A novel sRNA component of the carbon storage regulatory system of Escherichia coli


  • Thomas Weilbacher,

    1. Department of Molecular Biology and Immunology, University of North Texas Health Science Center at Fort Worth, Fort Worth, TX 76107-2699, USA.
    Search for more papers by this author
  • Kazushi Suzuki,

    1. Department of Microbiology and Immunology, Emory University School of Medicine, 3133 Rollins Research Center, 1510 Clifton Road NE, Atlanta, GA 30322, USA.
    Search for more papers by this author
  • Ashok K. Dubey,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
  • Xin Wang,

    1. Department of Microbiology and Immunology, Emory University School of Medicine, 3133 Rollins Research Center, 1510 Clifton Road NE, Atlanta, GA 30322, USA.
    Search for more papers by this author
  • Seshigirao Gudapaty,

    1. Department of Molecular Biology and Immunology, University of North Texas Health Science Center at Fort Worth, Fort Worth, TX 76107-2699, USA.
    Search for more papers by this author
      Biochemistry Department, Andhra University, Visakhapatnam 530-003, India.
  • Igor Morozov,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
      ‡Plant Science and Fungal Molecular Biology Research Group, School of Biological Sciences, Donnan Laboratories, The University of Liverpool, Liverpool L69 7ZD, UK.
  • Carol S. Baker,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
  • Dimitris Georgellis,

    1. Departamento de Genética Molecular, Instituto de Fisiología Celular, Universidad Nacional Autónoma de México, 04510 Mexico City, Mexico.
    Search for more papers by this author
  • Paul Babitzke,

    1. Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA 16802, USA.
    Search for more papers by this author
  • Tony Romeo

    Corresponding author
    1. Department of Microbiology and Immunology, Emory University School of Medicine, 3133 Rollins Research Center, 1510 Clifton Road NE, Atlanta, GA 30322, USA.
    Search for more papers by this author

  • Present addresses: Biochemistry Department, Andhra University, Visakhapatnam 530-003, India. Plant Science and Fungal Molecular Biology Research Group, School of Biological Sciences, Donnan Laboratories, The University of Liverpool, Liverpool L69 7ZD, UK.


Small untranslated RNAs (sRNAs) perform a variety of important functions in bacteria. The 245 nucleotide sRNA of Escherichia coli, CsrC, was discovered using a genetic screen for factors that regulate glycogen biosynthesis. CsrC RNA binds multiple copies of CsrA, a protein that post-transcriptionally regulates central carbon flux, biofilm formation and motility in E. coli. CsrC antagonizes the regulatory effects of CsrA, presumably by sequestering this protein. The discovery of CsrC is intriguing, in that a similar sRNA, CsrB, performs essentially the same function. Both sRNAs possess similar imperfect repeat sequences (18 in CsrB, nine in CsrC), primarily localized in the loops of predicted hairpins, which may serve as CsrA binding elements. Transcription of csrC increases as the culture approaches the stationary phase of growth and is indirectly activated by CsrA via the response regulator UvrY. Because CsrB and CsrC antagonize CsrA activity and depend on CsrA for their synthesis, a csrB null mutation causes a modest compensatory increase in CsrC levels and vice versa. Homologues of csrC are apparent in several Enterobacteriaceae. The regulatory and evolutionary implications of these findings are discussed.


Bacterial survival and competition under the feast or famine conditions of the natural environment require remarkable phenotypic plasticity. In Escherichia coli, the transition from exponential to stationary growth phase leads to increased stress resistance, decreased anabolic metabolism, altered cellular and subcellular morphology and enhanced ability to scavenge nutrients (Hengge-Aronis, 1996; Huisman et al., 1996). Acquisition of the stationary phase phenotype is brought about through changes in gene expression, which are co-ordinated by global regulatory systems (Gottesman, 1984; Neidhardt and Savageau, 1996).

The RNA-binding protein CsrA of E. coli is the key component of a global regulatory system that represses several stationary phase processes, whereas it activates certain exponential phase functions (reviewed by Romeo, 1998). Glycogen synthesis and catabolism (Romeo et al., 1993; Yang et al., 1996), gluconeogenesis (Sabnis et al., 1995) and biofilm formation (Jackson et al., 2002) are repressed by CsrA, whereas glycolysis (Sabnis et al., 1995), motility and flagellum synthesis (Wei et al., 2001) and acetate metabolism (Wei et al., 2000) are activated by this protein. The mechanism by which CsrA represses glycogen metabolism involves the binding of CsrA to the untranslated leader of the glgCAP transcript, which blocks translation and causes this transcript to be degraded rapidly (Liu et al., 1995; Liu and Romeo, 1997; Baker et al., 2002). Although not as extensively studied, positive control of flhDC expression by CsrA involves a similar post-transcriptional mechanism, whereby CsrA binding to the untranslated leader ultimately stabilizes this mRNA (Wei et al., 2001). Highly conserved CsrA (or RsmA) homologues are found in diverse eubacteria (reviewed by Romeo, 1998), and regulate virulence factors of animal and plant pathogens (e.g. Chatterjee et al., 1995; Altier et al., 2000; Heeb et al., 2002).

Purification of a His-tagged CsrA protein revealed that it binds to a 366 nucleotide (nt) untranslated RNA molecule, CsrB, to form a globular ribonucleoprotein complex containing ≈ 18 CsrA subunits and a single CsrB transcript (Liu et al., 1997). An imperfect repeat sequence (CAGGAUG) that is located primarily in the loops of predicted RNA hairpins may permit CsrA to bind to CsrB. CsrB functions as an antagonist of CsrA, apparently by sequestering this protein (Liu et al., 1997; Gudapaty et al., 2001). CsrA also binds to the glgC Shine–Dalgarno sequence and to a site further upstream in the untranslated glgC leader transcript, both of which are related in sequence to the repeated elements of CsrB (Baker et al., 2002). CsrA and CsrB levels accumulate as the culture approaches the stationary phase of growth (Gudapaty et al., 2001). Although CsrA binds to CsrB, it does not affect CsrB stability. Instead, CsrA indirectly activates csrB transcription (Gudapaty et al., 2001), via the BarA/UvrY two-component signal transduction system (Pernestig et al., 2001; Suzuki et al., 2002). Purified UvrY stimulates csrB–lacZ expression in vitro, revealing that UvrY resides immediately upstream from csrB in this signalling pathway (Suzuki et al., 2002).

