Unphosphorylated CsgD controls biofilm formation in Salmonella enterica serovar Typhimurium


  • Katherina Zakikhany,

    1. Department of Microbiology, Tumor and Cell Biology (MTC), Karolinska Institutet, FE 280, 17177 Stockholm, Sweden
    Search for more papers by this author
  • Carl R. Harrington,

    1. Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 7UA, UK
    Search for more papers by this author
  • Manfred Nimtz,

    1. Department of Structural Biology, Helmholtz-Centre for Infection Biology, Braunschweig, Germany
    Search for more papers by this author
  • Jay C. D. Hinton,

    1. Institute of Food Research, Norwich Research Park, Colney, Norwich NR4 7UA, UK
    2. Department of Microbiology, Moyne Institute of Preventive Medicine, School of Genetics and Microbiology, Trinity College, Dublin 2, Ireland
    Search for more papers by this author
  • Ute Römling

    Corresponding author
    1. Department of Microbiology, Tumor and Cell Biology (MTC), Karolinska Institutet, FE 280, 17177 Stockholm, Sweden
      E-mail ute.romling@ki.se; Tel. (+46) 8524 87319; Fax (+46) 8 300744.
    Search for more papers by this author

E-mail ute.romling@ki.se; Tel. (+46) 8524 87319; Fax (+46) 8 300744.


The transcriptional regulator CsgD of Salmonella enterica serovar Typhimurium (S. Typhimurium) is a major regulator of biofilm formation required for the expression of csgBA, which encodes curli fimbriae, and adrA, coding for a diguanylate cyclase. CsgD is a response regulator with an N-terminal receiver domain with a conserved aspartate (D59) as a putative target site for phosphorylation and a C-terminal LuxR-like helix–turn–helix DNA binding motif, but the mechanisms of target gene activation remained unclear. To study the DNA-binding properties of CsgD we used electrophoretic mobility shift assays and DNase I footprint analysis to show that unphosphorylated CsgD-His6 binds specifically to the csgBA and adrA promoter regions. In vitro transcription analysis revealed that CsgD-His6 is crucial for the expression of csgBA and adrA. CsgD-His6 is phosphorylated by acetyl phosphate in vitro, which decreases its DNA-binding properties. The functional impact of D59 in vivo was demonstrated as S. Typhimurium strains expressing modified CsgD protein (D59E and D59N) were dramatically reduced in biofilm formation due to decreased protein stability and DNA-binding properties in the case of D59E. In summary, our findings suggest that the response regulator CsgD functions in its unphosphorylated form under the conditions of biofilm formation investigated in this study.


Biofilm formation involving co-ordinated multicellular behaviour is a fascinating and complex phenomena spread widely amongst the bacterial kingdom. In Salmonella enterica serovar Typhimurium (S. Typhimurium), a form of multicellular/biofilm behaviour has been characterized by the expression of proteinaceous (curli fimbriae, BapA) and polysaccharide (e.g. cellulose) extracellular matrix components (Collinson et al., 1996; Zogaj et al., 2001; Solano et al., 2002; Latasa et al., 2005; Romling, 2005; Gibson et al., 2006). This multicellular behaviour can be visualized by the addition of Congo red (CR) to the agar medium, where cellulose and curli fimbriae interact with the dye to form a characteristic ‘rdar’ (red, dry and rough) morphotype. Components of the rdar morphotype are involved in biofilm formation on biotic and abiotic surfaces (Austin et al., 1998; Romling et al., 1998b; Boddicker et al., 2002; Solano et al., 2002; Annous et al., 2004; Hamilton et al., 2009). The stationary-phase expression of the rdar morphotype is regulated by the transcriptional regulator CsgD (previously called AgfD) a major regulator of biofilm formation (Romling et al., 2000; Gerstel and Romling, 2003). CsgD positively regulates the transcription of the csgBA operon leading to the biosynthesis of curli fimbriae (Hammar et al., 1995; Romling et al., 1998b). In addition, CsgD contributes indirectly to cellulose production by activating the transcription of adrA. AdrA activates cellulose biosynthesis on the post-transcriptional level through production of the allosteric activator bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) (Romling et al., 2000; Zogaj et al., 2001; Simm et al., 2004). Consequently, in a csgD deletion mutant cellulose and curli fimbriae are not expressed which leads to a ‘saw’ (smooth and white) morphotype on CR agar (Romling et al., 1998a).

A regulatory network that utilizes c-di-GMP as a secondary signalling molecule has been shown to be required for biofilm formation (Ross et al., 1990; Jenal and Malone, 2006; Romling and Amikam, 2006). The intracellular turnover of c-di-GMP involves the diguanylate cyclase and phosphodiesterase activity of GGDEF and EAL domain proteins, of which there are a total of twenty in S. Typhimurium. The expression of CsgD is regulated by at least six GGDEF/EAL domain proteins (Simm et al., 2004; Kader et al., 2006); however, the regulatory mechanisms of c-di-GMP signalling are not fully understood.

Furthermore, the expression of csgD is highly regulated at the transcriptional level. Environmental conditions such as temperature, oxygen tension, starvation, osmolarity, iron and pH, all regulate the expression of csgD (Gerstel and Romling, 2001). Maximal expression of csgD is observed in LB medium without salt under microaerophilic growth conditions in the liquid phase and on agar plates at temperatures below 28°C (Gerstel and Romling, 2001). However, a single point mutation in the promoter region can convert the highly regulated wild-type csgD promoter into a semi-constitutive promoter with a ≥ 3-fold increased level of csgD expression (S. Typhimurium MAE52) (Romling et al., 1998b).

The CsgD protein contains a LuxR-like helix–turn–helix (HTH) DNA-binding domain. Several members of the LuxR family, including CsgD, are response regulators (RR) of the FixJ/NarL family which act as transcriptional activators or repressors. The FixJ/NarL family is the second most abundant family of RR after the OmpR/PhoB family containing a winged HTH motif (reviewed in Gao et al., 2007). A common post-translational modification is phosphorylation at a conserved aspartate residue (D) in the N-terminal receiver domain performed by a cognate sensor histidine kinase (HK) or by autophosphorylation. A RR and its cognate HK form a ‘two-component system’, which is a widespread signal transduction system in bacteria (reviewed by Hoch, 2000; Stock et al., 2000; Mascher et al., 2006; Gao et al., 2007). Phosphorylation of the conserved aspartate in the receiver domain of the response regulator activates the protein by inducing conformational changes which facilitate interaction of the RR with the target DNA (Stock et al., 2000; Maris et al., 2002); examples are RRs such as FixJ of Rhizobium meliloti, NarL of Escherichia coli or DegU of Bacillus subtilis (Mukai et al., 1990; Walker and DeMoss, 1993; Birck et al., 1999). However, some RR are active in their unphosphorylated form (Msadek et al., 1990; Ma et al., 1998). Several studies have identified so-called ‘atypical response regulators’ (ARR), which are members of different families of RR, but all lack signature amino acids in the N-terminal receiver domain required for phosphorylation. Many of these ARR are orphan response regulators; some do not seem to accept phosphoryl groups, others are activated by binding of small molecules (Guthrie et al., 1998; O'Connor and Nodwell, 2005; Schar et al., 2005; Pflock et al., 2007; Ruiz et al., 2008; Wang et al., 2009).

