Global regulation by CsrA in Salmonella typhimurium



CsrA is a regulator of invasion genes in Salmonella enterica serovar Typhimurium. To investigate the wider role of CsrA in gene regulation, we compared the expression of Salmonella genes in a csrA mutant with those in the wild type using a DNA microarray. As expected, we found that expression of Salmonella pathogenicity island 1 (SPI-1) invasion genes was greatly reduced in the csrA mutant, as were genes outside the island that encode proteins translocated into eukaryotic cells by the SPI-1 type III secretion apparatus. The flagellar synthesis operons, flg and fli, were also poorly expressed, and the csrA mutant was aflagellate and non-motile. The genes of two metabolic pathways likely to be used by Salmonella in the intestinal milieu also showed reduced expression: the pdu operon for utilization of 1,2-propanediol and the eut operon for ethanolamine catabolism. Reduced expression of reporter fusions in these two operons confirmed the microarray data. Moreover, csrA was found to regulate co-ordinately the cob operon for synthesis of vitamin B12, required for the metabolism of either 1,2-propanediol or ethanolamine. Additionally, the csrA mutant poorly expressed the genes of the mal operon, required for transport and use of maltose and maltodextrins, and had reduced amounts of maltoporin, normally a dominant protein of the outer membrane. These results show that csrA controls a number of gene classes in addition to those required for invasion, some of them unique to Salmonella, and suggests a co-ordinated bacterial response to conditions that exist at the site of bacterial invasion, the intestinal tract of a host animal.


Originally identified in Escherichia coli, CsrA is a post-transcriptional regulator that alters the stability of its target messages. In E. coli, CsrA activates glycolysis, acetate metabolism and motility, while repressing gluconeogenesis and glycogen biosynthesis (Romeo et al., 1993; Liu et al., 1995; Liu and Romeo, 1997). CsrA has been shown to function as either a positive or a negative regulator, depending upon its target. The post-transcriptional regulation of two messages, those of glgCAP, which encodes glycogen synthetase, and flhDC, which encodes the central regulator of flagellar synthesis, has been studied in detail in E. coli (Romeo et al., 1993; Liu et al., 1995; Liu and Romeo, 1997; Wei et al., 2001; Baker et al., 2002). In the case of glgCAP, CsrA binds to the Shine–Dalgarno sequence and to a site in the glgCAP leader sequence, thereby blocking ribosomal binding and message translation (Baker et al., 2002). In contrast, CsrA binding to the flhDC transcript is thought to stabilize the message and allow its translation (Wei et al., 2001). The precise CsrA binding site within the flhDC message has not been determined fully, but it lies within the upstream untranslated region of the message. One hypothesis is that CsrA acts as a positive regulator by binding to its target message and protecting it from endonucleolytic attack (Wei et al., 2001). A second component of the csr control system is the untranslated RNA CsrB. In E. coli, ≈ 18 CsrA molecules can bind to a single molecule of CsrB, presumably sequestering the protein (Liu et al., 1997). CsrA is thought to bind CsrB at repeated sequence elements similar to the Shine–Dalgarno sequence (Liu et al., 1997). CsrA also activates transcription of csrB, possibly through indirect regulation of a transcription factor (Gudapaty et al., 2001). In E. coli, CsrB has a short half-life (≈ 2 min). The induction of CsrB by CsrA may serve to modulate CsrA activity by titrating the protein. This would result in tight control of the level of active CsrA and would provide an indirect mechanism by which CsrA could control its own activity, as studies have shown that CsrA does not directly control its own expression (Gudapaty et al., 2001).

Both csrA and csrB in Salmonella enterica serovar Typhimurium (hereafter S. typhimurium) share strong sequence homology with their counterparts in E. coli (Altier et al., 2000a,b). Although CsrA is a global regulator in E. coli that controls central functions such as carbon metabolism (Romeo et al., 1993; Liu et al., 1995; Sabnis et al., 1995; reviewed by Romeo, 1996; Wei et al., 2000; Baker et al., 2002), in S. typhimurium, it has been shown to regulate specialized virulence determinants not found in E. coli (Altier et al., 2000b). The ability of S. typhimurium to invade intestinal epithelial cells, an early step in pathogenesis, is encoded by a type III secretion system located at centisome 63 of the chromosome in Salmonella pathogenicity island 1 (SPI-1). Genetic and environmental control of SPI-1 invasion gene expression is multifactorial and is co-ordinated within the pathogenicity island by HilA (Bajaj et al., 1995; 1996). Regulators of HilA include those encoded within SPI-1, HilC (also known as SprA and SirC) and HilD, and regulators encoded outside SPI-1, PhoP/PhoQ and BarA/SirA (Behlau and Miller, 1993; Bajaj et al., 1995; Johnston et al., 1996; Ahmer et al., 1999; Altier et al., 2000a; Lucas and Lee, 2001). Both HilC and HilD are positive regulators of hilA expression (Schechter et al., 1999; Lostroh et al., 2000; Lucas and Lee, 2001). Altering the levels of CsrA results in decreased expression of SPI-1 invasion genes and a decreased ability of mutant bacteria to invade cultured epithelial cells (Altier et al., 2000b). Regulation of SPI-1 invasion genes by CsrAB is, however, complex. Loss of CsrA or a presumed reduction in free CsrA by the overexpression of CsrB decreases epithelial cell invasion and the expression of SPI-1 genes, but overexpression of CsrA or loss of CsrB have these same effects (Altier et al., 2000a,b). These findings indicate that the csr system of regulation can have both positive and negative effects on invasion and suggest that tight control of the level of active CsrA is required for optimal invasion. The csr system is also controlled by BarA/SirA, as barA and sirA are required for full expression of CsrB (B. Ahmer, personal communication; Lawhon et al., 2002).

