• Open Access

Nucleases in Bdellovibrio bacteriovorus contribute towards efficient self-biofilm formation and eradication of preformed prey biofilms


Correspondence: R. Elizabeth Sockett, Centre for Genetics and Genomics, School of Biology, Queen's Medical Centre, University of Nottingham, Nottingham NG7 2UH, UK. Tel.: +44 115 8230325; fax: +44 115 8230338; e-mail: liz.sockett@nottingham.ac.uk


Bdellovibrio bacteriovorus are predatory bacteria that burrow into prey bacteria and degrade their cell contents, including DNA and RNA, to grow. Their genome encodes diverse nucleases, some with potential export sequences. Transcriptomic analysis determined two candidate-predicted nuclease genes (bd1244, bd1934) upregulated upon contact with prey, which we hypothesised, may be involved in prey nucleic acid degradation. RT-PCR on total RNA from across the predatory cycle confirmed that the transcription of these genes peaks shortly after prey cell invasion, around the time that prey DNA is being degraded. We deleted bd1244 and bd1934 both singly and together and investigated their role in predation of prey cells and biofilms. Surprisingly, we found that the nuclease-mutant strains could still prey upon planktonic bacteria as efficiently as wild type and still degraded the prey genomic DNA. The Bdellovibrio nuclease mutants were less efficient at (self-) biofilm formation, and surprisingly, they showed enhanced predatory clearance of preformed prey cell biofilms relative to wild-type Bdellovibrio.


Bdellovibrio bacteriovorus are predatory bacteria that prey upon a wide range of Gram-negative bacteria, entering their periplasm and growing at the expense of the prey macromolecules. Once established in the periplasm, the Bdellovibrio rapidly kills the prey (Rittenberg & Shilo, 1970) and then begins to degrade cytoplasmic macromolecules, such as nucleic acids, in a controlled and stepwise manner requiring careful regulation of the hydrolytic enzymes involved (Matin & Rittenberg, 1972; Hespell et al., 1975). As replicating Bdellovibrio must generate an average of 3–5 genomes of 3.8 Mb from preying upon one cell of Escherichia coli with a single genome of similar size, efficient degradation and recycling of prey nucleic acids are predicted to be a pressing issue (in addition to de novo synthesis). Analysis of the genome of B. bacteriovorus HD100 (Rendulic et al., 2004) reveals many genes predicted to encode hydrolytic enzymes including 20 putative nucleases, some of which, we hypothesise, act to achieve this recycling. It would be expected that prey active nucleases would have secretion signals to allow their export from the Bdellovibrio into prey. Transcriptomic analysis of mRNA from the early stage of predation (Lambert et al., 2010) has highlighted two candidate-exported endonuclease-encoding genes upregulated at 30 min postintroduction of Bdellovibrio to prey cells; the products of which (Bd1244, Bd1934) may be involved in prey degradation. Here, we generate single and double mutants to investigate this possibility. Bdellovibrio have been shown to form biofilms themselves (Medina & Kadouri, 2009), and as biofilm matrices often contain many nucleic acids as an integral part of their structure (Whitchurch et al., 2002), we also examined the potential roles of the Bd1244 and Bd1934 nucleases in self-biofilms.

Materials and methods

Strains and growth media

The genome-sequenced type strain B. bacteriovorus HD100T (Stolp & Starr, 1963; Rendulic et al., 2004) was used throughout this study and grown by predation on E. coli S17-1 (Simon et al., 1983) in Ca/HEPES buffer using standard culturing methods described elsewhere (Lambert et al., 2003). Kanamycin-resistant exconjugants containing initially single crossovers of the pK18mobsacB plasmid for reciprocal recombination for gene deletion were maintained on YPSC overlay plates supplemented with 50 μg mL−1 kanamycin sulphate with kanamycin-resistant E. coli S17-1 (pZMR100) prey (Rogers et al., 1986). Host-independent Bdellovibrio derivatives were isolated on PY media as described elsewhere (Shilo & Bruff, 1965; Evans et al., 2007).

