonic counterparts, while another 61 were significantly upregulated in planktonic cells ( Table 2 and Table 3). To validate the results obtained by this RNA-seq, the relative abundance of selected RNA transcripts was quantified by quantitative reverse transcription-PCR (qRT-PCR). Using this approach, we detected the same gene expression trend between qRT-PCR and RNA-seq, thus, validating our results (Fig. 6). Below we discuss the most prominent of these genes and what they suggest about the H. pylori biofilm growth state.
Our data suggest that biofilm cells may be less metabolically active than planktonic cells, based on the decreased expression of several genes involved in translation and ribosomal structure ( Fig. 5 and Table 3). Similarly, genes involved in metabolism, biosynthesis of cofactors, and urease were also downregulated ( Fig. 5 and Table 3).
We found evidence that biofilm cells experience a stressful environment. Indeed, genes coding for several stress response-related proteins, such as hrcA, hspR, crdR, recR, and pgdA, were upregulated in biofilm cells ( Table 2). The hspR and hrcA genes code for transcriptional repressor proteins belonging to the heat shock protein family, and both were upregulated in biofilm cells. The crdR gene, which encodes a copper-related transcriptional response regulator, was also upregulated in H. pylori biofilm cells. Several transcripts encoding oxidative stress resistance were similarly upregulated in biofilm cells. These included recR, a gene encoding a DNA recombination protein, as well as pgdA, which encodes a peptidoglycan deacetylase. These have both been previously associated with oxidative stress in H. pylori (33). We found that the ATP-dependent protease hslV gene was among the most downregulated genes in H. pylori biofilm (Table 3). Although the HSlV protein has not yet been studied in the context of H. pylori biofilms, the orthologous Escherichia coli HslV protease has been previously associated with biofilm dispersal (34).
Our data suggest that biofilm cells may be less virulent in some ways, but more in others. Transcripts coding for some H. pylori virulence, colonization, or immunogenic factors were low in biofilm cells, including the UreA subunit of urease, the GroEL chaperone, and the HcpC cysteine-rich protein. These have each been shown to play roles in colonization or promoting inflammatory gene expression (35)(36)(37). On the other hand, only three genes carried within the cytotoxin-associated gene pathogenicity island (cagPAI) (38,39), cagL, cagW, and cagE, were significantly highly expressed in biofilm cells of H. pylori. These genes are in separate operons (38) and encode cagPAI protein CagL/Cag18, an integrin binding protein at the Cag pilus tip, cagPAI protein CagW/Cag10, and type IV secretion system protein CagE/VirB4, both part of the inner membrane protein transfer complex (39).
Many genes related to the cell envelope were upregulated in biofilm cells (Fig. 5). Indeed, genes coding for proteins involved in lipopolysaccharide synthesis, such as lpxB, which encodes a lipid A disaccharide synthase, and lptB, which encodes a lipopolysaccharide export system ATPase, were upregulated in biofilm cells (Table 2). Numerous transcripts encoding cytoplasmic and outer membrane proteins were also elevated in biofilm cells (i.e., homC, homD, and HPYLSS1_00450) ( Table 2).
Interestingly, the majority of the upregulated cell envelope genes in biofilm cells encoded flagellar structure and biosynthesis proteins, such as flgL, flgK, fliD, and flgE, which encode flagellar hook-associated proteins ( Table 2). Two known or putative flagellin genes, flaB and flaG, were also upregulated in the biofilms ( Table 2). These data suggested the intriguing idea that flagella might play a role in H. pylori biofilm.

FIG 3
Effect of enzymatic treatments on preformed biofilms. H. pylori SS1 was allowed to form biofilms for 3 days in BB2. The medium was then removed and replaced with either fresh medium or medium containing DNase I or proteinase K. Cells were reincubated for 24 h and then analyzed for the remaining biofilm using the crystal violet assay. The data shown here represent the percentage of remaining biofilm compared to the untreated control. Experiments were performed three times independently with at least 8 technical replicates for each. Statistical analysis was performed using ANOVA (*, P Ͻ 0.01 compared to the untreated control). Flagella are present and play a structural role in H. pylori biofilms. The transcriptomic data above suggested that flagellar components are upregulated in the biofilm cells, so we used SEM to gain insights into the biofilm architecture of H. pylori. This analysis demonstrated three-dimensional structures composed of bacterial cells adherent to one another and to the surface (Fig. 7A). Biofilms contained mainly coccoid cells along with some rod-shaped cells (Fig. 7A), as described previously for H. pylori biofilms (20)(21)(22)(23). Compared to planktonic populations, the proportions of both morphologies were similar at ϳ80% coccoid cells (data not shown).