Homologues of CsrA and CsrB are widespread among the eubacteria and regulate virulence in Pseudomonas and Erwinia species (White et al., 1996). The CsrA/CsrB homologues, RsmA (PrrB) and RsmB, respectively, in Pseudomonas fluorescens regulate the production of hydrogen cyanide, encoded by hcnABC, and the major exoprotease, encoded by aprA (Blumer et al., 1999; Heeb et al., 2002). In Pseudomonas aeruginosa, RsmA is a negative regulator of hydrogen cyanide production, stapholytic activity and N-acylhomoserine lactone synthesis required for quorum sensing (Pessi et al., 2001). The genes that encode these products are also under the control of the two-component regulator, GacS/GacA, homologous to BarA/SirA, which is required for the expression of PrrB, the CsrB homologue (Aarons et al., 2000). Consistent with this, overexpression of RsmA has effects similar to deletion of GacA, including increased expression of hcnABC in both P. fluorescens and P. aeruginosa (Blumer et al., 1999; Blumer and Haas, 2000; Pessi et al., 2001). Like its counterpart in E. coli, RsmA appears to act at the ribosomal binding site (RBS) of target sequences (Blumer et al., 1999; Pessi et al., 2001). In Erwinia carotovora, RsmA represses the production of virulence factors pectate lyase, cellulase and protease and regulates the production of N-acylhomoserine lactone (Chatterjee et al., 1995; Cui et al., 1995). Overexpression of RsmA in E. carotovora also inhibits motility and flagellar synthesis, although the effect of loss of rsmA on these functions has not been determined (Mukerjee et al., 1996).   Like   other   bacterial   species,   regulation   of   RsmA in Erwinia appears to be under the control of a two-component regulator, ExpS/ExpA (Cui et al., 2001; Hyytiainen et al., 2001). Unlike E. coli, however, in Erwinia, RsmA is proposed to degrade the message of RsmB rather than to stabilize it (Chatterjee et al., 2002).

Here, we investigate the global effects of CsrA in Salmonella. We use genomic analysis, using an S. typhimurium microarray, supported by phenotypic and genotypic characterization of a csrA mutant. As shown previously, we find that loss of csrA reduces the level of expression of SPI-1 invasion genes. We also find regulation of pathways not previously identified in other bacterial species, including those for maltose transport and ethanolamine utilization, and those not present in E. coli, such as propanediol metabolism, B12 synthesis and production of hydrogen sulphide.


Regulation of invasion genes by CsrA

To test the global effects on S. typhimurium gene expression induced by the loss of csrA, we used a DNA microarray that includes 4587 of the 4716 predicted open reading frames (ORFs; 97%) of S. typhimurium LT2, including its plasmid. We hybridized cDNA from a virulent wild-type S. typhimurium strain (ATCC 14028s) and the isogenic csrA mutant grown under the same conditions to the microarray and compared gene expression. Of the 4587 ORFs present on the array, 1814 were detected above the level of internal DNA controls. Of these, 142 were increased twofold or more in the mutant, and 223 were decreased twofold or more. We have also noted previously that the S. typhimurium csrA mutant exhibits a growth defect (Altier et al., 2000b). One concern in using microarray analysis was that differences in the growth rate between the csrA mutant and wild type might be responsible for differences in gene expression. To address this issue, we also tested a previously isolated suppressed csrA mutant that grows at a near normal rate but maintains its defect in epithelial cell invasion (Altier et al., 2000b). Microarray analysis showed that, in some cases, changes in the expression of individual genes differed between the suppressed and unsuppressed mutant, but that all gene classes described in this work as regulated by csrA were also affected in the suppressed csrA mutant. Therefore, our results do not merely reflect artifacts of altered growth rate.

