SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Biofilms are communities of surface-attached, matrix-embedded microbial cells that can resist antimicrobial chemotherapy and contribute to persistent infections. Using an Escherichia coli biofilm model we found that exposure of bacteria to subinhibitory concentrations of ribosome-targeting antibiotics leads to strong biofilm induction. We present evidence that this effect is elicited by the ribosome in response to translational stress. Biofilm induction involves upregulation of the polysaccharide adhesin poly-β-1,6-N-acetyl-glucosamine (poly-GlcNAc) and two components of the poly-GlcNAc biosynthesis machinery, PgaA and PgaD. Poly-GlcNAc control depends on the bacterial signalling molecules guanosine-bis 3′, 5′(diphosphate) (ppGpp) and bis-(3′-5′)-cyclic di-GMP (c-di-GMP). Treatment with translation inhibitors causes a ppGpp hydrolase (SpoT)-mediated reduction of ppGpp levels, resulting in specific derepression of PgaA. Maximal induction of PgaD and poly-GlcNAc synthesis requires the production of c-di-GMP by the dedicated diguanylate cyclase YdeH. Our results identify a novel regulatory mechanism that relies on ppGpp signalling to relay information about ribosomal performance to the Pga machinery, thereby inducing adhesin production and biofilm formation. Based on the important synergistic roles of ppGpp and c-di-GMP in this process, we suggest that interference with bacterial second messenger signalling might represent an effective means for biofilm control during chronic infections.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

In response to various stress conditions, bacteria like Escherichia coli can form communities of aggregated, surface-attached cells called biofilms. Cells in a biofilm typically express proteinaceous adhesive organelles, e.g. pili or fimbriae and secrete exopolysaccharides, including alginate, cholanic acid, cellulose or poly-β-1,6-N-acetyl-glucosamine (poly-GlcNAc) (Branda et al., 2005). These factors constitute a species-specific extracellular matrix, which serves as protective encasement against physical or chemical stress and against predation by the host immune system. Importantly, cells in a biofilm display a strongly decreased susceptibility to antibiotics and the host immune system (Mah and O'Toole, 2001; Furukawa et al., 2006). Resistance is mediated by the protective properties of the extracellular matrix and by subpopulations of metabolically dormant cells. These biofilm-associated persister cells are believed to be the base for latent and recurrent infections (Costerton et al., 1999; Lewis, 2007). While acute infections can be treated effectively in most cases, chronic infections like endocarditis, infections linked to prosthetic implants or recurring urinary tract infections, are notoriously difficult to eradicate and represent a public health problem of increasing importance (Fux et al., 2005).

In recent years it was shown that bacteria display species-specific, antibiotic-specific and dose-dependent transcriptional responses upon challenges with subminimal inhibitory concentrations (sub-MIC) of antibiotics (Goh et al., 2002; Tsui et al., 2004; Yim et al., 2006). These findings have led to the hypothesis that antibiotics can be intercellular or even interspecies signalling molecules and that the presence of low levels of antibiotics can evoke beneficial adaptational responses (Yim et al., 2007; Fajardo and Martinez, 2008). A number of bacterial species, including major human pathogens, respond to the presence of sub-MIC levels of antibiotics with increased biofilm formation (Bisognano et al., 1997; Rachid et al., 2000; Blickwede et al., 2004; Hoffman et al., 2005; Linares et al., 2006). In one report it was suggested that biofilm induction in response to antibiotic challenge is mediated by the intracellular signalling molecule cyclic di-GMP, a bacterial second messenger that is known to stimulate biofilm formation in a wide range of bacteria (Hoffman et al., 2005; Cotter and Stibitz, 2007). However, knowledge about the molecular details underlying bacterial adaptation to sub-MIC of antibiotics in general, and biofilm induction in particular is scarce (Fajardo and Martinez, 2008). In patients undergoing antimicrobial chemotherapy, pathogens can be exposed to subinhibitory concentrations of drugs for several hours (Craig, 1998). Also, widespread usage of antibiotics in farm animals and agriculture might lead to increasing exposure of individuals to low levels of antibiotics (Smith et al., 2002). Together, this suggests that biofilm formation and bacterial persistence can be a specific adaptation to antibiotic stress in the host. We sought to systematically analyse the effects of subinhibitory concentrations of antimicrobials on biofilm formation in order to define the cellular and molecular mechanisms involved in this phenomenon. As a model we chose an E. coli K-12 csrA::Tn5 mutant strain (Romeo et al., 1993; Timmermans and Van Melderen, 2008) that forms biofilms under laboratory conditions. These biofilms rely entirely on the polysaccharide adhesin poly-GlcNAc (Wang et al., 2004). Four proteins that reside in the cell envelope catalyse poly-GlcNAc biosynthesis and export. These include PgaA, which forms a pore across the periplasm and the outer membrane, and together with the N-acetyl-glucosamine deacetylase PgaB is required for export of the polymer (Itoh et al., 2008). The glycosyltransferase PgaC resides in the inner membrane and catalyses poly-GlcNAc polymerization from the precursor UDP-GlcNAc. The role of PgaD is less clear, but it is known to be an inner membrane protein (Daley et al., 2005) and is essential for poly-GlcNAc biosynthesis (Wang et al., 2004; Itoh et al., 2008). The pga genes are arranged in an operon, pgaABCD, which is negatively controlled on the mRNA level by the RNA binding protein CsrA (Wang et al., 2005). CsrA activity is governed by a complex signal transduction cascade that controls the levels of two small regulatory RNAs (CsrB and CsrC), which sequester CsrA and thereby prevent CsrA activity (Suzuki et al., 2006; Babitzke and Romeo, 2007). Poly-GlcNAc is utilized as an adhesin by a number of important bacterial human pathogens, including Yersinia (Bobrov et al., 2008), Staphylococcus (Gotz, 2002) and Bordetella (Parise et al., 2007). Importantly, a majority of clinical isolates of uropathogenic E. coli (UPEC) express poly-GlcNAc in the host environment, where it contributes to in vivo virulence (Cerca et al., 2007). Likewise, the response regulator UvrY that controls the levels of CsrB and CsrC has been shown to be a virulence factor in a uropathogenic E. coli-based bladder infection model (Tomenius et al., 2006). However, the host signals that feed into the regulatory cascade controlling pga expression are unknown. Therefore, we chose the csrA::Tn5 mutant as a biofilm model. This model system allows basal level expression of poly-GlcNAc and biofilm formation and thus represents a valid in vitro approximation of the situation in the host. Exploiting the simple biofilm readout provided by this strain in combination with the powerful genetic tools available for E. coli K-12, we set out to dissect the molecular mechanisms underlying biofilm induction by sub-MIC levels of antibiotics. We show that poly-GlcNAc-dependent biofilm formation is strongly induced by sublethal doses of all tested translation inhibitors. This effect is triggered by the ribosome itself and information about the ribosomal status is transmitted to the poly-GlcNAc machinery via the bacterial signalling molecule ppGpp. In addition, we show that poly-GlcNAc production and maximal biofilm formation require another bacterial signalling molecule, c-di-GMP. Together, these second messengers control biofilm formation by specifically regulating the cellular levels of two proteins of the poly-GlcNAc biosynthesis complex. Thus, our study identifies the sensory, signal transduction and output mechanisms underlying bacterial adaptation to antibiotic challenges.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Translation inhibitors induce biofilm formation

