The carbon storage regulator CsrA is an RNA binding protein that controls carbon metabolism, biofilm formation and motility in various eubacteria. Nevertheless, in Escherichia coli only five target mRNAs have been shown to be directly regulated by CsrA at the post-transcriptional level. Here we identified two new direct targets for CsrA, ycdT and ydeH, both of which encode proteins with GGDEF domains. A csrA mutation caused mRNA levels of ycdT and ydeH to increase more than 10-fold. RNA mobility shift assays confirmed the direct and specific binding of CsrA to the mRNA leaders of ydeH and ycdT. Overexpression of ycdT and ydeH resulted in a more than 20-fold increase in the cellular concentration of the second messenger cyclic di-GMP (c-di-GMP), implying that both proteins possess diguanylate cyclase activity. Phenotypic characterization revealed that both proteins are involved in the regulation of motility in a c-di-GMP-dependent manner. CsrA was also found to regulate the expression of five additional GGDEF/EAL proteins and a csrA mutation led to modestly increased cellular levels of c-di-GMP. All together, these data demonstrate a global role for CsrA in the regulation of c-di-GMP metabolism by regulating the expression of GGDEF proteins at the post-transcriptional level.
Successful adaptation of bacteria to different niches depends on their ability to adjust their life style according to the requirements of the environment. Bacteria have evolved numerous mechanisms to sense external signals, to translate them into complex cellular responses and, thereby, to mediate responses to physiological demands. The Escherichia coli carbon storage system, with the RNA binding protein CsrA as the central player, exemplifies such an adaptive regulatory cascade (see reviews by Babitzke and Romeo, 2007; Lucchetti-Miganeh et al., 2008). The Csr regulatory system is widely distributed among eubacteria (White et al., 1996) and has been found to control a variety of virulence-linked physiological traits (Lucchetti-Miganeh et al., 2008).
Regulation of CsrA activity is mediated in part by the action of the two small non-coding RNAs (sRNAs) CsrB and CsrC (Romeo, 1998; Weilbacher et al., 2003). During the past years sRNAs have been recognized as important players in gene regulation, in most cases by base pairing with target mRNAs (Majdalani et al., 2005; Storz et al., 2005; Romby et al., 2006). However, CsrB and CsrC RNAs antagonize the activity of CsrA by binding to and therefore sequestering this protein (Liu et al., 1997). Transcription of the Csr sRNAs is controlled by the two-component system BarA-UvrY (Suzuki et al., 2002; Weilbacher et al., 2003), thus permitting the integration of environmental signals into the Csr signalling network. Expression of csrB and csrC also requires CsrA. This regulation may be mediated indirectly through the BarA-UvrY system (Gudapaty et al., 2001; Suzuki et al., 2002). This auto-regulatory mechanism has been described as a homeostatic system, which leads to tight regulation of CsrA activity. Recently, a new regulatory factor, CsrD (YhdA) has been shown to influence the Csr system (Jonas et al., 2006; Suzuki et al., 2006). By targeting CsrB and CsrC for degradation by RNase E, CsrD acts positively on CsrA activity (Suzuki et al., 2006). The apparent membrane protein CsrD contains degenerate GGDEF and EAL domains. Such domains have been shown to be associated with the turnover of the second messenger cyclic di-GMP (c-di-GMP) (Simm et al., 2004; Ryjenkov et al., 2005; Schmidt et al., 2005), which can mediate the switch between a motile and sessile life style in diverse bacteria (see reviews by D'Argenio and Miller, 2004; Jenal, 2004; Romling, 2005; Romling and Amikam, 2006). In contrast to other GGDEF/EAL proteins, CsrD was demonstrated to lack both diguanylate cyclase (DGC) and phosphodiesterase activities, indicating that CsrD is involved neither in the production nor in the degradation of the second messenger (Suzuki et al., 2006). In contrast, CsrD was found to be an RNA binding protein, although its detailed mechanism of action in CsrB/C decay has not been resolved.
Despite the detailed knowledge about the molecular mechanisms of the Csr signalling system, limited information is available concerning the integration of the Csr cascade into other global networks. In order to identify novel direct targets for CsrA that might help us to better understand the global impact of the Csr network, we conducted a genome-wide search for genes, whose transcript levels rapidly change upon pulse overproduction of CsrA. Our search revealed that CsrA is a regulator for several GGDEF/EAL proteins, in particular of the two GGDEF proteins YcdT and YdeH. Both proteins produce c-di-GMP in vivo and control flagella-mediated swimming motility.
Identification of novel mRNA targets for CsrA by microarray
To screen for novel direct CsrA targets we decided to adopt a microarray-based approach, which has previously been used to identify direct sRNA targets (Papenfort et al., 2006; Tjaden et al., 2006; Vogel and Wagner, 2007). Our strategy involved the pulse overexpression of csrA, followed by the immediate analysis of changes in whole-genome expression patterns. The approach is based on the assumption that CsrA not only blocks the translation of many of its mRNA targets, but also secondarily destabilizes them. Hence, pulse overexpression of CsrA from an inducible vector is expected to lead to a rapid decrease in the transcript level of the directly regulated targets. Changes in the transcript levels of indirect CsrA targets are assumed to occur first at later time points after csrA induction. Such differential changes in the transcript level can be monitored over time by microarray analysis. To verify that our approach was working, we first monitored csrA expression as well as the expression of the known direct target pgaA and the known indirect target csrB in response to csrA overexpression by quantitative real-time reverse transcriptase PCR (RT-PCR). Our data show that addition of arabinose (at 0 min) to E. coli KJ157 (KSB837 csrA::kan), carrying the arabinose-inducible vector pBADcsrA, resulted in a strong upregulation of csrA expression within 2 min (Fig. 1A). Consistent with our prediction mRNA levels of the direct target pgaA dramatically decreased within less than 10 min (Fig. 1B). The expression of the sRNA CsrB, known to be indirectly and positively controlled by CsrA (Gudapaty et al., 2001), began to increase after a delay of ∼12 min (Fig. 1C). These data suggest that our approach was successful in discriminating between direct and indirect targets for CsrA.
