Gene expression profile and pathogenicity of biofilm-forming Prevotella intermedia strain 17

Background Prevotella intermedia (P. intermedia), a gram-negative, black-pigmented anaerobic rod, has been implicated in the development of chronic oral infection. P. intermedia strain 17 was isolated from a chronic periodontitis lesion in our laboratory and described as a viscous material producing strain. The stock cultures of this strain still maintain the ability to produce large amounts of viscous materials in the spent culture media and form biofilm-like structures. Chemical analyses of this viscous material showed that they were mainly composed of neutral sugars with mannose constituting 83% of the polysaccharides. To examine the biological effect of the extracellular viscous materials, we identified and obtained a naturally-occurring variant strain that lacked the ability to produce viscous materials in vitro from our stock culture collections of strain 17, designated as 17-2. We compared these two strains (strains 17 versus 17-2) in terms of their capacities to form biofilms and to induce abscess formation in mice as an indication of their pathogenicity. Further, gene expression profiles between these two strains in planktonic condition and gene expression patterns of strain 17 in solid and liquid cultures were also compared using microarray assays. Results Strain 17 induced greater abscess formation in mice as compared to that of the variant. Strain 17, but not 17-2 showed an ability to interfere with the phagocytic activity of human neutrophils. Expression of several genes which including those for heat shock proteins (DnaJ, DnaK, ClpB, GroEL and GroES) were up-regulated two to four-fold with statistical significance in biofilm-forming strain 17 as compared to the variant strain 17-2. Strain 17 in solid culture condition exhibited more than eight-fold up-regulated expression levels of several genes which including those for levanase, extracytoplasmic function-subfamily sigma factor (σE; putative) and polysialic acid transport protein (KpsD), as compared to those of strain 17 in liquid culture media. Conclusion These results demonstrate that the capacity to form biofilm in P. intermedia contribute to their resistance against host innate defence responses.

P. intermedia strain 17 was initially isolated from a chronic periodontitis lesion in our laboratory [12] and some of its phenotypic characteristics were determined. Among these included the ability of the organism to: (a) produce viscous materials in vitro [12]; (b) invade human oral epithelial cells [13]; and (c) stimulate CD4 + T cells expressing V8, V12 and V17 [14]. More recently, the whole genome sequence of strain 17 was determined by The Institute for Genomic Research (TIGR; Rockville, MD, USA) [15].
In our earlier study, we demonstrated that a clinical isolate of Prevotella nigrescens is able to produce extracellular viscous material that might contribute to its biofilm formation [16]. In this context, we hypothesized that the ability of P. intermedia strain 17 to produce viscous materials might be essential for its biofilm formation. In this study, we describe the chemical composition of the viscous materials as determined by means of high performance liquid chromatography (HPLC) and colorimetry. To define the role of the extracellular viscous materials in biofilm formation, we identified and obtained a naturally-occurring variant strain that lacked the ability to produce viscous materials in vitro from our stock culture collections of strain 17, designated as 17-2. We compared the ability of these two strains (strains 17 versus 17-2) in their ability to form biofilms and to induce abscess formation in mice as an indication of their pathogenicity. Further, we sought to determine the gene expression profiles associated with the biofilm formation by these two strains using microarray assays.

Viscosity of spent culture medium
Stock cultures of P. intermedia strain 17 were transferred to enriched-trypticase soy broth (enriched-TSB) and grown for 48 h. The viscosities of spent culture media were measured by a rotary viscometer. All tested P. intermedia strain 17 stocks, with the exception of one particular stock strain, designated as strain 17-2, produced materials in vitro that were highly viscous as compared to the control TSB medium. In contrast, the viscosity of spent culture medium obtained from strain 17-2 was similar to that of the control TSB medium (Fig. 1).

Cell surface associated structures
SEM observations on cells from colonies of these strains growing on blood agar plates revealed that strain 17 had dense meshwork-like structures around the cells ( Fig. 2A), but strain 17-2 lacked this phenotype (Fig. 2B). The lack of abilities to produce viscous materials in culture medium and to form meshwork-like structures around cells on strain 17-2 were stably maintained despite repetitive passages in vitro or in animals (data not shown).