In this study we investigated the properties of the transcriptional response regulator CsgD. First, we show that unphosphorylated CsgD binds to the promoter regions of its target genes csgBA and adrA and stimulates their transcription in vitro. Second, we show that CsgD is phosphorylated in the presence of acetyl phosphate (AcP) in vitro and that phosphorylated CsgD has reduced DNA binding ability. In vivo and in vitro studies indicate that the conserved aspartate D59 is important for the functionality of CsgD. Replacement of the conserved aspartate D59 by glutamate mimics phosphorylation, prevents binding to the target DNA and reduces stability of the CsgD protein in vivo. In summary, our findings suggest that biofilm formation in S. Typhimurium is regulated by unphosphorylated CsgD, and the phosphorylation of CsgD negatively affects the function of the protein.


CsgD binds directly to the csgBA and adrA promoter regions

In S. Typhimurium the expression of csgBA and adrA is positively regulated by the transcriptional regulator CsgD (Hammar et al., 1995; Romling et al., 2000; Gerstel and Romling, 2003). CsgD contains a HTH-DNA binding motif in its C-terminus; however, the direct interaction of CsgD with the promoter regions of its target genes has never been shown. In order to demonstrate the direct binding of CsgD to the csgBA and adrA promoter regions we performed electrophoretic mobility shift assays (EMSA) with purified CsgD-His6 tagged protein (Fig. S1). The mobility of DNA fragments covering the csgBA (−90/+89 relative to the transcriptional start site, fragment C0) or adrA (−222/+35, fragment A0) promoter regions was retarded in the presence of unphosphorylated CsgD-His6 (Fig. 1). While 100-fold excess of the unlabelled promoter fragments abolished CsgD-His6 binding, non-specific competitor DNA did not interfere with the binding of CsgD-His6. In addition, CsgD-His6 did not bind to a negative control (a 180 bp DNA fragment amplified from plasmid pBad30), which furthermore demonstrated the specificity of the binding of CsgD-His6 to the csgBA and adrA promoter regions (Fig. 1). Since CsgD is highly integrated into the c-di-GMP signalling network, we investigated whether c-di-GMP affects the binding of CsgD-His6 to the csgBA and adrA promoter region. However, no change of DNA binding by CsgD-His6 was detected after addition of 1 and 100 µM c-di-GMP (data not shown). C-di-GMP concentrations were chosen according to the range of physiological c-di-GMP levels in S. Typhimurium and other bacteria (Weinhouse et al., 1997; Kader et al., 2006; Waters et al., 2008; Simm et al., 2009).

Figure 1.

EMSA analysis of the csgBA and adrA promoter region. DIG-labelled DNA fragments were incubated with increasing amounts of purified CsgD-His6 (0–244 nM). Retarded protein–DNA complexes were detected with 28 nM CsgD-His6 (csgBA) and 56 nM CsgD-His6 (adrA). Control experiments were performed with: (I) no CsgD-His6, (II) 100× excess poly [dI-dC], (III) 10× excess BSA, (IV) negative control 180 bp DNA fragment amplified from plasmid pBAD30 and (V) 100× excess non-labelled specific competitor.

To narrow down the binding sites for CsgD, a series of DNA fragments covering different parts of the csgBA and adrA promoter regions were subjected to EMSA analysis (Fig. 2A). CsgBA fragments covering the regions downstream of the −35 region [fragments C1 (−28/+89) and C2 (−7/+89)] were not retarded in their mobility in the presence of CsgD-His6 (Fig. 2A). In contrast, fragment C3 (−90/−20) showed a clear shift (Fig. 2A).

Figure 2.

Determination of the CsgD binding sites in the csgBA and adrA promoter regions.
A and D. EMSA analysis of DNA fragments covering the csgBA (A) and adrA (D) promoter region.
B and E. DNase I protection assay of the csgBA promoter region [sense strand (B)] and the adrA promoter region [antisense strand (E)].
C and F. Location of the determined CsgD binding site within the csgBA (C) and adrA (F) promoter region. The CsgD-His6 binding site is indicated in bold, hypersensitive sites on both strands are indicated with asterisks and −10/−35 promoter regions are indicated in italics. The previously described CsgD recognition motif (Brombacher et al., 2003) is indicated with capital letters. The DNA fragments for EMSA analysis are indicated with black bars (C0: −90/+89; C1: −28/+89; C2: −7/+89; C3: −90/−20; wt: −90/+89) (A0: −222/+35; A1: −222/−123; A2: −75/−5; A3: −222/−60, wt: −222/+35). The position of the DNA fragments is given relative to the transcriptional start site (indicated in bold with arrow).

In the case of the adrA promoter region, no shift was detected for fragment A1 (−222/−123) (Fig. 2D). Fragment A2 (−78/−5) was slightly retarded in the presence of CsgD-His6 and increased amounts of CsgD-His6 led to a stronger retardation (Fig. 2D and data not shown). Fragment A3 (−222/−60) showed a retardation in the presence of CsgD-His6 (Fig. 2D). This suggested that the CsgD binding site(s) are located between −123 and −5 in the adrA promoter region.

Detection of the CsgD binding sites

To map the precise location of the CsgD binding sites for the csgBA and adrA promoter regions we performed DNase I footprint assays. For the csgBA promoter, a 179 bp 32P-labelled probe stretching from −90 to +89 was incubated with various amounts of purified CsgD-His6 protein prior to DNase I treatment. On the sense strand, the binding of CsgD-His6 resulted in a protection of nucleotides flanked by DNase I hypersensitive sites from positions −30 to −40 upstream of the transcriptional start site overlapping the −35 region of the csgBA promoter (Fig. 2B and C). No protection, and DNase I hypersensitive sites were observed on the antisense strand (Fig. 2B and C and data not shown). Previous work suggested a conserved 11 bp sequence (CGGGKGAGNKA) present in the csgBA and the yaiC (adrA) promoter region as the sequence recognized by CsgD in E. coli (Brombacher et al., 2003). A similar motif (TGGGTGAGTTA) is present in the csgBA promoter region of S. Typhimurium overlapping the −35 region. To confirm the importance of the binding sequence, its second GT motif, which was found protected by DNase I treatment upon binding of CsgD, was mutated to AA to create fragment C3-mut. Impaired binding of CsgD-His6 to C3-mut was observed, demonstrating the specificity of the binding and the importance of the 11 bp binding motif (Fig. 2A).