In the next step we screened for novel direct CsrA targets by using an Affymetrix whole-genome E. coli array. As pgaA mRNA was downregulated within less than 12 min, we compared the transcriptional profiles 4 and 12 min after arabinose addition with the profile before arabinose induction (0 min). To eliminate genes downregulated in a CsrA-independent manner, we normalized the observed signal ratios against the signal ratios resulting from induction of the vector control pBAD28.
Four of the genes showing the strongest repression (more than sevenfold) 12 min after pBADcsrA expression belonged to the pga operon (Fig. 1D). The mRNA levels of ycdT followed the same kinetics upon CsrA overexpression as the pga mRNAs, suggesting that ycdT may be regulated by CsrA in a similar manner. Database search revealed that ycdT is located directly adjacent to the pga operon but on the reverse strand. ycdT encodes a transmembrane protein with a C-terminal GGDEF domain (Fig. 1E and F). Among the most downregulated genes in response to CsrA overexpression we found another GGDEF protein encoding open reading frame, ydeH (Fig. 1D and G). ydeH is predicted to encode a cytoplasmic protein, not containing any known domains involved in signalling (Fig. 1G).
The two GGDEF proteins YcdT and YdeH are regulated by CsrA
To confirm the effect of CsrA overexpression on ycdT and ydeH transcripts, we determined the kinetics of CsrA-dependent downregulation by RT-PCR. In accordance with our array data, ycdT and ydeH mRNA levels decreased strongly upon arabinose addition in late exponential phase (OD600 1.5) (Fig. 2A). The mRNA level of ycdT was halved within 4 min and reached a minimum of 3% between 12 and 24 min. ydeH mRNA decreased to 50% within 5 min and continued to decrease to approximately 22% after 24 min. Similar results were observed when arabinose induction was performed earlier during growth at OD600 0.5 (Fig. 2B). In contrast, addition of arabinose to a strain carrying the empty vector pBAD28 did not affect the levels of ycdT and ydeH transcripts (Fig. 2C).
To test the effect of a csrA mutation on ycdT and ydeH expression, we measured the mRNA levels of ycdT and ydeH by RT-PCR along the entire growth curve in the wild-type and isogenic csrA mutant strains. In the wild-type strain, expression of ycdT slightly decreased within the first 8 h of growth (Fig. 2D), whereas ydeH mRNAs remained at constant levels (Fig. 2E). Between 8 and 24 h the expression of both genes strongly increased. In the csrA mutant, ycdT and ydeH mRNA levels were significantly elevated. ycdT expression was more strongly upregulated (up to 30-fold) during exponential growth compared with later time points (Fig. 2D), whereas the transcript levels of ydeH were approximately 10-fold higher throughout the growth (Fig. 2E). Monitoring csrA transcript levels over time in the wild-type strain shows that csrA expression rapidly decreased between 8 and 24 h (Fig. 2F), demonstrating that csrA is inversely regulated with ycdT and ydeH. However, the fact that CsrA activity is in large part under the control of CsrB and CsrC makes it difficult to correlate csrA mRNA levels with its activity. Nevertheless, these data confirm that CsrA is a negative regulator of ycdT and ydeH expression.
Effects of other components of the BarA-UvrY-Csr cascasde on ydeH and ycdT expression
CsrA is antagonized by the CsrB and CsrC sRNAs. These sRNAs are transcriptionally activated by the BarA-UvrY two-component system and negatively controlled by CsrD at the level of RNA stability (Suzuki et al., 2002; 2006; Weilbacher et al., 2003). By using genetic mutants we tested the contribution of these components on ycdT and ydeH expression. Disruption of csrB and csrC resulted in a slight decrease in ycdT and ydeH mRNA levels (approximately 70%) compared with the wild type (Fig. 3). Similar weak effects were observed in uvrY and barA mutants. These modest effects are consistent with the earlier finding that levels of CsrA protein normally exceeds the binding capacity of these small RNAs (Gudapaty et al., 2001). A more pronounced effect on ycdT and ydeH mRNAs was observed in a csrD mutant, in which the cellular levels of CsrB and CsrC are increased. Compared to the wild type the mRNA levels of ycdT and ydeH were approximately threefold increased (Fig. 3). All together, these results indicate that the entire Csr regulatory network is involved in the regulation of the expression of ycdT and ydeH.
CsrA directly interacts with the ycdT and ydeH transcripts
Previous studies have suggested that CsrA binds to the consensus sequence, ACA-GGAUG, with the GGA motif representing the most highly conserved nucleotides (Baker et al., 2002; Dubey et al., 2005). To make predictions about the binding of CsrA to the ycdT and ydeH mRNAs, we analysed the 5′ leader sequences of both transcripts for the existence of potential CsrA binding motifs. As no information about the ycdT promoter was available in the database, we determined the transcriptional initiation site of ycdT by Rapid Amplification of 5′-cDNA Ends (5′ RACE). A single band was observed for a PCR reaction, amplifying the 5′ non-translated region of the ycdT transript (Fig. 4A). Sequencing of the RACE PCR product identified the nucleotide A, 35 bp upstream of AUG, as the transcription start site. The −10 and −35 regions of ycdT[TATTAA (−10) and TTGACA (−35)], separated by a 19 bp spacing region, exhibited 4 and 6 bp of identity with respect to the consensus sequences for these promoter elements [TATAAT (−10) and TTGACA (−35)] (Hawley and McClure, 1983). We identified two potential CsrA binding sites with degenerate motifs in the 5′ non-translated region of the ycdT mRNA, one of them close to the transcription start and the other one overlapping the AUG translation initiation start codon (Fig. 4A). The ydeH transcript starts 29 nucleotides upstream of the initiation codon AUG (Yamamoto and Ishihama, 2006). Also in the 5′ non-translated region of the ydeH mRNA two potential CsrA binding sites were found, one of which overlaps the Shine–Dalgarno sequence and the other one is close to the 5′ end of the transcript (Fig. 4B).