Biofilm formation assay
The ability to form biofilm was investigated for strains 17 and 17-2 using crystal violet microtiter plate assay. Strain 17 was consistently able to form biofilm on flat-bottomed polystyrene microtiter plates, whereas strain 17-2 showed poorer biofilm formation (Fig. 3A). Quantitative analysis as measuring the optical density of destained biofilms at 570 nm revealed that the ability of strain 17 to form biofilm was significantly greater than that of strain 17-2 (p < 0.01) (Fig. 3B).

Morphology and chemical composition of the viscous materials
Negative staining of the viscous material isolated from strain 17 culture supernatants revealed that the viscous material was made up of fine fibrous structures formed in curly bundles (Fig. 4). Chemical analyses of this purified material showed that it primarily consisted of neutral sugars and small amounts of uronic acid and amino sugars (Table 1), with mannose constituting 83% of the polysaccharide ( Table 2).
Viscosities of the spent culture media of Prevotella intermedia strains 17 and 17-2

Gene expression profiles of P. intermedia strains 17 and 17-2
To see what kind of gene expression events induce phenotypic differences on P. intermedia, we compared gene expression patterns between strains 17 and 17-2, the respective viscous material producing and non-producing strains using microarray analysis. To determine the appropriate time point for isolating total RNA, we first observed the morphological changes of cell surface structures in each strain along with the bacterial growth. In general, the growth of strain 17-2 was faster than that of strain 17, entering into an exponential phase at around 12 h and reaching the plateau in 24 h (Fig. 5A, open rhombus). Strain 17-2 did not show the presence of cell-associated fibrous materials at any stage of the growth cycle (Fig. 5C).
By contrast, strain 17 showed a slower growth rate (Fig.  5A, hatched square) with a longer exponential growth phase. Morphological observation of cultures at different stages of growth revealed that strain 17 exhibited cell surface-associated meshwork-like structures at 12 h and the structures became denser with time ( Fig 5B). From these preliminary data, 12 h-old cultures of strains 17 and 17-2 were chosen for a comparison of gene expression patterns. When the microarray expression data for strains 17 and 17-2 were compared, a total of 11 genes were up-regulated by at least two-fold with statistic significance (p < 0.05) in biofilm-forming P. intermedia strain 17 ( Table 3). The expression data demonstrated that several heat shock protein (HSP) genes, such as dnaJ, dnaK, groES, groEL and clpB were up-regulated in strain 17 (Table 3). We also identified two genes down-regulated at least two-fold in strain 17 (PINA2115: hypothetical protein; PINA2117: sterolregulatory element binding protein (SREBP) site 2 protease family). The original raw data files have been deposited in Center for Information Biology gene Expression database (CIBEX; Mishima, Japan; CIBEX accession: CBX27) [17].

mRNA levels for HSPs validated by real-time RT-PCR
In the microarray analysis, we identified that several of the heat shock protein genes were up-regulated in strain 17 as compared with those of strain 17-2. The increased expression levels of these genes were validated in an independent experiment by real-time RT-PCR using the 16S rRNA gene as the endogenous control. Annotations of these genes (PIN0281, PINA1058, PINA1756, PINA1797, PINA1798, and PINA2006) on TIGR data base were described in Table 3. Except PIN0281, five out of six of tested genes showed an at least fivefold increased average expression levels in strain 17 as confirmed by the quantitative real-time RT-PCR. Although PIN0281 showed about a three-fold up-regulation in strain 17 by the microarray analysis, the average of increased expression level of PIN0281 was less than two-fold in the real-time RT-PCR analyses (Fig. 6).
The data obtained from the microarray analysis as well as the real time RT-PCR showed that several of HSP genes were up-regulated in strain 17 in 12 h-old cultures as compared with those of strain 17-2. Next, we addressed the question of whether the different expression levels of HSP genes between the two strains are due to a lag of growth because strain 17 showed a slower growth rate than that of strain 17-2 (Fig. 5). The relative expression levels of HSP genes through the culture period were obtained using real time RT-PCR by the strain. In strain 17, the expression levels of these genes were fluctuating; increased in early exponential phase (6 h to 12 h), decreased once in the middle of exponential phase (18 h to 24 h), and then slightly increased again in early stationary phase. By con- The specimen was prepared from a colony of each strain grown on a blood agar plate. Strain 17 had dense meshworklike structures surrounding the cell surfaces (A), but strain 17-2 lacked this phenotype (B). Bars = 2 m. Β Β Β Β Γ Γ Γ Γ trast, strain 17-2 did not show such fluctuated transcriptional levels in all HSP genes through the culture period (Fig. 7). Judging from the comparison between strains 17 and 17-2 at 12 h-old cultures (Fig. 6), strain 17-2 seems to keep the expression levels of these HSP genes very low.