We found genes of several regulons to have reduced expression in the mutant. One such class was the invasion genes of SPI-1. Of the 37 genes known to be part of this island, 27 were reduced in their expression by two- to 14-fold (Table 1). Among these were hilA (fivefold reduced), a central regulator of SPI-1, and invF (10-fold reduced), an SPI-1 regulator that is regulated both by hilA and independently of hilA. Also reduced 10-fold were SPI-1 genes encoding secreted effectors and secretion apparatus structural proteins, including sipABCD, invCBAG, sicP, sptP and prgHK. In addition to control of SPI-1 expression, genes outside the island but encoding proteins known to be translocated by the type III secretion system of SPI-1 were also reduced in their expression (Table 1). These included sopE2, reduced fivefold, and sopB, located in SPI-5 and reduced 10-fold. sopA, encoding another secreted effector protein required for enteropathogenesis, was also reduced in its expression, by threefold. An additional SPI-5 gene showed reduced expression, pipC, which exists in an operon with sopB. Conversely, pipA, required for enteropathogenesis and also encoded in SPI-5, had a threefold increase in its expression. Of the remaining 10 genes that are part of SPI-1, three were reduced in expression, but less than twofold, four were increased in their expression, and three were not detected above the level of non-specific controls on the array. The four genes with increased expression (sitABCD) encode a fur-regulated iron transporter known not to be under the control of SPI-1 regulatory elements. These results demonstrate that genome analysis using the Salmonella DNA microarray is supportive of our previous results showing that the loss of csrA greatly reduces the expression of SPI-1 invasion genes and epithelial cell invasion, and also show that csrA controls the expression of other virulence genes outside SPI-1.

Table 1. . Regulation of invasion genes by csrA.
GeneFunctionRatio of expression (csrA mutant/wild type)
 hilAInvasion genes transcription activator0.2
 hilCSPI-1 transcriptional regulator0.1
 hilDSPI-1 transcriptional regulator0.2
 iacPPutative acyl carrier protein0.4
 invAInvasion protein0.1
 invBSurface presentation of antigens; secretory proteins0.1
 invCSurface presentation of antigens; secretory proteins0.1
 invEInvasion protein0.2
 invFInvasion protein, transcriptional regulator0.1
 invGInvasion protein; outer membrane0.1
 invHInvasion protein0.5
 invISurface presentation of antigens; secretory proteins0.3
 invJSurface presentation of antigens; secretory proteins0.2
 orgAPseudogene; frameshift0.1
 prgHCell invasion protein0.1
 prgICell invasion protein; cytoplasmic0.2
 prgKCell invasion protein; lipoprotein, may link inner and outer membranes0.1
 sicASurface presentation of antigens; secretory proteins0.4
 sicPChaperone, related to virulence0.1
 sipACell invasion protein0.1
 sipBCell invasion protein0.1
 sipCCell invasion protein0.1
 sipDCell invasion protein0.1
 spaOSurface presentation of antigens; secretory proteins0.2
 spaPSurface presentation of antigens; secretory proteins0.5
 sprBTranscriptional regulator0.2
 sptPProtein tyrosine phosphate0.07
 sopB Salmonella outer protein: homologous to ipgD of Shigella0.1
 pipCPathogenicity island encoded protein; homologous to ipgE of Shigella0.2
 pipAPathogenicity island encoded protein3.4
Located outside Salmonella pathogenicity islands
 sopE2Type III-secreted protein effector: invasion-associated protein0.2
 sopASecreted effector protein of Salmonella dublin0.3

Regulation of flagellar synthesis and chemotaxis by CsrA

In Salmonella and E. coli, flagellar biosynthesis and chemotaxis/aerotaxis are co-ordinately regulated by flhDC (reviewed by Chilcott and Hughes, 2000). Flagellar genes are expressed in three stages, early, middle and late. The two early genes, flhD and flhC, form an operon through which environmental control of flagellar synthesis is co-ordinated. In E. coli, CsrA binds to and stabilizes flhDC mRNA (Wei et al., 2001), thus predicting a reduction in flhDC message in a csrA mutant. Microarray analysis showed no significant effect of the loss of csrA on flhD expression and only a slightly decreased level of flhC expression (a ratio of mutant to wild type of 0.7). However, genes regulated by flhDC, including flgM, flgK, flgL, fliAZ, fliDT, fliC, fljBA and motAB, were significantly decreased (at least threefold) in the csrA mutant (Table 2). Additional genes with twofold or less reduced expression included flgDEG, flgN, fliY, fliS and fliJL. These comprise both middle and late genes required for the synthesis and assembly of flagella, as well as transcriptional regulators of flagellar gene expression (flgM and fliA). Also regulated by flhDC are genes associated with chemotaxis/aerotaxis, cheA, tsr and aer. These were also decreased between three- and fivefold in the csrA mutant (Table 2). Mutants of the flagellar genes would be predicted to have profound defects in the synthesis of flagella and in motility. To confirm the findings of the microarray, we examined the wild type, the csrA mutant and the csrA mutant complemented with csrA on a low-copy-number vector for the presence of flagella by transmission electron microscopy (Fig. 1). The csrA mutant had no detectable flagella (Fig. 1B), whereas the complemented mutant had flagella indistinguishable from those of the wild type (Fig. 1C and A respectively). The csrA mutant was non-motile when grown on semi-solid LB agar (0.35% agar), a phenotype also restored to that of wild type in the complemented mutant (Fig. 1D). We also extracted outer membrane proteins from the wild type, the csrA mutant and the complemented mutant to assess the presence of flagellin. As shown in Fig. 2, two bands of apparent molecular weights 56 kDa and 53 kDa were missing in the mutant, but present in the wild type and complemented mutant. S. typhimurium phase 2 flagellin, encoded by fljB, and phase 1 flagellin, encoded by fliC, have these apparent molecular weights (Schmitt et al., 1996). Protein sequencing showed the proteins to have the amino-terminal sequence of AQXINTNSLS and XQVINTNS respectively. The predicted amino-terminal sequence of both phase 1 and phase 2 flagellin is MAQVINTNSLS, with no other Salmonella protein having a similar amino-terminal sequence. Thus, these outer membrane proteins missing in the csrA mutant represent the two phase variants of flagellin. Therefore, as in E. coli, S. typhimurium csrA is essential for the production of flagella and for motility.

