Introduction

A bacterial biofilm is a community of microorganisms attached to a solid surface using extracellular polymeric substances (EPS) (Branda et al. 2005). Scanning confocal laser microscopy and fluorescence markers such as the green fluorescence protein (GFP) have been used to show that biofilm communities are non-homogenous as a consequence of rhamnolipid synthesis, growth, or motility phenotypes. However, the relationship of these phenotypes to architecture has not been elucidated (Boles et al. 2004; Werner et al. 2004; Lequette and Greenberg 2005). This complex organization allows cells to respond to different conditions such as temperature (Pham et al. 2004) and the presence of antimicrobials (Szomolay et al. 2005). These robust communities cause health problems such as respiratory, urinary track, and ear infections (Potera 1999) as well as industrial problems such as corrosion (Elvers and Lappin-Scott 2000).

This non-homogeneity in biofilms led us to investigate the relationship between biofilm architecture and cell phenotype. Pseudomonas aeruginosa, for example, during biofilm development, exhibits mushroom-like structures which have been attributed to stress forces conditions (Stoodley et al. 1999), substrate gradients (Chang et al. 2003), and differences in gene expression (Espinosa-Urgel 2003). These mushroom structures form channels for nutrient and waste flow (Davey et al. 2003). The scrutiny of Escherichia coli biofilm structure is not as developed as that of P. aeruginosa because of its usually poor biofilm formation in the absence of conjugation plasmids (Ghigo 2001; Reisner et al. 2003).

In the absence of a conjugation plasmid, type 1 fimbriae expression have been reported to be important for biofilm formation in regard to initial cell attachment (Pratt and Kolter 1998), and motility is essential for biofilm formation, probably for overcoming the electrostatic repulsion of cells and surfaces (Pratt and Kolter 1998; Walker et al. 2004). However, the role of motility on biofilm maturation and its effect on architecture have not been elucidated, as there is not a clear report that relates motility and mature biofilm formation in E. coli (Reisner et al. 2003; Lazazzera 2005). Although studies on DNA microarrays of E. coli cells have not found to date a significant difference in flagella expression during biofilm development (Schembri et al. 2003; Beloin et al. 2004; Ren et al. 2004), other studies have shown that motility genes are repressed in P. aeruginosa in a 5-day biofilm and in Bacillus subtilis biofilms after 8, 12, and 24 h (Whiteley et al. 2001; Stanley et al. 2003).

The aim of this study was to investigate the role of motility in biofilm architecture using a series of E. coli strains with differences in motility. We chose to use continuous flow cells so that the biofilms could be compared rigorously on a consistent basis and used software that helps to avoid microscopic artifacts (Heydorn et al. 2002). This is the first report that relates motility and E. coli mature biofilm formation.

Materials and methods

Bacterial strains and growth media

The E. coli strains and plasmids are listed in Table 1. Biofilm flow chamber studies were performed using strains harboring the GFP plasmid pCM18 (Hansen et al. 2001) to visualize the biofilm. Minimal medium M9 with 0.4% casamino acids and 0.4% glucose (M9C glu) (Rodriguez and Tait 1983) was used to assay motility gene expression) for the flow chamber experiments. For plasmid selection, antibiotics were added to the cultures at 100 μg/ml for ampicillin, 50 μg/ml for kanamycin, and 30 μg/ml for chloramphenicol.

Table 1 E. coli strains and plasmids used. AmpR and EmR are ampicillin and erythromycin resistance, respectively
Full size table

Promoter transcriptional assays

MG1655, ATCC 25404, BW25113, and JM109 with the plasmid-based (Table 1) transcriptional reporters qseB::lacZ, flhD::lacZ, fliA::lacZ, fliC::lacZ, and motA::lacZ (Sperandio et al. 2002) were cultured overnight in LB ampicillin, diluted 1:100 in LB media, and grown to exponential phase to a turbidity of 1 at 600. Once cells reached this OD, they were centrifuged for 5 min at 4°C at 13,800×g, resuspended in 750 μl of Tris–ethyl diamine tetraacetic acid–phenyl methyl sulphonyl fluoride buffer, and sonicated in two intervals of 15 s at 10 W. The samples were assayed for β-galactosidase activity using o-nitrophenyl-β-galactopyranoside (Sigma, St Louis, USA) as described previously (Wood and Peretti 1991). All activities were calculated based on a protein concentration of 0.24 mg protein/ml/OD600 unit (Fishman et al. 2005). Each experiment was performed twice with two independent cultures for each strain in M9C glu.