Abscess induction in mice
To examine the influence of the biofilm phenotype on pathogenicity of P. intermedia, the abilities of strains 17 and 17-2 to induce abscesses in mice were compared. An injection of 500 l of strain 17 at a concentration of 10 7 CFU/ml induced abscesses in mice (Fig. 8, left panel). In contrast, injection of a similar amount of strain 17-2 at the same growth phase did not induce abscesses in mice. A much higher cell concentration (10 9 CFU/ml) of strain Biofilm formation on microtiter plates   17-2 was required to induce abscesses in mice (Fig 8, right  panel). However, an injection of a similar concentration of strain 17 was lethal for mice (data not shown).

Internalization of bacterial cells by human PMNLs
In the phagocytosis experiments, strain 17 cells were rarely internalized, though many of these cells were bound to the cell surface of PMNLs (Fig. 9A). In contrast, strain 17-2 cells were readily internalized by PMNLs after 90 min incubation. Many of these bacteria were found in cytoplasmic vacuoles (Fig. 9B).

Gene expression profiles of strain 17 in biofilm in vitro
We next attempted to compare gene expression patterns of strain 17 between in biofilm and in planktonic conditions in vitro. Total RNA was isolated from 12 h cultures of strain 17 on solid culture media as its biofilm-forming cells and liquid cultures as planktonic cells, respectively. When the microarray expression data were compared, a total of 25 genes were up-regulated by at least four-fold with statistic significance (p < 0.05) in solid culture condition ( Table 4). The expression of several genes which including those for a levanase (PINA0149), an extracytoplasmic function (ECF)-subfamily sigma factor (putative  E : PINA0299), a putative lipoprotein (PINA1510), and a putative polysialic acid transport protein (KpsD, PINA1911) were protruded. Among hypothetical proteins, PINA1526 (putative CpxP) showed extremely high levels of transcription.
Transmission electron microscopy of negatively stained exopolysaccharides isolated from Prevotella intermedia strains 17 cul-ture supernatants Figure 4 Transmission electron microscopy of negatively stained exopolysaccharides isolated from Prevotella intermedia strains 17 culture supernatants. Note the fine fibrous structures that are formed in bundles. Bar = 500 nm.

Discussion
It is well known that bacteria assuming biofilm-forming capacity have enormous advantages in establishing persistent infections even though they appear to be innocuous in their planktonic state [18][19][20]. Exopolysaccharide (EPS) is one of the main constituents of the biofilm extracellular matrix [21], and recent investigations have revealed that each biofilm-forming bacterium produces distinctive EPS components e.g. alginate and/or Psl found in Pseudomonas aeruginosa [22], acidic polysaccharide of Burkholderia cepacia [23], collanic acid, poly--1,6-GlcNAc (PGA) or cellulose found in Escherichia coli [24][25][26][27], cellulose of Salmonella [24,28], amorphous EPS containing Nacetylglucosamine (GlcNAc), D -mannose, 6-deoxy-Dgalactose and D -galactose of Vibrio cholerae [29], polysaccharide intercellular adhesin (PIA) of Staphylococcus [30], and glucose and mannose rich components found in Bacillus subtilis biofilm [31]. In this study we found that P. intermedia strain 17 produced a large amount of EPS, with mannose constituting more than 80% of the polysaccharides. Among oral bacteria, the production of mannoserich polysaccharide by Capnocytophaga ochracea has been reported [32]. This EPS provides a protection from attack by the human innate immune system [33]. We have also reported that a clinical isolate of Prevotella nigrescens can produce a copious amount of mannose-rich EPS [16].
In this study, biofilm-forming P. intermedia strain 17 showed stronger ability to induce abscesses in mice than that of strain 17-2, which was a naturally occurring variant of strain 17 that did not produce surface-associated fibrous material and therefore not capable of forming a biofilm. It is evidently shown that the slime/EPS production is critical for bacteria to exhibit the resistance to the neutrophil phagocytosis [33][34][35][36] . We found in our earlier study that exposing biofilm-positive P. intermedia to environmental stress such as animal passages of the organism resulted in the up-regulations of HSPs at a protein level with increased production of cell surface-associated meshwork-like structures. By contrast, animal passages induced

Conclusion
The data obtained in this study suggest that the Prevotella biofilms mainly composed of mannose-rich polysaccha- Relative expression

Culture period
Strain 17-2 J rides contribute to their resistance to host innate defence responses resulting in the development of chronic infections in vivo, and may also suggest that stress responsive systems of this organism might be behind its biofilm formation. To figure out a biofilm formation-gene expression relay system in P. intermedia requires the development of a suitable molecular tool that is capable of introducing specific targeted mutagenesis on genes highlighted in this study.