To experimentally determine whether CsrA directly binds to the ycdT and ydeH transcripts, quantitative RNA gel mobility shift assays were performed with a ycdT transcript, consisting of a 36 nt leader and the first 20 nt of the coding region, and a ydeH transcript containing the 29 nt untranslated leader and the first 25 nt of the coding sequence. CsrA bound strongly to both ycdT and ydeH transcripts (Fig. 4C and D). For the ycdT transcript, two distinct complexes were observed at 2.5 nM CsrA, and essentially all of the starting RNA was shifted at 80 nM CsrA (Fig. 4C). For the ydeH transcript, two distinct shifted complexes were formed at 5 nM CsrA. However, complete binding was not seen until 320 nM CsrA, and at this concentration essentially all of the RNA was present in the upper complex (Fig. 4D). These gel shift patterns suggested that two CsrA proteins were bound to each transcript at higher CsrA concentrations, although the stoichiometry of binding was not experimentally determined. A non-linear least-squares analysis of these data yielded an apparent equilibrium binding constant (Kd) of 2.6 ± 0.3 nM for ycdT and 2.3 ± 0.1 nM for ydeH.
The specificity of CsrA interaction with ycdT and ydeH transcripts was investigated by performing competition experiments with specific (ycdT or ydeH transcripts) and non-specific (Bacillus subtilis trp leader) unlabelled RNA competitors. Both ycdT and ydeH RNAs were able to compete for binding to CsrA while B. subtilis trp RNA did not effectively compete with the CsrA–ycdT or CsrA–ydeH interaction (Fig. 4E and F). These results establish that CsrA binds specifically to both ycdT and ydeH RNA.
In most cases CsrA downregulates its direct mRNA targets by binding to the leader, preventing translation and destabilizing the transcript (Baker et al., 2002; Dubey et al., 2003; Wang et al., 2005). However, in the case of hfq the binding of CsrA to the leader does not lead to mRNA destabilization, but to altered transcription (Baker et al., 2007). To test whether CsrA influences ycdT and ydeH mRNA levels by modulating promoter activity, we constructed plasmid-borne transcriptional ycdT– and ydeH–lacZ fusions, containing the upstream intergenic region and only 2 or 3 nt of each transcript (ycdT−547 to +2; ydeH−222 to +3). Measuring of β-galactosidase activity of these reporter fusions revealed that both promoters were highly active in the wild type but were not altered by a csrA mutation (Fig. 4G). In contrast, lacZ expression from a control plasmid carrying the csrB promoter, which has earlier been reported to be regulated in a CsrA-dependent manner (Gudapaty et al., 2001), was clearly decreased in the csrA mutant (Fig. 4G). This demonstrates that CsrA does not change transcription of ycdT and ydeH, but rather modulates the stability of the messages.
YcdT and YdeH regulate motility
Proteins with GGDEF and EAL domains have been demonstrated to be involved in the regulation of bacterial physiology, including motility, biofilm formation, cell morphology and virulence (see reviews by D'Argenio and Miller, 2004; Jenal, 2004; Romling et al., 2005; Romling and Amikam, 2006; Cotter and Stibitz, 2007). To characterize the phenotype of ycdT and ydeH in motility we analysed the swimming behaviour of strains, in which ycdT and ydeH were expressed from pBADycdT and pBADydeH respectively, as well as respective knock-out mutants. Overexpression of both pBADycdT and pBADydeH led to a strong repression of swimming behaviour (Fig. 5A). The same effect was observed in a Salmonella enterica serovar Typhimurium (S. Typhimurium) background, although S. Typhimurium do not contain orthologues of these proteins (Fig. 5B). Mutations in ydeH and ycdT led to slightly increased swimming ability compared with the wild type. A ydeH ycdT double mutant was, however, not more motile than the wild type (Fig. 5C).
Earlier studies have demonstrated that site-directed mutations in the GGDEF signature sequence of other proteins disrupt the function of this domain (Garcia et al., 2004; Paul et al., 2004; Simm et al., 2004). To test whether the repressing effect of ycdT and ydeH overexpression on motility was due to the activity of both proteins as DGCs, we engineered mutants, in which the two first glycine residues of the respective GGEEF motifs were replaced by two alanine residues (GGEEF→AAEEF). The swimming behaviour of E. coli or Salmonella expressing the plasmids encoding these mutant YdeH and YcdT variants (pBADycdT-mut, pBADydeH-mut) was indistinguishable from the bacteria carrying the empty vector (Fig. 5A and B). This strongly suggests that the effect of YcdT and YdeH on motility is mediated by the second messenger c-di-GMP.
Previous results have demonstrated that a csrA mutant is strongly impaired in motility and that CsrA upregulates flhDC expression by binding to and stabilizing this mRNA (Wei et al., 2001). To test whether deletions in ycdT or ydeH can compensate for the swimming defect of the csrA mutant we constructed csrA ycdT and csrA ydeH double mutants as well as a csrA ycdT ydeH triple mutant. All three strains were not more motile than the csrA mutant (Fig. 5D), suggesting complex regulation of motility by pathways within the Csr network.
Beside their impact on motility, many proteins with GGDEF domains have been shown to regulate biofilm formation. In accordance, results of a parallel ongoing study show that YdeH significantly affects biofilm formation (C. Goller and T. Romeo, unpublished). Furthermore, the effect on biofilm formation seems to be mediated through increased synthesis of the biofilm polysaccharide PGA. YcdT has earlier been characterized regarding its phenotype in biofilm formation (Wang et al., 2005). However, neither biofilm formation nor pgaA–lacZ expression was influenced by YcdT under the given conditions (Wang et al., 2005).
In Enterobacteriaceae several GGDEF proteins have been shown to control biofilm formation by regulating the expression of curli fibres (Romling, 2005). Here, we analysed the expression of curli by analysing the colony morphology on Congo Red (CR) agar plates as well as the ability to form pellicles and to adhere to glass culture tubes at the air–liquid interface. However, we were not able to detect distinct ycdT- or ydeH-dependent phenotypes with respect to CR binding, pellicle formation or glass adherence at the air–liquid interface, suggesting that neither YcdT nor YdeH influences curli production under the conditions tested (data not shown).