RT-PCR analysis of nuclease gene expression

Initial transcriptomic analysis had compared solely Bdellovibrio with no prey, to Bdellovibrio interacting with prey at 30 min after mixing (Lambert et al., 2010). Thus, to expand the transcriptomic ‘picture’, total RNA was extracted over the course of a semi-synchronous predatory infection and semi-quantitative RT-PCR was carried out as described previously to monitor Bdellovibrio nuclease gene expression (Lambert et al., 2006; Evans et al., 2007). The following primers were used: bd1244: Bd1244F CAATCAGTATGCGGTTCGTG Bd1244R GTTGATCACGGTGTTGTTCG, bd1934: Bd1934F AGCTTACGACAACCGTCTGG Bd1934R ACTGGATTTCTGCCCACTTG, bd1431: Bd1431F GAACGTCGAACTGCACAATG Bd1431R TAGGCATAGGCCAGGTTGTT.

Deletion of nuclease genes in B. bacteriovorus

Markerless deletions of the bd1244 and bd1934 genes were constructed by modifications of the methods of Steyert & Pineiro (2007; Santini et al., 2001). One kilobase of flanking DNA from either side of the genes was amplified and joined together to give an in-frame deletion of the orfs. This was then ligated into the kanamycin-resistant suicide vector pK18mobsacB (Schafer et al., 1994) and conjugated into Bbacteriovorus HD100 [as described in (Evans et al., 2007)]. The resulting merodiploid exconjugants were grown with kanamycin selection on YPSC overlay plates of E. coli lawns before sucrose suicide selection in 5% sucrose. The double mutant was made by conjugating the bd1934 deletion construct into a ∆bd1244 mutant. All mutants were confirmed by sequencing, Southern blot and RT-PCR to determine that the gene had been deleted in-frame as expected and that no transcript for it was present in the mutant. As these mutants were successfully isolated using the predatory HD100 strain, there was no need to attempt to rescue them by growing host-independent strains as is necessary for genes essential for predation (Hobley et al., 2012).

Nucleic acid staining during a predatory timecourse

A predatory Bdellovibrio prey lysate culture consisting of 10-mL Ca/HEPES buffer, 600 μL of a culture of E. coli S17-1 (c. 3 × 109 cells) previously grown in YT broth for 16 h at 37 °C with shaking at 200 r.p.m.) and 200 μL of Bdellovibrio (c. 1 × 109) from a previous prey lysate was incubated at 29 °C with shaking at 200 r.p.m. overnight. This was concentrated by centrifugation at 5100 g for 20 min, resuspended in 0.5-mL Ca/HEPES and added to 100 μL of a culture of E. coli S17-1 (grown in YT broth for 16 h at 37 °C with shaking at 200 r.p.m.) to give a semi-synchronous prey lysate with a MOI of > 3 as determined by plaque assay for the Bdellovibrio and colony-forming units for E. coli. At 15-min intervals, 9.5-μL samples was added to 0.5 μL Hoechst 33372 (a dye which stains DNA) and imaged by phase and fluorescent microscopy as described elsewhere (Altschul et al., 1997).

Self-biofilm formation assay

Biofilm formation assays for host/prey independent (HI) Bdellovibrio were carried out as described by Medina and Kadouri (2009), but with modifications to conveniently screen many HI derivatives of B. bacteriovorus simultaneously. Individual HI colonies were picked into 200-μL PY media (Lambert & Sockett, 2008) in 96-well microtitre dishes and grown for 48 h at 29 °C to an average OD600 of 0.4–0.6. Fifty microlitres from each well were transferred into 150-μL fresh PY in PVC 96-well microtitre dishes and incubated for 48 h at 29 °C for biofilm growth. Biofilm formation in the wells was measured by washing off planktonic cells and media with sterile distilled water, staining with 200-μL 1% (w/v in ethanol) crystal violet for 15 min, destaining with 200-μL 33% (v/v) acetic acid for 15 min and transferring 150 μL of this to a separate plate to measure OD600. At least three biological repeats were carried out. Student's t-test was performed on data to test statistical significance.