More than a dozen sRNAs have been studied in E. coli, and more than 20 sRNAs of unknown function have been identified by bioinformatics approaches and comprehensive transcript profiling (Argaman et al., 2001; Rivas et al., 2001; Wassarman et al., 2001). Several sRNAs regulate translation by basepairing with complementary segments of mRNAs, with assistance from Hfq protein (Sledjeski et al., 2001; Wassarman et al., 2001; Wassarman, 2002; Zhang et al., 2002). In contrast, CsrB (Liu et al., 1997) and 6S RNA (Wassarman and Storz, 2000) regulate gene expression by binding to proteins, CsrA and σ70-RNA polymerase respectively. In both cases, RNA binding reversibly inhibits the target protein and results in global changes in gene expression. Of the known E. coli sRNAs, only CsrB binds to a large number of subunits of its target protein, providing an efficient means of sequestering CsrA (discussed by Romeo, 1998).

In a previous attempt to identify regulators of the stationary phase synthesis of glycogen, an E. coli genomic library in the low-copy plasmid pLG339 was screened for effects on glycogen levels (Romeo et al., 1991). Two clones were found to increase both glycogen levels and the expression of a glgC′–′lacZ translational fusion. One of these clones, pMR221, contains the csrB gene (unpublished data). The activity of the second clone, pMR2113, was localized to an ≈ 360 bp region that lacks an open reading frame (ORF; unpublished data). This information, and the presence of an apparent Rho-independent terminator sequence located near one end of the functional region led us to hypothesize that this clone expresses a gene for a second regulatory RNA of the Csr system, CsrC. The present study confirms this notion and reveals that CsrC is similar to CsrB in that it binds to and antagonizes CsrA, possesses several of the imperfect repeat sequences that characterize CsrB RNA and is transcriptionally activated by CsrA and UvrY.


The csrC gene specifies an ≈ 245 nucleotide regulatory RNA

The original csrC clone, pMR2113, was isolated as a result of its stimulatory effects on glycogen accumulation and was shown to activate glgC′–′lacZ expression (Romeo et al., 1991). Subclones and deletions from this insert revealed a minimal functional region of 0.36 kb, which lacked an apparent ORF (unpublished data; summarized in Fig. 1). This suggested that pMR2113 might either contain a cis-acting element that sequesters a transcriptional repressor or may express a regulatory RNA. Furthermore, the nucleotide sequence of the functional region contains an apparent Rho-independent transcriptional terminator at one end, consistent with the latter hypothesis (Fig. 1A). Primer extension analysis revealed a single 5′-terminus for this RNA molecule (Fig. 1B). A putative σ70 promoter sequence is located immediately upstream from this site in the csrC gene (Fig. 1A), strongly suggesting that this 5′-terminus represents the initiation site for csrC transcription. Northern analysis revealed that this RNA accumulates as the culture approaches the stationary phase of growth (Fig. 1C) and is dependent on csrA, but not rpoS, which encodes a sigma factor needed for the expression of a variety of stationary phase genes (reviewed by Hengge-Aronis, 1996). Collectively, Northern hybridization, primer extension and sequence analysis reveal an RNA ≈ 245 nt in length, which we call CsrC (Fig. 1).

Figure 1.

Sequence of the csrC gene, Northern analysis of CsrC RNA during the growth curve and primer extension mapping.
A. The csrC gene is located in the intergenic region between the divergent open reading frames yihA and yihI, as shown. The 5′ end of CsrC RNA (+1) and the −10 and −35 elements of an apparent promoter are underlined. Repeat sequences representing putative CsrA binding sites are underlined. Inverted repeats of the apparent Rho-independent terminator are underlined with arrows. Asterisks flank a minimal functional region, deduced by subcloning and deletion analyses. The location of the initiating nucleotide of csrC on the E. coli genome (Blattner et al., 1997) is indicated.
B. A DIG-labelled riboprobe was used for Northern analysis of CsrC RNA harvested throughout the growth curve of MG1655 (WT) or isogenic csrA or rpoS mutants. The positions of RNA standards are shown. Stationary phase occurred at ≈ 6 h.
C. Primer extension analysis used the radiolabelled primer D1PrB, which was annealed to total RNA from MG1655 and extended using reverse transcriptase. The single product of this reaction (lane 1) was analysed with a Sanger sequencing ladder prepared using the same primer and plasmid pMR2113-D1. An asterisk (*) marks the 3′-terminus of the extension product.

Effects of csrC on glycogen levels, glgA′–′lacZ expression, biofilm formation and motility

To examine the effects of csrC on glycogen levels, a csrC mutant, TWMG1655 (Table 1), was generated using a linear transformation protocol (Datsenko and Wanner, 2000), in which the transcribed region of csrC was precisely replaced with a tetR marker. In addition, the transcribed region of the csrC gene, lacking the apparent promoter sequence, was amplified and subcloned into the multiple cloning site of pUC18 in both orientations. Glycogen levels in the csrC mutant were similar to those of the parent strain (Fig. 2A). However, glycogen was slightly decreased in a csrB csrC double mutant, relative to the wild-type strain or to either the csrB or the csrC single mutant. In a csrA mutant strain background, no change in glycogen was observed when csrC or both csrC and csrB were disrupted (data not shown), indicating that CsrC effects on glycogen, like those of CsrB (Liu et al., 1997), are mediated through CsrA. The plasmid pCSRC1, in which csrC is positioned downstream from the lac promoter, caused a substantial increase in glycogen levels (Fig. 2A), whereas no effect was observed for pCSRCre1, in which csrC lacks a promoter (not shown). This result indicated that csrC must be transcribed to be functional and provided genetic evidence that CsrC is a regulatory RNA molecule.

