The nucleoid-associated protein Fis directly modulates the synthesis of cellulose, an essential component of pellicle–biofilms in the phytopathogenic bacterium Dickeya dadantii

Authors

  • Claire Prigent-Combaret,

    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. CNRS UMR5557 Ecologie Microbienne, Villeurbanne, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Ouafa Zghidi-Abouzid,

    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. INSA-Lyon F-69621 Villeurbanne, France
    4. CNRS UMR5240 Microbiologie, Adaptation et Pathogénie, France.
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Géraldine Effantin,

    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. INSA-Lyon F-69621 Villeurbanne, France
    4. CNRS UMR5240 Microbiologie, Adaptation et Pathogénie, France.
    Search for more papers by this author
  • Philippe Lejeune,

    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. INSA-Lyon F-69621 Villeurbanne, France
    4. CNRS UMR5240 Microbiologie, Adaptation et Pathogénie, France.
    Search for more papers by this author
  • Sylvie Reverchon,

    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. INSA-Lyon F-69621 Villeurbanne, France
    4. CNRS UMR5240 Microbiologie, Adaptation et Pathogénie, France.
    Search for more papers by this author
  • William Nasser

    Corresponding author
    1. Université de Lyon, F-69622, Lyon, France
    2. Université Lyon 1, Villeurbanne, France
    3. INSA-Lyon F-69621 Villeurbanne, France
    4. CNRS UMR5240 Microbiologie, Adaptation et Pathogénie, France.
    Search for more papers by this author

Summary

Bacteria use biofilm structures to colonize surfaces and to survive in hostile conditions, and numerous bacteria produce cellulose as a biofilm matrix polymer. Hence, expression of the bcs operon, responsible for cellulose biosynthesis, must be finely regulated in order to allow bacteria to adopt the proper surface-associated behaviours. Here we show that in the phytopathogenic bacterium, Dickeya dadantii, production of cellulose is required for pellicle–biofilm formation and resistance to chlorine treatments. Expression of the bcs operon is growth phase-regulated and is stimulated in biofilms. Furthermore, we unexpectedly found that the nucleoid-associated protein and global regulator of virulence functions, Fis, directly represses bcs operon expression by interacting with an operator that is absent from the bcs operon of animal pathogenic bacteria and the plant pathogenic bacterium Pectobacterium. Moreover, production of cellulose enhances plant surface colonization by D. dadantii. Overall, these data suggest that cellulose production and biofilm formation may be important factors for surface colonization by D. dadantii and its subsequent survival in hostile environments. This report also presents a new example of how bacteria can modulate the action of a global regulator to co-ordinate basic metabolism, virulence and modifications of lifestyle.

Introduction

Biofilms are part of the range of structures used by bacteria to protect themselves against environmental challenges (Gerstel and Romling, 2001; Gualdi et al., 2008). Bacterial biofilms are defined as sessile communities characterized by cells that are attached to a substratum, to an interface, or to each other (Costerton et al., 1978). Large amounts of extracellular matrix material are often produced during biofilm formation. This matrix holds the cells in association with each other and with the surface, and it commonly contains exopolysaccharides (EPS), proteins, DNA, surfactants, lipids, glycolipids, and ions such as Ca2+ (Romling, 2002; Karatan and Watnick, 2009; Flemming and Wingender, 2010; Conover et al., 2012). The most extensively studied components of the biofilm matrices are EPS, followed by proteins and proteinaceous components, such as fimbriae and pili, and environmental DNA (eDNA). EPSs are a major component of most biofilm matrices and a commonly found EPS is cellulose, a linear polymer of (1-4)-β-linked glucose (Karatan and Watnick, 2009). It has been identified as the matrix component in biofilms produced by various plant-associated and environmentally isolated Pseudomonas strains (Ude et al., 2006). Furthermore, cellulose has been shown to play an important role in plant surface colonization by the plant interacting bacteria, Agrobacterium tumefaciens (Matthysse, 1983; Matthysse et al., 2005) and Rhizobium ssp. (Smit et al., 1992).

Recent analyses suggest the existence of two distinct types of cellulose synthesis gene clusters encoded by bacteria (Jahn et al., 2011) (Fig. 1). The reference for the first type (type A) is that of Gluconacetobacter xylinus, which is used for industrial cellulose production (Endler et al., 2010). In this bacterium, one operon, bcsABCD, is involved in the production of the cellulose polymer. BcsA–BcsB, responsible for the polymerization of glycan chains, constitutes a single modular protein with two domains. The two remaining subunits, BcsC and BcsD, mediate glucan chain extrusion and crystallization during cellulose assembly (Saxena et al., 1994). The reference for the second type (type B) of cellulose biosynthesis gene clusters is that of Escherichia coli or Salmonella enterica (Jahn et al., 2011). In this case, BcsA and BcsB are distinct proteins and an additional operon, bcsEFG, is required for cellulose production. Recently, BcsQ (YhjQ), which is encoded upstream of bcsA, was shown to be required for cellulose synthesis in E. coli (Le Quere and Ghigo, 2009). Moreover, two regulatory genes, adrA and csgD, control the cellulose synthesis (Romling, 2002). CsgD activates both the production of cellulose and curli. Its action on cellulose synthesis is mediated by AdrA, which activates cellulose production at the post-transcriptional level.

Figure 1.

Organization of the group A and group B cellulose synthesis gene clusters. The organization of the clusters is derived from the results of Jahn et al. (2011). The group A cluster (G. xylinus ATCC 23769, D. dadantii) is indicated with white arrows and group B (P. atrosepticum, E. coli and S. enterica) with grey arrows; the D. dadantii cluster is indicated with both grey (bcsQ and celY = bcsZ) and white arrows (bcsA, bcsB, bcsC and bcsD). The lengths of the bcsA genes and of the bcsQ-bcsA intergenic regions are indicated. Note that the bcsQ-bcsA intergenic region is longer (108 bp) in D. dadantii than those belonging to the three bacteria of the group B cluster. Numbers above the arrows indicate the amino acid identity of selected Bcs proteins compared with S. enterica; numbers under the arrows indicate the amino acid identity of these proteins compared with D. dadantii 3937. P. atrosepticum bcsA and E. coli bcsQ genes have nonsense mutations near the 5′ end, so amino acid identity was not calculated.