Prey biofilm depletion assay

Biofilm depletion assays for prey bacteria were carried out by modifications on the methods of Medina et al. (2008). Escherichia coli S17-1 cultures grown in LB broth at 37 °C with shaking at 200 r.p.m. for 16 h were backdiluted 1/100 in LB broth to give typical starting cell numbers of 1 × 107 CFU mL−1, and 200 μL was added to each well in PVC 96-well microtitre dishes and incubated at 29 °C for 24 h to produce a prey biofilm. The remaining planktonic cells were washed off with Ca/HEPES buffer. Planktonic cultures of predatory Bdellovibrio were grown on prey until they had fully lysed (with many attack phase Bdellovibrio and fewer than 1 prey cell per 1000 Bdellovibrio visible by phase-contrast microscopy) after incubation for 16 h at 29 °C with shaking at 200 r.p.m. before being filtered through a 0.45-μm filter to remove any remaining prey cells. Two hundred microlitres (c. 1 × 109) of these filtered predatory Bdellovibrio cells (in Ca/HEPES buffer) were added to the preformed E. coli biofilms and incubated at 29 °C for 24 h to test for predatory effects. Remaining biofilms were washed and quantified by crystal violet staining as described above. At least three biological repeats were carried out. Student's t-test was performed on data to test statistical significance.

Results and discussion

Choice of nuclease genes and transcriptional evaluation by RT-PCR

There are 20 genes annotated as encoding nucleases in the B. bacteriovorus HD100 genome (Rendulic et al., 2004); these are listed in Table 1 with the top blastp hits (http://blast.ncbi.nlm.nih.gov/Blast.cgi) and smart protein domains shown (http://smart.embl-heidelberg.de/smart/set_mode.cgi). This, despite the predatory degradation of prey nucleic acids by Bdellovibrio, is actually not an excessive number, relative to that of nonpredatory E. coli MG1655 (41 annotated nucleases in a genome of 4.6 Mb). This is also the case for other predatory bacteria such as Myxococcus xanthus (52 annotated nucleases in a genome of 9.1 Mb). Many of the Bdellovibrio nuclease genes encode products with homology to ‘housekeeping’ genes such as those encoding DNA repair mechanisms (e.g. the excinucleases and others in Table 1), but many are of unknown function. Transcriptomic analysis during early bdelloplast formation, 30 min after introducing Bdellovibrio to prey cells, revealed that of these predicted nucleases, two endonuclease genes were upregulated at this stage: bd1244 and bd1934 (Lambert et al., 2010). RT-PCR analysis across the predatory life cycle confirmed that expression of these genes was induced upon introduction to prey cells and peaked 30–45 min after this, implicating a role for these genes in the predatory process (Fig. 1), whereas in comparison, the predicted endonuclease bd1431 was constitutively expressed and hence is more likely to be a ‘housekeeping’ gene.

Table 1. Genes encoding proteins with homology to nucleases from the Bdellovibrio bacteriovorus HD100 genome listed in groups according to type. The top hit from blastp annotated as a nuclease is shown with its associated E-value. Motifs from the smart database are also listed with their E-value and predicted location as determined by the pSORT-B prediction program
GeneTop blastp HitExpect smart smart expectpSORT-B
uvrA bd0159Excinuclease ABC subunit A0.0ABC_tran6.80e-08Cytoplasm
uvrC bd0254Excinuclease ABC subunit C2e-70Exonuc_X-T2.60e-02Cytoplasm
uvrC bd2311Excinuclease ABC subunit C1e-164UvrC_HhH_N6.60e-51Cytoplasm
uvrA bd2442Excinuclease ABC, A subunit0.0ABC_tran1.50e-08Cytoplasm
xseA bd0197Exodeoxyribonuclease VII large subunit1e-134Exonuc_VII_L2.30e-80Cytoplasm
xseB bd0198Exodeoxyribonuclease VII small subunit1e-15Exonuc_VII_S5.00e-19Cytoplasm
bd1501ATP-dependent exoDNAse (exonuclease V)6e-18Nonen/aUnknown
recJ bd2232Single-stranded DNA-specific exonuclease RecJ1e-93DHH domain4.00e-07Cytoplasm
bd3524Exodeoxyribonuclease III3e-84Exo_endo_phos7.10e-41Cytoplasm
exoA bd3670Exodeoxyribonuclease III2e-87Exo_endo_phos2.30e-43Cytoplasm
bd0591Endonuclease III3e-88ENDO3c1.70e-62Cytoplasm
endA bd0934Endonuclease I2e-38Endonuclease_13.60e-51Extracellular
bd1244Secreted nuclease1e-106Endonuclease_14.40e-51Periplasmic
Bd1431Micrococcal nuclease-like protein3e-30SNase (staphylococcal nuclease)7.10e-09Unknown
bd1934Endonuclease YhcR4e-26SNc (staphylococcal nuclease)7.63e-13Unknown
bd3507Endonuclease I3e-34Endonuclease_11.60e-44Extracellular
tatD bd1042Putative deoxyribonuclease2e-73TatD_DNase2.50e-71Cytoplasm
bd3139UvrD/REP helicase subfamily5e-42UvrD-helicase2.10e-60Cytoplasm
bd3140Double-strand break repair protein AddB2e-08Nonen/aCytoplasm
bd3695Type I restriction-modification system 2C4e-53Methylase_S4.70e-11Cytoplasm
Figure 1.