Table 2. . Regulation of flagellar synthesis and chemotaxis by csrA.
GeneFunctionRatio of expression (csrA mutant/wild type)
Flagellar biosynthesis
 flgDInitiation of hook assembly0.4
 flgEHook protein (first module)0.4
 flgGCell-distal portion of basal body rod0.4
 flgKHook–filament junction protein 1 (second module)0.1
 flgLHook–filament junction protein0.2
 flgMAnti-FliA (antisigma) factor; also known as RflB protein0.3
 flgNBelieved to be export chaperone for FlgK and FlgL0.4
 fliASigma F (sigma 28) factor of RNA polymerase, transcription of late genes0.2
 fliCFlagellin, filament structural protein (second module)0.1
 fliDFilament-capping protein; enables filament assembly0.02
 fliTPossible export chaperone for FliD0.1
 fliZPutative regulator of FliA0.2
 fljARepressor of fliC0.3
 fljBPhase 2 flagellin (filament structural protein)0.02
 motAProton conductor component of motor0.1
 motBEnables flagellar motor rotation0.1
 aerAerotaxis sensor receptor, senses cellular redox state or proton motive force0.1
 cheASensory kinase, transduces signal between chemosignal receptors and CheB and CheY0.2
 cheBMethyl esterase0.5
 cheWPurine-binding chemotaxis protein; regulation0.3
 cheYChemotaxis regulator, transmits chemoreceptor signals to flagellar motor0.4
 cheZChemotactic response; CheY protein phosphatase0.4
 tsrMethyl-accepting chemotaxis protein I, serine sensor receptor0.2
Figure 1.

CsrA is required for production of flagella and for motility. Transmission electron microscopy was used to detect flagella on the wild type (A) and on the csrA mutant complemented with a plasmid carrying csrA(C), but the csrA mutant (B) had no detectable flagella.
D. The wild type (1) and the complemented csrA mutant were motile when grown on semi-solid (0.35%) LB agar, but the csrA mutant (2) was non-motile.

Figure 2.

Loss of csrA alters the expression of Salmonella outer membrane proteins. Three outer membrane proteins of apparent molecular weight 56, 53 and 48 kDa (arrowheads) were present in the wild type (lane 1) and the complemented csrA mutant (lane 3), but absent or in greatly reduced concentration in the csrA mutant (lane 2). Amino-terminal protein sequencing indicated these to be type 2 and type 1 flagellin and maltoporin respectively.

Regulation of genes associated with vitamin B12 synthesis and utilization

Unlike E. coli, S. typhimurium is able to synthesize vitamin B12de novo. B12 synthesis occurs under anaerobic conditions and requires the cob operon, including the cob and cbi genes that encode enzymes required for the production of the precursors to vitamin B12, adenosyl cobalamin and the corrin ring (reviewed by Roth et al., 1996). In the csrA mutant, we found two- to 10-fold reductions in the expression of genes required for the synthesis of vitamin B12, specifically genes within the cob operon, cbiACDFGHLKMPO and cobSTU(Table 3). We tested the regulation of cbi by csrA using a lacZ fusion in the cbi operon to measure β-galactosidase expression. We found a fourfold decrease in the expression of the cbi operon in the csrA mutant (Fig. 3C). This decrease in expression was restored to the wild-type level in the complemented mutant.