Atomic force microscopy analysis of YcdT- and YdeH-mediated phenotypes
To further investigate the phenotypes mediated by YcdT and YdeH, we employed atomic force microscopy (AFM), a technique recently shown to be a suitable tool for the study of bacterial morphology (Jonas et al., 2007). We allowed the bacteria to grow and to adhere to the substratum mica, which was submerged in the growth medium. For immobilization, the samples were air-dried at room temperature prior to AFM analysis. Images of the wild-type strain carrying pBAD28 showed rod-shaped bacteria expressing flagella and pili-like structures (Fig. 6A). Overexpression of pBADydeH resulted in a clear reduction in the abundance of flagella (Fig. 6A and B), suggesting a role for YdeH in the regulation of flagellum biosynthesis. We also noted that YdeH overexpression completely abolished the appearance of the pili-like structures (Fig. 6A and C), indicating that flagella and pili synthesis might be co-regulated by YdeH. In contrast to ydeH, overexpression of ycdT did not affect the occurrence of flagella or pili or another distinct phenotype (Fig. 6), indicating that YcdT might have functions in the cell different from YdeH, which cannot be visualized by AFM under the conditions we have tested.
YcdT and YdeH influence c-di-GMP levels in vivo
Both YcdT and YdeH contain GGDEF domains with consensus motifs, which are predicted to be dedicated to the synthesis of c-di-GMP (Ausmees et al., 2001; Paul et al., 2004; Simm et al., 2004; Schmidt et al., 2005). So far, proteins with highly conserved active site motifs have been shown to possess DGC activity (Ausmees et al., 2001; Kirillina et al., 2004; Paul et al., 2004; Simm et al., 2004; Weber et al., 2006). In contrast, CsrD with a degenerate motif (HRSDF) failed to produce c-di-GMP (Suzuki et al., 2006). The amino acid sequences of the GGDEF domain of YcdT and YdeH perfectly match the conserved GG(D/E)EF motif as well as additional more extended conserved amino acid signatures of other enzymatically active proteins (Fig. 7A). Together with the finding that site-directed mutations of the GGDEF domains of YcdT and YdeH disrupted the effect on motility (Fig. 5), this strongly suggests that both proteins synthesize c-di-GMP. To prove this experimentally, we measured the c-di-GMP concentrations produced by E. coli MG1655 containing plasmid-encoded ydeH (pBADydeH), ycdT (pBADycdT) or the empty vector using high-performance liquid chromatography (HPLC) and matrix-assisted laser desorption ionization-time of flight (MALDI-TOF). Expression of ydeH resulted in clearly increased c-di-GMP levels (2215.3 fmol mg−1 cells) compared with the low levels, close to the limit of detection, in the control strain carrying pBAD28 (94.6 fmol mg−1 cells). Even higher levels were detected when ycdT was overexpressed (7213.0 fmol mg−1 cells) (Fig. 7B). These data provide strong evidence that both of these proteins function as DGC in vivo.
The strong effect of CsrA on ycdT and ydeH transcript levels led us to analyse the overall effect of CsrA on the cellular c-di-GMP pool, by measuring the levels of the second messenger in the wild-type strain MG1655 and its csrA mutant. We were able to consistently detect slightly elevated c-di-GMP levels in the csrA mutant (120.8 fmol mg−1 cells) compared with the wild type (74.5 fmol mg−1 cells) (Fig. 7C). This finding demonstrates a net effect of CsrA in the regulation of c-di-GMP turnover and is consistent with the previously documented negative effect of CsrA on biofilm formation (Jackson et al., 2002).
Several genes encoding GGDEF and EAL proteins have previously been shown to cross-complement phenotypes (Garcia et al., 2004; Simm et al., 2004), even between different species (Simm et al., 2005). In Salmonella, a mutation in the GGDEF gene adrA results in deficiency in cellulose synthesis due to decreased c-di-GMP levels. Overexpression of enzymatically active GGDEF proteins in such a Salmonella mutant leads to the restoration of cellulose production, which can be visualized on agar plates containing the dyes calcofluor (CF) or CR. Thus, the ability to produce cellulose can be used as an indicator for DGC activity. We utilized this effect to study the enzymatic activities of YcdT and YdeH in S. Typhimurium. Strain MAE103, mutated in adrA and carrying pBADycdT, pBADydeH or the controls pBAD28 or pBADcsrA, was allowed to grow on CF and CR agar plates at 28°C. Overexpression of pBADydeH resulted in a strongly fluorescent colony appearance of Salmonella on the CF plates (Fig. 7D) and as pink and rough colonies on the CR plates (data not shown), suggesting that cellulose was produced due to the elevated production of c-di-GMP. However, no dye binding could be observed for the strain carrying pBADycdT, demonstrating that ycdT fails to cross-complement an adrA mutation in Salmonella under the given conditions. In agreement with these data, only subtle changes (less than twofold) in c-di-GMP levels were observed by HPLC and MALDI-TOF, when ycdT was overexpressed in Salmonella, grown at 28°C on Luria–Bertani (LB) agar without salt, whereas plasmid-encoded expression of ydeH in the same background strain resulted in strongly elevated c-di-GMP levels (> 200-fold) (Fig. 7E). Thus, in contrast to YdeH, which apparently possesses high DGC activity in plate-grown Salmonella at 28°C, YcdT appears to produce c-di-GMP at very low concentrations under the given conditions. A previous study has already demonstrated that in Salmonella most, but not all, GGDEF proteins with conserved sequence signatures could restore cellulose production in an adrA mutant (Garcia et al., 2004) and that the cross-complementation ability strongly depended on the experimental conditions.