To define the spectrum of compounds inducing a biofilm response, a comprehensive chemical library including more than 200 antimicrobials and related substances was screened. csrA::Tn5 mutant cells were grown in microtiter plates containing tryptone broth (TB) medium supplemented with four different concentrations of each of the various antimicrobials. After 24 h, the optical density (cell density) was recorded, the planktonic phase was discarded, the wells were washed vigorously and the surface-attached biomass was quantified. The ratio of attached biomass divided by the cell density is a measure for biofilm formation (see Experimental procedures). As expected, the vast majority of antimicrobial substances displayed a progressive growth-inhibitory effect with increasing concentrations (Table S1). The presence of many different individual antibiotics, targeting a wide range of cellular processes, led to induction of biofilm. However, whereas most classes of antibiotics, e.g. the β-lactams or the quinolones, had no coherent effect on biofilm formation (some members of a group induced biofilm while others inhibited biofilm), all antibiotics that target the ribosome strongly induced biofilm formation in a concentration-dependent fashion (Table S1). Because of this striking pattern and the prominent role of translation inhibitors as anti-infectives, we decided to focus on this group of antibiotics and to analyse the underlying molecular principles of biofilm induction. Towards this goal, four antibiotics representing the major chemical classes of translation inhibitors were tested. At increasing concentrations, all four drugs led to a strong increase of biofilm formation, with the strongest induction observed at concentrations that reduced cell density by 50–70% (Fig. 1A). As the antibiotics approached the MIC, biofilm formation rapidly declined, most likely as a consequence of a cumulative effect on cell growth.

image

Figure 1. Induction of biofilm formation by representative translation inhibitors. A. csrA::Tn5 cells were exposed to the indicated antibiotics for 24 h and their cell density and surface attachment was measured. Bars represent biofilm formation (surface-attached biomass divided by optical density of total cells) with standard errors of the mean. Biofilm values are indicated on the left y-axis. Curves represent relative optical density of total cells (optical density divided by the value of optical density in the absence of antibiotics) with standard errors. Values for normalized cell density are indicated on the right y-axis. B. Scanning electron micrographs of biofilms. csrA::Tn5 cells exposed to the indicated antibiotics are compared with a control without antibiotic. Scale bars are indicated. Arrows indicate cell surface-associated poly-GlcNAc spheres (see text).

Download figure to PowerPoint

To corroborate these findings we analysed the fine structure of the biofilms using scanning electron microscopy. In the absence of antibiotics, E. coli cells formed flat and fragile surface structures. Upon exposure to sub-MIC levels of translation inhibitors these developed into a thick, three-dimensional mesh of cells (Fig. 1B). Filamentous appendages and spherical, knob-like structures were prominently visible on the cell surface (Fig. 1B). The filamentous structures, which were identified as flagella, did not contribute to antibiotic-induced biofilm formation (Fig. S1). In contrast, the knob-like surface structures, which are reminiscent of poly-GlcNAc-associated surface structures in Staphylococcus epidermidis or Yersinia pestis (Vuong et al., 2004; Erickson et al., 2008), correlated with biofilm formation and increased in size upon exposure to antibiotics (Fig. 1B). Likewise, cells grown in the presence of translation inhibitors displayed a stronger signal when probed with an antibody raised against poly-GlcNAc (Fig. S2). Strains with deletions in the poly-GlcNAc biosynthesis genes (ΔpgaABCD) (Wang et al., 2004) showed no biofilm formation or induction (Fig. 3A and B), failed to display the knob-like surface structures (Fig. S3), and showed a background signal when probed with the poly-GlcNAc antibody (not shown). From this we concluded that the knob-like structures represent surface-exposed poly-GlcNAc and that antibiotic treatment induces biofilms through the upregulation of this amino-sugar polymer.

image

Figure 3. Poly-GlcNAc-mediated biofilm formation is modulated by c-di-GMP. A. Left: Biofilm formation upon plasmid-mediated overexpression of the foreign DGC dgcA (black bars) is compared with a vector control (grey bars) in the indicated csrA::Tn5 strain backgrounds. Error bars are standard errors of the mean. Right: Biofilm formation upon plasmid-mediated overproduction of two different c-di-GMP-specific phosphodiesterases (yliE and yjcC) is compared with a vector control in a csrA::Tn5 background. Error bars are standard errors of the mean. B. Scanning electron micrographs of biofilms. A csrA::Tn5 strain overexpressing the foreign DGC dgcA (middle) is compared with a vector control (top) and to a strain overexpressing dgcA but lacking the pga genes (bottom). Two different magnifications are shown. Scale bars are indicated. Arrows indicate characteristic poly-GlcNAc-associated surface structures. Such structures were never observed on the surface of cells lacking the pgaABCD genes (see also Fig. S3). C. The GGDEF domain protein YdeH is essential for aminoglycoside-mediated induction of biofilm formation. A csrA::Tn5ΔydeH mutant (grey) is compared with its ydeH+ ancestor (black). Bars represent biofilm formation (surface-attached biomass divided by optical density of total cells) with standard errors of the mean. Biofilm values are indicated on the left y-axis. Curves represent relative optical density of total cells (optical density divided by the value of optical density in the absence of antibiotics) with standard errors. Values for normalized cell density are indicated on the right y-axis. D. YdeH protein levels are controlled by CsrA. Western blots of strains carrying a C-terminal 3×Flag-tagged version of YdeH are shown. Relevant genotypes are indicated. Please note the presence of a faint band for the csrA+ sample as compared with a control lacking the 3×Flag epitope.

Download figure to PowerPoint

Translation interference induces biofilm formation

Next, we asked how the bacteria sense subinhibitory drug concentrations to induce biofilm formation. In principle, the chemicals per se could be perceived by one or several dedicated chemoreceptors. Alternatively, the drugs' effect(s) on ribosome functioning could constitute the signal leading to biofilm induction. The following observations support the latter, indirect mechanism. First, one would expect that potential chemoreceptors would only bind naturally occurring antibiotics (or derivatives thereof), but would not be able to sense artificial compounds that have been introduced only recently. However, the fully synthetic oxazolidinone antibiotic linezolid (Clemett and Markham, 2000) strongly induced biofilm formation (Fig. 2A). Second, to mimic a drug-induced drop in ribosome performance, we analysed an E. coli strain that produced a truncated version of SecM (secMΔN), which inhibits translation by jamming elongating ribosomes (Nakatogawa and Ito, 2002). Overexpression of secMΔN from a plasmid led to significant induction of biofilm formation, while an empty vector control showed no response (Fig. 2B). Similar effects were observed when different translation-targeting toxins (YoeB, MazE and RelE) were overproduced from plasmids (Fig. S4). Third, mutant strains with drug-resistant ribosomes showed an altered biofilm induction behavior. A streptomycin-resistant strain with a point mutation in the gene rpsL, coding for the S12 protein of the small ribosomal subunit (Ozaki et al., 1969), displayed no growth inhibition and showed no biofilm induction, even in the presence of high concentrations of streptomycin (Fig. 2C). In marked contrast, an rpsL mutant that requires the presence of high concentrations of streptomycin for optimal ribosome functioning (Timms and Bridges, 1993) showed induction of biofilm with decreasing concentrations of the drug (Fig. 2D). Thus, both classes of rpsL mutants showed a strict correlation between decreased ribosomal performance and increased biofilm formation. We also tested biofilm induction of the streptomycin-resistant rpsL mutant in response to tetracycline, which targets the ribosome in an RpsL-independent manner (Harms et al., 2003). The rpsL mutant was ‘hypersensitive’ to tetracycline-mediated biofilm induction, with significantly higher induction values at low drug concentrations as compared with the rpsL wild-type control (Fig. 2E). Although the molecular details of this ‘hypersensitive’ induction phenomenon are unclear, the synergistic effects of the rpsL mutation and tetracycline argue that at least two features of ribosomal functioning influence biofilm formation. Altogether, these findings strongly link ribosomal performance to biofilm induction and suggest that at sub-MIC concentrations of translation inhibitors, altered translation activity is responsible for biofilm induction.