To investigate the binding of CsgD to the adrA promoter region, a 257 bp 32P-labelled fragment from bp −222 to +35 was subjected to DNase I footprint analysis (Fig. 2E). In contrast to the csgBA promoter region, a larger area on both strands was protected in the presence of CsgD-His6 (Fig. 2E and data not shown). The protected region was located upstream of the −10 promoter region between nucleotides −61/−57 on the sense strand and between nucleotides −88/−28 on the antisense strand. Several DNase I-hypersensitive sites were located on sense and antisense strand adjacent and within the protected area (Fig. 2E and F). On the sense strand, the 11 bp binding motif is present at position −57 to −68 and partially protected by CsgD. An almost perfect inverted repeat of the binding motif is located 8 bp downstream, and is protected by CsgD. However, also other DNA regions outside the CsgD binding motif are protected by CsgD. Indeed, comparison of the binding affinity of CsgD to fragment A2 and A3 (Fig. 2D) showed that CsgD bound strongest to the sequences upstream of −73 which do not contain the 11 bp binding motif. Accordingly, mutation of the AGT motif at position −78 to GTG prevented CsgD-His6 binding to A3-mut (Fig. 2D).

Novel genes presumably directly regulated by CsgD

To identify novel genes directly regulated by CsgD, a bioinformatics approach was combined with global transcriptional analysis. We first searched for the CsgD-binding consensus sequence upstream of open reading frames (ORFs) in the S. Typhimurium genome by ‘Virtual Footprint Analysis’ (see Experimental procedures) (Munch et al., 2005) (Fig. 3A). To relate the bioinformatics data to CsgD expression, we only analysed the promoter regions of genes that were differentially expressed between S. Typhimurium strains UMR1 (wt), MAE52 (≥ 3-fold increased csgD expression compared with UMR1) and MAE50 (ΔcsgD) (Fig. 3B and C). To this end, we identified five genes potentially directly CsgD-regulated (Fig. 3C). Interestingly, the gene bapA and genes of the capsule operons, which are regulated by CsgD in S. enterica serovar Enteritidis strains (Latasa et al., 2005; Gibson et al., 2006) were not found to be CsgD-regulated in S. Typhimurium under the experimental conditions.

Figure 3.

Identification of novel CsgD-regulated genes in S. Typhimurium.
A. Position of the putative CsgD binding motifs are indicated relative to the experimentally determined transcriptional start site (*) or relative to the ORF with in silico determined promoter sequences (**).
B. Expression of CsgD-dependent genes in the S. Typhimurium strains UMR1 (wt), MAE52 (≥ 3-fold increased csgD expression) and MAE50 (ΔcsgD). Transcriptomic data show that the three CsgD target genes csgB, csgA and adrA are upregulated in MAE52, and strongly downregulated MAE50 as compared to UMR1 (P ≤ 0.05 for all genes).
C. Expression of genes carrying a putative CsgD binding site in their promoter region in S. Typhimurium strains UMR1, MAE52 and MAE50. Transcriptomic data show that flgB, fliC and STM3670 are upregulated in MAE52; STM3670 and sipB are downregulated in MAE50 in comparison to UMR1. fimC is upregulated in MAE50 compared to UMR1 and MAE52 (P ≤ 0.05 for STM3670 and sipB). The expression of selected genes was confirmed using qRT-PCR analysis (Fig. S2).

CsgD is crucial for the transcription of csgBA and adrA in vitro

With csgBA and adrA being the main targets of CsgD, we hypothesized that CsgD may be necessary and sufficient to stimulate transcription from these promoters. To study in vitro transcription, we cloned the csgBA (−99/+132) and adrA (−104/+129) promoter fragments into the vector pJDC1 where the fragments are flanked on both sides by strong transcriptional terminators (Marschall et al., 1998). For the linearized csgBA promoter fragment, no RNA transcript was detected in the absence of CsgD-His6 using E. coli RNA polymerase saturated with either the house keeping sigma factor 70 (RpoD) or sigma factor S (RpoS) (Fig. 4A). Addition of CsgD-His6 led to the detection of strong run-off transcripts for RpoD and RpoS loaded RNA polymerase of approximately equal intensity. With the adrA promoter, a weak RNA transcript was detected with E. coli RNA polymerase saturated with RpoS, but transcription was significantly enhanced in the presence of CsgD-His6 (Fig. 4B). Almost no in vitro transcription activity could be detected with RNA polymerase saturated with RpoD (Fig. 4B), while transcription was stimulated by the presence of CsgD-His6. Similar results were achieved with supercoiled csgBA and adrA promoter fragments (data not shown).

Figure 4.

In vitro transcription assay with csgBA and adrA.
A. With the csgBA promoter, no RNA transcript was detected solely with the RNA polymerase holoenzyme with RpoS (EσS) (lane 1) or RpoD (Eσ70) (lane 4) in the absence of CsgD-His6. Addition of CsgD-His6 led to the detection of a run-off transcript for both EσS (lanes 2 and 3) and Eσ70 (lanes 5 and 6).
B. Detection of an adrA-specific run-off transcript in the absence of CsgD-His6 (lane 1) with EσS the intensity of which increased after addition of CsgD-His6 (lanes 2 and 3). Almost no adrA expression was detected when Eσ70 was incubated without CsgD-His6 (lane 4); however, addition of CsgD-His6 led to transcription activity and the detection of a RNA transcript (lanes 5 and 6).
No RNA transcript was detected with csgBA or adrA DNA, CsgD-His6 and the RNA polymerase core enzyme only (A and B, lane 7) (CsgD-His6 concentrations: + = 100 nM; ++ = 130 nM).

To investigate whether c-di-GMP plays a role in the transcriptional activation of csgBA and adrA by CsgD, we added c-di-GMP to the in vitro transcription assays. None of the concentrations tested (1–100 µM) influenced the transcription efficiency (data not shown).

Effects of AcP on CsgD activity

Typically, phosphorylation of the conserved aspartate in the N-terminal receiver domain enhances or is essential for DNA interaction of a RR (Birck et al., 1999; Zhang et al., 2003). As some response regulators have been shown to autophosphorylate in the presence of low-molecular-weight phosphodonors such as AcP in vitro (Lukat et al., 1992; Schroder et al., 1994; Cheng and Walker, 1998), we investigated whether the presence of AcP affects the DNA binding of CsgD-His6. Pre-incubation of CsgD-His6 with 10 mM AcP led to a decreased binding of CsgD to the adrA and the csgBA promoter fragments, compared with CsgD alone (Fig. 5A).