Bacterial strain and cultures
A viscous material producing clinical isolate of P. intermedia, which was isolated from a periodontitis lesion and designated as strain 17 [12], was used in this study. A total of 10 frozen culture stocks of isolated strain 17 were used in this study. Stock cultures of strain 17 in each vial were grown on trypticase soy blood agar plates (BAP) supplemented with 0.5% yeast extract (Difco Laboratories, Detroit, MI), hemin (5 mg/l), L-cystine (400 mg/l) and vitamin K 1 (10 mg/l) or grown in the enriched-TSB: trypticase soy broth (TSB; BBL Microbiology Systems, Cockeysville, ND) supplemented with 0.5% yeast extract, hemin (5 mg/l), L-cystine (400 mg/l) and vitamin K 1 (10 mg/l). Bacterial cultures were grown anaerobically in an anaerobic chamber (ANX-3, Hirasawa, Tokyo, Japan) at 37°C in a 5% CO 2 , 10% H 2 , 85% N 2 atmosphere.

Biofilm phenotype on strain 17 stock cultures
The ability to produce viscous materials in culture media and form meshwork-like structures on cell surfaces were used as criteria for distinguishing between "biofilm-forming" and "biofilm-non-forming" as described previously [16]. We first examined whether strain 17 met the criteria for being a biofilm-forming bacterium, since more than a decade has passed when we first described the unique phenotypic characteristic of strain 17 for its ability to produce viscous material [12]. Ten culture stocks were plated on BAP respectively and grown for 48 h anaerobically. Single colony from each culture stock was transferred to enriched-TSB and grown for 24 h as the seed culture. One hundred and fifty l of this seed culture was transferred to enriched-TSB (15 ml) and grown for 48 h. The spent culture medium (550 l) was put into a rotor, and the viscosity was measured as shearing stress between a rotor and a rotor shaft at 50 rpm, 20°C using a rotary viscometer (Toki-sangyo, Tokyo, Japan).
To examine cell surface structures, scanning electron microscopy (SEM) was performed. Bacteria grown on BAP for 48 h were collected on a piece of filter paper (Glass fiber GA55, Toyo Roshi, Tochigi, Japan), fixed with 2% glutaraldehyde in 0.1 M phosphate buffer for 2 h and 1% OsO 4 in 0.1 M phosphate buffer for 1 h at 4°C, and dehydrated through an ethanol series and 2-methyl-2-propa- During the evaluation for the ability of our stock strain 17 cultures to form biofilms, one of the 10 stocks that we tested was a naturally-occurring variant that lacked the ability to form biofilms. A stock strain, designated as strain 17-2, produced neither viscous materials in culture medium nor cell surface-associated meshwork-like structures was obtained and considered as a biofilm-negative variant. These phenotypes were stably maintained in this variant strain despite repetitive passages in vitro or in animals (data not shown). The enzymatic activities of strains 17 and 17-2 were examined using the API ZYM system (bioMerieux, Marcy l'Etoile, France) and there was no significant difference regarding the production of enzymes (data not shown).

Biofilm formation assay
The ability to form biofilm was investigated for strains 17 and 17-2 using crystal violet microtiter plate assay. Briefly, the seed cultures of both strains were prepared as described above and diluted to an OD of 0.1 at 620 nm in the same medium. Next, 150 l diluted culture was transferred to each of eight sterile polystyrene microtiter plate wells (IWAKI, Tokyo, Japan) per strain. Sterile enriched-TSB was used as a control. The plates were prepared in duplicate and incubated at 37°C for 24 and 48 h, respectively. Biofilm formation was quantified according to Mohamed et al. [60]. This assay was repeated three times. A statistical analysis was performed using Student's t-test.