Table 1. . Bacterial strains, plasmids and phages used in this study.
Strain, plasmid or phageDescriptionSource or reference
 BW25113 Δ(araD-araB)567 ΔlacZ478,lacIp-4000 rpoS396 rph-1Δ(rhaD-rhaB)568 rrnB-4
Datsenko and Wanner (2000)
 BW3414 ΔlacU169 rpoS(Am)B. Wanner
 CF7789MG1655 ΔlacI-Z(mluI)M. Cashel
 DH5α supE44 ΔlacU169(Φ80lacZΔM15) hsdR17 relA1 endA1gyrA96thi-1 Ausubel et al. (1989)
 DHB6521λInCh1 (Kanr) lysogen Boyd et al. (2000)
 GS1114CF7787 Δ(λatt-lom)::bla Φ(csrC-lacZ)1(hyb) AmpR KanSThis study
 KSB837CF7787 Δ(λatt-lom)::blaΦ(csrB-lacZ) 1(hyb) AmpR KanS Gudapaty et al. (2001)
 KSGA18CF7789 Φ(glgA-lacZ) (λplacMu15)KanR Gudapaty et al. (2001)
 MG1655PrototrophicM. Cashel
 RGGS1114GS1114 csrB::camRThis study
 RGKSB837KSB837 csrB::camR Gudapaty et al. (2001)
 RGKSGA18KSGA18 csrB::camR Gudapaty et al. (2001)
 RGMG1655MG1655 csrB::camR Gudapaty et al. (2001)
 RGTWKSGA18KSGA18 csrB::camR csrC::tetRThis study
 RGTWMG1655 csrB::camR csrC::tetR This study
 RHMG1655 rpoS::tn10 Wei et al. (2000)
 TR GS1114GS1114 csrA::kanRThis study
 TRKSB837KSB837 csrA::kanR Gudapaty et al. (2001)
 TRMG1655 csrA::kanR Romeo et al. (1993)
 TRTWMG1655 csrA::kanR csrC::tetR This study
 TWKSB837KSB837 csrC::tetRThis study
 TWKSGA18KSGA18 csrC::tetRThis study
 TWMG1655 csrC::tetR This study
 TWRG1113GS1114 csrC::tetRThis study
 UYKSB837KSB637 uvrY::camR Suzuki et al. (2002)
 UYMG1655MG1655 uvrY::camR Suzuki et al. (2002)
 UYRGS1114GS1114 uvrY::camRThis study
 pBR322Cloning vector, source of tetR marker Ausubel et al. (1989)
 pUY14 uvrY in pBR322, TetR Suzuki et al. (2002)
 pCCZ1pGE593 Φ(csrC-lacZ)This study
 pCRA16 csrA in pBR322, TetR Suzuki et al. (2002)
 pCSRC1Transcribed region of csrC oriented downstream from the lac promoter in pUC18This study
 pCSRCRe1Transcribed region of csrC oriented opposite of the lac promoter in pUC18This study
 pEL1Partial csrC clone (+1 to +209) in pT218UThis study
 pGE593Vector for lacZ transcriptional fusions, AmpR Eraso and Weinstock (1992)
 pIM5Partial csrB clone (+1 to +337) in pT218U Baker et al. (2002)
 pLG339Low-copy cloning vector TetR KanR Stoker et al. (1982)
 pMR2113Clone of csrC region in low-copy plasmid pLG339, KanR Romeo et al. (1991)
 pMR2113-D1 csrC HindIII fragment from pMR2113 in the HindIII site of pUC19This study
 pPB77 B. subtilis trp leader clone in pT218U Babitzke et al. (1994)
 pSPT18Transcription vector with SP6 and T7 promoters, AmpRBoehringer Mannheim
 pSPT18-CsrB csrB cloned into pSPT-18 behind SP6 promoter Gudapaty et al. (2001)
 pSPT18-D1 csrC cloned into pSPT-18 behind T7 promoterThis study
 pTZ18UCloning vector for generating in vitro transcripts with T7 RNA polymeraseUS Biochemical
 pUC18Cloning vector, AmpR Ausubel et al. (1989)
 pCR-XL-TOPOCommercial cloning vector (Invitrogen)Sigma Chemical
 λInCh1For genomic insertions, KanR Boyd et al. (2000)
 P1virStrictly lytic P1C. Gross
Figure 2.

Regulatory effects of csrC on glycogen levels, glgCA′∠′lacZ expression and biofilm formation.
A. Glycogen levels in WT E. coli K-12 (MG1655) and isogenic csrA (TRMG1655), csrB (RGMG1655), csrC (TWMG1655) and csrB csrC (RGTWMG1655) mutants, determined by iodine staining. Also shown is the csrC mutant containing the plasmid pCSRC1 (csrC+++), which overexpresses csrC.
B. Expression of a chromosomal glgCA′∠′acZ translational fusion. Cultures of KSGA18 (WT), RGKSGA18 (csrB), TWKSGA18 (csrC ), RGTWKSG18 (csrB csrC) and KSGA18[pCsrC1] were grown in Kornberg medium and assayed for specific β-galactosidase activity (A420 mg−1 protein). Error bars denote standard deviation and are not visible if this value was less than the area occupied by a given symbol. This experiment was conducted twice with essentially the same results.
C and D. Effects of csr genes on biofilm formation. Bars depict the means and the standard errors from two independent experiments with eight replicates per strain in each experiment. Asterisks denote statistically significant differences with respect to isogenic control strains, WT (MG1655) in (C) or its csrC mutant (TWMG1655) containing the vector pUC18 in (D) (P < 0.01). Plasmids contain a promoterless csrC gene in the forward (pCSR1) or reverse (pCSRre1) orientation with respect to the lac promoter of the vector.

Disruption of csrC in a csrB wild-type strain did not affect expression of a glgCA′–′lacZ translational fusion (Fig. 2B). However, csrC disruption in a csrB mutant strain background decreased the expression of this gene fusion up to ≈ 50%, consistent with its modest effect on glycogen levels in a csrB mutant background. Finally, pCSRC1, which expressed csrC from the lac promoter, stimulated glgCA expression two- to threefold (Fig. 2B).