Dickeya dadantii is described as a necrotrophic Gram-negative bacterium that causes disease in a wide range of plant species (Perombelon, 2002; Lebeau et al., 2008). Soft rot, the visible symptom, is mainly due to the production of secreted degradative enzymes, mostly pectate lyases (Pels), proteases and the cellulase CelZ (or CelV), that can destroy the plant cell wall (Hugouvieux-Cotte-Pattat et al., 1996). The D. dadantii 3937 genome (Glasner et al., 2011) contains an active cellulose biosynthesis cluster (Yap et al., 2005; Jahn et al., 2011) which displays specific features. While at the level of the primary sequence of proteins the D. dadantii cellulose synthesis cluster is closer to that of G. xylinus, the genetic organization of the D. dadantii cluster is intermediate between type A and type B, with distinct bcsA and bcsB genes and a bcsQ gene (characteristics of type B) located upstream of the bcsABCD operon (organization characteristics of type A) (Fig. 1). The D. dadantii genome also contains the adrA gene, but its impact on cellulose production is less important than that described in E. coli (Yap et al., 2005; Jahn et al., 2011). To date, the involvement of cellulose in the D. dadantii life cycle has not been explored. Furthermore, no information about the expression of the bcs operon has yet been reported. This is particularly relevant since the main regulator (CsgD) involved in the modulation of cellulose production in animal pathogen enterobacteria is absent in D. dadantii.

Bacteria are traditionally thought to use transcriptional regulation to adapt to changing environmental conditions, such as the presence of a new carbon or energy source, nutrient starvation, a change in temperature or pH, or introduction to a host environment (Sepulchre et al., 2007; Hommais et al., 2008; Kepseu et al., 2010; Reverchon et al., 2010). It is now clearly evident that expression of virulence and adaptation genes is co-ordinated with global cellular functions (Kelly et al., 2004). Abundant Nucleoid-Associated Proteins (NAPs), which are thought to influence chromatin structure, have been shown to also influence transcription. Among these abundant chromatin proteins, Fis has been identified as a master regulator of cellular growth phase-dependent and/or virulence gene expression in several pathogenic enterobacteria (Dorman and Deighan, 2003; Kelly et al., 2004; Lautier and Nasser, 2007; Lenz and Bassler, 2007; Saldana et al., 2009). Although Fis binds DNA non-specifically at very high concentrations, many of its functions require that it binds and bends specific DNA targets (Shao et al., 2008). The level of Fis in the cell is subject to complex and multifactorial control. Fis is synthesized in large amounts during the early exponential phase of cells grown in rich medium (Ball et al., 1992; Nasser et al., 2001; 2002), decreases soon after, and becomes nearly undetectable as cells enter the stationary phase. Fis is a transcriptional activator of the genes and operons associated with primary metabolism, such as those encoding biosynthetic enzymes and stable RNAs (Ross et al., 1990; Gonzalez-Gil et al., 1996), and it was first shown to be involved in modulating virulence in various pathogenic bacteria of animals (Goldberg et al., 2001; Sheikh et al., 2001; Kelly et al., 2004; Lenz and Bassler, 2007). Recently it was suggested that Fis might be involved in the colonization of plant roots by Pseudomonas putida, probably by promoting biofilm formation in this bacterium (Jakovleva et al., 2012). However, the Fis target genes involved in this phenotype have not yet been identified in P. putida. We have previously shown that Fis regulates virulence in D. dadantii (formerly Erwinia chrysanthemi) and that the fis mutant is impaired in plant infections (Lautier et al., 2007; Lautier and Nasser, 2007; Zghidi-Abouzid et al., 2012). In this bacterium, Fis was shown to be essential both for an optimal production of virulence factors and for the systems involved in the neutralization of antibacterial compounds produced by host defence mechanisms (Lautier and Nasser, 2007). In this paper, we reveal that synthesis of cellulose increases the resistance of D. dadantii to chlorine treatments and enhances its ability to colonize plant surfaces. We also show that expression of the bcs operon is induced in biofilms and is directly regulated by Fis.

Results

The aggregative phenotype of the D. dadantii fis mutant is related to the functional bcsABCD operon

While characterizing the D. dadantii fis mutant, we noticed that this strain displayed a strong aggregation capacity in liquid media. Since this phenotype was strongly increased in the double celZ-fis mutant, inactivated in the production of both Fis and the main cellulase (CelZ or CelV) (Lautier and Nasser, 2007), we suspected that, in D. dadantii, there was a link between Fis and the synthesis of cellulose, an important constituent of biofilms in enterobacteria. The D. dadantii bcsA mutant has no specific phenotype in liquid growth conditions. By contrast, the double mutant fis-bcsA loses the aggregation phenotype previously described for the fis mutant (Lautier and Nasser, 2007) (Fig. 2A). This result suggests that the aggregation phenotype of the fis mutant is directly related to the bcs operon. To determine whether the aggregation phenotype of the fis mutant could be correlated with pellicle/biofilm structure formation, standing cultures in 24-well plates were initiated. After a period of 36–48 h of growth, pellicle–biofilm structures were observed both for the wild-type and for the fis strains, albeit with a thicker biofilm for the fis mutant. On the other hand, no pellicle–biofilm was observed for the bcsA strain or for the double mutant bcsA-fis (Figs 2B and S1). Moreover, the quantities of biofilm obtained with strains celZ and fis-celZ were higher than those obtained with the parental strain and the fis mutant respectively (Fig. S2). These results, therefore, suggest that there is a correlation between the amount of cellulose produced and the degree of aggregation of D. dadantii in liquid medium. We were unable to consistently complement the bcsA mutation. So we used two different mutants (A5105, A5112, Table S1 and Fig. 2 legend) carrying each a uidA-Kan insertion in an opposite orientation. These two mutants have the same phenotype. Furthermore, we have quantified by quantitative reverse transcription polymerase chain reaction (qRT-PCR) the expression of two genes bordering the bcsA-bcsB-bcsC-bcsD-celY operon (bcsQ in the upstream region and dctA in the downstream region) in the parental strain and in the derivatives bcsA, fis and fis-bcsA. These experiments showed that the transcriptional bcsA::uidA-Kan fusion (A5105) does not affect the expression of genes bordering the bcs operon (data not shown). To assess the role of cellulose, we examined pellicle–biofilm formation by the fis mutant in the presence of purified cellulase from Aspergillus niger. Under this condition, the fis mutant displayed a pattern similar to that of the double fis-bcsA (Fig. 2C). In addition, the treatment of a formed biofilm with pure cellulase totally disrupted the biofilm (data not shown). These results clearly establish that the high production of pellicle–biofilm in the fis mutant is correlated with increased synthesis of cellulose in this strain, compared with that of the parental strain. Since previous studies have shown that D. dadantii pellicle formation also depends on the type III secretion system, we analysed the impact of a hrcC mutation on pellicle–biofilm formation. The HrcC protein, an essential constituent of the type III secretion system, did not appear to be required for biofim formation in our experimental conditions (Fig. S2), which are different from those used in previous works (Yap et al., 2005; Jahn et al., 2011). Similarly in S. enteritidis, biofilm structure was seen to differ depending on the growth media used (Solano et al., 2002). In our experimental conditions (M63 minimal medium), the biofilm structures are practically undetectable when cellulose production is impaired (i.e. in bcsA and bcsA-fis mutants). Thus, without cellulose no pellicle and surface-attachment occurred. The term ‘pellicle–biofilm’ is used in this work to designate both a floating pellicle growing at the air/liquid interface and the liquid/solid surface of biofilm structures. Attempts to further characterize the pellicle–biofilm structures by using ‘flow cell experiments’ were unsuccessful, probably because D. dadantii pellicle–biofilms cannot adhere strongly to abiotic surfaces.