Expression patterns of genes encoding proteins with homology to exported endonucleases across the predatory cycle studied by RT-PCR. RT-PCR with transcript-specific primers on total RNA prepared from identical volumes of Bdellovibrio bacteriovorus HD100 predator with Escherichia coli S17-1 prey infection culture as the predatory infection proceeds across a timecourse. L-NEB 100-bp ladder, AP – attack phase 15–45: 15–45 min of predation, respectively, 1–4 h: 1, 2, 3 and 4 h of predation, respectively, S17: E. coli S17-1 only RNA as template, −ve: no template control, +ve: B. bacteriovorus HD100 genomic DNA as template positive control. Primers designed to predicted nuclease gene bd1431 give a product in every sample, thus act as a positive control for the RNA, validating the lack of expression of bd1244 and bd1934 in the earlier and later parts of the infectious cycle. The expression peaking at 30–45 min implicates a role in interaction with the prey cells, whereas bd1431 is constitutively expressed and hence less likely to have such a role. Above is a cartoon of the stages of predation represented by each timepoint.

Analysis of the sequences using the SignalP program indicated that bd1244 and bd1934 products (and also bd0934, bd1431 and bd3507) were likely to have a signal sequence for sec-dependent transport across the inner membrane and are therefore predicted to be at least extracytoplasmic (and possibly exported beyond out of the cell by other transport systems such as PulD), and therefore, we considered them unlikely to be housekeeping genes for chromosome maintenance. Table 1 shows that the five endonucleases with predicted signal sequences have conserved domains placing them in the endonuclease I superfamily (the staphylococcal nuclease is part of this family), the biological functions of which are unclear in most other bacteria. In pathogenic bacteria such as Streptococcus pneumoniae, an extracellular nuclease of this family degrades DNA meshes of neutrophil extracellular traps to escape host immune responses (Altschul et al., 1997), whilst conversely, in Vibrio cholerae, extracellular nucleases of this family are involved in biofilm formation (Seper et al., 2011), suggesting a diversity of functions for these. Figure 2 shows a multiple sequence alignment of different predicted endonuclease I protein sequences using the ClustalW program. This shows that the two metal-binding residues known in the Vibrio vulnificus protein (Li et al., 2003) are conserved amongst the gamma and epsilon proteobacteria, but only one (Asn127-Vibrio numbering, arrow on Fig. 2) is conserved, whilst the other (Glu79) is not amongst the delta proteobacteria, suggesting that they have a different structure. Similarly, the cysteine residues that form disulphide bridges in the Vibrio protein (Li et al., 2003; below the asterisks in Fig. 2) are also conserved in the gamma and epsilon proteobacteria and not the delta proteobacteria. Figure 3 shows a multiple sequence alignment of different predicted SNase protein sequences and in contrast to the endonuclease I alignment shows that all of the conserved residues known to be involved in metal or substrate binding in Staphylococcus aureus (Ponting, 1997) are completely conserved in the delta proteobacteria, suggesting that these are likely a similar structure and mode of action.

Figure 2.

Multiple sequence alignment of predicted endonuclease I proteins created with the ClustalW program. The metal-binding residues known in the Vibrio vulnificus protein are indicated by arrows and are conserved amongst the Gamma- and Epsilonproteobacteria, but only one (Asn127-Vibrio numbering) is conserved, whilst the other (Glu79) is not amongst the Delta proteobacteria, suggesting that they have a different structure. Asterisks are above the cysteine residues that form disulphide bridges in the Vibrio protein are also conserved in the Gamma- and Epsilonproteobacteria and not the Deltaproteobacteria. Bd number refer to the predicted Bdellovibrio proteins in Table 1, Bacteriovorax sequences are predicted proteins from Bacteriovorax marinus (Crossman et al., 2012), Vvn is the sequence of the V. vulnificus protein (Seper et al., 2011) and the others are from genera as named.