CsrA is a repressor of biofilm formation, whereas CsrB activates this process (Jackson et al., 2002). The effect of csrC on biofilm formation was measured using a microtitre plate assay (Jackson et al., 2002). Disruption of csrC modestly decreased biofilm formation, an effect that was determined to be statistically significant (Fig. 2C). As observed previously (Jackson et al.,2002), biofilm formation was reduced in a csrB mutant. A csrB csrC double mutant was further compromised for biofilm formation and produced only 10% of the biofilm of the wild-type strain. In contrast, increased gene dosage of csrC led to several-fold greater accumulation of biofilm (Fig. 2D), an effect that was dependent upon csrC transcription. Recall that csrC lacks a promoter for expression from pCSRCre1. Thus, CsrC RNA activates biofilm formation in E. coli, similar to CsrB.

CsrA is required for the motility of E. coli under a variety of conditions, e.g. in tryptone medium (Wei et al., 2001). The effects of csrC disruption and increased copy number on motility were examined in tryptone medium. Similar to its isogenic parent, the csrC null mutant of MG1655 was fully motile (data not shown). In contrast, increasing the copy number of csrC by introducing pCSRC1, which contains csrC downstream from the lac promoter, completely inhibited motility, whereas no effect was observed using pCSRre1, in which csrC lacked a promoter (data not shown). These results revealed that overexpression of csrC inhibits motility, an effect that is opposite to that of CsrA (Wei et al., 2001).

In vitro interactions of CsrA with CsrB and CsrC RNAs

CsrB was discovered when it was co-purified with His-tagged CsrA (Liu et al., 1997). Subsequent genetic experiments demonstrated that CsrB functions as an antagonist of CsrA action (Liu et al., 1997; Gudapaty et al., 2001). CsrB contains 18 repeated sequences that are presumed to be CsrA binding sites (Liu et al., 1997). To characterize further the interaction of CsrA with CsrB transcript, we performed quantitative gel mobility shift assays with a CsrB transcript containing all 18 of the putative CsrA binding sites (+1 to +337 relative to the start of csrB transcription). CsrA binding was detected as a distinct band in native gels between 0.1 and 0.8 nM CsrA (Fig. 3A). As the concentration of CsrA was increased further, additional shifted species were observed. This gel shift pattern suggested that multiple CsrA molecules were bound to each CsrB transcript at the higher CsrA concentrations. Furthermore, as small twofold increases in CsrA concentration gave rise to new shifted species, our results suggest that the formation of these complexes is co-operative. As a total of four distinct CsrA–CsrB RNA complexes were observed before all the free RNA was shifted (lane with 3.2 nM CsrA), we were unable to use a non-linear least-squares analysis of these data to derive an equilibrium binding constant for this reaction. However, visual inspection of the gel suggested that the Kd value is between 0.8 and 1.6 nM CsrA. Although we presume that the first shifted species contained one CsrB transcript and one molecule of CsrA, and that the additional shifted species contained multiple CsrA molecules, the stoichiometry of these species has not been examined. The specificity of the CsrA–CsrB interaction was investigated by performing competition experiments with specific (CsrB RNA) and non-specific (Bacillus subtilis trp leader) unlabelled RNA competitors (Fig. 3B). The finding that unlabelled CsrB transcripts competed for CsrA binding while B. subtilis trp leader RNA did not compete demonstrates that CsrA binds specifically to CsrB.

Figure 3.

RNA gel mobility shift analyses of CsrA–CsrB and CsrA–CsrC binding. 5′ end-labelled CsrB (10 pM) or CsrC (0.2 nM) was incubated with the concentration of CsrA indicated at the bottom of each lane. Gel shift assays were performed in the absence or presence of various competitor RNAs. The concentration of competitor RNA is shown at the bottom of each lane.
A. Labelled CsrB without competitor.
B. Labelled CsrB with either specific (E. coli CsrB) or non-specific (B. subtilis trp leader) RNA competitors.
C. Labelled CsrC without competitor.
D. Labelled CsrB with either specific (E. coli CsrC or CsrB) or non-specific (B. subtilis trp leader) competitors. The positions of free (F) and bound (B) RNA are shown.

Similar gel mobility shift and competition experiments were conducted with CsrC. The CsrC transcript used in this analysis contained all nine of the putative CsrA binding sites (+1 to +209 relative to the start of csrC transcription). CsrA binding to this transcript was detected as a distinct band in native gels between 2.5 and 10 nM CsrA (Fig. 3C). A non-linear least-squares analysis of these data yielded an apparent Kd value of 8.7 ± 0.6 nM CsrA with a co-operativity coefficient (n) of 2.0. The positive co-operativity value indicates that interaction of CsrA with the CsrC transcript is co-operative. These results also demonstrate that the affinity of CsrA for CsrB is ≈ 10-fold higher than its affinity for CsrC. As observed for CsrB, additional shifted complexes were formed as the concentration of CsrA increased. This gel shift pattern suggested that multiple CsrA molecules were bound to each CsrC transcript at the higher CsrA concentrations. Although the stoichiometry of these species has not been examined, we presume that the first shifted species contained one CsrC transcript and one molecule of CsrA, and that the additional shifted species contained multiple CsrA molecules. The specificity of the CsrA–CsrC interaction was investigated by performing competition experiments with specific (CsrB or CsrC) and non-specific (B. subtilis trp leader) unlabelled RNA competitors (Fig. 3D). The finding that CsrB was a more effective competitor than CsrC is consistent with previous observations that CsrA has a higher affinity for CsrB. CsrA–CsrC RNA complex formation was not competed by B. subtilis trp leader RNA. These results establish that CsrA binds specifically to CsrC.

Regulation of CsrC levels and stability

As CsrC is related to CsrB in its ability to bind to and antagonize CsrA, we were interested to determine whether the expression of the csrC and csrB genes also shares regulatory features. CsrC and CsrB levels were measured at 2 h post-exponential phase in a series of isogenic strains varying in csrA, csrB, csrC or uvrY (Fig. 4). This experiment showed that the CsrB and CsrC riboprobes were specific, as no signal was detected from the csrB mutant with the CsrB probe or from the csrC mutant with the CsrC probe. Furthermore, CsrC accumulation was found to depend upon both CsrA and UvrY (Fig. 4B), although CsrC levels were somewhat less sensitive than those of CsrB (Fig. 4A). Interestingly, an ≈ 30% increase in CsrB levels was noted in cells lacking CsrC RNA (Fig. 4A). Likewise, in the absence of CsrB, CsrC transcript levels were similarly elevated (Fig. 4B). Although these effects were modest, they were reproducible in two independent experiments. We suspect that the compensatory effects of the two sRNAs result from increased intracellular availability of CsrA in the absence of one or the other of its sRNA antagonists.