Global role of CsrA in the regulation of other GGDEF/EAL proteins
To test whether CsrA controls the expression of additional GGDEF/EAL proteins we analysed our array data for the expression patterns of all genes, annotated to contain a GGDEF and/or EAL domain, 4 and 12 min after CsrA pulse overproduction. For 4 of the 29 selected genes the signals were too low for reliable detection on the microarray (yeaI, yaiC, yhjH and ycgG). Most of the other GGDEF/EAL genes were relatively weakly expressed, but strongly enough for detection on the array. Interestingly, beside ycdT and ydeH several other genes showed changes in their transcript levels upon CsrA overproduction with an additive effect between 4 and 12 min: yddV (GGDEF), yliF (GGDEF), dos (GGDEF-EAL), yhjK (GGDEF-EAL), csrD (GGDEF-EAL), yliE (EAL) and yjcC (EAL) (Fig. 8A and E). CsrA-dependent repression of these genes was, however, not as strong (1.5–2.5 fold) as repression of ycdT and ydeH. RT-PCR analysis of the kinetics of CsrA-dependent expression confirmed that mRNA levels of yliE, yliF, yddV, dos and csrD were indeed downregulated upon induction of pBADcsrA (Fig. 8B), but remained constant or increased upon induction of the vector control (Fig. 8C). In the csrA mutant strain, expression of these genes was moderately increased (between two- and sixfold) compared with the wild type (Fig. 8D). Repression of yhjK and yjcC by CsrA overproduction could not be confirmed (data not shown). Noticeably, yddV and dos (Mendez-Ortiz et al., 2006) as well as yliE and yliF are organized as polycistronic units in operons (Fig. 8F and G). The expression patterns of yliE and yliF as well as yddV and dos followed almost identical kinetics (Fig. 8B and C), indicating that these genes are co-regulated at the mRNA level by CsrA. The observation that CsrD is negatively regulated by CsrA agrees with the earlier finding that expression of a chromosomal csrD–lacZ translational fusion was modestly repressed (twofold) by CsrA (Suzuki et al., 2006). These data confirm that CsrD is part of an additional autoregulatory loop within the Csr system. In summary, our data demonstrate that beside ycdT and ydeH, genes for several other GGDEF and GGDEF-EAL proteins as well as one EAL protein are negatively regulated by CsrA. This finding suggests a global role for CsrA in the regulation of c-di-GMP metabolism.
Post-transcriptional regulation of GGDEF/EAL proteins by CsrA
The present study was initiated with a genome-wide search for novel targets for the post-transcriptional regulator CsrA to better understand its role in bacterial adaptation and the cross-talk between the Csr system and other regulatory systems. Our search led to the finding that CsrA controls the expression of several GGDEF/EAL proteins, in particular the GGDEF proteins YcdT and YdeH, by physically binding to and changing their mRNA levels. To our knowledge this is the first example of GGDEF/EAL proteins being regulated at the mRNA level by a global post-transcriptional regulator. This supports the idea that c-di-GMP signalling is a multilayer process, including transcriptional, translational and post-translational levels. The array data also indicated that CsrA controls other mRNAs, some of them with unknown functions, but these effects need to be confirmed and were not the focus of this study.
With binding constants (Kd) of approximately 2.5 nM, CsrA binding to the ycdT and ydeH transcripts was remarkably strong. The affinities to the other known mRNA targets pgaA, glgC, cstA and hfq were approximately 10-fold lower (22, 39, 40 and 38 nM respectively) (Baker et al., 2002; 2007; Dubey et al., 2003; Wang et al., 2005). Noteworthy, for each of the ycdT and ydeH transcripts only two GGA boxes were found in the 5′ leader sequences and binding of two CsrA proteins per transcript was observed at higher concentrations. For comparison, pgaA, glgC and cstA contain four to six potential CsrA binding sites. Moreover, in the case of ycdT, the sequence signature of both sites showed relative poor similarity to the consensus sequence. Thus, in addition to the primary sequence conservations, other determinants seem to largely influence the affinity of CsrA to its targets.
Our array revealed that beside YcdT and YdeH, two additional GGDEF proteins (YddV, YliF), two GGDEF-EAL proteins (Dos, CsrD) and one EAL protein (YliE) were regulated by CsrA. Together with the finding that increased overall levels of cellular c-di-GMP were measured in a csrA mutant, this implicates a global role for CsrA in the regulation of c-di-GMP metabolism. It is plausible that under other experimental conditions CsrA might control the expression of additional GGDEF/EAL proteins. For most of the GGDEF/EAL genes, relatively weak signals were detected on the microarray, suggesting that these genes require specific conditions for enhanced expression, different from the standard conditions used in our experiment. As CsrA homologues are present in many different Gram-negative bacteria (White et al., 1996) the role of CsrA in the regulation of GGDEF/EAL proteins might be a conserved feature. With the exception of CsrD, no other of the CsrA-regulated GGDEF/EAL genes have homologous genes in S. Typhimurium. Therefore, CsrA might act on other GGDEF/EAL proteins in other bacteria. Furthermore, we cannot exclude that our microarray approach, which requires the destabilization of the CsrA mRNA targets upon its binding, failed to identify other important CsrA targets, in which translation is regulated without a corresponding alteration in mRNA stability, similar to the previous findings for hfq (Baker et al., 2007).
Interplay between Csr and c-di-GMP signalling
While c-di-GMP-mediated phenotypes and the molecular mechanisms governing c-di-GMP synthesis and turnover have received much attention, the role of the c-di-GMP network in signal transduction, including its linkage to external signals of specific adaptive responses and its interconnection with other global networks, is relatively unexplored. Nevertheless, in E. coli genes encoding GGDEF/EAL domains were recently reported to be overrepresented in the σS (RpoS) regulon, suggesting a role for c-di-GMP during the general stress response (Weber et al., 2006). In Vibrio cholerae quorum-sensing signalling was recently demonstrated to be connected to c-di-GMP signalling through the action of the major quorum-sensing regulator HapR (Waters et al., 2008). Furthermore, some GGDEF/EAL proteins, exemplified by the response regulator PleD (Aldridge and Jenal, 1999), contain phospho-receiver domains or other signalling domains, facilitating cross-talk and the integration into other signal cascades (Paul et al., 2008).