Figure 2.

Behaviour of the D. dadantii bcsA mutants. A. Aggregation phenotype of the D. dadantii wild-type (WT) strain (3937) and its derivatives, fis and fis-bcsA; strains were grown in minimal M63 medium supplemented with glycerol. B. Quantification of the cells present in the pellicle–biofilm and in the planktonic fractions for the different strains; each value represents the mean of three experiments and bars indicate the standard deviation. The asterisk indicates a P value of < 0.05 (Student's t-test). C. Cellulase digestion of bacterial pellicle–biofilm. Cellulase was added to the growth medium at the beginning of the experiment, just prior to inoculation of the bacteria. These images are representative of the observations made from three independent experiments. D. Difference in resistance to the bactericidal activity of chlorine between WT, fis and cellulose-deficient mutants. The surviving bacteria were enumerated by viable plate counts and their numbers were compared with those of control bacteria that had not been incubated with NaOCl. This number was defined as 100% survival. Each value represents the mean of three experiments; bars indicate the standard deviation. Similar results were obtained with two different bcsA mutants, A5105 and A5112 (Table S1). The asterisk indicates a P value of < 0.05 (Student's t-test). The white arrowheads indicate pellicle–biofilm structures in the fis mutant not treated with cellulase on panel C.

The D. dadantii pellicle–biofilm mediates chlorine survival

To determine whether cellulose production and biofilm formation might be responsible for chlorine resistance, as described in E. coli and S. enteritidis (Solano et al., 2002), we carried out survival experiments on the wild-type strain and on its derivatives bcsA, fis and fis-bcsA (see Experimental procedures). After a 15 min exposure period to NaOCl 5 p.p.m. (parts per million), 70–85% of the wild-type and fis cells survived. By contrast, only 10% and 25% of the bcsA and bcsA-fis mutants survived under these conditions respectively (Fig. 2D). These results clearly show that cellulose is involved in D. dadantii resistance to chlorine exposure and it may contribute to D. dadantii survival in water supplies and other environments.

Overproduction of cellulose improves plant surface colonization by D. dadantii

The binding of bacteria to a surface is the first step in the formation of a bacterial biofilm. The impact of cellulose synthesis on D. dadantii binding to chicory leaves was examined using microscopy. For this purpose, a green fluorescence tag was used to mark the wild-type strain and its derivatives, bcsA, celZ, fis and bcsA-fis (i.e. an egfp harbouring plasmid was introduced into the strains, see Experimental procedures). After an incubation period of 24 h, colonization of chicory leaves was observed for the wild-type as well as for the fis and celZ strains, although it was more extensive in the fis and celZ mutants (Fig. S3A). Furthermore, in the fis and celZ mutants a thick biofilm (typically around 28 µm) was often observed on the leaves (data not shown). In the bcsA strain and in the double bcsA-fis mutant, colonization of the chicory leaves was strongly reduced (Fig. S3B), indicating that cellulose production is required for an optimal development of biofilms on chicory leaves. Semi-quantitative analysis of the percentage area covered by green fluorescence cells was used to quantify the plant surface colonized by the five strains. This reveals that colonization of plant surfaces by the wild-type strain was at least 2.3- to 3.7-fold lower than colonization by the fis and celZ mutants (Fig. 3A). Furthermore, a 3.7- to sixfold decrease in chicory leaf colonization was observed for the bcsA and bcsA-fis mutants, compared with that of the wild-type strain. These results strongly suggest that cellulose production does promote plant surface colonization by D. dadantii. Overall, a good correlation was observed between the amount of cellulose produced and the degree of plant surface colonization by the different strains.

Figure 3.

Comparison of plant surface colonization and of virulence between various D. dadantii strains. A. Semi-quantitative analysis of the percentage of chicory leaf surface colonization by EGFP-labelled D. dadantii WT, bcsA, celZ, fis and fis-bcsA strains. For each strain, confocal images were acquired using identical parameter settings on the CLSM510 microscope to allow for a semi-quantitative comparison of fluorescence between treatments. The bioImage_L software 1.0 was used to quantify the percentage area covered by green fluorescence cells. For each treatment, means were calculated from five images for every four biological replicates (n = 4). Bars indicate the standard deviation and letters indicate significant differences between the treatments (P < 0.05, Fisher LSD). B and C correspond to the weight of macerated chicory leaves and potato tuber respectively; each value represents the mean of three experiments and bars indicate the standard deviation. D. Enumeration of the strains in the macerated tissues. Each value represents the mean of three independent experiments; ± standard deviations. Similar results were obtained with two different bcsA mutants, A5105 and A5112 (Table S1). cfu, count forming units.