Figure 4.

Regulation of CsrC levels and stability.
A. Northern analysis of CsrB RNA in MG1655 (WT) or isogenic csrA, csrB, csrC or uvrY mutant.
B. Northern blot of CsrC RNA from the same strains. Results of phosphorimager analysis are indicated as a percentage of the wild-type strain (%WT) or as barely detectable (BD). Essentially identical results were obtained in a second independent experiment (data not shown). The uvrY mutant was UYRMG1655, and other strain designations were given in the legend to Fig. 2A.
C. CsrC RNA chemical decay in MG1655 (triangles) and its isogenic csrA mutant (squares) was monitored by Northern blot and phosphorimager analysis, as described in Experimental procedures, and plotted on a semi-log scale. This experiment was performed twice with essentially identical results.

The half-life (≈ 2 min) of CsrC was not altered in a csrA mutant (Fig. 4C). Because the levels of any RNA molecule are determined by its rates of synthesis and turnover, this result strongly suggested that CsrA affects CsrC synthesis.

The results of the Northern hybridization were confirmed using csrC–lacZ and csrB–lacZ gene fusions (Fig. 5). The csrC and csrB (Gudapaty et al., 2001) fusions were designed to contain the upstream sequences and only 4 bp of the DNA templates of these two genes. Thus, transcripts synthesized from these fusions only contain four CsrB or CsrC nucleotides, and therefore lack CsrA binding elements. The expression of each gene fusion was reduced in csrA and uvrY mutant strains, indicating that the effects of CsrA and UvrY on both csrC and csrB expression were mediated at the level of transcript initiation. These experiments also revealed a modest, but reproducible, increase in csrC–lacZ expression in a csrB mutant, as well as an increase in csrB–lacZ expression in the csrC mutant. Thus, the compensatory effects of these two RNAs, which were first noted in Northern hybridization analyses (Fig. 4), were mediated at the level of transcription.

Figure 5.

Regulation of csrB–lacZ and csrC–lacZ expression.
A. Effects of mutations in csrA (TRKSB837), csrB (RGKSB837), csrC (TWKSB837), both csrB and csrC (csrBC; RGTWKSB837) or uvrY (UYKSB837) on csrB–lacZ expression in the KSB837 (WT) genetic background.
B. Effects of mutations in csrA (TRGS1114), csrB (RGGS1114), csrC (TWGS1114), both csrB and csrC (csrBC; RGTWGS1114) or uvrY (UYGS1114) on csrC–lacZ expression in the GS1114 (WT) genetic background. Error bars denote standard deviation and are not visible if this value was less than the area occupied by a given symbol.
C. Effects of ectopic expression of csrA (pCRA16) and uvrY (pUY14) on a csrC–lacZ transcriptional fusion in isogenic csrA (TRGS1114) or uvrY (UYRGS1114) mutants. The vector control in each case was pBR322. Cultures were grown to 2 h post-exponential phase, and specific β-galactosidase activity was determined as the average of duplicate samples. This experiment was repeated, with essentially identical results.

Purified CsrA had no effect on the transcription–translation of the pCCZ1-encoded csrC–lacZ transcriptional fusion in S-30 extracts, whereas recombinant UvrY stimulated expression twofold (data not shown). Expression of csrC–lacZ exhibited a linear dose–response up to 13 µM UvrY protein, which was the highest concentration that was tested because of technical constraints. The latter response was weaker than that of csrB–lacZ expression (Gudapaty et al., 2001; data not shown), consistent with the in vivo results (Figs 4 and 5).

Epistasis studies were conducted to determine whether the effect of CsrA on csrC expression depends upon UvrY. Ectopic expression of csrA or uvrY from a multicopy plasmid restored the defect in csrC–lacZ expression that was caused by a csrA mutation (Fig. 5C). In contrast, only uvrY restored a uvrY defect; csrA did not stimulate csrC–lacZ expression in the uvrY mutant (Fig. 5C). In conjunction with the results from Northern hybridization (Fig. 4B), gene fusion assays (Fig. 5), CsrC transcript stability (Fig. 5B) and in vitro transcription–translation studies, these experiments demonstrate that activation of csrC transcription by CsrA depends at least in part on UvrY. In addition, UvrY is positioned immediately upstream from csrC in the signalling pathway from CsrA to CsrC.

Predicted CsrC secondary structure and apparent homologues in related species

A striking feature of CsrB RNA is the presence of 18 conserved sequences, located primarily in the loops of predicted hairpins or other single-stranded regions (Liu et al., 1997). The CsrA–CsrB complex contains ≈ 18 subunits of CsrA (Liu et al., 1997). Furthermore, the repeated sequences of CsrB resemble a high-affinity CsrA binding site of glgC mRNA (Baker et al., 2002), strongly suggesting that these sequences are binding sites for CsrA. Not surprisingly, CsrC RNA contains nine similar repeated sequence elements (Fig. 1A), which tend to be located in predicted single-stranded loops or bulges of the molecule (Fig. 6). The apparent terminator of CsrC RNA contains a 16 nt inverted repeat, in which the oligo-U sequence of the terminator is complementary to an oligo-A sequence (Fig. 6), a feature distinct from that of CsrB (Liu et al., 1997).

Figure 6.

Predicted secondary structure of CsrC RNA. Secondary structure was predicted using mfold, as described in Experimental procedures. Imperfect repeat sequences that resemble the CsrB-type repeats are shown in red; predicted basepairing interactions are indicted by filled circles.

blast analysis revealed apparent csrC homologues in several Enterobacteriaceae, Klebsiella pneumoniae, E. coli O157:H7 and Salmonella enterica, typhi and paratyphi (data not shown). The CsrC homologue of K. pneumoniae was the most distinct from that of E. coli K-12 and exhibited 75% identity. CsrC homologues were not apparent in other eubacterial families, perhaps because CsrC function requires limited sequence conservation.