Interestingly, extensive networks of bundles of filaments were visible in the biofilms. In some cases, these appeared to be connected to the bacterial pole, as would be expected for flagella (Fig. 7A, arrowheads). We measured the dimensions of the filaments to see if they were consistent in size with flagella. The width and the length measured 20 to 30 nm and 3 to 4 m, respectively, and were in agreement with those reported previously for H. pylori flagella (40). These data, especially when combined with transcriptomics, suggested these structures could be flagella. We therefore analyzed a mutant strain that lacks a key component of the flagellar basal body, FliM, and is aflagellated (41). SEM analysis of the ΔfliM mutant showed a complete loss of flagella    7A). This mutant displayed significantly less biofilm biomass (Fig. 7B), but we were able to find few microcolonies. Within these microcolonies, the filaments were completely lacking (Fig. 7A). These results suggest that these filaments are flagella and, furthermore, that flagella and/or motility is important for biofilm formation.
To further dissect the roles of motility and flagella in biofilm formation, we analyzed biofilm formation in a nonmotile but flagellated strain created by disruption of the motor protein MotB. As reported previously, this mutant still expresses flagella (Fig. 7A). Biofilm formation, however, was severely impaired compared to that in the wild-type strain, which suggests that a lack of motility might contribute to the biofilm defect. However, the ΔmotB mutant produced significantly more biofilm than the ΔfliM mutant, suggesting that the flagellar structure even in the absence of motility also contributes to biofilm formation in H. pylori.
To examine whether other strains of H. pylori similarly use flagella in biofilms, we imaged the biofilm of H. pylori strain G27 and similar flagellar mutants to those used above. Wild-type H. pylori G27 biofilm cells also contained filaments consistent with flagella (Fig. 8A). As with strain SS1, mutants lacking flagella (flgS or fliA) formed very weak biofilms, while strains that had flagella but no motility (motB) retained partial biofilm formation (Fig. 8B).
Taken together, these data suggest that flagella are produced by H. pylori when in a biofilm and appear to play roles in addition to simple motility, promoting biofilm integrity by holding cells together and to the surface.

DISCUSSION
In this report, we present the first transcriptomic characterization of the H. pylori biofilm. This study demonstrated clearly distinct expression profiles between planktonic and biofilm cells. The biofilm cells were characterized by low metabolic activity and triggering of several stress responses. Among the upregulated genes in the biofilm cells, we found several genes associated with cell membrane proteins, outer membrane proteins, and stress response, as well as, surprisingly, genes related to the flagellar apparatus. SEM analysis confirmed that flagella are present in a mature H. pylori biofilm and appear to play a role in maintaining solid biofilm structures. This result was somewhat surprising, as typically flagella are proposed to be turned off during the sessile biofilm growth mode (42)(43)(44). Recent work, however, discussed below has suggested that flagella in E. coli biofilms may play a structural role. Our studies with H. pylori thus build on an emerging theme that flagella are not always turned off in mature biofilms and indeed may play important functions in biofilm structure.
To gain insights into the mechanisms behind the biofilm formation in H. pylori, we   (43,45,46). In our experimental design, we compared a static biofilm mode of growth, where attached cells adhered to the bottom of the wells, with planktonic nonattached cells in the same wells. This approach was used to maintain the same growth conditions as much as possible between biofilm and planktonic samples and likely contributed to the relatively small number of differentially expressed genes. However, since biofilm formation is a dynamic process with frequent switching between planktonic and biofilm  modes occurring frequently, we likely have some contamination between the biofilm and planktonic populations. Therefore, our method may have missed some genes that are expressed in either population. One of the findings from our transcriptomic analysis was that several flagellar protein transcripts were significantly elevated in the biofilm. Notably, these were not for Quantification of biofilm formation by the H. pylori SS1 wild type and ΔmotB and ΔfliM mutants. Strains were grown in BB2 medium for 3 days, followed by biofilm evaluation using the crystal violet assay. Experiments were performed three times independently with 6 to 9 technical replicates for each. Statistical analysis was performed using ANOVA (**, P Ͻ 0.01; and *, P Ͻ 0.05).