image

Figure 2. Translation interference leads to biofilm induction. In all panels, bars represent biofilm formation (surface-attached biomass divided by optical density of total cells) with standard errors of the mean. Biofilm values are indicated on the left y-axis. Curves represent relative optical density of total cells (optical density divided by the value of optical density in the absence of antibiotics) with standard errors. Values for normalized cell density are indicated on the right y-axis. A. The artificial translation inhibitor linezolid induces biofilm formation of a csrA::Tn5 strain. B. Jamming the ribosome by overproduction of SecMΔN induces biofilm formation. IPTG-mediated overproduction of a truncated version of SecM (grey) is compared with a vector control (black) in a csrA::Tn5 strain. C. Streptomycin-resistant mutants do not induce biofilm formation upon exposure to streptomycin. A streptomycin-resistant csrA::Tn5 rpsL(K43N) mutant (grey) is compared with its streptomycin-sensitive rpsLwt ancestor (black). D. A streptomycin-dependent mutant displays biofilm induction with decreasing streptomycin concentrations. A streptomycin-dependent csrA::Tn5 rpsL(R54C P91L) mutant (grey) is compared with its streptomycin-sensitive rpsLwt ancestor (black). E. A streptomycin-resistant mutant displays ‘hypersensitive’ biofilm induction in response to tetracycline. Normalized biofilm values of a streptomycin-resistant csrA::Tn5 rpsL(K43N) mutant (grey) and its streptomycin-sensitive rpsLwt ancestor (black) are compared.

Download figure to PowerPoint

The diguanylate cyclase YdeH is required for full biofilm upregulation in response to translation inhibition

Next, we investigated how information about the status of the ribosome is communicated to the poly-GlcNAc system in the cell envelope. Recently, an almost ubiquitous bacterial intracellular signalling molecule – bis-(3′-5′)-cyclic di-GMP (c-di-GMP) – was identified as a key factor controlling biofilm formation in pathogenic and non-pathogenic bacteria (Jenal and Malone, 2006; Tamayo et al., 2007). The cellular levels of c-di-GMP are controlled by two antagonistic enzyme families, diguanylate cyclases (DGCs) harbouring a GGDEF domain to produce c-di-GMP; and phosphodiesterases harbouring an EAL domain to degrade the compound (Jenal and Malone, 2006). To test if biofilm formation in our model system responds to perturbations of the cellular c-di-GMP pool, c-di-GMP signalling proteins were overproduced. Ectopic expression of the Caulobacter crescentus DGC dgcA induced biofilm formation and led to a marked increase of both number and size of the knob-like poly-GlcNAc surface structures (Fig. 3A and B). A strain lacking the poly-GlcNAc genes showed no biofilm formation and no poly-GlcNAc-associated surface structures, even when DgcA was overproduced (Fig. 3A and B). Conversely, ectopic expression of either of two predicted c-di-GMP-specific phosphodiesterase genes from E. coli, yliE and yjcC, strongly reduced biofilm formation (Fig. 3A). The latter result is consistent with the observed reduction of biofilm formation in a csrA::Tn5 strain upon overexpression of the phosphodiesterase yhjH (Suzuki et al., 2006). These findings strongly support a model where c-di-GMP signalling controls poly-GlcNAc production and thereby biofilm formation in E. coli.

According to the SMART database E. coli K-12 possesses 29 potential c-di-GMP-specific diguanylate cyclases or phosphodiesterases (Letunic et al., 2006). To identify components involved in poly-GlcNAc regulation, 29 mutant strains were constructed, each carrying a deletion of one of the respective genes. Analysis of this mutant pool identified a single strain with significantly altered biofilm formation (Fig. S5). This mutant had a deletion in the ydeH gene, which encodes a soluble GGDEF domain protein with a short 117-residue N-terminal domain of unknown function. The ΔydeH mutant not only showed a significant reduction in surface attachment (see also Fig. S3C at zero μg ml−1 streptomycin) but also a very weak signal when probed with anti-poly-GlcNAc antibodies (Fig. S2). A similar phenotype was observed for a strain harbouring a YdeH active site mutant protein (GGEEF[RIGHTWARDS ARROW]GGQEF) (Fig. 4A). The attachment defect of the ΔydeH strain was fully restored upon expression of the heterologous DGC DgcA (Fig. 4B).

image

Figure 4. YdeH is a DGC. A. A ydeH active site mutation behaves like a null allele. Biofilm formation of a csrA::Tn5 ydeH+ strain is compared with a csrA::Tn5ΔydeH mutant and a csrA::Tn5 mutant harbouring a point mutation in ydeH, leading to a defective active site motif (GGEEF to GGQEF). B. A foreign DGC can compensate the biofilm defect of a ydeH mutant. Biofilm formation of a csrA::Tn5ΔydeH mutant is compared with the csrA::Tn5 ydeH+ ancestor in the presence (black) or absence (grey) of a plasmid encoding for the foreign DGC DgcA. C. YdeH is a bona fide DGC. Rate of c-di-GMP formation as a function of substrate (GTP) concentration fitted with a simple Michaelis–Menten model (see equation) in a Hanes representation. YdeH was present at 2 μM. Error bars are standard deviations. D. The DGC activity of YdeH is product-inhibited. V0 of c-di-GMP production is plotted over the c-di-GMP concentration present at the start of the experiment. YdeH was present at 2 μM. Error bars are standard deviations. Please note that product inhibition was found to be independent of the substrate (GTP) concentration and must therefore be allosteric (data not shown).

Download figure to PowerPoint

The ydeH gene was recently identified as a member of the CsrA regulon (Jonas et al., 2008). Consistent with this, YdeH protein levels were higher in a csrA::Tn5 mutant compared with a csrA+ control (Fig. 3D). Jonas et al. (2008) also provided genetic data indicating that YdeH is a DGC. To test if YdeH possesses DGC activity in vitro, a hexahistidine-tagged version of the protein was purified by Ni-affinity and subsequent size exclusion chromatography. Based on static light scattering measurements the protein eluted from the gel filtration column as a stable dimer at a concentration of 2 μM (not shown). Biochemical characterization of YdeH revealed kinetic properties similar to other bona fide DGCs. GTP was converted into c-di-GMP (Fig. S6) with a specific activity of approximately 1.6 ± 0.2 (μM c-di-GMP) min−1 (μM YdeH)−1 and a Km for GTP of about 17 ± 3 μM (Fig. 4C). The enzyme was subject to product inhibition with a relatively large Ki for c-di-GMP of about 44 ± 9 μM, but exhibited residual activity even at high c-di-GMP concentrations (Fig. 4D). Together, these data strongly argue that YdeH is a DGC and that the ydeH mutant biofilm phenotype is caused by a reduction of cellular c-di-GMP levels.

Importantly, exposure to aminoglycosides, including streptomycin (Fig. 3C), kanamycin (Fig. S7), tobramycin, dihydrostreptomycin, apramycin, gentamicin, sisomicin or amikacin (data not shown), completely failed to induce biofilm of the ΔydeH mutant strain. This suggested that YdeH is not only required for basal level surface attachment, but is also involved in aminoglycoside-mediated induction of biofilm formation. This response is not mediated through upregulation of ydeH expression, as YdeH protein levels were unaltered in the presence of sub-MIC concentrations of streptomycin or other antibiotics (data not shown). In contrast to aminoglycosides, addition of tetracycline or chloramphenicol still led to biofilm induction of the ΔydeH mutant, although at a much lower level compared with the ydeH+ strain (Fig. S7). Thus, YdeH is essential for biofilm induction by aminoglycosides and contributes to the response to other classes of translation inhibitors. Although the molecular details underlying the differential requirement of YdeH for the response to different drugs are not clear, aminoglycosides are known to evoke a different adaptational response from ribosomes as compared with tetracycline or chloramphenicol (VanBogelen and Neidhardt, 1990).