In order to test whether a lack of cellulose production might confer a virulence defect on cellulose-minus mutants, we used chicory leaf and potato tuber assays to compare the virulence of the parental strain and its bcsA derivative on wounded organs. Under this condition, no significant difference was observed between these two strains (Fig. 3B and C). Furthermore, both the wild-type strain and the bcsA mutant displayed similar growth in planta (Fig. 3D). These data indicate that, at least under our experimental conditions, the production of cellulose is not involved in the virulence of D. dadantii.

Fis controls transcription of bcs operon

To determine the mechanism of action of Fis, we investigated the expression of bcsA through the entire growth period in a fis background. This was achieved using the chromosomal gene fusion bcsA::uidA. The fis mutation increases the expression of the bcsA gene throughout growth, with a stronger effect observed during the exponential growth phase (at least a fourfold increase at the beginning of the exponential growth phase versus a two- to 1.4-fold increase in stationary phase cells) (Fig. 4A). Furthermore, a slight growth-phase regulation of the bcs operon was observed in both wild-type and fis backgrounds. The expression of the bcs operon was subsequently analysed in planktonic and pellicle–biofilm phases using the qRT-PCR. These analyses showed that deletion of fis substantially increased the amount of bcsA mRNA in both planktonic and pellicle–biofilm phases. A stronger expression of the bcs operon was also observed in the pellicle–biofilm phase, compared with planktonic cells, in both wild-type and fis backgrounds (Fig 4B). From these observations, we infer that Fis represses the expression of the bcs operon but it does not seem to be involved in the specific induction of bcs operon transcription by the growth of D. dadantii in the biofilm phase.

Figure 4.

Expression of the bcs operon. A. Growth curves of the strains and time-dependent expression of the bcsA::uidA transcriptional fusion. Bacteria were grown, at 30°C, in liquid minimal M63 medium supplemented with glycerol. Each value represents the mean of four independent experiments and the bars indicate the standard deviation. β-Glucuronidase-specific activity is expressed as nmol of p-nitrophenol produced per min per mg of bacterial dry weight. B. Quantification of bcs operon expression in the pellicle–biofilm and in the planktonic fractions, for the different strains, using qRT-PCR analysis. Each value represents the mean of four independent experiments. Bars indicate the standard deviation. The asterisk indicates a P value of < 0.05 (Student's t-test).

Fis specifically interacts with the bcsA regulatory region

The regulation of bcs operon expression by Fis could be via a direct mechanism or by indirect control. To elucidate further, we performed interaction experiments between the bcsA promoter and purified Fis. Prior to these experiments, we needed to identify the promoter region of the bcs operon. A comparison of the organization of the bcs operon of D. dadantii with similar operons of E. coli, S. enterica and P. atrosepticum revealed that the bcsA gene of D. dadantii is shorter than those of the three other bacteria mentioned (2091 bp for D. dadantii versus 2619 bp for E. coli, 2625 bp for S. enterica and 2700 bp for Pectobacterium) (Fig. 1). Further comparisons between the BcsA protein from E. coli and D. dadantii revealed that the 146 N-terminal amino acids of the E. coli BcsA protein are missing in the D. dadantii protein (Fig. S4A). In addition, the intergenic region of bcsQ-bcsA in D. dadantii is longer (108 bp) than that of the three other bacteria belonging to group B (2 bp) (Fig. 1). This prompted us to look for the transcription initiation site upstream of the D. dadantii bcsA gene.

Primer extension analysis, with RNA extracted from D. dadantii 3937, revealed that bcsABCD operon transcription was initiated at the G nucleotide, at position −47 upstream of the ATG translation initiation codon of the bcsA gene (Fig. 5A). At a position six bases further upstream from the transcriptional start, there is a potential −10 element (TTCAGC) separated, by 17 bases, from a potential −35 element (TTGATA) (Fig. 5). Thus, the bcs transcription initiation site mapped here seems to be located in an appropriate place for transcription by RNA polymerase associated with σ70 (σ70RNA polymerase). The −10 element appears weak, compared with the canonical E. coli promoters, and this may account for the moderate expression of the bcsA gene.

Figure 5.

Determination of the D. dadantii bcsA transcription start site and sequence of the bcs operon promoter. A. Primer extension reactions were performed in the presence of 8 µg RNA extracted from the WT strain and the labelled primer bcsAPext2C (Table S1). DNA sequencing ladders were generated with the same primer (lane A, C, G and T); the arrow indicates the transcription initiation site +1. The sequence of the zone corresponding to the transcription initiation site is not easily readable and we have tried to solve this problem, without success, by using two different oligonucleotides (bcsAPext1C and bcsAPext2C; Table S1) and by varying the hybridization temperature. Finally, we read this sequence with the help of the published sequence of the bcsA gene (Glasner et al., 2011). Similar results were obtained in four independent experiments. B. The binding sites for Fis, indicated by a dashed line, have been arbitrarily numbered in ascending order from gene proximal to gene distal. The KMnO4-sensitive bases are indicated by closed circles. The probable −10/−35 and the putative Shine–Dalgarno sequence regions are underlined. The ATG start codon of the BcsA protein and the TAA stop codon for the BcsQ protein are indicated by open boxes. Fis consensus sequences (CNtYAaWWWtTRaNC; Shao et al., 2008) are indicated by grey boxes. The arrow indicates the transcription initiation site +1.