Structural Role of Flagella in H. pylori Biofilm the entire flagellum but instead specific genes encoding the rod, hook, and filament. Specifically, we saw biofilm cell overexpression of genes encoding the FlgB rod protein, the FlgE flagellar hook protein, the FlgK and FlgL hook-filament junction proteins, the FliK hook length control protein, and two flagellins (FlaB and the putative flagellin encoded by flaG). Notably absent was the gene for the major flagellin FlaA and genes for the motor and stator. We also saw the upregulation of flgM, which encodes an anti-sigma factor that interacts with flagellar sigma factor FliA and therefore would be expected to decrease expression of flaA (47).
Historically, flagella have been typically viewed as important only for initial biofilm attachment and later cell dispersion (44,48). In fact, it has often been suggested that genes encoding flagella are turned off in mature biofilms (42)(43)(44). However, other reports have shown that some microbes express flagella during all stages of biofilm development, not only during the attachment and dispersion processes (49,50). In E. coli, several flagellar biosynthesis genes were induced in mature biofilms, and around 20 flagellar genes were regulated throughout all stages of biofilm development and not simply turned off (49). E. coli flagella were proposed to have a structural role along with other matrix components (i.e., eDNA and extracellular proteins), acting to cement and hold cells together and to the surface (50,51). Our data furthermore showed that aflagellated mutants are poor biofilm formers, supporting that these filaments could play a structural role. Taken together, these findings suggest that flagella of H. pylori may play a structural role during biofilm formation to help bacteria attach to each other and to surfaces. Interestingly, we found that the HspR and HrcA transcriptional repressor proteins are upregulated in biofilm cells. These proteins had previously been shown to positively correlate with flagellar expression (52), providing candidate regulatory proteins that function in biofilm cells. HspR and HrcA belong to the heat shock protein family and have been shown to respond to heat shock temperature conditions, although the nature of their "true" signal is not yet clear (52,53). A previous comparative transcriptomic analysis of wild-type H. pylori along with ΔhspR, ΔhrcA, and double mutants revealed a set of 14 genes that were negatively regulated and 29 genes that were positively regulated by these transcriptional regulators (52). The regulated genes include those for chaperones, urease enzyme activity, adhesion to epithelial cells, and flagella. Interestingly, among the 29 positively regulated genes, nearly half (14) encoded flagellar proteins, including the flgM, flaG, fliD, flgK, flgB, flgE, and fliK transcripts we identified here. Thus, our data suggest that biofilm conditions activate expression of HrcA and HspR, which in turn upregulate a subset of flagellar genes.
Experiments suggest that HrcA and HspR regulators do not directly regulate the flagellar genes (52). However, they do directly repress expression from several promoters, including those upstream of the groESL, hrcA-grpE-dnaK, and cbpA-hspR-hp1026 operons. These gene products encode the major chaperones of H. pylori (52)(53)(54). Heat shock conditions relieve the repression and allow expression of these operons. Consistent with elevated expression of HspR and HrcA, we found the genes coding for the heat shock protein GroEL to be downregulated in biofilm. Our data suggest that some yet-to-be-determined conditions occurring during biofilm formation trigger the expression of HspR and HcrA regulators.
Other genes associated with stress responses were also upregulated in biofilm cells, including the pgdA and recR genes. These genes encode a peptidoglycan deacetylase and DNA recombination protein, respectively. RecR has been shown to be involved in repairing in DNA double-strand breaks induced by oxidative stress (55), and the recR mutant was highly sensitive to DNA-damaging agents and oxidative stress and had a reduced ability to colonize mouse stomachs (55). pgdA has been reported to be highly induced by oxidative stress (33,56). Upregulation of oxidative stress genes has previously been reported in biofilm cells of other organisms, including E. coli (57), Pseudomonas aeruginosa (42), Neisseria gonorrhoeae (58), and Clostridium perfringens (48).
As reported for other microorganisms, H. pylori biofilm cells have altered metabolism, typically thought to be associated with the restricted availability of nutrients (48,59). H. pylori biofilm cells were characterized by downregulation of the expression of multiple genes involved in metabolism and translation, including atpC, atpE, and nifU, and several ribosomal protein genes. This low-metabolism phenotype seems not be related simply to the presence of coccoid cells but rather to the microenvironment generated during biofilm formation since the proportions of rods and coccoid forms did not differ between planktonic and biofilm populations.