SpoT-mediated reduction of ppGpp triggers biofilm upregulation in response to translation inhibition

Because the ΔydeH mutant showed residual biofilm induction in response to tetracycline or chloramphenicol, we reasoned that an additional signal transduction mechanism must exist to respond to non-aminoglycoside inhibitors. A candidate for such a redundant function is the signalling molecule guanosine-bis 3′, 5′(diphosphate) (ppGpp). ppGpp is involved in the response to nutrient starvation-induced translational stress in bacteria (Cashel et al., 1996) and has been previously linked to biofilm formation in E. coli and Campylobacter jejuni (Balzer and McLean, 2002; McLennan et al., 2008). In E. coli, the cellular ppGpp concentration is controlled by two enzymes, RelA and SpoT (Ramagopal and Davis, 1974; Xiao et al., 1991). RelA has GDP diphosphokinase activity and uses ATP and GDP to produce ppGpp. SpoT is bifunctional and comprises both diphosphokinase and ppGpp hydrolase activity. To test whether RelA or SpoT are involved in biofilm formation mutants lacking either RelA (ΔrelA) or RelA and SpoT (ΔrelAΔspoT) were analysed. Whereas the ΔrelA single mutant exhibits (SpoT-derived) residual levels of ppGpp, the double mutant is completely devoid of the signalling compound and is therefore also referred to as ppGpp0 mutant (Xiao et al., 1991). As shown in Fig. 5A, the ΔrelA mutant displayed slightly higher relative surface attachment as compared with the isogenic relA+ strain. In contrast, biofilm formation was strongly increased in the ΔrelAΔspoT double mutant, with biofilm values reaching levels similar to those observed upon antibiotic induction of a relA+spoT+ strain. Increased attachment of the ΔrelAΔspoT mutant was accompanied by an upregulation of poly-GlcNAc-associated surface structures (Fig. 5C, Fig. S2) and was entirely dependent on the genes encoding the poly-GlcNAc synthesis machinery (data not shown). Strikingly, increased biofilm formation of the ppGpp0 mutant was also fully dependent on the presence of YdeH (Fig. 5A), arguing that c-di-GMP and ppGpp together control biofilm formation through poly-GlcNAc synthesis. This notion is further supported by the finding that the increased biofilm formation of a ppGpp0 strain was diminished by overproduction of either of two c-di-GMP-specific phosphodiesterases (Fig. 5B). YdeH protein levels were not altered in a ppGpp0 strain, indicating that ppGpp does not influence biofilm formation by modulating ydeH expression (Fig. 5D).

image

Figure 5. ppGpp controls biofilm formation in a c-di-GMP- and poly-GlcNAc-dependent fashion. A. Bars represent biofilm values with standard errors of the mean. All strains are csrA::Tn5. Relevant genotypes are indicated. The spoT(D259N) allele confers a ppGpp synthase-negative, ppGpp hydrolase-positive phenotype. The spoT(D73N) allele confers a ppGpp synthase-positive, ppGpp hydrolase-negative phenotype. Note that in the presence of a wt relA allele, spoT cannot be deleted because accumulation of ppGpp is toxic (Cashel et al., 1996). B. Bars represent normalized biofilm values. A csrA::Tn5ΔrelAΔspoT (ppGpp0) strain harbouring plasmids encoding for c-di-GMP-specific phosphodiesterases (YliE, YjcC) or a control plasmid are shown in comparison. C. Scanning electron micrographs of a csrA::Tn5ΔrelAΔspoT (ppGpp0) strain (left) and a relA+spoT+ control (right) are compared. Arrows indicate surface structures associated with poly-GlcNAc. D. ppGpp does not influence YdeH protein levels. Western blot of strains harbouring a 3×Flag-tagged version of YdeH are shown. All strains are csrA::Tn5. Relevant genotypes are indicated. E. A ppGpp0 strain shows aberrant biofilm induction in response to chloramphenicol. A csrA::Tn5ΔrelAΔspoT strain (grey) is compared with its relA+spoT+ ancestor (black). Bars represent biofilm formation with standard errors of the mean. Biofilm values are indicated on the left y-axis. Curves represent relative optical density of total cells with standard errors. Values for normalized cell density are indicated on the right y-axis. F. Biofilm formation of a ppGpp0ΔydeH strain is diminished and cannot be induced by chloramphenicol. A csrA::Tn5ΔrelAΔspoTΔydeH strain (grey) is compared with its relA+spoT+ydeH+ ancestor (black). Bars represent biofilm formation with standard errors of the mean. Biofilm values are indicated on the left y-axis. Curves represent relative optical density of total cells with standard errors. Values for normalized cell density are indicated on the right y-axis. G. Treatment with chloramphenicol leads to reduction of the cellular ppGpp pool. Bars indicate cellular ppGpp levels (pmol mg−1 dry weight) of a csrA::Tn5 strain that has been grown in the presence or absence of chloramphenicol (1.5 μg ml−1). Values are derived from HPLC measurements of ppGpp in cell extracts (see Experimental procedures).

Download figure to PowerPoint

To determine whether SpoT-derived ppGpp synthase or hydrolase activity is responsible for biofilm control, we introduced mutations in spoT that specifically affected one of the two enzymatic activities by replacing invariant residues in the enzyme's ppGpp synthase (Asp259) or hydrolase (Asp73) active centres (Hogg et al., 2004). The ΔrelA spoT(D259N) synthase mutant showed strongly increased biofilm formation, similar to the ΔrelAΔspoT strain (Fig. 5A). In contrast, the ΔrelA spoT(D73N) hydrolase mutant, which constitutively produces ppGpp, showed moderate biofilm formation, comparable to the relA+spoT+ ancestor. Importantly, both the ppGpp0 and the spoT hydrolase mutants were severely impaired in biofilm induction in response to chloramphenicol, tetracycline or streptomycin. The already very high biofilm level of the ΔrelAΔspoT (ppGpp0) mutant was only weakly induced with translation inhibitors (Fig. 5E, Fig. S8A). This weak induction was accompanied by a marginal increase of the attached biomass (biofilm values not normalized to cell density) and is thus mainly based on a decreased antibiotic susceptibility of the cells in the biofilm compared with the cells in the planktonic phase (Fig. S9A, see also Experimental procedures). Likewise, the ΔrelA spoT(D73N) hydrolase mutant was unable to respond to translation inhibitors with full biofilm induction (Fig. S9B). Importantly, relA does not appear to play a role in the antibiotic induction phenomenon: a ΔrelA strain shows slightly higher basal biofilm values compared with the relA+ control strain and is not impaired in biofilm induction when challenged with translation inhibitors (Fig. S10). Altogether, these findings suggest that ppGpp inhibits biofilm and that the SpoT hydrolase activity is critical for induction of E. coli surface attachment. The data also support a model where sub-MIC concentrations of translation inhibitors cause a SpoT-dependent decrease of the cellular ppGpp pool, leading to the derepression of poly-GlcNAc production and biofilm induction. This was confirmed by the finding that cellular levels of ppGpp were indeed strongly reduced in the presence of chloramphenicol (Fig. 5G). This result is fully consistent with a series of reports demonstrating that cells exposed to translation inhibitors display markedly decreased levels of ppGpp, even under conditions that would normally lead to a stringent response (Gallant et al., 1972; Muto et al., 1975; Baracchini and Bremer, 1988; Hernandez and Bremer, 1990; Murray and Bremer, 1996). Together, this suggests that SpoT-mediated reduction of ppGpp is necessary for maximal biofilm induction. However, drug-induced biofilm formation was not completely abolished in a ppGpp0 mutant (Fig. 5E), arguing that translation inhibition does not influence biofilm formation solely through ppGpp reduction. In agreement with this, a strain lacking relA, spoT and ydeH showed no significant biofilm formation, even when challenged with optimal concentrations of chloramphenicol or tetracycline (Fig. 5F, Fig. S8B). In summary, these data suggest that the guanosine-based second messengers c-di-GMP and ppGpp together control biofilm formation in response to translational stress.