Sugar composition of viscous materials from strain 17 cultures
The exopolysaccharide was prepared from culture supernatants by the method of Campbell et al. [61]. Briefly, P. intermedia strain 17 was grown at 37°C in enriched-TSB for 24 h. Supernatants were separated by centrifuging the liquid culture at 12,000 × g for 30 min, and sodium acetate was added to a final concentration of 5%. The mixture was stirred for 30 min at 22°C and the exopolysaccharide was isolated by ethanol precipitation from the reaction mixture. The ethanol-precipitated material was collected by centrifugation (18,200 × g for 15 min at 22°C), resolved in 5% sodium acetate, and treated with chloroform: 1-butanol (1: 5 by volume). Water-soluble and chloroform-butanol layer were separated by centrifugation, an equal amount of ethanol was added to the watersoluble layer (this procedure was repeated twice), and the ethanol-precipitated material was freeze-dried and stored at -80°C until use. Contaminated lipopolysaccharides (LPS) were removed from preparations according to the . The freeze-dried material was dissolved in distilled water (0.5 mg/ml), and Triton X-114 (MP Biomedicals, Eschwege, Germany) stock solution (lower detergent rich phase) was added to a final concentration of 1% (v/v). After cooling on ice for 30 min, the solution was stirred at 4°C for 30 min and incubated at 37°C until the separation into two layers was complete. The upper aqueous phase was recovered by centrifugation for 30 min (1,000 × g, 30°C). This Triton X-114 treatment was performed twice. The upper aqueous phase was extracted three times with 3 vol CHCl 3 /CH 3 OH (2: 1 by volume) to remove detergent. The aqueous phase was concentrated under reduced pressure and freeze-dried. The contaminated-LPS level was measured by Limulus Amebocyte Lysate test according to the manufacturer's protocol (Seikagaku-kogyo, Tokyo, Japan).
The sugar composition of the purified viscous material were determined by means of HPLC for neutral and amino sugars and colorimetry for uronic acid. Briefly, neutral monosaccharides were released from purified exopolysaccharide (5 mg) by hydrolysis in a sealed tube with 2 N trifluoroacetic acid (200 l) at 100°C for 6 h. The hydrolysate was concentrated in vacuo and dissolved in 500 ml of distilled water. The sugars were identified by HPLC (LC-9A, Shimadzu, Kyoto, Japan) with a TSK-gel sugar AXG column (15 cm × 4.6 mm) (Tosoh, Tokyo, Japan) using 0.5 M potassium tetraborate buffer (pH 8.7) as a carrier at a flow rate of 0.4 ml/min and a column temperature of 70°C. Amino sugars were released from purified exopolysaccharide (5 mg) by hydrolysis in a sealed tube with 4 N HCl (200 l) at 100°C for 6 h. The hydrolysates were analyzed by HPLC (LC-9A, Shimadzu).

Transmission electron microscopy of purified viscous materials
For negative staining, the ethanol precipitated viscous material was dissolved in distilled water (1 mg/ml). Fifteen microliters of the sample was deposited onto a formvar-coated and carbon-stabilized copper grid. After 1 min, excess fluid was removed with filter paper strips, stained with 2% uranyl acetate for 1 min, and examined in a transmission electron microscope (TEM) (H7100, Hitachi, Tokyo, Japan) at 100 kV.

RNA isolation
To determine an appropriate time point for total RNA isolation from the cultures of strains 17 and 17-2, morphological changes of cell surface structures associating with growth were examined by SEM. Single colony of Strains 17 and 17-2 grown on BAP for 24 h were inoculated into enriched-TSB and grown for 24 h as the seed culture. Five ml of this seed culture was used to inoculate 500 ml of enriched-TSB. The growth of the culture was monitored by measuring the absorbance at the wavelength of 600 nm. The morphology of cultured cells at a different stage of growth was examined by SEM as described above. RNA isolation was performed at a time point (12 h To compare gene expression profiles of strain 17 in solid and liquid culture conditions, seed cultures of strain 17 were newly prepared as described above. Five ml of this seed culture was transferred to enriched-TSB (500 ml) and 200 l of the seed cultures was transferred to each of 50 BAPs. Both cultures were incubated for 12 h anaerobically. Total RNA was isolated from the liquid cultures as described above. Two hundred l of PBS was added to BAPs to harvest growing cells using cell scrapers (IWAKI). Cell suspensions were washed twice with PBS and total RNA was isolated as described above. When the individual probes met the criteria that the aver-age signals from the culture of biofilm-positive strain versus the average signals from biofilm-negative strain were different by at least twofold with statistic significance, probes selected were used to find up-regulated regions.