Figure 3.

Multiple sequence alignment of predicted SNase proteins created with the ClustalW program. Arrows show all of the conserved residues known to be involved in metal or substrate binding in Staphylococcus aureus, which are completely conserved in the Deltaproteobacteria, suggesting that these are likely acting in the same way. Bd numbers refer to the predicted Bdellovibrio proteins in Table 1, and the others are from genera as named.

Nuclease deletion mutants

The B. bacteriovorus HD100 genome has many copies of genes with similar functions (Rendulic et al., 2004) and often display redundancy with single mutants sometimes having no obvious lack of function (Lambert et al., 2006; Morehouse et al., 2011). Therefore, to investigate the roles of the two nucleases implicated in prey interaction, in addition to generating deletion mutations of bd1244 and bd1934, we also generated a double ∆bd1244bd1934 mutant with both genes deleted. The genes were deleted such that only the first two and last three codons remained, so that the deletion was in-frame and therefore unlikely to affect surrounding genes. Deletion was confirmed by sequencing the regions in the mutants, Southern blot and RT-PCR to confirm the absence of transcript (data not shown). The mutants showed no obvious morphological differences and could prey upon E. coli in a manner apparently identical to wild type, with semi-synchronous, planktonic, predatory cultures (MOI > 3) showing Bdellovibrio swimming rapidly, attaching to and entering, then growing within and lysing prey cells within 3–5 h as did the wild-type HD100 strain. A predation assay using luminescent prey was performed as described previously (Lambert et al., 2003) and showed no difference in rate of predation between the mutants and wild type (data not shown).

Hoechst 33372 staining of nucleoid material throughout the predation process

To observe any differences in prey DNA degradation, semi-synchronous cultures of wild-type and double ∆bd1244bd1934 mutant were set up (MOI > 3) on E. coli prey, and samples were taken and stained with Hoechst 33372, which stains DNA and can be observed under fluorescence and phase-contrast microscopy. The Bdellovibrio cells showed a very intense signal as their large 3.8 Mb genome is compacted into a small space, and the E. coli prey had a more diffuse signal reflecting their less concentrated DNA. No obvious differences in the disappearance of the prey DNA fluorescence upon predation by the different strains were observed. It may be that this assay was not sensitive enough to monitor minor differences in the rate of DNA breakdown or that the nucleases have an altogether different role in the Bdellovibrio lifestyle, which is induced upon signals from detecting the presence of prey, such as biofilm formation. We tested this idea further for the single and double nuclease mutants.

Biofilm formation by HI Bdellovibrio

Medina and Kadouri (2009) describe the ability of HI Bdellovibrio to form biofilms and it is known that extracellular DNA (eDNA) is often important in biofilm formation of other bacteria (Whitchurch et al., 2002), so we tested the ability of the nuclease mutants to form biofilms. As HI derivatives of Bdellovibrio have, without additional deliberately introduced mutations, widely varying phenotypes of morphology, growth rate and cell coloration (Seidler & Starr, 1969; Barel & Jurkevitch, 2001); many different HI derivatives of each mutant were tested in these experiments by growth in PVC microtitre plates in PY for 48 h by the method of Medina and Kadouri (2009). This was to ensure that differences between strains were not merely a result of normal variation of HI derivatives. Figure 4a shows that both the ∆bd1244 mutant and the double ∆bd1244bd1934 mutant formed Bdellovibrio HI biofilms less efficiently than wild-type and the ∆bd1934 mutant. These differences were significant as determined by Student's t-test (P > 0.001), implying that the Bd1244 nuclease is involved in biofilm formation. Extracellular nucleases of this endonuclease I family have been shown to alter biofilm formation in Vibrio cholerae (Seper et al., 2011) in that instance deletion of the two nuclease genes, both individually and together, resulted in greater biofilm production with the nucleases implicated in mechanisms of biofilm architecture, dispersal and nutrient acquisition. This disparity could be a result of different growth media used as this is known to have a profound effect on biofilm formation (Bininda-Emonds, 2005). As HI Bdellovibrio have a stringent requirement for nutrient-rich peptone–yeast extract-based (PY) media, this media effect could not be tested further. Different species of bacteria produce widely varying amounts of eDNA in biofilms (Steinberger & Holden, 2005) with a diversity of roles, so it is easily conceivable that the Bdellovibrio nucleases have different roles to that in Vibrio In diverse bacteria, eDNA has been shown to have different roles, for example, DNaseI detached biofilms of Staphylococcus aureus, but not Staphylococcus epidermidis; in Caulobacter, eDNA of biofilms binds to the holdfast of stalked cells and inhibits cell attachment to the biofilm (Berne et al., 2010). Bdellovibrio also have specialised attachment processes (to prey) allowing them to distinguish between prey and nonprey cells [although the details of this are elusive (Varon & Shilo, 1969) and may not be the same as for Caulobacter]. Therefore, it is possible that an analogous process is happening here, and in the absence of nucleases, a build-up of eDNA inhibits further planktonic cells from joining the biofilm that results in slower biofilm formation by the Bd1244 mutant.