Figure 5.

CsgD-His6 shows altered DNA binding affinities in the presence acetyl phosphate (AcP) and is phosphorylated by AcP in vitro.
A. EMSA analysis with CsgD-His6 and the csgBA and adrA promoter fragments in the presence of 10 mM AcP. For both, the csgBA and adrA DNA fragments, the presence of AcP led to decreased CsgD-His6 binding in vitro.
B. CsgD-His6 and CsgD59N-His6 are phosphorylated in the presence of AcP in vitro, while CsgDD59E-His6 is not. CsgD-His6, CsgDD59N-His6, CsgDD59E-His6 (25 kDa) and OmpR (27 kDa, positive control) were incubated with [32P]-AcP. (1) Coomassie-stained SDS-PAGE of OmpR and CsgD-His6 wild-type and mutant proteins; (2) detection of phosphorylated OmpR and CsgD-His6 wild-type and mutant proteins.
C. ΔptaackA mutant strains in S. Typhimurium UMR1 and MAE52 background. AcP does not contribute to rdar morphotype expression. Cells were grown on CR agar plates for 24 h at 28°C.

The response regulator CsgD contains a highly conserved aspartate at position 59 (D59) in its N-terminal region, but lacks several other conserved amino acids required for phosphorylation (Fig. S3). Since the binding of CsgD to the csgBA and the adrA promoter regions is altered in the presence of the phosphodonor AcP, we investigated whether CsgD can be phosphorylated in vitro. As shown in Fig. 5B, CsgD-His6 was phosphorylated in the presence of [32P]-AcP along with the positive control OmpR (McCleary and Stock, 1994). No in vitro phosphorylation was detected for CsgD when the aspartate at position 59 was replaced by glutamate (D59E) (Fig. 5B).

Acetyl phosphate, a high-energy intermediate of the phosphotransacetylase-acetate kinase (Pta/AckA) pathway, has been proposed to be a global signal in vivo (McCleary et al., 1993). A transcriptomic study revealed that AcP acts as a global signal during biofilm development in E. coli (Wolfe et al., 2003) and as a high-energy phosphodonor for RR in vivo (Pruss, 1998). In Listeria monocytogenes, AcP levels have been shown to regulate motility and chemotaxis (Gueriri et al., 2008). Since CsgD is phosphorylated by AcP in vitro, we investigated whether AcP may act as a phosphodonor and alter the activity of CsgD in S. Typhimurium in vivo. We blocked AcP synthesis by creation of Δpta/ΔackA deletion mutants in S. Typhimurium strains UMR1 and MAE52 and screened the mutant strains for rdar morphotype expression. No change in rdar morphotype expression was detected for both mutant strains in comparison with the respective wild-type strains (Fig. 5C) suggesting that under our experimental conditions, AcP does not phosphorylate CsgD and consequently does not affect rdar morphotype expression in S. Typhimurium.

D59 of CsgD is important for functionality and stability

To determine the role of the conserved D59 in CsgD in S. Typhimurium, we created single amino acid substitutions in CsgD. Exchange of the conserved aspartate (D) to asparagine (D→N) or alanine (D→A) prevents phosphorylation (Dahl et al., 1991; Ma et al., 1998), while the substitution to glutamate (D→E) has been shown to mimic the phosphorylated state (Klose et al., 1993; Lan and Igo, 1998; Lauriano et al., 2004; Gueriri et al., 2008). We constructed mutant alleles of csgD encoding the amino acid substitutions D59N or D59E in the UMR1 and MAE52 background (see Experimental procedures). Compared with CsgD, expression of the rdar morphotype was dramatically reduced upon introduction of the alleles CsgDD59N and CsgDD59E in UMR1 and MAE52 (Fig. 6A). Introduction of the csgD allele encoding CsgDD59N partially retained rdar morphotype expression, while introduction of CsgDD59E abolished rdar morphotype expression almost completely. Similarly, expression of CsgD with the respective amino acid substitutions led to a lower expression of the rdar morphotype compared with wild-type CsgD when expressed from the plasmid pBAD30 in the csgD-negative strain MAE50 (data not shown). Consistent with the reduction of the rdar morphotype, with the exception of the CsgDD59N allele in the MAE52 background, the expression of the CsgD target gene csgBA was dramatically reduced in the mutant strains indicating a lower CsgD activity (Fig. 6B).

Figure 6.

D59 is important for CsgD functionality and stability.
A. Rdar morphotype expression of the CsgDD59N and CsgDD59ES. Typhimurium mutant strains and their related parental strains. Strains were incubated for 20 h (upper line) and 48 h (lower line) on CR agar plates at 28°C.
B. qRT-PCR analysis of csgBA expression in S. Typhimurium CsgDD59N and CsgDD59E mutant strains in relation to their parental strains. For determination of csgBA expression, RNA samples were prepared from bacterial cells after 20 h growth on LB without salt plates, incubated at 28°C. Relative csgD expression levels were determined in comparison with the parental strains UMR1 or MAE52. Shown are the mean values from three independent experiments ± standard deviation.
C. Western blot analysis of the CsgDD59N and CsgDD59E mutant strains with anti-CsgD antibody. Samples were harvested from LB without salt plates after 20 h and the total protein was separated on SDS-PAGE.
D. Protein stability assay. To determine protein stability upon induction with 0.1% l-arabinose, chloramphenicol (220 µg ml−1) was added. Samples were taken at the time points indicated, separated with SDS-PAGE and analysed by Western blot analysis.
E. EMSA analysis of the csgBA and adrA promoter regions with CsgDD59N-His6 and CsgDD59E-His6 in the presence or absence of 10 mM AcP. pc, positive control with wild-type CsgD.

To identify the cause of the reduced CsgD activity, we first determined the csgD transcription levels by qRT-PCR analysis. CsgD expression was similar in the strains containing the mutant alleles, compared with the respective wild-type strains (data not shown).

In contrast, significantly reduced levels of CsgD protein were observed in the CsgDD59E mutant strains (∼40% compared with wild-type expression) (Fig. 6C). The results suggested that a D59E exchange in CsgD could affect the overall protein stability of CsgD. This was confirmed by a CsgD stability test where CsgD was expressed under the control of the PBAD promoter (Fig. 6D). The D59N mutant strains, however, showed protein levels comparable to their respective wild-type strains (Fig. 6C and D).