Table 3. . Regulation of vitamin B12 synthesis and of metabolic pathways requiring B12 by csrA.
GeneFunctionRatio of expression (csrA mutant/wild type)
Synthesis of vitamin B12 precursor
 cbiAAdenosyl cobalamide precursor0.4
 cbiCAdenosyl cobalamide precursor0.5
 cbiDAdenosyl cobalamide precursor0.5
 cbiFAdenosyl cobalamide precursor0.3
 cbiGAdenosyl cobalamide precursor0.2
 cbiHAdenosyl cobalamide precursor0.4
 cbiKAdenosyl cobalamide precursor0.3
 cbiLAdenosyl cobalamide precursor0.5
 cbiMAdenosyl cobalamide precursor0.4
 cbiPAdenosyl cobalamide precursor0.4
 cbiOAdenosyl cobalamide precursor0.4
 eutAChaperonin in ethanolamine utilization0.5
 eutBEthanolamine ammonia-lyase, heavy chain0.1
 eutCEthanolamine ammonia-lyase, light chain0.1
 eutDPutative phosphotransacetylase0.3
 eutEPutative aldehyde oxidoreductase0.1
 eutHPutative transport protein,0.3
 eutJParallel putative heat shock protein (Hsp70)0.1
 eutKPutative carboxysome structural protein0.2
 eutLPutative carboxysome structural protein0.1
 eutTPutative cobalamin adenosyltransferase0.1
 eutMPutative detoxification protein0.3
 eutNPutative detox protein0.2
 eutQPutative ethanolamine utilization protein0.1
 eutPPutative ethanolamine utilization protein0.3
 eutRPutative regulator (AraC/XylS family)0.3
 pduAPolyhedral bodies0.4
 pduCDehydratase, large subunit0.4
Hydrogen sulphide production
 phsAMembrane anchoring protein0.3
 phsBIron-sulphur subunit; electron transfer0.4
 phsCMembrane-anchoring protein0.3
Figure 3.

CsrA regulates the utilization of ethanolamine and propanediol and vitamin B12 synthesis. β-Galactosidase production from lacZ fusions to pdu (A), eut (B), cbi (C) and ttr (D) were used to measure expression in the wild type (black bars), the csrA mutant (grey bars), the complemented csrA mutant (hatched bars) and the csrA mutant with the appropriate cloning vector without csrA (white bars). Error bars represent the standard error of the mean. Asterisks indicate a significant difference in β-galactosidase production compared with wild type.

In addition to genes required for the synthesis of B12, microarray analysis established csrA regulation of genes in two operons, eut and pdu, that function in carbon metabolism and require B12. The 17 genes in the eut operon are required for ethanolamine usage as a carbon and nitrogen source. Ethanolamine is a phospholipid component of both prokaryotic and eukaryotic cell membranes. We found a two- to 10-fold reduction in the expression of eutPQTDMNEJHABCKL, members of the operon required for ethanolamine degradation, and of eutR, a regulator that induces the eut operon in the presence of ethanolamine and vitamin B12 (Table 3). The remaining genes of the eut operon, eutS and eutG, were not detected above the level of non-specific controls on the array. We also tested the regulation of the eut operon by csrA using a lacZ fusion to the eut operon and found a fourfold reduction in the expression of the fusion in the csrA mutant compared with wild type. The decreased expression seen in the mutant was restored to wild-type levels by plasmid complementation of the csrA mutant (Fig. 3B). We also found that genes required for the degradation of propanediol were more poorly expressed in the csrA mutant. Propanediol is a by-product of catabolism of rhamnose and fucose, both of which are found in the mammalian gastrointestinal tract (Badia et al., 1985; Obradors et al., 1988), with rhamnose being a breakdown product of cellulose, and fucose a glycoconjugate present on the cell surface of intestinal epithelial cells (Bry et al., 1996). Propanediol usage requires B12 and the expression of the genes of the pdu operon, which mediate the conversion of propanediol to propionyl-CoA and are located adjacent to the cob operon. Using microarray analysis, we found a 2.5-fold decrease in the expression of pduA and pduC, proposed to encode a shell protein of the polyhedral organelles that are involved in the degradation of propanediol and the large subunit of dehydratase respectively (Table 3). We tested a lacZ fusion to the pdu operon and found a 10-fold decrease in β-galactosidase production in the csrA mutant compared with wild type. Complementation fully restored pdu expression to the csrA mutant (Fig. 3A). The anaerobic utilization of ethanolamine and propanediol as carbon and energy sources by S. typhimurium requires tetrathionate as a terminal electron acceptor (Price-Carter et al., 2001). Tetrathionate is reduced to thiosulphate by tetrathionate reductase, encoded by the ttr operon, located in SPI-2, and thiosulphate is further reduced to sulphite and hydrogen sulphide by thiosulphate reductase activity encoded by the phs operon. We found a threefold decrease in the expression of phsA and phsC in the csrA mutant by microarray analysis (Table 3). To examine the effects of the reduced expression of the phs operon, we grew strains on triple-sugar iron agar, which turns black with bacterial production of hydrogen sulphide. The csrA mutant failed to produce sufficient hydrogen sulphide to change the colour of the medium, but the complemented mutant had restored production of hydrogen sulphide (Fig. 4). By microarray analysis, we found a slight increase in expression (less than twofold) of the ttr operon in the csrA mutant (Table 3). To investigate further the effect of csrA on ttr, we measured β-galactosidase production from a lacZ fusion to ttr and found a threefold increase in the expression of ttr in the csrA mutant (Fig. 3D). This increase was reduced to wild-type levels in the csrA mutant carrying the complementing plasmid (Fig. 3D). These results, taken together, show that csrA regulates the production of vitamin B12, as well as two metabolic pathways that require B12, ethanolamine and propanediol utilization.