To further elucidate the bcsA promoter, we performed deletion analysis using the bcsA-uidA fusion construct in pNB4 (pNB4-pbcsA1, Table S1). Due to plasmid instability observed in D. dadantii, these experiments were performed in E. coli CSH50 (Miller, 1972). Indeed, it has been previously shown that most of the global regulatory systems, particularly that of Fis, are conserved between E. coli and D. dadantii (Lautier and Nasser, 2007). A deletion from +159 to position −22 (pbcsA2, Table S1), removing the −10 consensus sequence, led to a sevenfold decrease in the bcsA::uidA fusion expression (Fig. 6). An extended deletion to position −57 (pbcsA3, Table S1), removing both the −10 and −35 consensus sequences, led to a stronger decrease in β-glucuronidase production (10-fold). The residual β-glucuronidase activity observed with the promoter pbcsA3 may be due to a weak secondary promoter present in the upstream region of the bcsA promoter, not detectable in the primer extension experiments. A 5′ deletion from −353 to position −57 (pbcsA4, Table S1) had no significant impact on bcsA::uidA expression (Fig. 6). These results, therefore, provide additional evidence for the position of the bcsA promoter detected from primer extension experiments. When we analysed, in more detail, the deletion mutants in the E. coli CSH50 fis derivative (Koch et al., 1988) we observed a global increase, from two- to fourfold, in β-glucuronidase activity with the four bcsA promoters. This, therefore, indicates the presence of Fis binding site(s) in these four DNA segments. The fold change of Fis regulation, with the full length pbcsA1 regulatory region, is higher than that obtained with the deleted pbcsA4 region (3.4 versus 1.9), suggesting that the upstream Fis V, VI and VII, located in the bcsQ gene, participate in bcsA repression by Fis. Finally, given that the level of regulation by Fis on the promoter pbcsA1 (3.4-fold) is similar to that observed on the bcs operon in D. dadantii (at least fourfold in exponentially growing cells), it is reasonable to assume that the control of bcsA operon expression by Fis is exerted via the promoter pbcsA1.

Figure 6.

Deletion analysis of the bcsA-uidA promoter fusion. The extended promoter construct pbcsA1 (−353 to +159) and the deletion mutants are represented on the left of this figure. The arrow indicates the transcription initiation site +1. The position of the end of the bcsQ gene is indicated, together with those of the bcsA::uidA transcriptional fusion in the different constructs. Quantification of the bcsA::uidA promoter fusions in E. coli CSH50, and in its fis derivative, is shown on the right of this figure. The β-glucuronidase activities of the constructs were measured in cells from the mid-exponential (OD600 = 0.4) growth phase in LB medium. Each value represents the mean of four independent experiments. Bars indicate the standard deviation. For the different constructs, values from WT and fis strains are different, with P values of < 0.05 (Student's t-test). Expression from pbcsA2 and pbcsA3 are different from the expression from pbcsA1, with P values < 0.05. Expression from pbcsA4 is similar to expression from pbcsA1. β-Glucuronidase-specific activity is expressed as nmol of p-nitrophenol produced per min per mg of bacterial dry weight.

In vitro gel retardation assays were performed in the presence of purified Fis and the 510 bp pbcsA1 DNA fragment (−353 to +159, versus the +1 transcription initiation site) containing the regulatory regions of the bcs operon. Typical results are shown in Fig. 7. Fis is bound, with high affinity, to the DNA fragment tested since a complex is observed in the presence of 10 nM of Fis. With increasing Fis concentration, highly retarded complexes appeared at the expense of the lower complexes. This suggests the existence of several Fis binding sites in the bcsA operator. Up to a 25-fold molar excess of unlabelled specific probe was needed to completely prevent binding of Fis to the labelled probes, whereas no significant difference in Fis binding was observed in the presence of a similar amount of a non-specific probe (Fig. 7). Furthermore, in similar experiments, no binding of Fis was observed on the two negative control promoters (prtC and lacZ) previously identified in D. dadantii (Lautier and Nasser, 2007) (data not shown). These results suggest that Fis regulates the expression of bcsA by specifically binding to the promoter region.

Figure 7.

Representative result of Fis binding at the bcs operon promoter region. The concentrations of Fis used are indicated at the top. The positions of free DNA (F) and of the main Fis/DNA complexes (C) are indicated. The non-specific competitor corresponds to a 250 bp fragment derived from the coding region of a transcriptional regulator, MfhR. This non-specific fragment was amplified using the primers mfhR5′ and mfhR3′, described in Hommais et al. (2008). Similar results were obtained in five independent experiments.

The regulatory regions of the bcsA gene that interact with Fis were identified by DNase I footprinting analysis. For these experiments we retained the same fragment used for the gel retardation experiments, although its relatively large size (510 bp) could be a handicap for obtaining high-resolution footprinting pictures. However, this is relevant for the identification of the different Fis binding sites on the bcsA promoter region, as suggested by the gel retardation assays. At a Fis concentration of 10 nM, only a narrow area, extending from −3 to +33 (site III), was protected (Fig. 8). Increasing the Fis concentration up to 25 nM resulted in the appearance of an additional binding site (IV), extending from −68 to −6. A further gradual increase in the Fis concentrations (50–200 nM) resulted in the appearance of five additional binding sites, located between −347 and +159. Thus, Fis specifically interacts with seven sites in the bcsA operator (Figs 5 and 8). Apart from the protected regions, I and III, the Fis consensus sequence (CNtYAaWWWtTRaNC; Shao et al., 2008) were identified in the five other binding sites (Fig. 5). The Fis-protected regions I and III may contain several relatively degenerated Fis consensus that are not detectable by computer analysis. Indeed, it is known that Fis could also bind DNA with little or no obvious sequence preference (Stella et al., 2010). The areas protected by Fis binding, on pbcsA1, are consistent with the in vivo promoter studies data (Fig. 6).

Figure 8.

Representative result of DNase I footprinting of Fis on the bcs promoter region. Digestions were performed on both bottom and top strands. Lanes 1 and 7, no protein; lanes 2 and 8, 10 nM Fis; lanes 3 and 9, 25 nM Fis; lanes 4 and 10, 50 nM Fis; lanes 5 and 11, 100 nM Fis; lanes 6 and 12, 200 nM Fis. The areas protected by Fis are indicated on the left. The stars indicate hypersensitivities induced by the binding of Fis. Only the regions containing Fis protected areas were presented on both DNA strands. No binding sites were detected between −121 bp and −68 bp. DNA sequencing ladders, generated from plasmid RRbcsA3.2 in the presence of primer bcsAPext2C, were used to localize the Fis binding sites. Similar results were obtained in four independent experiments.