Finally, we investigated whether the D59E and D59N substitutions affect the DNA-binding properties of CsgD by EMSA analysis of the csgBA and adrA promoters using purified CsgDD59E-His6 and CsgDD59N-His6 (see Experimental procedures). No binding was detected for CsgDD59E-His6 in the absence and presence of AcP (Fig. 6E) suggesting that reduced DNA binding and protein stability contributed to the phenotype of CsgDD59E. CsgDD59N-His6 bound with affinities comparable to wild-type CsgD to both the csgBA and adrA promoter fragments (compare Fig. 6E with Fig. 1). Interestingly, however, the addition of AcP significantly reduced the affinity of CsgDD59N-His6 to its designated DNA targets (Fig. 6E). In addition, CsgDD59N-His6 was phosphorylated in the presence of AcP despite the presence of the non-phosphorylatable asparagine at position 59 (Fig. 5B). We conclude that CsgDD59N contains an alternative phosphorylation site in addition to D59 or a newly emerged site of phosphorylation occurred upon the D59N exchange. This phosphorylation site might have significance in vivo and thereby contributes to the reduced rdar morphotype expression upon CsgDD59N expression.


In vivo studies had shown an essential role for CsgD in the transcriptional activation of the operon encoding the structural subunits of curli fimbriae csgBA, and the adrA gene, encoding a di-guanylate cyclase (Hammar et al., 1995; Romling et al., 1998a; 2000). In this study, we demonstrated for the first time that CsgD interacts directly and activates transcription of the promoter regions of csgBA and adrA. CsgD binds to the csgBA and adrA promoter regions, which both include a previously suggested binding motif (Brombacher et al., 2003). However, the mode of CsgD binding and the location of the binding sites relative to the promoter are different. At the csgBA promoter, the CsgD binding site is short and precisely overlaps the 11 bp CsgD binding motif located in the −35 promoter element. In the adrA promoter region, CsgD also binds to the 11 bp binding motif centred at position −65 on the sense strand and position −48 on the antisense strand. However, significant CsgD binding is also observed up- and downstream of these regions in the absence of a consensus binding motif. Indeed, CsgD bound with highest affinity to the region upstream of the −65 binding motif. Binding of CsgD to the adrA promoter indicates a more complex mechanism of target recognition by CsgD. Some RR have been demonstrated to bind to sequences having dyad symmetries (Zhu and Winans, 1999; Urbanowski et al., 2004), and the intensively characterized RR NarL has been shown to bind to a heptameric sequence which is found in various numbers and arrangements (e.g. repeat, inverted repeat) (Li et al., 1994; Darwin et al., 1996; 1997). Other studies focusing on selected RR have identified complex protein–DNA recognition mechanisms involving physical rearrangements such as DNA bending or distortion, rather than simple sequence recognition (Maris et al., 2002; Feng et al., 2004).

The different binding patterns and the position of binding sites of CsgD in the csgBA and adrA promoter regions imply different mechanisms of transcriptional activation. At the csgBA promoter region, the DNA binding site for CsgD overlaps the core-promoter −35 element and resembles a class II promoter. At other class II promoters, the transcriptional activator has more than one contact with the RNA polymerase and more than one function. The transcriptional activator does not only facilitate binding of the RNA polymerase to the promoter to yield the RNA polymerase–promoter closed complex, but also isomerizes the RNA polymerase–promoter closed complex to yield the RNA polymerase–promoter open complex (Lawson et al., 2004). It remains to be determined if CsgD recruits the RNA polymerase and whether CsgD is involved in post-recruitment processes at the csgBA promoter.

The binding specificity of the RNA polymerase is determined by formation of complexes with different sigma subunits (Gruber and Gross, 2003). Although the RNA polymerase associated with RpoD or RpoS essentially recognizes the same promoter sequences (Gruber and Gross, 2003), the alternative sigma factor RpoS is responsible for the expression of many genes transcribed in the stationary phase and/or under stress conditions (Ishihama, 2000). At the stationary-phase activated csgBA promoter, however, RNA polymerases loaded with RpoS or RpoD are able to bind to the csgBA promoter (Bougdour et al., 2004), but these holoenzymes cannot initiate transcription (Robbe-Saule et al., 2006). Although trans-acting transcription regulators can contribute to transcriptional selectivity (Gruber and Gross, 2003), we showed that CsgD activated transcription for the RNA polymerase associated with RpoS and RpoD with similar efficiency. This finding is consistent with in vivo data which showed that expression of the csgBA genes does not require RpoS (Romling et al., 1998b).

In agreement with in vivo findings (Romling et al., 2000), CsgD-activated transcription at the adrA promoter occurred with RNA polymerase loaded with RpoS and RpoD. However, low-level CsgD-independent transcriptional activity was detected for adrA with RNA polymerase associated with RpoS. This result indicates a more efficient recognition of the adrA promoter by the RNA polymerase loaded with RpoS in the absence of CsgD. Sequence features of the adrA promoter such as a C nucleotide immediately upstream and a TG motif located four nucleotides upstream of the −10 TATAAT hexamer might favour promoter recognition by RNA polymerase loaded with RpoS (Becker and Hengge-Aronis, 2001). Although this is an unusual location for a TG motif, RpoS seems to have a less stringent requirement in terms of position of the TG dinucleotide relative to the −10 box (Lacour and Landini, 2004). The uncoupling of expression of the adrA promoter from CsgD may have in vivo significance under certain environmental conditions.

CsgD expression is a major target of c-di-GMP signalling (Kader et al., 2006; Simm et al., 2007) and CsgD regulates the expression of the di-guanylate cyclase AdrA (Romling et al., 2000). However, we discovered that c-di-GMP, an integral part of the regulatory network promoting CsgD expression in S. Typhimurium, is not required for binding of CsgD to the csgBA and adrA promoters or for CsgD-activated transcription. In contrast, the CsgD homologue VpsT of Vibrio cholera binds c-di-GMP, and requires c-di-GMP for binding to the target promoter sequences (Krasteva et al., 2010). However, the authors also predicted that CsgD does not bind c-di-GMP, as the binding site for c-di-GMP is not conserved between VpsT and CsgD. Our experiments indicated that c-di-GMP indeed did not bind to CsgD, highlighting the rapid evolution of c-di-GMP binding sites in homologous proteins. At the moment, we also have no evidence of another small molecule that could potentially enhance the activity of CsgD as it has been postulated (Chirwa and Herrington, 2003).