ppGpp and c-di-GMP post-transcriptionally regulate the levels of PgaA and PgaD

To address the molecular basis of c-di-GMP- and ppGpp-mediated control of poly-GlcNAc synthesis, we sought to test if either of these factors influences the expression of the pga genes. To be able to monitor Pga components, we constructed 3×Flag-tagged versions of PgaA and PgaD, which are encoded by the most proximal and most distal genes of the pga operon. Surprisingly, levels of PgaD, but not PgaA, were controlled by c-di-GMP. Deletion of the DGC coding gene ydeH or overproduction of the phosphodiesterase YjcC reduced PgaD levels (Fig. 6A), while ectopic expression of the heterologous DGC dgcA led to strongly elevated levels of PgaD, both in the presence or absence of ydeH (Fig. 6A). In contrast to PgaD, PgaA levels were not altered in a mutant lacking the DGC YdeH (Fig. 6A). Conversely, cellular levels of PgaA, but not PgaD, were controlled by ppGpp. Whereas PgaA levels were strongly increased in a strain unable to produce ppGpp, PgaD levels were unaltered under these conditions (Fig. 6A). These data argue that ppGpp negatively regulates PgaA levels, while YdeH through its product c-di-GMP stimulates PgaD protein levels. Next, we tested if translation inhibition influences PgaA and PgaD levels. PgaD showed a small but reproducible increase in response to tetracycline (Fig. 6B) or chloramphenicol (data not shown). Surprisingly, this increase was not dependent on YdeH, as PgaD levels still increased under these conditions in a strain lacking YdeH (Fig. 6B). In contrast, PgaA levels were strongly induced in response to tetracycline in the control strain, while they were constitutively upregulated and insensitive to the drug in a mutant unable to produce ppGpp (Fig. 6C). Because pgaD and pgaA are encoded in one operon, but their cellular levels were influenced differentially by c-di-GMP and ppGpp, it appeared likely that the second messengers influence the PgaA and PgaD levels post-transcriptionally. To test this idea, a translational lacZ fusion to the pga promoter, including the 5′ untranslated region of the pga operon, was used to measure pga promoter activity in response to perturbations of ppGpp or c-di-GMP levels, or in response to translation inhibitors. As expected, β-galactosidase activity of the pgaA–lacZ assay strain was dependent on the transcription factor NhaR, known to be essential for pga transcription (Goller et al., 2006), and was negatively controlled by CsrA, which is known to inhibit pga operon translation (Wang et al., 2005) (Fig. S11). However, neither deletion of ydeH nor overexpression of the DGC dgcA led to significant alteration of the LacZ activity (Fig. S11 and data not shown). Likewise, deletion of relA and spoT or exposure to subinhibitory concentrations of translation inhibitors did not change the specific LacZ activity (Fig. S11 and data not shown). Thus, none of these parameters has any measurable influence on the pga promoter or the pga 5′ untranslated region. To corroborate the above findings we constructed a complementary strain in which the native pga promoter was replaced with the l-arabinose-dependent Para promoter. The resulting strain harbours an araB–pgaA translational fusion with the promoter and the 5′ untranslated region of the pga operon being replaced with the corresponding regions of the araBAD operon. As expected, biofilm formation of strains harbouring such an araB–pga fusion is dependent on the presence of l-arabinose in the medium, but independent of nhaR and csrA (data not shown). Biofilm formation of an araB–pga (csrA+) strain was induced by tetracycline, chloramphenicol or streptomycin in an l-arabinose-dependent fashion (Fig. S12). This corroborates the above notion that poly-GlcNAc-dependent biofilm induction by translation inhibitors is independent of the pga promoter and the 5′ untranslated region. Moreover, because the araB–pga (csrA+) strain shows strong biofilm induction in response to translation inhibitors (Fig. S12), it can be ruled out that antibiotic induction depends on the presence of the csrA::Tn5 mutant allele. Together, these data demonstrate that induction of poly-GlcNAc-dependent biofilm formation by subinhibitory concentrations of ribosome inhibitors involves upregulation of at least two components of the Pga machinery, PgaD and PgaA. Our data further indicate that induction of PgaA is mediated by ppGpp signalling, while c-di-GMP specifically influences cellular levels of PgaD. Because upregulation of both proteins is independent of the promoter and the untranslated leader sequence of the pga message, drug-mediated control takes place on the post-transcriptional level.

image

Figure 6. Control of the PgaA and PgaD protein levels by ppGpp, c-di-GMP and tetracycline. A. PgaD levels are controlled by c-di-GMP, whereas PgaA levels are controlled by ppGpp. Western blots of strains harbouring a 3×Flag-tagged version of PgaD or PgaA (indicated on the right side of each panel). All strains are csrA::Tn5. Relevant genotypes are indicated. pdgcA and pyjcC represent overexpression of a foreign DGC or of a phosphodiesterase from E. coli respectively. B. Tetracycline induces PgaD protein levels. PgaD protein levels are compared by Western blotting at increasing tetracycline concentrations in csrA::Tn5 ydeH+ (top) and a csrA::Tn5ΔydeH (bottom) strain. C. Tetracycline induces PgaA protein levels. PgaA protein levels are compared by Western blotting at increasing tetracycline concentrations in a csrA::Tn5 ppGpp+ (top) and a csrA::Tn5 ppGpp0 (bottom) strain.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

With their potential to withstand antimicrobial therapy and the host immune system, bacterial biofilms represent a major problem for human health. Several reports indicated that the presence of certain antibiotics influences bacterial biofilm formation (Rachid et al., 2000; Hoffman et al., 2005; Linares et al., 2006). However, in these studies only a few selected antibiotics were tested and the mechanistic details remained largely unexplored. To analyse this phenomenon in a more comprehensive way and to decipher the underlying cellular and molecular mechanisms, we used an established E. coli laboratory biofilm model system. The model strain harbours a csrA::Tn5 transposon insertion mutation causing derepression of the primary surface adhesin poly-GlcNAc (Romeo et al., 1993; Wang et al., 2004; Itoh et al., 2008). This system was chosen because it allowed the employment of a commercially available comprehensive chemical library – the Biolog system (Bochner et al., 2001), which is ideally suited for use with E. coli K-12. The relevance of our model system is underscored by recent reports demonstrating that uropathogenic E. coli form biofilms and express poly-GlcNAc during host colonization (Anderson et al., 2004; Cerca et al., 2007). Also, CsrA homologues from a variety of pathogenic bacteria have been shown to be involved in host–pathogen interaction (Lucchetti-Miganeh et al., 2008).

When grouped according to their mode of action, it became obvious that all translation inhibitors induced biofilm formation. The few exceptions (see Table S1) most likely failed to induce biofilm formation because the screening strain encodes a resistance factor for kanamycin and closely related aminoglycosides, or because the drug concentrations present in the Biolog plates were outside of the effective range. Other classes of antibiotics, e.g. compounds targeting cell wall biosynthesis or gyrase, showed an indistinct picture with some representatives inducing biofilms (e.g. cefotaxime, enoxacin, see Table S1), while others inhibited biofilm (e.g. cefmetazole, novobiocine, see Table S1). In this study, we focused on representatives of the major classes of translation inhibitors because they constitute one of the biggest groups of antibiotics, are of great clinical relevance and, as a group, behaved very homogeneously in our initial screening.