Figure 4.

Regulation of hydrogen sulphide production by csrA. Wild type (tube 1) and the complemented csrA mutant (tube 3) produced hydrogen sulphide on triple-sugar iron agar, but the csrA mutant (tube 2) failed to do so.

Regulation of maltose operon by CsrA

Maltose and maltodextrins are present in high concentrations in the intestinal tracts of animals as by-products of starch metabolism. Maltose and maltodextrin are transported through a pore, consisting of maltoporin encoded by lamB, which serves as a channel for sugar migration across the outer membrane. Both transport and utilization of these compounds are regulated by MalT, a regulator required for transcription at mal promoters (reviewed by Boos and Shuman, 1998). Genomic analysis using the microarray demonstrated decreased expression of the maltose system in the csrA mutant (Table 4). The ratio of expression of malT in the csrA mutant compared with wild type was reduced only modestly, to 0.7. Expression of genes under the control of MalT was, however, significantly reduced in the csrA mutant, approximately three- to 10-fold below the levels of the wild type (Table 4). Loss of csrA reduced expression of malE, encoding the maltose-binding protein, and malFGK2, encoding the translocation complex. Genes used for maltose and maltodextrin metabolism were also reduced in expression, malP, encoding maltodextrin phosphorylase, and malS, which encodes a non-essential maltodextrin metabolizing enzyme, periplasmic α-amylase, as well as malM, a periplasmic protein of unknown function. Expression of lamB, encoding maltoporin, the specific pore for maltodextrins and the receptor for phage λ in E. coli, was reduced threefold in the csrA mutant. To confirm the effect of csrA on maltose acquisition, we purified outer membrane proteins from the csrA mutant to be compared with those of the wild type and the complemented mutant (Fig. 2). The csrA mutant showed a greatly reduced amount of a protein with an apparent molecular weight of 48 kDa. Complementation of the mutant with a low-copy-number plasmid carrying csrA restored the protein to the wild-type level. Production of maltoporin in E. coli is known to be induced by maltose and repressed by glucose. We found that a protein of this apparent molecular weight was present in outer membrane preparations of E. coli DH5α when grown in maltose, but not in glucose, but that S. typhimurium produced the protein during growth with either sugar (data not shown). To identify this outer membrane protein, we determined its amino-terminal sequence to be MDFHGYAR. The amino-terminal sequence of processed S. typhimurium maltoporin, lacking its signal sequence, is VDFHGYAR, differing only at the initial methionine. No other Salmonella protein is predicted to have a similar sequence. Therefore, the combination of sequence analysis, molecular weight and maltose induction in E. coli suggests that the protein is indeed maltoporin. Although reduced in its expression of maltoporin, the csrA mutant was still able to grow on minimal medium with 0.2% maltose provided as the sole carbon source (data not shown). Thus, although it produced less maltoporin, the csrA mutant remained able to import sufficient maltose for growth.

Table 4. . Regulation of the maltose operon by csrA.
GeneFunctionRatio of expression (csrA mutant/wild type)
malE Maltose transport protein, substrate recognition for transport and chemotaxis0.2
malF Maltose transport protein0.2
malG Maltose transport protein0.1
malS Alpha-amylase0.1
malK Maltose transport protein; phenotypic repressor of the mal regulon0.2
malP Maltodextrin phosphorylase0.1
malM Periplasmic protein of mal regulon0.3
lamB Phage lambda receptor protein; maltose high-affinity receptor facilitates diffusion of maltose and maltose oligosaccharides0.3

Regulation of carbon metabolism by CsrA

CsrA was originally described as a regulator of carbon metabolism in E. coli. The Salmonella microarray showed no genes of central metabolic pathways to be altered in their expression by threefold or more in the csrA mutant. Expression of genes in the glycogen biosynthetic pathway, specifically glgA, glgP and glgB, all known to be negatively regulated by CsrA in E. coli, was in fact slightly decreased in the S. typhimurium csrA mutant, whereas glgC was not detected above the level of non-specific controls present on the array. To confirm these results, we tested the csrA mutant for glycogen accumulation when grown on Kornberg medium by iodine vapour staining (Romeo et al., 1993). Unlike E. coli, the mutant was indistinguishable in this test from the wild type (not shown), suggesting that Salmonella csrA does not play a role identical to that of E. coli in glycogen biosynthesis. Conversely, two glyoxyate shunt genes under the positive control of CsrA in E. coli, aceA and aceB, were induced twofold in the S. typhimurium csrA mutant. Thus, the effect of Salmonella CsrA on carbon metabolism appears to be less pronounced and not always consistent with that of E. coli.