Fis acts at the transcription initiation step

We investigated further the Fis regulatory effects on the bcsA promoter. The effect of Fis on RNAP activity was initially investigated, using potassium permanganate (KMnO4) footprinting, on supercoiled plasmids containing the bcsA regulatory region (pNB4-pbcsA1). KMnO4 preferentially targets the pyrimidine residues in the untwisted regions of DNA and, thus, allows the extent of the promoter opening to be measured. Following the addition of RNAP, we observed that two bases located around the transcription initiation position (+6 and +7) of the bcsA gene are sensitive to KMnO4 (Fig. 9A). The addition of increasing concentrations of Fis substantially decreased KMnO4 reactivity at the bcsA promoter. From these data, we can infer that binding of Fis in the bcs operon regulatory region inhibits open complex formation by RNAP and provides evidence of a direct negative action by Fis.

Figure 9.

Fis prevents transcription initiation at the bcs promoters. A. The KMnO4 and in vitro transcription experiments were performed on supercoiled templates, and the reactions were carried out in the presence of 100 nM σ70RNAP with the following concentrations of Fis: lanes 2 and 7, no Fis protein; lanes 3 and 8, 10 nM Fis; lanes 4 and 9, 25 nM Fis; lanes 5 and 10, 50 nM Fis; lanes 6 and 11, 100 nM Fis. Lane 1 corresponds to the KMnO4 reaction without protein. Similar results were obtained in four independent experiments. B. Quantitative analysis of the expression of the bcs promoter. The data were normalized to the value obtained for the bla promoter (an internal control, see A) and are expressed as a percentage (100% bcs relative expression corresponds to the expression obtained in the absence of Fis). Each value represents the mean of four independent experiments. Bars indicate the standard deviation. DNA sequencing ladders, generated from plasmid RRbcsA3.2 in the presence of primer bcsAPext2C, were used to localize the position of the bases +1, +6 and +7.

We next used in vitro transcription to follow precisely the effect of Fis on RNAP activity. For this purpose, we monitored bcs operon transcription using pNB4-pbcsA1 DNA, with RNAP and Fis added either alone or in combination. As expected, the addition of increasing Fis concentrations decreased the accumulation of transcripts from the bcs operon promoter (Fig. 9A and B). Under the same conditions, the transcription of the reference bla promoter, located on the same plasmid, was not noticeably affected (Fig. 9A). These results demonstrate that Fis specifically inhibits the bcs promoter. This is, to our knowledge, the first report in which Fis is shown to play a direct role in cellulose production.

Discussion

It is generally thought that, in most natural environments, growth as a biofilm is the prevailing microbial lifestyle. The ability to form aggregates is an important colonization strategy for bacteria in a wide range of different environments. Cellulose is one of the main EPSs involved in the formation of the matrix of biofilms (Romling, 2002; Karatan and Watnick, 2009). D. dadantii is a plant pathogenic bacterium that causes disease in a wide range of plant species and there are many stages in its life cycle where biofilms could play a role, including survival of this pathogen in water and on plant waste, and during the colonization of plant surfaces. In this report, we have shown that cellulose is involved in surface colonization by D. dadantii and in the resistance of this bacterium to a chlorine agent. Hence, cellulose production seems to constitute a favourable phenotype for the microbial life of D. dadantii.

As previously described in G. xylinum (Wong et al., 1990), the D. dadantii bcs operon is driven by one unique promoter. Its expression is submitted to growth phase-regulation and induced by growth of the bacterium in a pellicle–biofilm structure.

Gel shift and DNase I footpring assays demonstrated that purified Fis specifically binds to the regulatory region of the bcs operon via seven binding sites, two of them (sites III and IV) fully overlapping the σ70RNA polymerase binding sites and the transcription initiation site of the operon (Figs 5 and 8). Finally, KMnO4 and in vitro transcription experiments revealed that Fis directly represses bcsA transcription initiation by σ70RNAP, even at a low concentration (Fig. 9). Overall these data provide a good correlation between the pattern of control on the bcs operon and the growth-phase cellular content of Fis: a strong repression during the exponential growth phase, when the Fis concentration is relatively high, followed by a decreased effect in the stationary growth phase correlated with a reduction in the cellular content of Fis. In the context of natural environments, it is likely that once bacteria face nutrient limitation, the reduction of cell growth and, thus, the decrease in the Fis cellular concentration might be perceived by bacteria as a signal for the establishment of a pellicle–biofilm structure. Furthermore, the transcription of the bcs operon, by the housekeeping σ70 factor, suggests that cellulose production may be linked to basic metabolism.

A phylogenetic analysis, performed on BcsA proteins, shows that the cluster of BcsA proteins from Dickeya/Erwinia species is closer to that of G. xylinus (type A) than to the E. coli/S. enterica type B-BcsA proteins (Fig. S4). Regulation of the expression of the first type cellulose synthesis cluster is poorly understood, whereas extensive studies have been carried out on E. coli and S. enterica concerning the transcription of the second type cluster. In these enterobacteria, CsgD is a master regulator of the production of cellulose and curli. The csgD expression is itself modulated by a myriad of transcriptional regulators, including OmpR/EnvZ (Ishihama, 2010) and CpxA/CpxR (Jubelin et al., 2005), as well as by the global regulators H-NS, IHF, FIS and RpoS (Barnhart and Chapman, 2006). Thus, in these animal pathogenic bacteria, Fis regulates cellulose production by an indirect mechanism through CsgD and AdrA. The D. dadantii genome does not contain the csg gene cluster. Furthermore, we have shown in this work that the intergenic region of bcsQ-bcsA in D. dadantii is longer (108 bp) than that of E. coli/S. enterica (2 bp). This intergenic region contains the D. dadantii bcs operon transcription initiation site and several Fis binding sites. So far, no direct regulation of the bcs operon by Fis has been reported in any bacteria other than D. dadantii. It is, therefore, reasonable to propose that the regulation of cellulose synthesis by Fis, described in this current study, might be the result of an evolutionary adaptation in order to establish a link between the synthesis of this EPS and the basic physiology and virulence functions of the cell. In other words, this study highlights the evolutionary remodelling of the regulatory circuits that link the global regulator, Fis, to the bcs genes in certain animal pathogenic bacteria (E. coli, S. enterica) and in the phytopathogenic bacterium D. dadantii. This functional adaptability of Fis to the bacterial genetic background may be widespread in Enterobacteriaceae since it was recently reported that regulation of DNA supercoiling by Fis, in response to environmental challenges, is different in the two taxonomically related bacteria E. coli and S. enterica (Cameron et al., 2011). By integrating the results described here with the data accumulated over recent years in D. dadantii (Lautier et al., 2007; Lautier and Nasser, 2007), it is tempting to speculate, as previously described in animal pathogenic bacteria (Kelly et al., 2004; Saldana et al., 2009), that Fis regulates critical genetic determinants of D. dadantii that are required for basic metabolism, for adherence to plant organ surfaces, and for the colonization/invasion of the host tissues. It remains to be determined whether pellicle–biofilm production could contribute to the persistence and survival of D. dadantii in soil and within environmental surface habitats.