In principle, cells could sense translation inhibitors directly via dedicated receptors or indirectly through their effect on ribosomal function. Several observations make a strong case for the latter, indirect mechanism. First, for all substances tested, biofilm upregulation correlated with a progressive effect on cell growth. Second, linezolid, a fully synthetic translation inhibitor effectively stimulated biofilm formation. Third, experiments with secMΔN and ribosome-specific toxins confirmed that interference with ribosome functioning can induce biofilm formation independently of the presence of antibiotics. Fourth, mutations in rpsL that lead to streptomycin-insensitive ribosomes completely abolished streptomycin-mediated biofilm induction, while streptomycin-dependent mutants induced biofilm with decreasing streptomycin concentrations. Interestingly, although blind to streptomycin induction, rpsL mutants displayed a higher basal level of biofilm formation as compared with the streptomycin-sensitive strains (Fig. 2C and E). Streptomycin-resistant rpsL mutants are known to exhibit ‘restrictive’ or hyperaccurate translation (Bilgin et al., 1992). It is possible that ribosome hyperaccuracy might contribute to biofilm induction. However, because most other conditions that were shown here to induce biofilms are not linked to hyperaccurate ribosomes, the mechanisms involved might be more complex and the exact nature of the signal remains unclear.

Several experiments suggested that the signal emanating from drug-affected ribosomes stimulates biofilm formation through upregulation of the cell surface-exposed poly-GlcNAc adhesin. The presence of the pga genes was absolutely essential for biofilm formation under all conditions tested and poly-GlcNAc itself, as well as components of the Pga machinery were upregulated when cells were challenged with translation inhibitors. While scanning EM demonstrated that Pga-associated surface structures increased under inducing conditions, these experiments failed to provide evidence for the upregulation of other cell surface structures like pili or fimbriae. This does not rule out that additional factors that were not visualized by EM contribute to the observed biofilm induction. Although flagella are involved in biofilm formation in a different E. coli strain, they did not play a role in the induction phenomenon (Pratt and Kolter, 1998).

How is the information about translation performance relayed from ribosomes to the Pga machinery? The second messenger c-di-GMP was considered a good candidate because of its implication in biofilm control in a wide range of bacteria (Jenal and Malone, 2006) and because the pga operon is linked to and co-regulated with the ycdT gene, encoding a DGC (Jonas et al., 2008). While biofilm formation was unaffected in a strain lacking YcdT in the presence and absence of translation inhibitors (Fig. S5 and data not shown) (Wang et al., 2005), a systematic analysis of all potential genes involved in the turnover of c-di-GMP in E. coli revealed that the DGC YdeH is essential for aminoglycoside-mediated biofilm induction and is involved in the chloramphenicol- and tetracycline-mediated response. YdeH, via its product c-di-GMP, specifically upregulates poly-GlcNAc and the levels of at least one Pga component, PgaD. The underlying mechanism of this regulation is unclear. However, the observation that upregulation of PgaD in the presence of translation inhibitors is independent of YdeH (Fig. 6B) suggests that c-di-GMP-dependent stimulation of PgaD is merely a precondition for full biofilm induction by antibiotics and that c-di-GMP signalling is not the main inducing principle.

Conversely, the direct involvement of ppGpp signalling in antibiotic-mediated biofilm induction is supported by a number of observations. First, ppGpp inhibits poly-GlcNAc-dependent biofilm formation, and strains with lesions in ppGpp signalling proteins show aberrant biofilm induction. Second, surface-exposed poly-GlcNAc as well as PgaA protein levels are negatively controlled by ppGpp. Third, the already derepressed PgaA levels in a ppGpp0 strain cannot be induced further by tetracycline. Fourth, treatment of E. coli cells with subinhibitory concentrations of chloramphenicol results in reduced levels of ppGpp. Because basal biofilm formation as well as biofilm induction by antibiotics is similar in a ΔrelA mutant and the relA+ ancestor, reduction of the ppGpp pool in response to drug-elicited translational stress must be mediated by SpoT. These data support a model where in response to partial inhibition of ribosome functioning, a SpoT-mediated reduction of ppGpp leads to the upregulation of Pga components and increased production of poly-GlcNAc.

It remains unclear which parameters of ribosome function are measured and linked to SpoT activity. Although it is well documented that the ribosome or ribosome-coupled factors can function as sensory devices (VanBogelen and Neidhardt, 1990), a direct association of SpoT with the ribosome is controversial (Gentry and Cashel, 1995; Potrykus and Cashel, 2008). However, based on recent reports one cannot rule out the possibility that under certain conditions SpoT associates with the ribosome (Wout et al., 2004). Possibly, translation inhibitors influence the GTP hydrolysis rate of ribosome-associated GTPases, which in turn might govern the balance between SpoT-mediated ppGpp synthesis and hydrolysis (Jiang et al., 2007). At first sight, the notion that slow growth of E. coli in the presence of antibiotics results in reduced levels of ppGpp appears at odds with a central dogma of ppGpp signalling, which inversely correlates ppGpp concentration with cell division rate (Cashel et al., 1996). Nevertheless, a rapid and strong decrease of the cellular ppGpp pool in response to treatment with translation inhibitors is well documented in the literature (Gallant et al., 1972; Muto et al., 1975; Baracchini and Bremer, 1988; Hernandez and Bremer, 1990; Murray and Bremer, 1996). Furthermore, global transcription analysis of relA or relA+ cells exposed to subinhibitory concentrations of the translation inhibitor puromycin (a strong inducer of biofilm, see Table S1) revealed a pattern that can be characterized as an inverse stringent response, e.g. repression of RpoS-dependent genes and amino acid biosynthesis genes and induction of ribosomal genes (Sabina et al., 2003). It has been known for a long time that ppGpp0 mutants display a very distinctive physiology, but little information is available about environmental conditions that might lead to a reduction of the cellular ppGpp pool (Xiao et al., 1991; Cashel et al., 1996; Traxler et al., 2008). Our data open up the possibility that the physiology of ppGpp0 cells represents a specific adaptation to ribosomal stress conditions that do not originate from nutritional stress or a shortage of charged tRNA species. We propose that ppGpp signalling is involved in the decision between two mutually exclusive adaptational programmes; nutrient starvation leads to an increase of the cellular ppGpp pool and evokes a stringent response, while ribosomal stress caused by the presence of translation inhibitors diminishes the cellular ppGpp pool and induces poly-GlcNAc-dependent biofilm formation. In this context it should be noted that the formation of a different type of E. coli biofilm based on curli fimbriae and cellulose expression might actually require elevated levels of ppGpp. This requirement is based on ppGpp dependence of rpoS expression, which in turn is instrumental for curli and cellulose production (Lange et al., 1995; Weber et al., 2006).

The exact molecular mechanism through which ppGpp and c-di-GMP control biofilm formation, in particular the cellular receptors that bind these second messengers, remain to be elucidated. The fact that pgaA and pgaD are encoded in the same operon, together with the finding that PgaA levels are controlled by ppGpp but not by c-di-GMP, while PgaD levels are controlled by c-di-GMP but not by ppGpp, argues for post-transcriptional regulation. This idea is strongly supported by the finding that the pga promoter and 5′ untranslated region are dispensable for antibiotic-mediated biofilm induction (Fig. S12). In addition, neither the pga promoter nor the 5′ untranslated region of the pga message appear to respond to perturbations of c-di-GMP levels, ppGpp levels or to the presence of translation inhibitors (Fig. S11). Further support for a post-transcriptional mechanism comes from the finding that DksA, a factor known to enhance the effects of ppGpp on transcription is not involved in biofilm induction by translation inhibitors (data not shown). Irrespective of the molecular mechanisms, the observation that strong biofilm formation of a ppGpp0 strain is fully dependent on the presence of YdeH (Fig. 5A) argues for a model where the effect of the two signalling molecules is not merely additive. Instead, it appears that maximal poly-GlcNAc expression depends on the exact ratio between ppGpp and c-di-GMP. It is noteworthy that c-di-GMP and ppGpp not only influence biofilm formation in an antagonistic fashion, but also virulence properties of pathogenic bacteria: while c-di-GMP is implicated in the downregulation of virulence traits (Cotter and Stibitz, 2007), basal ppGpp levels are required for full virulence of a number of bacterial species (Braeken et al., 2006). Moreover, because both guanosine second messengers are structurally related, it is conceivable that they might compete for some cellular target(s). Thus, it appears possible that the connections between c-di-GMP and ppGpp signalling are even more intricate.