Flow chamber biofilm experiments and image analysis

M9C glu medium was supplemented with erythromycin to maintain the green fluorescent plasmid pCM18 at 120 μg/ml for MG1655, ATCC 25404, DH5α, and MG1655 motA and 300 μg/ml for JM109 and BW25113. Biofilms were formed at 37°C in a continuous flow chamber (BST model FC81, Biosurface Technologies, Bozeman, MT) with chamber dimensions of 47.5×12.7 mm with a 1.6-mm gap between the surfaces by inoculating 1.1 to 4.35×107 cells/ml. The flow chambers were cleaned with 95% alcohol and sterilized by autoclaving. The flow cell contains a standard glass microscope slide on one side and a plastic cover slip on the other side. An overnight culture was diluted 1/100 in the M9C glu medium. A diluted overnight culture (turbidity of 0.05) was allowed to flow into the flow chamber for 2 h at a flow rate of 10 ml/h (in laminar flow) before fresh medium was added at the same flow rate; this culture was aerated by stirring the feed flask. The biofilm development was monitored at 24 h (MG1655 was monitored at 24 and 48 h) using a TCS SP2 scanning confocal laser microscope with a 40× N PLAN L dry objective with correction collar and numerical aperture of 0.55 (Leica Microsystems, Heidelberg, Germany). The ATCC 25404 and BW25113 flow cells were performed in duplicate.

Color confocal flow chamber images were converted to gray scale using Image Converter (Neomesh Microsystems, Wainuiomata, Wellington, New Zealand). Biomass, surface coverage, surface roughness, and mean thickness were determined using COMSTAT image-processing software (Heydorn et al. 2000) written as a script in Matlab 5.1 (The MathWorks) equipped with the image processing toolbox. Thresholding was fixed for all image stacks, and 25 planar images were processed for each position; in total, nine positions with 25 images were obtained for each strain so a total of images were analyzed. Values are means of data from 1800 different positions at the same time point, and standard deviations were calculated based on these mean values for each position. Simulated three-dimensional images were obtained using IMARIS (Bitplane, Zurich, Switzerland). Twenty-five planar images were processed for each three-dimensional image.

Motility assays

LB overnight cultures were used to assay motility in plates containing 1% tryptone and 0.25% NaCl and 0.3% agar. The motility halos were measured at 8 h (Sperandio et al. 2002). Three plates were used to evaluate the motility in each strain, and the experiments were conducted with two independent cultures.

Results

Biofilm architecture

As different strains have been used to study E. coli biofilm formation (Pratt and Kolter 1998; Ren et al. 2001; Beloin et al. 2004; Ren et al. 2004), we investigated the impact of strain on biofilm architecture using flow chambers. The two wild-type strains ATCC 25404 and MG1655 formed good biofilms, as their architecture was that of a mature biofilm (Reisner et al. 2003), and displayed few differences for the COMSTAT parameters evaluated (Table 2, Figs. 1a and 2a). For these two strains, the most common parameter used for describing biofilm literature (Heydorn et al. 2000), mean thickness, was very similar (42±6 and 44±6 μm), as were surface coverage (21±8 and 34±13%) and the amount of biomass (24±10 and 11±8 μm3/μm2). The roughness coefficient, which provides a measure of how much the thickness of the biofilm varies (Heydorn et al. 2000), is higher in ATCC 25404 compared with MG1655 (0.51±0.10 vs 0.20±0.09); therefore, MG1655 is more heterogeneous than ATCC 25404. ATCC 25404 and MG1655 display cylindrical towers and scattered structures on the surface (Figs. 1a and 2a) which resemble the mushroom structures reported for P. aeruginosa (Lequette and Greenberg 2005) showing liquid channels inside the biofilm. However, different from mushrooms, the structures displayed by ATCC 25404 and MG1655 are not flat at the top. A more detailed view of the biofilm structure for each strain studied here can be seen on http://www.cheweb.tamu.edu/orgs/groups/Wood/biofilm%20architecture.htm.