The majority of RR are activated by phosphorylation at a conserved aspartate in the N-terminal receiver domain (Roy and Falkow, 1991; Holman et al., 1994; Liu and Hulett, 1997; Gusa et al., 2006). In contrast, our results suggest that phosphorylation of CsgD reduces the activity of CsgD, making it less stable in vivo. However, we cannot rule out the possibility that phosphorylated CsgD activates or represses the transcription of other genes. As an example, the RR DegU is active both in its phosphorylated and non-phosphorylated form; this is considered to be a ‘regulatory switch’ that activates expression of different sets of genes depending on environmental changes (reviewed in Murray et al., 2009).

It is not clear how CsgD is phosphorylated in vivo. Under the conditions investigated in this study AcP does not play a role in vivo, and other mechanisms, such as phosphorylation by cognate or non-cognate histidine kinases or other low-molecular-weight phosphodonors, may be relevant (Lukat et al., 1992; McCleary et al., 1993). We have evidence for an alternative phosphorylation site besides D59, as CsgDD59N was phosphorylated in the presence of AcP in vitro. Furthermore, the incubation of the CsgDD59N protein with AcP completely prevented the protein:DNA interaction. This phosphorylation could be relevant in vivo as expression of CsgDD59N caused decreased rdar morphotype expression even though the protein concentrations and DNA binding affinities were comparable with the wild-type protein. D59N replacement abolishes phosphorylation of many response regulators (Dahl et al., 1991; Ma et al., 1998), but for some RR, phosphorylation remained detectable (Bourret et al., 1990; Moore et al., 1993; Reyrat et al., 1994; Appleby and Bourret, 1999). In one study a D to N replacement still allowed phosphotransfer to an adjacent amino acid, but the activity of the protein was not necessarily maintained (Appleby and Bourret, 1999).

Few CsgD homologues have been functionally characterized, but all are involved in biofilm formation. In V. cholera, VpsT has been found to be required for the transcription of the Vibrio polysaccharide (vps) genes (Casper-Lindley and Yildiz, 2004). CsgD homologues might also act as repressors, because the CsgD homologue of Vibrio parahaemolyticus CpsS represses expression of the capsular polysaccharide genes (Guvener and McCarter, 2003). In E. coli, CsgD also represses genes that inhibit biofilm formation (Brombacher et al., 2006)

In this work, we have demonstrated for the first time that CsgD is directly required for gene transcription. However, the identification of in total eight genes potentially directly regulated by CsgD does not account for the 21 genes that showed altered expression in a csgD mutant as demonstrated by transcriptomic analysis (data not shown). In E. coli, CsgD also regulates genes indirectly as it enhances the expression of the alternative sigma factor RpoS through transcription activation of iraP encoding a RpoS stablilization factor (Gualdi et al., 2007). A similar scenario might occur in S. Typhimurium as RpoS expression is also enhanced upon CsgD protein levels (U. Römling et al., unpubl. result). Another mode of gene regulation by CsgD could be via c-di-GMP produced by the di-guanylate cyclase AdrA. Subsequent studies will dissect the direct and indirect regulation of gene expression mediated by CsgD. Such studies are now possible since this study describes the first purification of functional CsgD.

Experimental procedures

Bacterial strains, culture conditions and chemicals

Bacterial strains used in this study are listed in Table S1. S. Typhimurium and E. coli strains were routinely grown at 37°C or 28°C on either Luria–Bertrani (LB) agar plates (with or without sodium chloride) or in LB liquid medium (pH 7.0) supplemented with ampicillin (Amp, 100 µg ml−1) or chloramphenicol (Cm, 20 µg ml−1), when indicated. E. coli strain DH5α was routinely used for cloning experiments and E. coli BL21 (DE3) was used for protein overexpression. For phenotypic analysis, S. Typhimurium cells were cultivated at 28°C on LB agar plates without sodium chloride and supplemented with CR solution (CR: 40 µg ml−1, 20 µg ml−1 Coomassie Brilliant Blue in ethanol). l-arabinose (up to 0.1%) or isopropyl-β-d-thiogalactopyranoside (IPTG) (1 mM) was used to induce gene expression from appropriate vectors. All chemicals and oligonucleotides were obtained from Sigma Aldrich (Germany) unless stated otherwise. The source of c-di-GMP was described (Simm et al., 2004).

DNA manipulations and PCR analysis

Genomic DNA and plasmid preparations were performed using the GenElute™ Bacterial Genomic DNA Kit (Sigma Aldrich) and the GenElute™ Miniprep Kit (Sigma Aldrich) according to manufacturer's instructions. T4 DNA Ligase (Roche) and restriction enzymes (New England Biolabs) were used according to manufacture's instructions. DNA sequencing was performed using the StarSEQ (Germany) or GATC-Biotech (Germany) sequencing service. For standard PCR analysis, DNA TaqPolymerase (New England Biolabs) or Phusion High-Fidelity DNA Polymerase (Finnzymes) was used. For colony PCR single bacterial colonies were picked, resuspended in 15 µl of sterile water, boiled at 95°C for 5–10 min and an aliquot was directly used as template for PCR. Primers used in this study are listed in Table S2.

Construction of mutant strains

Construction of chromosomal deletions was performed using the one-step gene inactivation method (Datsenko and Wanner, 2000), and point mutations were introduced with the tetRA system (Karlinsey, 2007). All S. Typhimurium mutant strains were constructed in UMR1 (wt) and in the MAE52 background. Strain MAE52 carries a point mutation in the csgD promoter region leading to increased csgD expression levels (≥ 3-fold) (Romling et al., 1998b). Mutant alleles of csgD encoding the amino acid substitutions D59N or D59E in the UMR1 and MAE52 background were created through PCR-based site-directed mutagenesis (SDM primers, Table S2). The ΔptaackA mutant strain was created using the Datsenko and Wanner method. In brief, both genes were deleted in one approach by replacing their ORFs except 40 nucleotides in the beginning of ackA and the end of pta with a chloramphenicol resistance cassette (ackA primer, Table S2). Approximately 300 ng PCR-Product was electroporated into S. Typhimurium strains UMR1 and MAE52 containing pKD46. Recovered colonies were purified at least twice on LB medium containing the appropriate antibiotics. All mutants constructed were verified with PCR and sequencing. To excise the resistance cassette, selected clones were transformed with the temperature-sensitive helper-plasmid pCP20. Clones were purified at least twice on LB agar without antibiotics at 42°C. The tetRA system was used to create point mutations on the chromosome, creating scar-less chromosomal mutant strains. In the first step, the complete csgD ORF was replaced with a tetRA element which provides tetracycline resistance (TetR). TetR clones carrying a correct insertion were purified and transformed with DNA fragments encoding csgD containing the respective amino acid substitutions. TetS agar (Karlinsey, 2007) was used to screen for clones carrying the correct insertion and candidate clones were purified at least twice at 42°C. All mutants constructed were verified with PCR and sequencing and different clones of each mutant strains were created and tested. As a control, the wt-csgD fragment in plasmid pGSA1 was reintroduced and screened for rdar morphotype expression.