Experimental procedures

Bacterial strains, plasmids, culture conditions and DNA manipulation techniques

The bacterial strains, phages and plasmids used are described in Table S1. D. dadantii and E. coli were grown at 30°C and 37°C respectively. The rich medium used was Luria broth (LB) and we also used M63 minimal salts medium (Miller, 1972), supplemented with a carbon source [polygalacturonate (PGA) at 0.4% (w v−1) and sucrose or glycerol at 0.2% (w v−1)]. M63 minimal medium was employed for all the experiments concerning pellicle–biofilms because the majority of our previous studies, particularly those concerning regulation of gene expression, were carried out in this growth medium. When required, antibiotics were used as follows: ampicillin (Ap), 100 µg ml−1; kanamycin (Km), 50 µg ml−1, chloramphenicol (Cm), 25 µg ml−1 and gentamicin (Gm), 100 µg ml−1. Liquid cultures were grown in a shaking incubator (220 r.p.m.). Media were solidified by the addition of 1.5% agar (w v−1). DNA manipulations were performed using standard methods (Sambrook and Russell, 2001).

Genetic techniques

To construct the bcsA mutant, the corresponding gene was specifically amplified using the primers bcsAleft and bcsAright (Table S1). The resulting 2017 bp PCR fragment was cloned into the pGEMT plasmid using the TA cloning kit from Promega. Inactivation of the bcsA gene was carried out by ligation of a uidA-KmR cassette (Bardonnet and Blanco, 1992) at the unique BamHI site, located in bcsA. Insertion of a uidA-KmR cassette into a gene, in the correct orientation, generates a transcriptional fusion. In this construction the uidA reporter gene, encoding β-glucuronidase, is expressed under the bcsA promoter. The insertion was introduced into the D. dadantii chromosome by marker exchange recombination between the chromosomal allele and the plasmid-borne mutated allele. The recombinants were selected after successive cultures in low phosphate medium, in the presence of kanamycin, conditions in which pBR322 derivatives are very unstable (Roeder and Collmer, 1985). Correct recombination was confirmed by PCR. Transduction of the mutation from one strain to another was performed using phage phiEC2 (Resibois et al., 1984).

To construct the bcsA-uidA promoter fusion, the full bcsA promoter (pbcsA1, −353 to + 159 relative to the the transcription initiation +1 of bcsA) was amplified by PCR using primers bcsA1 and bcsA2 (Table S1), so that the resulting fragment contained EcoRI and HindIII sites at the 5′ and 3′ end respectively. This 510 bp EcoRI-HindIII bcsA PCR product was inserted into the pNB4 plasmid containing the uidA promoterless gene (Bardonnet and Blanco, 1992), digested by the same enzymes to generate pNB4-pbcsA1 (Table S1). Various derivative fragments, carrying different deletions, were amplified by PCR and cloned in EcoRI-HindIII sites of pNB4 (Fig. 6; Table S1). In these constructs, the uidA reporter gene is expressed under the bcsA operators.

Plate and enzyme assays

The detection of protease activity was performed on medium containing Skim Milk (12.5 g l−1) and detection of cellulase activity was performed using the Congo red procedure (Teather and Wood, 1982). Detection of pectinase activity was performed on medium containing PGA, using the copper acetate procedure described by Reverchon et al. (1994).

The assay of pectate lyase was performed on toluenized cell extracts. Pectate lyase activity was determined by the degradation of PGA to unsaturated products that absorb at 235 nm (Moran et al., 1968). Specific activity is expressed as µmol of unsaturated products liberated min−1 mg−1 (dry weight) bacteria. Bacterial concentration was estimated by measuring turbidity at 600 nm, given that an optical density at 600 nm (A600) of 1.0 corresponds to 109 bacteria ml−1 and to 0.47 mg of bacteria (dry weight) ml−1. β-Glucuronidase activity was measured by monitoring the degradation of p-nitrophenyl-β-D-glucuronide into p-nitrophenol, which absorbs at 405 nm (Bardonnet and Blanco, 1992).

RNA isolation, primer extension and qRT-PCR analysis

Total RNA was extracted from D. dadantii by the frozen-phenol method (Maes and Messens, 1992). Primer extension experiments were essentially performed as described previously (Rouanet et al., 2004) using the RevertAid™ first strand cDNA synthesis kit (Fermentas) suitable for GC-rich RNA templates. The primer bcsAPext2C (Table S1), used for the specific detection of bcsA mRNA, was 5′ end-labelled and annealed to bcsA mRNA molecules at positions +142 to +158.

For RT-PCR analysis, cDNA was synthesized, using random hexamers and Fermentas reverse DNA polymerase, and qPCR was performed using the LightCyclerR faststart DNA masterplus SYBR Green I kit from Roche (Roche Applied Science), as previously described (Lautier et al., 2007). Expression of bcsA was quantified using the oligonucleotides bcsAqPCRf and bcsAqPCRr (Table S1) and target gene expression was defined using the method described by Pfaffl (2001). The statistical program used to analyse the data was the Relative Expression Software Tool (REST) (Pfaffl et al., 2002). The lpxC and hemF genes were selected as the reference genes for real-time RT-PCR (Hommais et al., 2011) to provide an accurate normalization.