Table 2 COMSTAT analysis of the E. coli biofilms formed in continuous flow cells with M9C glu containing erythromycin (concentrations in Materials and methods section) after 24 h, and motility of the E. coli strains at 8 h. One standard deviation is shown from two independent cultures
Full size table
Fig. 1
figure 1

Biofilm architecture for ATCC 25404 (a), BW25113 (b), JM109 (c), and DH5α (d) obtained in continuous flow chambers with M9C glu after 24 h. Scale bar represents 5 μm

Full size image
Fig. 2
figure 2

Biofilm architecture for MG1655 (a), MG1655 motA (b), MG1655 qseB (c), and MG1655 fliA (d) obtained in continuous flow chambers after 24 h with M9C glu and erythromycin for mutants. Scale bar represents 5 μm

Full size image

BW25113 (Datsenko and Wanner 2000), whose genotype (Table 1) does not exhibit mutations in genes related to biofilm formation reported to date, was not able to form a robust biofilm in minimal medium, as there was a reduction of thickness (threefold), biomass (5.4-fold), and surface coverage (ninefold) compared with ATCC 25404 (Table 2, Fig. 1b). JM109 and DH5α displayed a similar biofilm architecture with each other but different from that of ATCC 25404 and MG1655 (Fig. 1c, d) in that these two strains cover the glass surface more efficiently (substratum coverage is twofold higher), and JM109 and DH5α displayed smoother colonies compared with ATCC 25404 and MG1655 (Figs. 1a and 2a). The mean thickness (15±8 and 10±4 μm) and total biomass values (13±6 and 8±3 μm3/μm2) show fewer cells accumulated for JM109 and DH5α, respectively, compared with ATCC 25404 (42±6 μm and 24±10 μm3/μm2, respectively) and MG1655 (44±6 μm and 11±8 μm3/μm2, respectively).

The ATCC 25404 and BW25113 flow cell experiments were duplicated and we obtained consistent COMSTAT and architecture results for both. For example, the COMSTAT mean thickness for ATCC 25404 (38±12 μm vs 47±5 μm) and for BW25113 (16±2 μm vs 10±2 μm) were reproducible.

Motility

It is established that motility plays an important role in biofilm development of E. coli (Pratt and Kolter 1998) although the presence of a conjugation plasmid overrides the dependence on motility (Reisner et al. 2003). We therefore tested the motility of the strains and found that MG1655 and ATCC 25404 displayed the highest motility (MG1655 is slightly more motile than ATCC 25404) as MG1655 displayed six- and eightfold more motility than JM109 and BW25113, respectively (Table 2). There was no motility for DH5α, and the paralyzed strain MG1655 motA was non-motile, as expected.

Using transcriptional reporters, we found that the differences in motility are due to the differences in flagella gene expression (Fig. 3). MG1655 had the highest quorum sensing flagella regulon expression (qseB) (Sperandio et al. 2002) which was 28-fold greater than ATCC 25404, 139-fold greater than BW25113, and 209-fold greater than JM109. This difference in qseB expression was also reflected in the expression of motA (encodes the motor torque generator for flagella movement) (Chilcott and Hughes 2000) as motA expression for MG1655 was higher compared with ATCC 25404 (24-fold), BW25113 (123-fold), and JM109 (68-fold).