Our study revealed a direct link between the global Csr network and c-di-GMP signalling, placing both pathways in a broad cellular context. CsrA activity is controlled by the sRNAs CsrB and CsrC, whose expression levels are regulated by the BarA-UvrY two-component system and the probable inner membrane protein CsrD (Fig. 9). CsrA has previously been shown to control motility and biofilm formation by directly targeting the flhDC and pgaA mRNAs respectively. Here, we show that in addition to the regulation of biosynthesis and global regulators CsrA regulates bacterial physiology in a c-di-GMP-dependent pathway by directly controlling the expression of ycdT and ydeH, which cause c-di-GMP accumulation and thereby favour the sessile life style. The combination of c-di-GMP-dependent and c-di-GMP-independent regulatory pathways allows CsrA to regulate biofilm-related processes at various levels and thus to trigger the switch between a motile and a sessile life style. The CsrA- and c-di-GMP-specific adaptive responses are controlled by environmental signals, integrated at multiple sites within the signalling cascade. Although the nature of the signal sensed by BarA is not known, it is proposed to reflect the energy/growth status of the cell (Pernestig et al., 2003). In addition, BarA-UvrY signalling was recently demonstrated to be pH-dependent (Mondragon et al., 2006). The prediction that CsrD and YcdT are membrane-bound suggests that their activity is controlled from the outside. In addition, transcription of ydeH was previously demonstrated to depend on the CpxAR two-component system, responding to cell envelope stress and external copper (Yamamoto and Ishihama, 2005; 2006). A future challenge will be the identification of the nature of different input signals controlling Csr and c-di-GMP signalling.
The roles of YcdT and YdeH in bacterial physiology
Numerous studies have shown that c-di-GMP controls bacterial behaviour (reviews by D'Argenio and Miller, 2004; Jenal, 2004; Romling, 2005; Romling and Amikam, 2006). High levels of this second messenger favour sessility whereas low levels of c-di-GMP promote a motile life style. In accordance, YdeH and YcdT were found to repress swimming behaviour. YdeH seems to act at the level of flagellum synthesis while YcdT seems to modulate flagella function, raising the possibility that individual GGDEF proteins are dedicated to specific functions in the cell. We also observed that in the strain overexpressing YdeH the occurrence of pili was abolished. A recent study proposed a link between c-di-GMP signalling and type 1 pili and flagella expression in the Crohn disease-associated adherent-invasive E. coli strain LF82 (Claret et al., 2007). While similar pathways might exist in E. coli K12, to this date we have no evidence for this hypothesis.
Our data show that overexpression of ydeH led to highly elevated c-di-GMP levels and pronounced cellulose production in Salmonella. Consitent with these data, results from another parallel study suggest a significant role for YdeH in biofilm formation by regulating PGA synthesis (C. Goller and T. Romeo, unpublished). Although overexpression of ycdT resulted in a strong accumulation of cellular c-di-GMP, we did not observe a distinct biofilm-related phenotype neither in E. coli nor in S. Typhimurium. In addition, in an earlier study biofilm formation and pgaA–lacZ expression were not affected in the ycdT mutant XWMGΔT (Wang et al., 2005). Nevertheless, we suspect that not only YdeH, but also YcdT might have an impact on biofilm formation under other experimental growth conditions. The ycdT gene and the pga operon are divergently organized (Fig. 1E). A comprehensive bioinformatics study has recently demonstrated that chromosomal proximity indicates gene co-regulation in prokaryotes independent of relative gene orientation and that adjacent bidirectionally transcribed genes with conserved gene orientation are strongly co-regulated (Korbel et al., 2004). Furthermore, the ycdT homologue in Yersinia pestis, called HmsT, has been reported to be required for biofilm formation (Kirillina et al., 2004). Likewise, there is evidence that E. coli and Y. pestis produce the PGA polysaccharide as biofilm matrix component (Itoh et al., 2005; Bobrov et al., 2008). Synthesis of this polysaccharide was in a recent study shown to be positively regulated by HmsT, which was suggested to control c-di-GMP levels in close proximity to the glycosyltransferase HmsR, responsible for the production of the polysaccharide (Bobrov et al., 2008). Thus, regulation of PGA synthesis in Yersinia seems to occur in a c-di-GMP-dependent fashion, similar to the production of the biofilm polysaccharide cellulose in Salmonella. Moreover, another recent study showed that synthesis of the PEL polysaccharide in Pseudomonas aeruginosa is regulated by c-di-GMP. Here, the PelD protein serves as the c-di-GMP receptor, activating the production of the PEL polysaccharide by a yet to be defined mechanism (Lee et al., 2007). These data suggest that there are related c-di-GMP-dependent processes for controlling synthesis of the PGA exopolysaccharide in E. coli.
Noticeably, the pel genes in Pseudomonas, necessary for PEL synthesis, have been suggested to be regulated by the GacS-GacA-Rsm cascade, which is homologous to the BarA-UvrY-Csr pathway in E. coli (Goodman et al., 2004), further suggesting that the role of the Csr regulatory network in the regulation of biofilm components may be a conserved feature among γ-proteobacteria.
Bacterial strains and growth conditions
All strains used in this study are listed in Table 1. Chromosomal ydeH::cat and csrB::cat mutations were constructed using the Datsenko method (Datsenko and Wanner, 2000). The cat gene was amplified from pKD3 by PCR using primers ydeHKOFor2 and ydeHKORev2 or csrBKOFor and csrBKORev (Table 2) respectively, and introduced by electroporation into arabinose-treated BW25141 carrying pKD46. Transformants were selected on chloramphenicol plates, and their insertion sites were confirmed by PCR using the primer pairs ydeHKOtestFor/ydeHKOtestRev and csrBKOtestFor/csrBKOtestRev (Table 2). Mutations were transferred among strains by P1 transduction. For construction of the csrB csrC double mutant KJ230, the csrC::tet allele from strain TWMG1655 was moved into the csrB mutant KJ227, from which the chloramphenicol cassette had been flipped out using the FLP recombinase. Strain KJ311 was generated by removing the chloramphenicol cassette from KJ295 by using the FLP recombinase and subsequent infection with a P1 lysate containing the ycdT::cat mutation from XWMGΔT. For construction of strain KJ157, csrA::kanr was moved from TRMG into KSB837. To generate the csrD mutant KJ205 the yhdA::catr cassette from KJ27 was transduced into MG1655. To generate KJ331, KJ330 and KJ369, csrA::kanr was transduced from TRMG into KJ295, XWMGΔT or KJ311 respectively. In most of the experiments, bacteria were grown in LB medium at 37°C with shaking at 200 r.p.m. If necessary, antibiotics were added: ampicillin 100 μg ml−1, kanamycin 50 μg ml−1 and chloramphenicol 30 μg ml−1.