It is well known that antibiotic chemotherapy can cause severe side-effects (Walker, 1996). Besides direct effects on the patients' physiology, unwanted side-effects of antibiotics are also attributed to imbalances of the commensal flora that are brought about by a strong selection for bacterial species (or life styles) that are less susceptible to the growth inhibitory effect of the drug (Dancer, 2004). Our work raises the question whether some side-effects of antimicrobial chemotherapy might be attributed to biofilm formation of host-associated bacteria that experience subinhibitory concentrations of translation inhibitors. Finally, our findings imply that pharmacological interference with ppGpp and/or c-di-GMP signalling, possibly in combination with translation inhibitors or antibiotics that have a different mode of action, might represent promising avenues for the development of novel antimicrobial strategies.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Biofilm assay

Attachment assays were carried out essentially as described (O'Toole et al., 1999). Freshly grown LB overnight cultures were diluted 1:40 into 200 μl LB medium in 96-well polystyrene microtiter plates (Falcon, ordering number 353072). When necessary, ampicillin was present at 100 μg ml−1 to select for plasmids. Other antibiotics were present at the indicated concentrations. The 96-well plates were incubated for 24 h at 30°C without shaking and cell density was recorded at 600 nm with the help of a plate reader. Subsequently, medium containing non-attached cells was discarded and the wells of the microtiter plates were washed vigorously with deionized water from a hose. After air-drying, wells were filled with 200 μl of a crystal violet solution [0.1% in H2O, 1-propanol, methanol (96.7:1.66:1.66)] and incubated with moderate shaking for at least 30 min at room temperature (RT). The staining solution was discarded and wells were washed and dried as before. Retained crystal violet was redissolved in 200 μl of 20% acetic acid and quantified at 600 nm in a plate reader. If measurements were outside the dynamic range of the plate reader, crystal violet solutions were diluted in 20% acetic acid. Normalized attachment values are ratios of the optical density of dissolved crystal violet (corresponding to the attached biomass) divided by the cell density. In general, a single data point is derived from at least six replicates. Error bars for normalized attachment values are standard errors of the mean. For antibiotic titration curves, normalized cell density values are displayed. These were calculated by dividing the mean optical density measured for a specific concentration of antibiotic by the mean optical density measured in the absence of antibiotics. Error bars for relative cell densities were calculated as follows: (X/Y)^2*((SE(x)/X)^2 + (SE(y)/Y)^2), where X and Y are the mean optical densities with and without antibiotics, and SE(x) and SE(y) are the standard errors of the mean densities. These error bars correspond to the standard errors of the ratios.

Because cells in the biofilm display a decreased susceptibility towards the action of antibiotics and other forms of stress (Costerton et al., 1995), it is possible that certain conditions lead to a selective decrease of the cell density in the planktonic phase, while the attached biomass (crystal violet value) remains unchanged. In these instances, normalized biofilm values suggest that there is a (usually weak) biofilm induction. However, to rule out the possibility that this is a mere artefact of selectively decreasing the cell density in the planktonic phase, a condition or compound can only be scored as biofilm inducing or inhibiting if it has an effect on both, the attached biomass (not normalized to cell density) as well as on the normalized value.

Screening of a chemical compound library

Strain AB400 was grown over night in TB (10 g l−1 Bacto tryptone, 5 g l−1 NaCl). The optical density at 600 nm was adjusted to 0.1 with fresh TB and 150 μl cells were inoculated in individual wells of Biolog phenotype microarray plates (Bochner et al., 2001), containing the various chemicals in freeze-dried form. Plates were incubated for 24 h at 30°C and attachment was quantified as described above. Antimicrobials and related substances are present at four increasing concentrations. Compounds were scored as growth inhibitory or growth promoting if cell density readings (see above) decreased with increasing antibiotic concentrations (indicated by ‘−’ in Table S1), or if cell density readings increased with increasing antibiotic concentrations (indicated by ‘+’ in Table S1) respectively. Normalized attachment (biofilm formation) was calculated as outlined in the section above. Compounds were scored as biofilm inducing (indicated by ‘+’ in Table S1) or inhibiting (indicated by ‘−’ in Table S1) if normalized attachment values increased or decreased with increasing antibiotic concentration respectively. Factors for cell density, attached biomass and normalized attachment values are ratios of the highest value for a given chemical divided by the lowest value for the same chemical respectively. If factors were below an arbitrarily chosen threshold of 1.5, the compound was scored as neither inhibiting nor inducing (indicated by ‘0’ in Table S1).

Bacterial strains and plasmids

All strains are derivatives of AB400 (csrA::Tn5kan) and are listed in Table S2. AB400 was constructed from E. coli K-12 MG1655 by P1 transduction with TR1-5 as donor (Romeo et al., 1993). To obtain a csrA::Tn5 mutant that harbours no antibiotic resistance cassette, the kanamycin cassette of AB400 was replaced by a chloramphenicol cassette with the help of λRED-mediated gene replacement and subsequent removal of the chloramphenicol cassette by site-specific recombination according to (Datsenko and Wanner, 2000). The resulting kanamycin- and chloramphenicol-sensitive strain (AB958) harbours a ‘gutted’ Tn5 (a Tn5 lacking the kan cassette) and was used for all subsequent strain constructions. Deletion mutations of genes coding for c-di-GMP signalling proteins and other genes were moved from a comprehensive gene deletion library [the ‘Keio collection’ (Baba et al., 2006)] into recipient strains by P1 transduction. In cases where deletion mutants were not present in the Keio collection, they were generated according to Datsenko and Wanner (2000). Resistance cassettes used as selection markers were generally removed by Flp recombinase-mediated site-specific recombination (Datsenko and Wanner, 2000). Strains AB1029 and AB1000 are spontaneous streptomycin-resistant mutants, which were selected on LB plates containing 100 μg ml−1 streptomycin and screened for streptomycin dependence on LB plates without antibiotic. Sequencing of the rpsL gene confirmed the presence of mutations leading to the indicated amino acid exchanges (see Table S2). To obtain the spoT(D259N) and spoT(D73N) alleles, spoT was first replaced with the help of λRED technology by a counter-selectable marker (the toxin ccdB under control of the l-rhamnose promoter plus a linked kanamycin resistance cassette; kind gift of K. Datsenko and B. Wanner, Purdue University) in a ΔrelA background. In a second recombineering step, this marker was replaced by splice overlap extension PCR (Higuchi et al., 1988) products encoding for the two spoT alleles by selecting for growth on minimal plates containing 0.2% l-rhamnose and 0.1% casamino acids. To confirm the presence of desired mutations and absence of undesired mutations the spoT alleles of the final strains AB1132 [spoT(D259N)] and AB1134 [spoT(D73N)] were sequenced (there is a second unrelated mutation in spoT leading to I158N, which was found to be present in our copy of MG1655 and is thus present in all our strains). Due to the constitutive high level of ppGpp in AB1134, this strain has a severe growth deficit (Cashel et al., 1996). Therefore, spontaneous fast-growing suppressors (most of which are predicted to be ppGpp0) arise with high frequency. To monitor the emergence of these suppressor mutants in liquid cultures of AB1134, aliquots from cultures were routinely plated on LB agar plates. Suppressors are easily distinguishable as larger colonies and data were only considered to be meaningful, if less than approximately 5% suppressors were present at the end of an experiment (e.g. a biofilm assay). To obtain the ydeH active site mutant allele in strain AB1299 the same strategy as for the spoT active site mutant alleles was employed (see above). The Para -pgaA fusion strain AB1028 was constructed by fusing the first codon of the araB open reading frame with the second codon of the pgaA open reading frame with the help of λRED technology at the native pga locus. This was carried out in a way that replaces the entire pga promoter plus 5′ untranslated region of the pga message with the corresponding regions of Para. The final strain harbours a copy of araC at the pga locus, which is transcribed divergently to the Para promoter. A kanamaycin cassette that was used for selection during intermediate steps of the construction of AB1028 was removed by Flp-mediated site-specific recombination (Datsenko and Wanner, 2000). Chromosomal 3×Flag-tag encoding sequences at the 3′ ends of genes were constructed according to (Uzzau et al., 2001) with the help of pSUB11. The translational pgaA–lacZ fusion strain was constructed in two steps with the help of λRED technology. In a first step, a chromosomal region comprising the native lacZYA promoter, lacI and the upstream genes mhpR and a part of mhpA was replaced by the same counter-selectable marker as mentioned above. This procedure removes any promoter that could read into lacZ and thereby cause undesired basal activity of the lacZ fusion. The mhp operon is not expressed under laboratory conditions and therefore the removal of the mhp genes can be considered a neutral mutation (Torres et al., 2003). In a second step, the counter-selectable element was replaced by the entire ycdT–pgaA intergenic region in a way that fuses the 5′ untranslated region directly to the start codon of lacZ. The kanamaycin resistance cassette (which stems from the first recombineering step and reads into the opposite direction relative to lacZ) was left intact during this procedure and was used to move this lacZ fusion into any desired recipient strain with the help of P1 transduction. The advantage of this method is that it does not involve any molecular cloning and that it is not necessary to remove the native lac locus (in a second, time-consuming step) when the fusion is introduced into a lac+ target strain. In cases where l-arabinose was used to drive gene expression from Para (Guzman et al., 1995), host strains were deleted for the araB gene, which yields a strain that allows for uptake but not metabolism of l-arabinose. Plasmid psecMΔN was constructed according to standard PCR-cloning procedures and encodes for SecM lacking the first 40 amino acids. Expression is driven from a lac promoter, which is under control of lacIq. Plasmids pyliE and pyjcC were isolated in a parallel study (A.B and U.J., unpublished) from a chromosomal expression library. pyliE harbours a 1107 bp fragment (position 873528–874635 of the genome according to the ‘Colibri’ database: http://genolist.pasteur.fr/Colibri/) inserted in the BamH1 site of pCJ30 (see Table S2). The plasmid encodes the C-terminal part of YliE starting from amino acid 443 with the sequence MLQD (derived from the vector) at the N-terminus. This peptide comprises the entire EAL domain plus a stretch of 83 N-terminal amino acids from the N-terminal domain of YliE. Expression of yliE can be induced with IPTG, but basal expression was found to be sufficient to observe the phenotypes described in the results section. pyjcC harbours a 2443 bp fragment (position 4272495–4274938 of the genome according to the ‘Colibri’ database: http://genolist.pasteur.fr/Colibri/) inserted in the BamH1 site of pCJ30 (see Table S2). The insert contains the entire yjcC gene including the native promoter (same orientation as the plasmid encoded lac promoter), plus very short truncated versions of both genes that are adjacent to yjcC (yjcB and soxS).