The CsrA–CsrB system of E. coli is a global regulatory system that has profound effects on metabolism, physiology and multicellular behaviour. We now show that a second sRNA molecule, CsrC, interacts with the RNA-binding protein CsrA (Fig. 3), albeit at lower affinity than CsrB, and antagonizes the regulatory effects of CsrA (Fig. 2). Although somewhat smaller than CsrB RNA, CsrC appears to use a similar mechanism for antagonizing CsrA activity. It contains nine related repeat sequences mainly in predicted single-stranded loops and bulges, which resemble the repeated elements of CsrB (Fig. 6). Other than these repeated elements, CsrB and CsrC sequences exhibit no striking similarity. The riboprobe for each RNA was completely specific (Fig. 4), and blast analyses with each RNA failed to identify the other (data not shown). Interestingly, CsrC RNA was detected in a comparative genomics approach search for sRNAs (Wassarman et al., 2001). That study found that, unlike many sRNAs, notably those that regulate gene expression by basepairing mechanisms, CsrC RNA does not bind to Hfq protein.

CsrA, CsrB and CsrC accumulate as the culture enters the stationary phase of growth (Gudapaty et al., 2001; Fig. 1). Although the sigma factor RpoS (σs) activates many genes that are induced in the stationary phase (Hengge-Aronis, 1996), it does not affect the levels of Csr components (Gudapaty et al., 2001; Fig. 1). In contrast, csrB and csrC transcription depends indirectly upon CsrA (Gudapaty et al., 2001; Figs 4 and 5). Epistasis studies revealed that the response regulator UvrY mediates these effect of CsrA (Suzuki et al., 2002; Fig. 5C). CsrC levels are less dependent upon uvrY than those of CsrB, and csrC–lacZ expression exhibits weaker activation by UvrY (Figs 4 and 5; data not shown). The significance of this minor regulatory distinction is uncertain. Furthermore, the  physiological stimulus for UvrY phosphorylation (Pernestig et al., 2001) remains to be determined.

Activation of CsrB and CsrC synthesis by CsrA defines an autoregulatory mechanism for CsrA, as each sRNA also antagonizes CsrA activity. One apparent consequence of this mechanism is that CsrB and CsrC exhibit compensatory effects on each other (Fig. 4). Interestingly, CsrC effects on glycogen levels and glgCA′–′lacZ expression were only observed in a csrB mutant background and were modest (Fig. 2). In contrast, csrC overexpression had more dramatic effects on both processes. Biofilm formation also exhibited greater effects on csrC overexpression versus disruption (Fig. 2C). At least two factors may contribute to these quantitative discrepancies. First, the compensatory effects of CsrB and CsrC on each other would tend to minimize the effects of either single mutation. Secondly, under the culture conditions used for these experiments, there is sufficient CsrB in the cell to sequester only ≈  30% of the CsrA protein (Gudapaty et al., 2001). Although we did not determine the absolute intracellular levels of CsrC in the present study, previous characterization of the purified CsrA ribonucleoprotein complex revealed that CsrB is the major RNA component (Liu et al., 1997). Thus, we expect intracellular CsrC to bind < 30% of the CsrA in the cell and, on this basis, a single disruption of csrC should be expected to have minimal effects on CsrA-regulated gene expression. Of course, it is possible that csrC disruption might exhibit more pronounced regulatory effects under other growth conditions.

Why E. coli and related Enterobacteriaceae should express two highly similar sRNAs is not known. Although csrB and csrC were regulated similarly in our experiments, it is possible that they may be differentially expressed under certain conditions. If so, the involvement of two sRNAs would increase the flexibility of the Csr signalling circuitry. The half-lives of both CsrB and CsrC are relatively short (≈ 2 min), and are more similar to those of mRNAs than other sRNAs (reviewed by Wassarman et al., 1999). As discussed previously (Gudapaty et al., 2001), a short half-life should allow CsrA activity to respond rapidly to conditions that alter CsrB or CsrC levels. CsrA plays an important role in directing central carbon flux (Sabnis et al., 1995; Yang et al., 1996; Wei et al., 2000; Tatarko and Romeo, 2001), thus CsrB and CsrC must help to fine-tune carbon flow. The sophisticated role of the Csr system in controlling carbon flux is exemplified by the observation that CsrA differentially regulates the individual isozymes of the phosphofructokinase and pyruvate kinase reactions (Sabnis et al., 1995). Because these two reactions are each catalysed by pairs of isozymes that respond individually to distinct allosteric ligands, such findings indicate that the Csr system fine-tunes the allosteric regulation of glycolysis (Sabnis et al., 1995). Although no hard evidence exists, it is conceivable that CsrB or CsrC possesses a CsrA-independent function. Some sRNAs use distinct structural domains to permit interaction with different messages, e.g. DsrA interactions with rpoS and hns messages (Majdalani et al., 1998; Lease and Belfort, 2000) and OxyS interactions with rpoS and flhA mRNAs (Altuvia et al., 1998). The apparent CsrA binding sites of CsrC tend to be clustered towards the 5′ segment of CsrC RNA (Fig. 1). Perhaps the 3′ segment of CsrC may possess a regulatory function that is distinct from sequestration of CsrA. In any case, our findings indicate that there is selective pressure to maintain two CsrA-binding sRNAs in E. coli and related bacteria, which argues that CsrA activity in the cell is tightly regulated.