Table 1. Bacterial strains and plasmids used in this study.
All plasmids used in this study are listed in Table 1. For construction of pBADcsrA, pBADydeH and pBADycdT, the genes for csrA, ydeH and ycdT were amplified from the MG1655 chromosome by PCR using the primer pairs CsrAForBAD/CsrARevBAD, pBADydeHFor/pBADydeHRev or pBADycdTFor2/pBADycdTRev2 respectively (Table 2). The PCR products of csrA and ydeH were cleaved with the enzymes HindIII and XbaI, while the product of ycdT was cut with SacI and XbaI. After cleavage of the pBAD28 vector at the corresponding sites followed by dephosphorylation (Shrimp Alkaline Phosphatase, Roche Diagnostics), the cleaved PCR fragments were inserted using the Rapid DNA Ligation Kit (Roche Diagnostics). For construction of pPYCDT and pPYDEH the upstream intergenic regions of the ycdT gene and the ydeH gene, including 2 or 3 nt of the respective transcripts (ycdT−547 to +2; ydeH−222 to +3) were amplified using the primer pairs PycdTFor-EcoRI/PycdTRev-BamHI or PydeHFor-EcoRI/PydeHRev-BamHI (Table 2) respectively, and subsequently digested by BamHI and EcoRI. After removing the PcsrB insert from vector pCBZ1 (Gudapaty et al., 2001) by BamHI and EcoRI cleavage, the empty linearized vector was dephosphorylated and ligated with the respective ycdT or ydeH fragments to create pPYCDT and pPYDEH. Sequencing verified the integrity of all plasmid constructs.
To engineer the mutated ycdT and ydeH alleles, plasmids pBADycdT and pBADydeH were subjected to site-directed mutagenesis using the high-performance liquid chromatography-purified primer pairs YcdT-Mut-For/YcdT-Mut-Rev and YdeH-Mut-For/YdeH-Mut-Rev (Table 2) and the QuikChange II site-directed mutagenesis kit (Stratagene) to create plasmids pBADycdT-mut and pBADydeH-mut. Mutations introduced into ycdT and ydeH led to the replacement of the two glycines at positions 359 and 360 (ycdT) or 206 and 207 (ydeH) in the GGEEF motif by alanine (YcdT G359A, G360A; YdeH G206A, G207A). The mutations were confirmed by sequencing.
Bacterial cultures were mixed with 2 vols of RNAprotect Bacterial Reagent (Qiagen) and incubated for 5 min at room temperature. Total cellular RNA was subsequently prepared by using the RNeasy Mini Kit with on-column DNA digestion (Qiagen). RNA concentrations were determined using the NanoDrop ND-1000 UV-Vis Spectrophotometer (NanoDrop Technologies, Wilmington, DE). The quality of the RNA used for the microarray was assessed using the Agilent Bioanalyser.
Microarray analysis was performed at the Bioinformatics and Expression Analysis Core Facility at the Karolinska Institute (http://www.bea.ki.se) using the GeneChip E. coli Genome 2.0 Array (Affymetrix, P/N 900551, Santa Clara, CA). This array includes approximately 10 000 probe sets for all 20 366 genes present in four strains of E. coli. Affymetrix analysis was conducted according to the Affymetrix manual (http://www.affymetrix.com). The absolute signals from the samples, taken at 0 min (before arabinose induction), were compared with the signals from the 4 and 12 min samples. The signal ratios resulting from pBADcsrA overexpression were then normalized with the ratios resulting from overexpression of the empty vector pBAD28. Genes, whose expression levels were too low for reliable detection, or whose expression levels were decreased in response to induction of the empty vector pBAD28, were excluded from the analysis.
Quantitative real-time RT-PCR
Five hundred nanograms of total RNA was used to synthesize cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Primers were designed using the Primer Express Software v3.0 (Applied Biosystems). All RT primers used in this study are listed in Table 2. 0.1 ng template was used for the real-time PCR reaction using the Power SYBR Green PCR Master Mix (Applied Biosystems). Analysis was performed with an ABI 7500 Real Time PCR System (Applied Biosystems) using the standard run mode of the instrument. For detection of primer dimerization or other artefacts of amplification, a dissociation curve was run immediately after completion of the real-time PCR. Individual gene expression profiles were normalized against the rrnD gene, serving as an endogenous control. All results were analysed using the 7500 SDS Software v1.3.1 (Applied Biosystems) and further prepared using Excel (Microsoft). The data values presented in all figures represent the mean expression level of quadruplicates from one real-time PCR assay, relative to a calibrator value (time point 0 min or wild type). The error bars represent the standard error of the mean expression level calculated by the SDS software using the confidence value 95%. Each experiment was repeated independently and representative data were chosen for presentation.
5′ ends of the ycdT transcripts were determined using the 5′ RACE System for Rapid Amplification of cDNA Ends (v2.0, Invitrogen). Three micrograms of total RNA was reverse-transcribed using the primer YcdTGSP1 and the superscript II RT. cDNAs were purified, C-tailed with a terminal deoxynucleotidyl transferase and used as template in a PCR with an anchor primer (AAP), specific for the C-tail, and the gene specific primer YcdTGSP2HindIII (GSP2), complementary to a region upstream of the binding site of GSP1. To increase specificity, a nested PCR was carried out using the nested anchor primer UAP and the gene-specific nested primer YcdTGSP3HindIII (GSP3). The PCR products were visualized on a 2% agarose gel in TBE buffer and subsequently sequenced using the Big Dye Terminator Cycle Sequencing Kit (v3.1).