H. pylori biofilm cells may also actively block the translational machinery, as suggested by the upregulation of the gene encoding RsfS, a ribosomal silencing factor. This protein was previously described in E. coli and Mycobacterium tuberculosis to slow or block the translation machinery during stationary phase and/or nutrition deficiency stress (60). It interacts with the 50S large ribosomal subunit, prevents its association with the 30S ribosomal submit, and thus blocks formation of functional ribosomes (60). Whether it functions similarly in H. pylori remains to be determined.
These observations that biofilm cells may have decreased translation are relevant because at least two of the main antibiotics used to treat H. pylori infection, clarithromycin and tetracycline, inhibit the 50S and 30S ribosomal subunits, respectively. Thus, these antibiotics may have less impact on biofilm cells. In fact, recent in vitro studies have shown that clarithromycin is 4-to 16-fold less effective on H. pylori biofilm cells than on planktonic ones (26,61).
Taken together, our study has shown that H. pylori biofilm cells display a distinct transcriptomic profile compared to their planktonic counterparts. Lower metabolism and stress responses, likely associated with the microenvironment generated in the H. pylori biofilm, could be determinants of antimicrobial tolerance and involved in the persistence and survival of H. pylori. However, the upregulated and downregulated genes identified in this study are not specific for biofilm cells, and stress response genes have been previously observed under other conditions when both planktonic and biofilm cells were exposed to various stresses. Therefore, our data do not support the existence of a biofilm-specific genetic program. Additionally, our data show that flagellar filaments are upregulated in biofilm cells and form an integral part of the biofilm matrix. Indeed, H. pylori cells without flagella form weak biofilms. These results thus contribute to correcting the idea that flagella are only involved during the first and last steps of biofilm formation and instead support their importance throughout the biofilm process.
Biofilm assays. Biofilm formation assays were carried out as described previously, with slight modification (27). H. pylori SS1 was grown overnight in BB10 as described above, diluted to an optical density at 600 nm (OD 600 ) of 0.15 with fresh BB10, BB2, BB6, or Ham's medium as desired, and then used to fill triplicate wells of a sterile 96-well polystyrene microtiter plate (Costar no. 3596). Following static incubation for 1, 2, 3, or 5 days under microaerobic conditions, culture medium was removed by aspiration, and the plate was washed twice with phosphate-buffered saline (PBS). The wells were then filled with 200 l of crystal violet (0.1% wt/vol), and the plate was incubated for 2 min at room temperature. After removal of the crystal violet solution by aspiration, the plate was washed twice with PBS and dried for 20 min at room temperature. To visualize biofilms, 200 l of ethanol (70% vol/vol) was added to the wells, and the absorbance at 590 nm was measured.
Biofilm dispersion assays. To evaluate the composition of SS1 biofilm matrix, we assessed the response of preformed biofilms to different enzymatic treatments. DNase I and proteinase K (both from Sigma-Aldrich) were used to target extracellular DNA and extracellular proteins, respectively. Biofilms were grown as described above, and after 3 days of growth, the old media were replaced by fresh media containing different concentrations of DNase I (from 380 g/ml to 95 g/ml) or proteinase K (from 200 g/ml to 50 g/ml). The cells were then incubated for a further 24 h. Control wells were exposed to media without enzyme. After treatments, the biofilm was stained with crystal violet as described above. Results are presented a percentage of the untreated control.
Confocal laser scanning microscopy. Biofilms of H. pylori SS1 were prepared as described above using BB2; however, for confocal laser scanning microscopy (CLSM), -Slide 8-well glass bottom chamber slides (ibidi, Germany) were used instead of 96-well microtiter plates. Three-day-old biofilms were stained with FilmTracer FM 1-43 (Invitrogen), BOBO-3 (Invitrogen), Filmtracer SYPRO Ruby biofilm matrix stain (Invitrogen), or the FilmTracer LIVE/DEAD biofilm viability kit (Invitrogen) according to the manufacturer's instructions. Stained biofilms were visualized by CLSM with an LSM 5 Pascal laser scanning microscope (Zeiss) and images were acquired using Imaris software (Bitplane). Biomass analysis of biofilm was carried using FM 1-43-stained z-stack images (0.1-m thickness) obtained by CLSM from randomly selected areas. The biomass of biofilms was determined using COMSTAT (63).