Microarray image acquisitions and data analyses
Pertinent information on raw data containing experimental designs and hybridization results for specific oligonucleotide sets is available in CIBEX database [17].

Quantitative real-time RT-PCR
To confirm the up-regulation of several genes in strain 17 recorded by the microarray, a real-time RT-PCR strategy was employed. Twelve hours cultures of strains 17 and 17-2 were prepared again and total RNA was isolated as described above. Real-time RT-PCR was performed according to the one-step RT-PCR protocol of iScript™ One-Step RT-PCR Kit with SYBR ® Green (BIO-RAD Laboratories, Tokyo, Japan). Briefly, 50 ng of total RNA, 200 nM of forward and reverse primers for a target gene, and 25 l of SYBR ® Green RT-PCR Reaction Mix (BIO-RAD Laboratories) were added into a PCR tube containing one l of iScript Reverse Transcriptase for One-Step RT-PCR.
To observe how the expression levels of these genes fluctuate through the culture period, single colony of strains 17 and 17-2 grown on BAP for 24 h were inoculated into enriched-TSB and grown for 24 h as the seed culture. One hundred and fifty l of this seed culture was used to inoculate 15 ml of enriched-TSB. Total RNA samples were extracted from 6, 12, 18, 24 and 30 h cultures of strains 17 and 17-2 using RNeasy Midi Kit (QIAGEN) and applied to the real-time RT-PCR as described above. Changes of the target mRNA levels through the culture period were recorded by the strain.

Animal studies
The virulence of biofilm-forming strain 17 was compared with that of biofilm-non-forming variant strain 17-2 regarding abscess formation in mice. Bacterial strains were cultured in enriched-TSB for 24 h for strain 17-2 and 36 h for strain 17, respectively (early stationary phase; see Fig.  5). Five hundred l of bacterial suspensions (10 6 to 10 10 CFU/ml) was injected subcutaneously into the inguen of each BALB/c mouse (male, 4 weeks; 3 mice per strain). Changes of abscess lesions were recorded photographically using a camera (Nikon FIII, Nikon, Japan) set at a fixed magnification for five consecutive days.

Phagocytosis assay
To compare anti-phagocytic activity of strain 17 with that of strain 17-2, bacterial cells were co-cultured with polymorphonuclear leukocytes (PMNL) obtained from healthy human volunteers (n = 3; age 20-23 years) in accordance with institutional approved procedures. PMNLs were purified by a PolymorphprepTM (Axis-Shield PoC AS, Oslo, Norway) gradient and cultured in RPMI 1640 medium (Gibco BRL, Grand Island, NY) supplemented with 10% heat-inactivated fetal calf serum (FCS) (MultiSer, Trace Scientific Ltd, Melbourne, Australia) and 2 mM L-glutamine at 37°C in 5% CO 2 . Bacterial cell suspensions (1.5 × 10 6 CFU/ml) were prepared from strains 17 and 17-2 cultures as described in the animal studies. Three hundred l of PMNLs (10 6 cells/ml) was dispensed into the wells of 24-well tissue culture plates (Becton Dickinson, Franklin Lakes, NJ). To these wells, 100 l of bacterial suspension of different tested strains was added. After incubation for 60-90 min at 37°C, PMNLs co-cultured with bacterial cells were centrifuged at 8,000 × g at 4°C for 5 min and processed for transmission electron microscopy to determine the internalization of tested strains by PMNLs. Cell pellets were fixed with 2% glutaraldehyde in 0.1 M phosphate buffer for 2 h at 4°C, post-fixed with 1% OsO 4 in 0.1 M phos-phate buffer for 1 h at 4°C, and dehydrated through an ethanol series. Samples were embedded into Epon resin and ultrathin sections were prepared by a ultramicrotome (Ultracut, Leica, Tokyo, Japan). Ultrathin sections were placed on a copper grid, stained with uranyl acetate and lead citrate, and observed in a TEM (H7100, Hitachi).

Authors' contributions
TY, TF and CM carried out the phenotype characterization and microarray analysis, and drafted the manuscript. KY and CS performed RT-PCR. NM and HN screened a culture collection of strain 17 for the ability to produce viscous material. TN participated in the analysis of microarray data. CBW, KPL, and HF participated in the design of this study and drafted the manuscript.