Fig. 3
figure 3

Transcription of qseB::lacZ, flhD::lacZ, fliA::lacZ, fliC::lacZ, and motA::lacZ in MG1655, ATCC 25404, BW25113, and JM109 in M9C glu. The experiment was performed with two independent cultures, and one standard deviation is shown. Note the different Y-axis scales used in the two panels

Full size image

ATCC 25404 had high qseB transcription compared with BW25113 and JM109, which explains the differences in the motility phenotype between these strains. These differences in the motility phenotype were also reflected in the biofilm architecture. The strains that were highly motile had high flagella expression (MG1655 and ATCC 25404) and displayed vertical biofilm structures, whereas the strains with low motility, as a consequence of lower flagella gene expression (BW25113, JM109, and DH5α), displayed smoother microcolonies (Fig. 1).

Motility vs flagella

To determine if the difference in biofilm architecture is related to motility or flagella assembly, qseB and fliA isogenic mutants deficient in flagella assembly were compared with the paralyzed motA mutant which has an intact flagella assembly (Chilcott and Hughes 2000) using flow cells (Table 2; Fig. 2b–d). The deletion of qseB and fliA leads to a dramatic decrease in biomass (sixfold for qseB and eightfold for fliA), surface coverage (11-fold for qseB and sevenfold for fliA), and mean thickness (sevenfold for qseB and ninefold for fliA) compared with the isogenic MG1655 wild-type strain. In contrast, the reductions in biomass, surface coverage, and mean thickness were less severe for motA mutant. The biofilm architecture of the motA mutant (Fig. 2b) was similar to the structures exhibited by E. coli JM109 and E. coli DH5α (Fig. 1c,d), which also display poor motility (Table 2). This similarity was corroborated with COMSTAT, as biomass, surface coverage, and mean thickness were similar for these three strains (Table 2); for example, mean thickness was similar for JM109 (15±8 μm), DH5α (10±4 μm), and the motA mutant (16±6 μm). Hence, E. coli biofilm architecture is impacted both by motility and the presence of flagella. Growth was not altered after deleting qseB, fliA, and motA so the changes in the biofilm are not a result of growth rate differences. The specific growth rates in M9C glu were 0.96±0.004/h for MG1655, 0.99±0.01/h for MG1655 qseB, 1.00±0.04/h for MG1655 fliA, and 0.97±0.01/h for the motA mutant.

Discussion

In this work, we show that there are differences in biofilm architecture for the E. coli strains MG1655, ATCC 25404, BW25113, JM109, and DH5α. Differences in flagella expression were also clearly demonstrated for the strains, and the level of expression of the quorum sensing master flagella regulon qseB (MG1655>ATCC 25404>BW25113>JM109, Fig. 3) agrees with the observed cell motility (MG1655>ATCC 25404>BW25113>JM109>DH5α, Table 2). We hypothesize that the differences in mature biofilm architecture are caused by these differences in motility. High motility leads to vertical structures displayed by MG1655 and ATCC 25404 (Figs. 1a and 2a), and when motility is reduced, smoother colonies are formed like the structures exhibited by JM109 and DH5α (Fig. 1c,d). We also determined that the differences in architecture are caused by flagella rotation as well as the by the presence of the flagella themselves. The deletion of qseB or fliA affects cell attachment and motility with the result that there is a dramatic decrease in biofilm formation, while deletion of motA (which affects only motility) results in less severe defects in the biofilm (Table 2, Fig. 2). Hence, we report a strong relationship between motility, flagella, and biofilm architecture in mature E. coli biofilms.

Unexpectedly, the difference in qseB expression was not reflected in the expression of flhD and fliA for BW25113 and of flhD, fliA, and fliC for MG1655 (Fig. 3). Out of the genes whose expression was measured, motA is the one that reflected best the expression of qseB (Fig. 3). This also suggests that the movement of flagella via motA may play the most important role in tower formation for mature biofilm architecture (for these five motility genes) (cf., MG1655 vs MG1655 motA, Fig. 2a vs b).