CsrA (RsmA)-binding RNAs are the only examples of bacterial sRNAs that bind to multiple copies of and antagonize a regulatory protein. What factors might have predisposed the Csr system to evolve in this way? First, CsrA must have existed before its two sRNA antagonists and thus influenced their design. Features of CsrA that may have favoured the evolution of multiple binding sites within CsrB and CsrC are its own small size (61 amino acids) and its relatively small RNA target sequence (Baker et al., 2002). Thus, a limited target site for CsrA binding could be repeated at relatively short intervals to produce a small RNA capable of sequestering many protein subunits. In addition, CsrA binding exhibits co-operatively (Fig. 3), which may facilitate sequestration of CsrA within compact ribonucleoprotein complexes (e.g. Liu et al., 1997). The Csr system is noted for regulating stationary phase processes, which operate when resources are limiting. The synthesis of a small RNA regulator requires less energy than translation of a regulatory protein, and the Csr system further amplifies this advantage, as a single sRNA transcript can determine the availability of many CsrA molecules (discussed by Romeo, 1998). Interestingly, many sRNAs of E. coli accumulate during the stationary phase of growth (Argaman et al., 2001; Wassarman et al., 2001), suggesting that preservation of resources may provide selective pressure for using sRNAs in regulation. Only additional studies will determine whether similar selective conditions have permitted the independent evolution of other sRNAs that function by sequestering regulatory proteins.

Experimental procedures

Strains, plasmids and phage

Bacterial strains, plasmids and phage used in this study are listed in Table 1.

Media and growth conditions

LB medium (Miller, 1972) with 0.2% glucose was used for routine cultures. SOC medium (Miller, 1972) was used for recovery of transformed cells. Kornberg medium (1.1% K2HPO4, 0.85% KH2PO4, 0.6% yeast extract containing 0.5% glucose for liquid or 1% glucose for agar) was used for gene fusion assays, Northern blot and RNA stability studies and assessment of the glycogen phenotype by iodine staining. Semi-solid tryptone medium (pH 7.4) containing 1% tryptone, 0.5% NaCl and 0.35% agar was used for motility studies (Wei et al., 2001). Colonization factor antigen (CFA) medium (pH 7.4) (Evans et al., 1997) contained 1% casamino acids, 0.15% yeast extract, 0.005% MgSO4 and 0.0005% MnCl2, and was used to grow cultures for biofilm studies. Antibiotics were added at the following concentrations: chloramphenicol, 20 µg ml−1; kanamycin, 100 µg ml−1; ampicillin, 100 µg ml−1; tetracycline, 10 µg ml−1; rifampicin, 200 µg ml−1, except that ampicillin and kanamycin were used at 50 µg ml−1 and 40 µg ml−1 during the construction of the chromosomal csrC–lacZ fusion. Liquid cultures were grown at 37°C with rapid shaking, unless otherwise noted.

Molecular biology

Standard procedures were used for plasmid isolation, restriction digests, ligations, transformation and transduction of antibiotic markers (Miller, 1972; Ausubel et al., 1989).

RNA isolation

Total cellular RNA was isolated using the Masterpure™ RNA purification kit (Epicentre), quantified by UV absorbance and suspended in 70% ethanol at −80°C.

Primer extension

Total RNA was harvested from a culture grown to the transition to stationary phase in Kornberg medium. The oligonucleotide primer D1prB (Table 2) was labelled using T4 polynucleotide kinase and [γ-32P]-ATP (3000 Ci mmol−1; NEN Life Science Products) according to standard procedures (Ausubel et al., 1989). Approximately 15 ng of labelled primer was annealed to 10 µg of RNA. cDNA was synthesized using 15 U of ThermoScript RT (Invitrogen) in a 20 µl reaction mixture, incubated for 60 min at 48°C and terminated for 5 min at 85°C. RNA was degraded with 2 U of RNase H for 20 min at 37°C. The same labelled primer was used to prepare a DNA sequencing ladder, with pMR2113-D1 as the template, which served as a standard for the RT product. Products were analysed on 6% polyacrylamide gels containing 6 M urea (e.g. Romeo and Preiss, 1989).

Table 2. . Oligonucleotide primers used in this study. a
PrimerSequence (5′ to 3′)
  • a

    . Primers were purchased from Integrated DNA Technologies.


Riboprobe synthesis

The csrB riboprobe was produced from plasmid pSPT18-CsrB as described elsewhere (Gudapaty et al., 2001). The plasmid for the production of the csrC riboprobe was generated by subcloning a 209 bp NsiI–KpnII fragment from pMR2113-D1 into the multiple cloning site of pSPT18. The resulting plasmid, pSPT18-D1, was used to generate digoxigenin (DIG)-labelled riboprobe, using T7 RNA polymerase and the DIG-RNA labelling kit (SP6/T7), according to the manufacturer's instructions (Roche Diagnostics). The synthesis reaction was carried out for 2 h at 37°C, followed by 15 min incubation with 2 µl of RNase-free DNase I. The reaction was subsequently terminated with the addition of 2 µl of 0.2 M EDTA. Probes were stored at −80°C.

Northern hybridization

Total cellular RNA (5 µg) was separated on formaldehyde agarose (1%) gels. Before blotting, gels were stained with ethidium bromide, photographed, and 23S RNA was digitally quantified using molecular analyst software (version 2.1.2). RNA was transferred overnight onto positively charged nylon membranes (Boehringer Mannheim) in 20× SSC, and immobilized by baking at 120°C for 30 min (Sambrook et al., 1989). Prehybridization, hybridization to DIG-labelled riboprobes (2 µl of probe per 10 ml of prehybridization buffer) and membrane washing were conducted using the DIG luminescent detection kit for nucleic acids (Roche Diagnostics), according to the manufacturer's instructions, except that the membrane was incubated for 10 h in blocking solution. The resulting chemiluminescent signals were detected using Kodak X-OMAT-AR film and quantified by phosphorimaging using a GS-525 Phosphor Imager (Bio-Rad) with a chemiluminescent screen. Phosphorimaging data were analysed using molecular analyst software and Microsoft excel. The 23S rRNA signal was used to normalize for minor loading differences between samples.