Biofilm methods

For a visual assessment of biofilm formation, 2 ml of minimal M63 medium, supplemented with a carbon source, in 24-well culture plates (Nunc) was inoculated with an overnight culture grown in the same medium to a final density of 106 cells ml−1. The plate was incubated at 30°C for 36–48 h. The planktonic cells were collected and the pellicle/biofilm structures were washed once with 2 ml of M63 minimal medium. The pellicle/biofilm structures were then dissolved in 1 ml of M63 medium by pipetting up and down. Next, the free-living and sessile fractions were vortexed for 20 s and the bacteria were quantified spectrophotometrically, as described by Dorel et al. (1999). The role of cellulose in the biofilm formation process was analysed as described by Solano et al. (2002). Briefly, the biofilm-forming assay was performed in the presence of 0.1% (w v−1) of cellulase from Aspegillus niger (Sigma), and biofilms already formed in M63-glycerol were digested with 0.1% (w v−1) cellulase in 0.05 M citrate buffer, pH 4.6, for 18–24 h at 45°C.

Chlorine lethal assay

Chlorine survival analysis was performed as described by Solano et al. (2002) with minor modifications. D. dadantii and its derivatives were assayed for biofilm formation in M63-glycerol. Then, the supernatant of the wild-type strain and of the fis mutants was discarded, leaving just the biofilm. For the bcsA and bcsA-fis mutants, the sedimented cells at the bottom of the wells were used. M63 medium (2 ml), containing 5 p.p.m. sodium hypochlorite (chlorine 10–15%; Sigma-Aldrich), was added to all the wells and incubated for 15 min at 30°C. Control wells were incubated with 2 ml M63 medium. Surviving bacteria were enumerated by viable plate counts of the bacterial suspension.

Witloof chicory leaf surface colonization assays

A green fluorescence tag was used to mark the different D. dadantii strains by introducing the pMP2444 (pLac-egfp) plasmid (Bloemberg et al., 2000). Witloof chicory leaves were inoculated, without wounding, by depositing 5 × 106 bacteria in 5 µl of 50 mM KH2PO4 pH 7 buffer. Four leaves were inoculated with each strain and then incubated, at 28°C, in a dew chamber at 100% relative humidity. Microscopic observations were performed, 24 h post-inoculation, by mounting thin (< 2 mm thick) leaf surface sections in AquaPoly/Mount (Polysciences, Eppelheim, Germany). Before mounting the leaf samples between coverslip and slide, their surface was washed twice with water to eliminate unattached cells. Samples were examined immediately using a confocal laser scanning microscope (CLSM) (510 Meta microscope; Carl Zeiss S.A.S.) equipped with an argon–krypton laser, a detector and filter sets for green fluorescence (i.e. 488 nm for excitation and LP 505 for detection). Transmitted light (in white) was also used. The green fluorescence and transmitted light images were overlaid, to form a single image, using LSM software release (Carl Zeiss S.A.S). For each strain, at least one section from four independently inoculated leaves was observed. The whole experiment was repeated twice. Semi-quantitative analysis of chicory leaf colonization by the wild-type strain 3937, and by its mutant derivatives, was performed on confocal green images using the bioImage_L software 1.0. All images were acquired using identical parameter settings on the CLSM510 microscope (including the same detector amplification gain) to allow for a semi-quantitative comparison of the percentage area covered by green fluorescent cells. For each treatment, means were calculated from five images for every four biological replicates (n = 4).

Virulence assays

Pathogenicity assays were performed, as described in Lautier and Nasser (2007), with 106 or 2.5 × 106 bacteria in 5 µl of 50 mM KH2PO4 pH 7 buffer. Assays were carried out at least in triplicate. Negative controls were performed using sterile buffer.

Statistical analysis

All statistics were performed using a Student's t-test, except for plant surface colonization for which the Fischer LSD test was used. Tests were determined to be significant at a P value of < 0.05.

Proteins

Dickeya dadantii Fis protein was purified as described by Lautier and Nasser (2007). E. coliσ70RNA polymerase was obtained from GE Healthcare and the protein molarity was determined based on the concentration of the batches (mg ml−1).

In vitro DNA/protein interaction

Band-shift assay and DNase I footprinting were performed as previously described (Nasser et al., 1997). The bcsA regulatory region was specifically amplified using the primers BcsAOpDL and bcsAPext2C. The resulting 510 bp PCR fragment was cloned into the pGEM-T plasmid, using the TA cloning kit from Promega, to generate plasmid RRbcsA3.2. For in vitro DNA/protein interaction experiments, the regulatory region of the bcsA DNA fragments was recovered from plasmid RRbcsA3.2 by digestions with HindIII-PstI or with SalI-SacII. The DNA fragments obtained were further end-labelled by filling up the HindIII and SalI extremities in the presence of (α-32P)dCTP (3000 Ci mmol−1, GE Healthcare) and the Klenow fragment of DNA polymerase. The labelled DNA fragments were purified, after electrophoresis, on agarose gels using the Qiagen gel extraction kit. About 1 nM [at least 50 000 counts per min (c.p.m.)] and 2 nM (at least 100 000 c.p.m.) of labelled probe were used for band-shift assays and DNase I footprinting experiments respectively.

Potassium permanganate reactivity assay and in vitro transcription

The reactions for the potassium permanganate (KMnO4) reactivity assay and in vitro transcription experiments were performed with supercoiled templates containing the bcsA regulatory region (pNB4-RRbcsA3.2), as previously described (Castang et al., 2006; Lautier et al., 2007). The reaction products were solubilized in water, divided into equal parts and then submitted to primer extension, with radioactively end-labelled primers uidAdeb for bcsA mRNA and bla3B4 for the bla transcript (Table S1). The extension, with primers uidAdeb and bla3B4, yields 183 bp and 100 bp fragments respectively. The amplification products were analysed on a 6% sequencing gel. The signals obtained were detected by autoradiography on Amersham MP film and quantified using ImageMaster TotalLab version 2.01 software (GE Healthcare).

Acknowledgements

We are grateful to Valerie James for the English corrections and to our colleagues G. Condemine and F. Hommais for their support and advice. We are grateful to J. Wawrzyniak, P. Grangette and A. Vaught for their technical support. This work was supported by grants from the Centre National de la Recherche Scientifique (CNRS), from the FR 41 (Université Lyon 1), and from the French ‘ANR blanc Régupath 2007 Program, N°ANR-07-BLAN-0212’. In particular, O.Z. has a post-doctoral contract supported by this funding. The funders were not involved in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. This work made use of the technical platform ‘Centre Technologique des Microstructures’ at FR 41 (Université Lyon 1).

Ancillary