Expression and purification of CsgD

CsgD coding region was amplified from genomic DNA. Using PCR and appropriate primers, an XbaI site and a Shine-Dalgarno sequence were introduced to the 5′ end of the fragment. Furthermore, a SphI site and nucleotides encoding for six histidines (6× HIS) were introduced to the 3′ end. After digestion, the DNA fragment was cloned into the low-copy expression vector pPD100 (Dersch et al., 1994) to result in pZKA1. For CsgD-His6 expression, 600 ml of LB culture medium was inoculated (1:100) with an overnight culture of BL21 (pZKA1), incubated at 28°C with shaking until an OD600 of 0.4–0.5 was reached. Expression of CsgD-His6 was induced by adding 1 mM IPTG and the cultures incubated for 2 additional hours at 28°C. After harvesting, cells were resuspended in lyses buffer containing 1 mg ml−1 lysozyme and incubated for 30 min on ice. Cells were disrupted by sonication on ice (six bursts of 10 s each followed by 30 s on ice) and then a 30 min centrifugation (10 000 g, 4°C). The supernatant was incubated with 0.5 ml of Ni-NTA resin (Qiagen) and pre-incubated with gentle shaking for 1 h at 4°C prior to loading on the column. Two washing steps were performed (Wash buffer 1 and 2) and fractions were collected. CsgD-His6 was eluted with Elution buffer; purified CsgD-His6 protein was dialysed against storage buffer and stored at −20°C. The purity and nature of the protein was confirmed with SDS-PAGE and mass spectrometry. All steps of the protein purification were performed at 4°C in the cold room. Lyses buffer: 50 mM NaH2PO4, 300 mM NaCl, 10 mM imidazole, pH 8.0; Wash buffer 1: 50 mM NaH2PO4, 300 mM NaCl, 20 mM imidazole, pH 8.0; Wash buffer 2: 50 mM NaH2PO4, 300 mM NaCl, 80 mM imidazole, pH 8.0; Elution buffer: 50 mM NaH2PO4, 300 mM NaCl, 250 mM imidazole, pH 8.0; Storage buffer: 10 mM Tris-HCl, 10 mM MgCl2, 100 mM KCl, 1 mM DTT, 0.1 mM EDTA, 50% Glycerol.

SDS-PAGE, Western blot analysis and antibodies

For SDS-PAGE and Western blot analysis, 5 mg of cells (wet weight) were harvested from agar plates (LB agar without salt), directly resuspended in 2× Laemmli Buffer and heated at 95°C for 10 min. Total protein samples were separated on SDS-PAGE (12% separation, 4% stacking gel) and analysed by colloidal Coomassie staining or transferred to PVDF membranes (Millipore). Detection of CsgD was carried out as described (Romling et al., 2000) using a polyclonal anti-CsgD antibody and goat-anti-rabbit antibody (horseradish peroxidase-conjugated) (Jackson Immunoresearch). HIS-tags were detected using an anti-His antibody (Qiagen).

To determine protein stability, ΔcsgD strain MAE50 transformed with pGSA1, pGSA2 or pGSA3 was used. Cells were grown for 2 h and protein synthesis was induced with 0.1% arabinose. To stop protein synthesis, chloramphenicol (220 µg ml−1) was added (time point 0) and aliquots were taken after 30, 60 and 90 min. Samples were frozen immediately and kept at −20°C until usage.

Electrophoretic mobility shift assay (EMSA)

Gel shift analyses were performed using the DIG Gel Shift Kit (Roche) according to manufacturer's instructions. DNA fragments were 3′ end labelled with DIG-ddUTP and recombinant Terminal Transferase. Labelled DNA fragments and CsgD-His6 were incubated for 30 min at 30°C in a water bath to allow binding. On some occasions, c-di-GMP (1–100 µM) was added to the sample. For reactions containing AcP, 10 mM AcP, phosphorylation buffer and CsgD-His6 were incubated 30 min at 30°C prior to the binding reaction. Samples were analysed on a 6% native acrylamide gel in 0.5× TBE buffer in the cold room. After electrophoresis, samples were electro-blotted on nylon membranes (Hybond N+; Amersham Pharmacia) and fixed (UV-Crosslinker, Stratagene) according to manufacturer's instructions. Anti-DIG antibody was used at a 1:10 000 working concentration and luminescence was detected with the LAS-1000 system (Fujifilm).

RNA extraction

Extraction of total RNA was performed using the SV Total RNA Isolation System (Promega) with minor modifications. Prior to RNA extraction, bacterial cells were incubated in ice-cold 5% (v/v) phenol/95% (v/v) ethanol to stabilize the RNA (≥ 30 min, on ice). The samples were centrifuged at 4000 r.p.m. for 10 min and the supernatant discarded. Subsequently, the pellet was resuspended in 100 µl of lyses buffer (50 mg ml−1 lysozyme in TE buffer) and incubated for 5 min at room temperature. The following steps of RNA purification were performed according to the manufacturer's instructions which include an on-column DNase I treatment. Quality of RNA samples was assessed via gel electrophoresis and RNA concentrations were determined using the Nanodrop System (THERMO Scientific). All RNA samples were subjected to PCR analysis to ensure no DNA remained. DNA-positive samples were treated with RQ1 RNase-Free DNase (Promega) according to manufacturer's protocol. RNA samples were stored at −70°C.

cDNA synthesis and quantitative real-time PCR

Expression of target genes was determined by two-step real-time RT-PCR using the Power SYBR Green PCR Master Mix (Applied Biosystems) and the 7500 Real-Time PCR System (Applied Biosystems). First-strand cDNA synthesis from total RNA was performed with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) following the manufacturer's instructions. Relative transcript abundance was determined with the 2−ΔΔCt method (Livak and Schmittgen, 2001; Schmittgen and Livak, 2008) using the 7500 SDS Software v1.3.1 (Applied Biosystems). The recA gene was used for internal normalization. All experiments were performed as quadruplicates per real-time PCR assay relative to a calibrator value (UMR1). Values are the mean values from three independent experiments ± standard deviation.