Construction of a csrC null mutant

A linear DNA fragment was amplified by polymerase chain reaction (PCR), which contained the tetR gene from pBR322 flanked on either side by 40 nt homologous to the upstream and downstream regions of csrC. Primers used in generating the fragment were D1KOF and D1KOR (Table 2). Linear recombination of this fragment into the E. coli genome was performed using the protocol described elsewhere (Datsenko and Wanner, 2000). Cells were then plated on selective Kornberg medium containing tetracycline. Overnight colonies were chosen for confirmation by PCR and Northern blot. PCR was performed using 15 µl of whole cells harvested from overnight cultures. Cells were resuspended in 10 µl of deionized water and heated for 5 min at 94°C (http:www.protocol-online.org). The resulting lysate was used as the DNA template to generate a PCR product using primers D1serB and D1Check1 (Table 2). The identity of the resulting PCR product was confirmed by restriction digestion with EcoRV, BamHI and SalI.

Construction of a minimal csrC clone

The transcribed region of csrC was amplified by PCR using primers D1Pr1 and D1Pr2 (Table 2). This PCR product was T-A cloned into the pCR-XL-TOPO cloning vector (Invitrogen). The csrC region was excised from this clone using EcoRI and subcloned into the EcoRI site of pUC18. Clones containing csrC in the forward or reverse orientation with respect to the lacZ promoter of pUC18 were designated pCSRC1 and pCSRCre1 respectively. Plasmid DNA inserts were sequenced at the University of Arizona core facility using primers D1Pr1 and D1Pr2.

Construction of a chromosomal csrC–lacZ transcriptional fusion

A 243 bp PCR product was prepared, which contained the 3′ end of the yihA gene, the upstream untranscribed region of csrC and the first four transcribed nt of csrC. The primers csrC-UP and csrC-DN (Table 2) were used for generating this PCR product. The PCR product was gel purified, treated with T4 DNA polymerase to create blunt ends and cloned into SmaI-treated and dephosphorylated pGE593 plasmid. The resulting plasmid, pCCZ1, was partially sequenced and found to be free of PCR-generated mutations. The csrC–lacZ fusion in pCCZ1 was moved into the E. coli CF7789 chromosome using the λInCh1 system, as described elsewhere (Boyd et al., 2000). The resulting strain that was chosen for subsequent studies, GS1114, was Ampr Kans and no longer temperature sensitive. The presence of the csrC–lacZ transcriptional fusion in this strain was confirmed by PCR analysis, as recommended (Boyd et al., 2000).

RNA gel mobility shift assay

Plasmid pEL1, which contains nt +1 to +209 relative to the start of csrC transcription, was constructed by cloning a chromosomally generated PCR product into the EcoRI and BamHI sites of the pTZ18U polylinker. Quantitative gel shift assays were performed according to previously published procedures (Yakhnin et al., 2000; Baker et al., 2002). RNA was synthesized in vitro using the Ambion MEGAscript kit and linearized plasmids pEL1, pIM5 (Baker et al., 2002) or pPB77 (Babitzke et al., 1994) as templates (Table 1). Gel-purified RNA was 5′ end-labelled with [γ-32P]-ATP as described previously (Yakhnin et al., 2000). RNA suspended in TE was renatured by heating to 85°C followed by slow cooling. Binding reactions (10 µl) contained 10 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 100 mM KCl, 32.5 ng of yeast RNA, 7.5% glycerol, 20 mM dithiothreitol (DTT), 4 U of RNase inhibitor (Ambion), 10 pM CsrB RNA or 0.2 nM CsrC RNA, purified CsrA (various concentrations) and 0.1 µg ml−1 xylene cyanol. Assays were also carried out in the presence of various unlabelled RNA competitors (see text for details). Reaction mixtures were incubated for 30 min at 37°C to allow CsrA–RNA complex formation. Samples were then fractionated on native 8% polyacrylamide gels for CsrB RNA and 10% gels for CsrC RNA. Radioactive bands were visualized using a phosphorimager. Free and bound RNA species were quantified using imagequant, version 5.2 (Molecular Dynamics), and the apparent equilibrium binding constants (Kd) of CsrA–RNA complexes were calculated as described previously (Yakhnin et al., 2000).

RNA decay analysis

The transcription inhibitor rifampicin was added to cultures at 2 h post-exponential phase. Cultures were harvested at regular intervals thereafter, and total cellular RNA was isolated. CsrC RNA was analysed by Northern blot and phosphorimaging analysis (Gudapaty et al., 2001).

Glycogen, β-galactosidase, total protein and motility assays

Glycogen accumulation was examined by staining colonies with iodine vapour (Liu et al., 1997). β-Galactosidase activity was assayed as described previously (Romeo et al., 1990). Total protein was assayed with bicinchoninic acid using bovine serum albumin as a standard (Smith et al., 1985). Motility was assessed on tryptone semi-solid medium (Wei et al., 2001).

In vitro transcription–translation

Effects of CsrA and UvrY proteins on csrC–lacZ expression were examined using S-30 extracts prepared from a uvrY mutant strain (UYCF7789), as described previously (Romeo and Preiss, 1989; Suzuki et al., 2002), except that reaction volumes were reduced to 28 µl. Radiolabelled proteins were separated by SDS-PAGE and detected by fluorography using sodium salicylate (Chamberlain, 1979). Methionine incorporation into the LacZ polypeptide was determined by densitometry with the aid of molecular analyst (version 2.1.2) software and Microsoft excel.

Quantitative biofilm assay

Cultures were grown for 24 h at 26°C in microtitre plates, and biofilm formation was monitored using crystal violet staining, as described previously (Jackson et al., 2002). The experiment was conducted twice, with eight replicates per sample in each trial. The data were analysed by Tukey multigroup analysis (StatView-SAS Institute).

Sequence and secondary structure analysis of CsrC

The location of csrC on the E. coli K-12 genome (Blattner et al., 1997) and csrC homologues were recognized by blast  analyses (Altschul et al., 1990), courtesy of the National Center for Biotechnology Information (http:www.ncbi.nlm.nih.gov). Secondary structure predictions for CsrC RNA were generated with mfold, which uses an algorithm for free energy minimization (Zuker et al., 1999; http:bioweb.pasteur.frseqanalinterfacesmfold-simple.html). De-fault parameters were used in all fields for the predictions, and the structure with the lowest predicted free energy is presented here.


We thank Eric Lazer for constructing pEL1. This work was supported by the National Institutes of Health (GM-59969), National Science Foundation (MCB-9726197) and CONACyT (37342-N).