Anti-poly-GlcNAc immunoblots

Bacterial cultures were grown as described for biofilm assays in 96-well microtiter plates. Cells from the planktonic phase and surface-associated cells were harvested by scraping them off the surface of individual wells with a pipette tip followed by vigorous up and down pipetting. Cellular material from six wells was pooled and adjusted to the same OD. Sample processing was done according to reference (Cerca et al., 2007). Anti-poly-GlcNAc antibody raised against poly-GlcNAc from S. epidermidis was a kind gift from R. Landmann (University of Basel). Blots were quantified by scanning the blots and dividing the signal by the background intensity. The results from five independent experiments were combined and different treatments were compared with an analysis of variance (with the treatment as fixed effect and the experiment as random effects; procedure GLM in SPSS 13.0.0, SPSS, Chicago, IL). Differences between the control strain grown in the absence of antibiotics and other treatments were tested with a Dunnett posthoc analysis that corrects for multiple testing.

Anti-Flag-tag immunoblots

Bacteria were grown as for biofilm assays, total cells of several wells were pooled and adjusted to the same optical density. Samples were boiled for 5 min in SDS sample buffer and gel electrophoresis and blotting onto PVDF membrane, which were carried out according to standard protocols (Laemmli, 1970; Towbin et al., 1979). For immunodetection of 3×Flag-tagged proteins mouse monoclonal α-M2 antibody and HRP-conjugated rabbit α-mouse (DakyCytomation, Denmark) were used at 1:10000 dilutions. Blots were developed with the ECL Kit and photographic films.

Measurements of ppGpp in total cellular nucleotide extracts

Total cellular nucleotides were extracted according to the procedure by Little and Bremer (1982). Cells were grown in minimal medium A (Miller, 1972) containing 0.4% glycerol and 0.1% casamino acids in the presence or absence of 1.5 μg ml−1 chloramphenicol under conditions that closely mimic the conditions used for biofilm assays. Biofilm formation in this medium was found to be similar to biofilm formation in LB and can be induced with translation inhibitors. To rapidly prevent any turnover of nucleotides during cell harvesting, formaldehyde was added to the culture to a final concentration of 0.19% and cells were chilled on ice for 20 min. Forty microlitres of cells were spun down, the pellet was resuspended in ice cold 0.1 M KOH (1 ml), incubated on ice for 30 min and the samples were acidified with 5 μl of 88% H3PO4. Cellular debris was removed by centrifugation (1 h, 20 000 g, 4°C) and supernatants containing total cellular nucleotide extracts were stored at −80°C. Samples were analysed on a Prostar HPLC system (Varian) equipped with a nucleosil-4000 PEI column (Macherey-Nagel) by anion exchange chromatography according to Ochi (Ochi, 1986). To identify ppGpp in elution profiles representative samples were spiked with the authentic compound purchased from Trilink (http://www.trilinkbiotech.com/).

Purification and activity tests of YdeH

YdeH was expressed at 37°C from a pET28 vector with a C-terminal 6×His-tag (without intervening linker amino acids) in Rosetta cells (Novagen). Ni-affinity chromatography was carried out according to standard protocols (Novagen) with the help of FPLC equipment. Elution occurred at 300 mM imidazol. YdeH containing fractions were pooled and chromatographed over a Sephadex S75 column. Oligomerization of purified and concentrated YdeH was determined on a Sephadex S200 column in a buffer containing 20 mM Tris pH 7.6 and 150 mM NaCl with the help of an online refractometer (Optilab rEX, Wyatt technology). To test for DGC activity 2 μM protein was incubated with 100 μM GTP in 100 μl at RT for 30 min and the sample was analysed by LC/MS. Detailed protocols are available upon request.

Scanning electron microscopy

Cells were grown essentially as for attachment assays in 2 ml LB in 24-well plates in the presence of a sterile glass slide. After growth, glass slides were removed, rinsed gently with 1×PBS and fixed in 2.5% glutaraldehyde in 1×PBS for 1 h at RT. Glutaraldehyde was washed out with 1×PBS and the sample was dehydrated with an acetone step gradient (30%, 50%, 70%, 90%, 100%; 10 min each). Samples were critical point-dried and sputter-coated with a 3–5 nm Pt layer. Micrographs were recorded on a Hitachi S-4800 field emission scanning electron microscope. Acceleration voltage was generally between 1.5 and 5 kV.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We gratefully acknowledge sharing of unpublished material by K. Datsenko and B. Wanner and thank M. Düggelin, D. Mathys and M. Dürrenberger from the ZMB for excellent service, R. Landmann for the gift of anti-poly-GlcNAc antiserum, M. Folcher for assistance with HPLC and R. Hallez for strains and discussion. This work was supported by Swiss National Science Foundation Fellowship 3100A0-108186 to U.J. and a grant from the Werner Siemens-Foundation (Zug) to S.S.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
MMI_6739_sm_Tables_S1-S3_and_Figures_S1-S12.pdf5205KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.