DNase I footprint analysis

To generate 32P-labelled fragments, primers were labelled with T4 polynucleotide-kinase (NEB) and [γ-32P]-ATP (6000 Ci mmol−1, Perkin Elmer). 32P 5′ end-labelled DNA fragments [adrA (AO: −222/+35) and csgBA (CO: −90/+89)] were generated with PCR using a combination of one labelled and one unlabelled primer respectively. Fragments were purified with Qiaquick PCR purification columns (Qiagen). Protein–DNA binding reactions were carried out in a 15 µl reaction volume containing 10 mM Tris-HCl (pH 7.5), 10 mM MgCl2, 1 mM DTT, 50 mM KCl, 0.5 µg of poly [dI-dC], 0.5 mg ml−1 BSA and 2 ng of labelled DNA fragment. CsgD was added in the amounts indicated and the reaction was incubated for 30 min at 30°C. DNase I treatment was performed by adding 2 µl of 0.002 U µl−1 DNase I (in 10 mM MgCl2) (Sigma Aldrich) for 1 min at room temperature. The reaction was stopped by adding 10 µl of stop buffer [200 µl of sodium acetate pH 5.2, 75 mM EDTA, 1 µg µl−1 sonicated salmon sperm DNA (Stratagene)] followed by phenol/chloroform extraction. After ethanol precipitation and drying, samples were resuspended in 3 µl of loading buffer [80% v/v formamide, bromophenolblue (BBP), xylene cyanol (XC)], boiled for 5 min at 80°C and loaded onto a 7% acrylamide/7M Urea sequencing gel. To evaluate the protection pattern, a sequencing ladder (Sequenase Cycle Sequencing Kit, Amersham Biosciences) generated with the same primers was loaded on the gel. After electrophoresis (1400 V, 90 min), footprinting patterns were visualized using a phosphoimager (STORM 820 Amersham Biosciences).

In vitro transcription assay (run-off assay)

In vitro transcription assay was performed by using linearized pJDCadrA and pJDCcsgBA (Table S2). In vitro transcription reactions were carried out in a 10 µl reaction volume containing 10 mM Tris-HCl, 10 mM MgCl2, 1 mM DTT, 50 mM KCl, 0.5 mg ml−1 BSA, 200 µM each of ATP, GTP and CTP, 10 µM UTP (GE Healthcare), 2.5 µCi of [α-32P]-UTP (3000 Ci mmol−1, Perkin Elmer), 10 U of RNase Inhibitor (NEB) and 1 nM linearized plasmid DNA. The mixture was incubated for 5 min at 30°C (CsgD 0–130 nM) to allow protein binding. Reconstitution of active E. coli RNA polymerase holoenzyme was achieved by incubating the core enzyme (Epicentre) with either σS (1:2.5) or σ70 (1:1) for 30 min at 37°C. Transcription was initiated by adding 0.2 U of RNA polymerase holoenzyme or core enzyme and the reaction was further incubated for 20 min at 28°C. The reaction was stopped by adding 5 µl of formamide buffer (80% formamide, 0.1% BPB, 0.1% XC) and loaded directly on a pre-warmed 6% polyacrylamide/7 M urea sequencing gel. Following electrophoresis (1400 V; 90 min.), the signal was detected on a phosphoimager screen (STORM 820, Amersham Biosciences).

Phosphorylation assay of CsgD

[32P]-AcP was synthesized with E. coli acetate kinase (Sigma Aldrich) and [γ-32P]-ATP (6000 Ci mmol−1, Perkin Elmer) (Cheng and Walker, 1998). The reaction mixture contained 25 mM Tris-HCl (pH 7.4), 60 mM NaAc, 10 mM MgCl2, 300 µCi of γATP and 0.3 U of acetate kinase (in 100 mM triethanolamine buffer, pH 7.6). The reaction was incubated at 25°C for 30 min and the AcP was separated from the enzyme using a Microcon-10 microconcentrator (Millipore). Two micrograms of CsgD-His6, CsgDD59E-His6, CsgDD59N-His6 or OmpR protein, respectively, were incubated with [32P]-AcP for 30 min at 30°C. The reaction was stopped by addition of equal volumes of 2× SDS loading buffer. Samples were subjected to SDS-PAGE and analysed with a phosphoimager (STORM 820 Amersham Biosciences). Proteins were visualized with colloidal Coomassie. Size determination was with the Novex®Sharp protein ladder (Invitrogen). To further verify protein phosphorylation, Coomassie-stained protein bands were cut from the gel and re-exposed (> 24 h). All phosphorylation experiments were repeated at least three times.

Transcriptomic analysis

For transcriptomic analysis S. Typhimurium strains UMR1 (wt), MAE50 (ΔcsgD) and MAE52 (≥3-fold increased csgD expression) were used. S. Typhimurium strains were incubated for 20 h on LB agar without sodium chloride at 28°C in the incubator. After 20 h cells were scraped off the plates and immediately frozen in liquid nitrogen. Extraction of total RNA is described under RNA extraction. Microarray hybridization and scanning were performed according to protocols described on the Institute of Food Research (IFR), Norwich, microarray website http://www.ifr.ac.uk/safety/microarrays/#protocols (Clements et al., 2002; Ygberg et al., 2006). Briefly, RNA samples from three biological replicates were labelled with Cy5-dCTP and hybridized to the IFR SALSA microarrays. Cy3-dCTP-labelled S. Typhimurium genomic DNA was used as a common reference in an indirect comparison type experimental design (Yang and Speed, 2002). Microarrays were scanned and fluorescence intensities quantified using GenePix Pro™ software, version 6.0 (Axon Instruments). Microarray features showing a reference signal lower than background plus 2 standard deviations were discarded (see http://www.ifr.ac.uk/safety/microarrays/#analysis). GeneSpring™ 7.2 (Silicon Genetics) microarray analysis software was used for data visualization, analysis and statistical analysis of differentially expressed genes [parametric Student's t-test, Benjamini and Hochberg false discovery rate (FDR)]. In general, transcriptomic data were filtered to only include equal to or greater than twofold differences; however, changes of less than twofold can also be biologically significant (Hughes et al., 2000; Ichikawa et al., 2000). The expression of selected genes was confirmed using qRT-PCR analysis (Fig. S2).

In silico analysis

Virtual footprints were performed using the ‘Virtual Footprints 3.0’ software (http://www.prodoric.de/vfp/) (Munch et al., 2005). The ‘Regulon prediction’ was applied using S. Typhimurium LT2 (complete chromosome) as genome. As search pattern, the reported CsgD binding site in E. coli was used [CGGGKGAGNKA (IUPAC-code)], as well as the protected regions identified in this study (for both the adrA and csgBA promoter regions). Sequence alignments were performed using the ‘MULTIALIGN’ interface (Corpet, 1988).


Alexandra Sasse contributed to the project during an internship. This work was supported by the Karolinska Institutet (‘Elitforskartjänst’ to U.R.) and Vetenskapsrådet (621-2004-3979). J.C.D.H. received a BBSRC Core Strategic Grant.