The csr system for post-transcriptional regulation is found in a variety of bacterial species. In E. coli, in which the system has been best studied, it functions to regulate central carbon metabolism (hence the name csr for carbon storage regulator) and motility (reviewed by Romeo, 1998). In pathogens such as Erwinia and Pseudomonas species, homologues of csr have been shown to regulate a number of virulence functions. We have shown previously that CsrA and CsrB control an important virulence trait in S. typhimurium, the ability to penetrate intestinal epithelial cells by the action of the SPI-1 type III secretion apparatus (Altier et al., 2000a,b).

In this work, we have studied the global effects of CsrA by examining genome-wide changes in expression caused by the loss of csrA. We found that, in S. typhimurium, CsrA has a number of novel regulatory roles. As expected, CsrA regulated the expression of SPI-1 genes, including those that encode transcriptional regulators, secreted effector proteins and components of the type III secretion apparatus. However, it also controlled the expression of genes outside SPI-1 that produce secreted proteins translocated into eukaryotic cells by the SPI-1 secretion system. Among these was sopB, which is known to be controlled by the SPI-1 regulator InvF (Eichelberg and Galán, 1999), and so probably comes under CsrA control indirectly. Expression of genes for two other secreted effectors, sopA and sopE2, was also reduced in the csrA mutant. Control of these genes by SPI-1 regulatory elements has not been demonstrated, but our work suggests the co-ordinated control of these secreted effectors with the apparatus by which they are secreted. We found only one gene required for enteropathogenesis, pipA, to have significantly increased expression in the csrA mutant. The function of PipA has not been elucidated, and the significance of this finding is currently unknown.

Although the classes of genes induced by CsrA in S. typhimurium at first appear to be unrelated, they may be linked by their use within the intestinal tract of an animal host. The invasion genes of SPI-1 are obviously important in this environment, being required for bacterial penetration of the intestinal epithelium. Flagella are also likely to be important in the intestinal milieu, as flagellar synthesis and the flagellar export apparatus contribute to enteritis caused by S. typhimurium (Schmitt et al., 2001). Our analysis has shown CsrA to be required for both production and secretion of flagella. In addition to these virulence functions, CsrA controls genes required for the metabolism of three nutrients found in the intestinal tract: maltodextrins, propanediol and ethanoloamine. Maltodextrins are present in high concentration in the intestinal tract as by-products of starch metabolism. They are processed in Salmonella by the same means as maltose, and CsrA appears to be required for both their transport and metabolism through the control of maltoporin, maltose-binding protein, the maltose translocation complex and an enzyme needed for their metabolism, maltodextrin phosphorylase. The utilization of propanediol is also regulated by CsrA. The degradation of propanediol to propionaldehyde and then to propionyl-CoA to yield carbon and energy is catalysed by genes of the pdu operon, which require CsrA for full expression. Propanediol is the fermentation product of rhamnose or fucose (Badia et al., 1985), whereas rhamnose is itself derived from the breakdown of dietary polysaccharides, and so is created by the intestinal microflora. Fucose is a product of the intestinal epithelial cells themselves. Fucosylated glycans on the surface of intestinal epithelial cells can be cleaved by bacterial fucosidase to liberate fucose. It is clear that members of the normal microflora are capable of producing fucosidases and can also induce the production of fucosylated glycoconjugates by intestinal epithelial cells (Bry et al., 1996). The similar catabolism of ethanolamine to acetaldehyde and then acetyl-CoA uses the eut genes, also induced by CsrA. Like propanediol, ethanolamine is a carbon and energy source common in the intestinal tract, as it is a component of the membranes of both bacteria and eukaryotic cells. The utilization of both propanediol and ethanolamine also requires vitamin B12, its synthesis being encoded by the cob operon, which is controlled by CsrA as well. Thus, CsrA induces the enzymes required for the degradation of common intestinal nutrients, as well as for the cofactor, B12, required for their action. In addition to B12, the anaerobic degradation of propanediol and ethanolamine requires tetrathionate as the terminal electron acceptor (Price-Carter et al., 2001). We have shown here that CsrA controls genes required for tetrathionate metabolism and for the production of hydrogen sulphide, the final product of tetrathionate reduction. The importance of tetrathionate as a terminal electron acceptor in the intestinal tract is, however, unknown. Our results, taken together, therefore suggest that CsrA in Salmonella co-ordinates the regulation of functions valuable for bacterial life in the intestinal tract of an animal host.

As in E. coli, we also found that CsrA was required for motility. The csrA mutant was deficient in the expression of flagella and chemotaxis genes, lacked flagella and was non-motile. Among Salmonellae, control of flagellar synthesis and chemotaxis is co-ordinated with the expression of SPI-1 genes (Goodier and Ahmer, 2001; Iyoda et al., 2001; Lucas and Lee, 2001). In Salmonella typhi, invasion gene expression and penetration of epithelial cells is greatly reduced with the loss of FliA, the flagellar sigma factor required for the expression of late flagellar operons. In S. typhimurium, however, the effect of a fliA mutation is much less pronounced (Eichelberg and Galán, 2000). A second flagellar gene, fliZ, which exists in an operon with fliA, is a positive regulator of two SPI-1 regulators, hilA and hilC (Lucas et al., 2000; Iyoda et al., 2001; Lucas and Lee, 2001). It is possible therefore that some portion of the effect on invasion by CsrA results from its control of flagellar genes, including fliA and fliZ. It is unlikely, however, that control of flagellar genes provides the only mechanism by which CsrA regulates invasion: FliZ does not alter the expression of hilD (Lucas and Lee, 2001), an SPI-1 regulator controlled by CsrA, and the effect of fliA on invasion genes in S. typhimurium is mild (Eichelberg and Galán, 2000), whereas the loss of csrA causes a profound defect in invasion (Altier et al., 2000b).