Figure 4.

(a) Self-biofilm formation (as crystal violet staining intensity) by Bdellovibrio HI isolates. HI isolates from each of the nuclease-mutant and wild-type Bdellovibrio bacteriovorus HD100 were incubated in PVC microtitre plates in PY media for 48 h. The resulting biofilms were washed and stained with crystal violet. The stain was then removed and quantified by OD600 readings. The bd1244 mutant and double mutant formed less self-biofilm than the wild-type and bd1934 mutant. (b) Elimination of preformed Escherichia coli prey biofilms by Bdellovibrio HD strains. Escherichia coli biofilms were pregrown in PVC microtitre plates for 24 h, and then, filtered cultures of Bdellovibrio (log8 ± log0.5 PFU) were added and incubated for a further 24 h. The remaining prey biofilms were washed and stained with crystal violet. The stain was then removed and quantified by OD600 readings. The nuclease-mutant Bdellovibrio strains surprisingly clear the biofilm more efficiently than the wild type. Ca/HEPES buffer was added in place of Bdellovibrio as a negative control.

Several attempts to clone the Bdellovibrio nuclease genes to complement the mutants were unsuccessful. This is likely due to the genes being lethal expressed in trans in E. coli. As the cysteine residues that form disulphide bridges in endonuclease I (Li et al., 2003) are not conserved in Bd1244 of Bdellovbrio, it is possible that at least some of the protein can fold into an active conformation before export and this could result in lethal degradation of the cloning strain genome.

Breakdown of preformed prey biofilms by Bdellovibrio

Bdellovibrio have been shown to eliminate preformed biofilms of E. coli (Kadouri & O'Toole, 2005). To investigate any possible role of nucleases in this, we tested the mutants using the methods of Medina et al. (2008) by pregrowing a biofilm of E. coli on PVC microtitre plates for 24 h, washing and then applying filtered Bdellovibrio for 24 h. Figure 4b shows that surprisingly, all of the mutant strains eliminated preformed biofilms more efficiently than wild-type strain HD100. These differences were significant as determined by Student's t-test (P ≫ 0.001). Plaque enumerations confirmed that there were similar numbers (log8 ± log0.5) of viable Bdellovibrio in the samples initially added.

One interpretation of this interesting result is that for each of the nuclease mutants, reduced modification of the eDNA in the prey biofilm partially restricts Bdellovibrio escape from the biofilm matrix. This could then cause increased localised predation and hence more prey death in the biofilm compared to wild-type Bdellovibrio predation.

The disparity between the ∆bd1934 mutation causing an effect here, but not in the HI self-biofilms, may be due to different Bdellovibrio gene expression caused by the different media conditions. Alternatively, it may be that the disruption of bd1934 has resulted in overproduction of another nuclease in that strain, which breaks down biofilm nucleic acids. RT-PCR with primers annealing to bd1244, bd1934 and bd1431 (as a sample of putatively exported endonucleases that may have compensated in the other two mutants) showed no evidence of this in any of the mutants (data not shown), but there is a possibility that abnormal expression of other nucleases may be overcompensating for the mutant defect. This counter-intuitive result, of increased predation in nuclease-defective mutants, is interesting to note should Bdellovibrio ever fulfil its potential as a biocontrol agent. It may be possible to engineer strains that are more potent than wild-type strains at eradicating prey in specific situations such as biofilms.


This work was funded by grants to R.E.S. for C.L. from Wellcome Trust AL077459/Z/05/Z and then from BBSRC G013632/1.