RNA extraction and library construction. Biofilms of H. pylori SS1 were grown in 6-well plates (Costar) in BB2 as described above. After 3 days of incubation, medium containing nonattached planktonic bacteria (the planktonic fraction) was removed by pipetting, and the cells were harvested by centrifugation and washed twice with PBS. The attached bacteria, representing the biofilm fraction, were washed twice with PBS to remove any remaining planktonic cells. Attached cells were scrapped off the plate using a cell scraper. Both planktonic and biofilm fractions were subject to total RNA extraction using the TRIzol Max bacterial enhancement kit (Ambion, Life Technology, Carlsbad, CA) as described by the manufacturer. RNA was further purified and concentrated using an RNeasy kit (Qiagen). rRNA was removed using the RiboZero magnetic kit (Illumina). Sequencing libraries were generated using NEBNext Ultra Directional RNA library prep kit for Illumina (NEB, USA). cDNA library quality and amount were verified using the Agilent Bioanalyzer 2100 system (Agilent Technologies, CA) and then sequenced using Illumina NextSeq Mid-Output (University of California Davis Genome Center).
Transcriptomic analysis. RNA-seq data were analyzed using CLC Genomics Workbench (version 11.0; CLC Bio, Boston, MA). After adapters were trimmed, forward-and reverse-sequenced reads generated for each growth state (biofilm versus planktonic, with three biological replicates for each condition) were mapped against the SS1 reference genome (32) to quantify gene expression levels for each experimental condition. The expression value was measured in reads per kilobase per million mapped reads (RPKM). Genes were considered differentially expressed when the log 2 fold change was above 1 and the P value was less than 0.05.
Construction of H. pylori mutants. The SS1 ΔfliM mutant was created by natural transformation of the SS1 wild type with plasmid pBS-fliM::catmut (40), which replaces most of the fliM gene, corresponding to amino acids 1 to 105, with cat. The G27 motB mutant was created by natural transformation of the G27 wild type with plasmid pKO114K and selection for kanamycin resistance. pKO114K was made as described for pKO114i (65), but instead of insertion of an aphA3-sacB allele, only an aphA3 allele was inserted. This allele inserts the aphA3 gene at the position corresponding to amino acid 113 of 257.
Scanning electron microscopy. H. pylori biofilms were grown on glass coverslips (12 mm; Chemglass Life Sciences, Vineland, NJ) by dispersing 4 ml of a culture diluted to an OD of 0.15 in BB2 into wells of a 6-well plate (Costar). The plate was incubated as described above. After 3 days of growth, biofilms formed on the surface of the coverslips and planktonic cells were washed twice with PBS and fixed with 2.5% (wt/vol) glutaraldehyde for 1 h at room temperature. Samples were then dehydrated with a graded ethanol series, critical point dried, sputter coated with ϳ20 nm of gold (Hammer IV, Technics, Inc., Anaheim, CA), and imaged in an FEI Quanta 3D Dualbeam scanning electron microscope operating at 5 kV and 6.7 pA.
Statistical analysis. Biofilm data were analyzed with GraphPad Prism (version 7.0) software (Graph-Pad, Inc., San Diego, CA) using one-way analysis of variance (ANOVA) followed by Dunnett's multiplecomparison test.

ACKNOWLEDGMENTS
We thank Fitnat Yildiz (University of California, Santa Cruz) and Davide Roncarati (University of Bologna, Italy) for helpful suggestions and comments on the study and the manuscript. We thank Aaron Clarke (University of California, Santa Cruz) for assistance with portions of the SEM experiments. We thank Ben Abrams (University of California, Santa Cruz) for CLSM assistance. We also acknowledge Tom Yuzvinsky (University of California, Santa Cruz) for assistance with sample preparation and electron microscopy and the W. M. Keck Center for Nanoscale Optofluidics for use of the FEI Quanta 3D Dualbeam microscope. We also thank David Bernick and the METX Department (University of California, Santa Cruz) for holding and financially supporting the RNA-seq workshop.
The work described here was supported by National Institutes of Health National Institute of Allergy and Infectious Disease (NIAID) grant RO1AI116946 (to K.M.O.). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.