RNA gel mobility shift assays
Quantitative gel mobility shift assays followed a previously published procedure (Yakhnin et al., 2000). E. coli CsrA-His6 protein was purified as described previously (Mercante et al., 2006). DNA templates for generating ycdT and ydeH RNA transcripts were produced by annealing primers ycdT-T7 (−36 to +20) and GC ycdT-T7 (−36 to +20) and ydeH-T7 (−29 to +25) and GC ydeH-T7 (−29 to +25) in TES buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 100 mM NaCl). RNA was synthesized in vitro using the MEGAshortscript kit (Ambion) using the annealed DNA primers (for ydeH and ycdT) or linearized plasmid pPB77 (Babitzke et al., 1994) as templates. After gel purification, transcripts were 5′ end-labelled using T4 polynucleotide kinase and [γ-32P]-ATP. Radiolabelled RNA was gel-purified and re-suspended in TE (10 mM Tris-HCl pH 8.0, 1 mM EDTA), heated to 85°C and chilled on ice. Increasing concentrations of purified CsrA-His6 recombinant protein were combined with 80 pM radiolabelled RNA in 10 μl of binding reactions [10 mM Tris-HCl pH 7.5, 10 mM MgCl2, 100 mM KCl, 3.25 ng total yeast RNA, 20 mM DTT, 7.5% glycerol, 4 U SUPERasin (Ambion, Austin, TX)] for 30 min at 37°C to allow for CsrA–RNA complex formation. Competition assays were performed in the absence or presence of unlabelled RNA specific and non-specific competitors. Binding reactions were separated using 12% native polyacrylamide gels, and radioactive bands were visualized with a Molecular Dynamics phosphorimager. Free and bound RNA species were quantified with ImageQuant Software (Molecular Dynamics), and an apparent equilibrium binding constant (Kd) was calculated for CsrA–RNA complex formation according to a previously described cooperative binding equation (Mercante et al., 2006). The mean values and standard errors from two independent experiments were determined for each transcript. Graphpad Prism version 3.02 for Windows (San Diego, CA) software was used for calculations.
β-galactosidase activity was measured in 10 min reactions using the Miller protocol (1972). Twenty microlitres of bacterial culture, grown to an OD600 of 1.5, was used for each reaction. Each measurement was carried out independently at least two times.
Quantification of c-di-GMP
Nucleotide extracts were prepared essentially as previously described (Simm et al., 2004). For c-di-GMP extraction from liquid cultures, bacteria were grown in LB medium to OD600 1.5 at 37°C, treated with formaldehyde (0.19% final concentration) and pelleted by centrifugation. The pellet was re-suspended in ice-cold water, heated to 95°C for 10 min, before nucleotides were extracted by ethanol treatment. For c-di-GMP extraction from plate-grown bacteria, approximately 100 mg of cells was harvested and immediately suspended in ice-cold 0.19% formaldehyde, before being boiled for 10 min and treated with ethanol. Nucleotide extracts of 10 or 50 mg of cells (wet weight) were fractioned by HPLC using a reversed-phase column (Hypersil ODS 5 μ; Hypersil-Keystone). Runs were carried out with a multistep gradient using 0.1 M triethyl ammonium acetate (pH 6.0) at 1 ml min−1 with increasing concentrations of acetonitrile. Relevant fractions were collected, lyophilized and re-suspended in 10 μl of water. Fractions containing c-di-GMP were pinpointed by MALDI-TOF analysis and pooled. Synthetic c-di-AMP was added to the pooled fractions at a suitable concentration to be used as an internal standard. A standard curve was established using fractions spiked with known amounts of c-di-GMP, using a fixed amount of synthetic c-di-AMP as internal control. The isotope areas of c-di-GMP and c-di-AMP were calculated, and the ratio was determined. Each c-di-GMP measurement was carried out independently at least two times.
Sample preparation and AFM imaging were performed as earlier described (Jonas et al., 2007). Bacteria were allowed to grow for 24 h at 28°C on mica slides Grade V-4 (SPI Supplies, USA) submerged in Petri dishes containing 3 ml of LB medium without NaCl. After incubation the mica slides were dipped three to four times into double-distilled water, air-dried at room temperature in a dust-free environment for several hours and mounted onto glass microscope slides. Bacteria were imaged with the BioScope SZ (Veeco Instruments, Woodbury, NJ) operated in the contact mode using V-shaped silicon nitride nanoprobe cantilevers MLCT-AUHW (Veeco) with a spring constant of 0.05 N m−1. Images were captured using NanoScope v6.13 (Veeco) and further analysed with the scanning probe software WSxM (Nanotec Electronica, Spain) (Horcas et al., 2007). To quantify flagella expression, the number of flagella and the number of bacteria were counted at five different locations on the microscope slide for each strain. The ratio of flagella per 10 bacteria was calculated and the mean and the standard deviation determined. To quantify pili expression, the number of bacteria expressing pili per total number of bacteria was calculated at five different locations for each strain, from which the mean value and the standard deviation were calculated.
To analyse the swimming behaviour of the bacteria, 0.3% motility agar plates, if necessary supplemented with 0.1% arabinose, were inoculated with 4 μl of overnight culture and incubated at 37°C. The diameter of the swimming zone was measured over time. For analysis of colony morphology, bacteria from an overnight culture were streaked onto LB agar plates with or without NaCl supplemented with CR (40 μg ml−1) and Coomassie brilliant blue (20 μg ml−1) or CF (fluorescence brightener 28; 50 μg ml−1). Plates were incubated at 28°C or 37°C for 20 h or 24 h respectively. The colony morphology and dye binding were analysed over time. Glass adherence was measured by culturing the bacteria in standing glass culture tubes containing LB medium with or without salt at 28°C or 37°C for 24 h. After analysing the formation of pellicles visually, the culture liquid was discarded by decanting and the bacteria, adherent to the glass tubes, were stained with crystal violet solution. The tubes were subsequently rinsed with water, allowed to air-dry in the upside-down position and adherence of the bacteria to the glass was analysed visually.
We thank Dr Jörg Vogel for helpful discussions and the Bioinformatics and Expression Analysis Core Facility at the Karolinska Institute for carrying out the microarray experiment. This work was supported by a grant from the Marie Curie Early Stage Research Training Fellowship of the European Community's Sixth Framework Program under Contract Number MEST-CT-2004-8475 and by National Institutes of Health grants to T.R. (GM059969, GM066794).