Similarly, our studies leave open the possibility that control of the cob, eut and pdu operons by CsrA occurs through regulation of FlhDC, which is itself a global regulator (Prüßet al., 2001). However, genomic analysis comparing gene expression in an S. typhimurium flhDC mutant with that of the wild type indicates no control of these operons by flhDC (K. Hughes, J. Karlinsey, J. Frye, S. Porwollik, and M. McClelland, unpublished results). It is therefore clear that CsrA affects these genes largely independently of its role in flagellar regulation.

Genome-wide analysis using DNA microarrays leads to the identification of genes with the most pronounced reproducible changes in their expression. It is therefore likely that our work has failed to identify important targets of CsrA control. Specifically, transcriptional regulators might go unnoticed, as only small changes in their expression could cause detectable alterations in their regulons. Similarly, the activity of regulators that are the direct targets of CsrA is likely to be controlled post-translationally via message stability. One potential example of an   undetected   regulator   is   flhDC,   known   in   E.   coli  to be controlled by CsrA, but showing little change in expression in the S. typhimurium csrA mutant. It remains possible that CsrA controls flagella through flhDC in Salmonella, but investigation of this question will require other methods. Similarly, DNA microarrays leave open the question of the direct targets of regulation. CsrA controls post-transcriptionally by altering message stability or by blocking the Shine–Dalgarno sequence from recognition by ribosomes. It is likely that most of the genes that we have identified are components of regulons under the control of CsrA and are not themselves the direct targets of its action. Identification of the targets, and elucidation of the pathways by which CsrA controls these genes, will also require further study. Another caveat of the microarray used in this study is that its construction was based upon the sequence of the S. typhimurium type strain LT2, which is a laboratory strain. Because the experiments were performed in a fully virulent wild-type strain, 14028s, any genes unique to 14028s or highly divergent from LT2 were not assayed. Therefore, some genes regulated by CsrA in 14028s may not have been detected by our LT2 microarray.

Although CsrAB and their homologues exist in a number of bacterial species, the functions that they regulate are quite diverse. In fact, the only role of S. typhimurium CsrA that we have shown to be in common with that of E. coli is the regulation of flagellar synthesis. Beyond this, several of the regulated functions in S. typhimurium do not exist in the closely related E. coli. Invasion genes are part of an acquired pathogenicity island, SPI-1. The utilization of propanediol and the synthesis of B12 are encoded by the linked cob and pdu operons in a region that was acquired through horizontal transfer and is absent in E. coli (Price-Carter et al., 2001). We have also thus far found no evidence of regulation by S. typhimurium CsrA of carbon metabolism genes similar to that in E. coli. Thus, this ubiquitous regulator appears to have been adapted in Salmonella to the control of specific functions necessary for bacterial life in the intestinal tract, including those required for virulence.

Experimental procedures

Strains and growth conditions

Strains and plasmids used in this study are listed in Table 5. Strains were grown standing at 37°C in Luria–Bertani (LB) broth buffered to pH 8.0 with 100 mM Hepes throughout, except where otherwise noted. To detect changes in glycogen synthesis, strains were grown on Kornberg medium agar plates (1.1% K2PO4, 0.85% KH2PO4, 0.6% yeast extract, 0.2% glucose) and stained with iodine vapour by inverting the plates over iodine crystals.

Table 5. . Strains and plasmids used.
GenotypeResistanceSource or reference
  1. Cam, Kan, Str and Sp, resistance to ampicillin, chloramphenicol, kanamycin, streptomycin and spectinomycin respectively.

 14028s Salmonella typhimurium ATCC
 Δ(csrA)::CamCam Altier et al. (2000b)
 cbiD24::MudJKan Andersson and Roth (1989)
 ttrB123::MudJKan Price-Carter et al. (2001)
 metE205 ara-9 pdu12::MudJKan Walter et al. (1997)
 eut-38::MudJKanJ. Roth
 Δ(csrA)::Cam, cbiD24::MudJCam, KanThis work
 Δ(csrA)::Cam, ttrB123::MudJCam, KanThis work
 Δ(csrA)::Cam, metE205, ara-9, pdu12::MudJCam, KanThis work
 Δ(csrA)::Cam, eut-38::MudJCam, KanThis work
 pFF584Str, Sp Altier et al. (2000a)
 pCA132 0.7 kb csrA fragment on pFF584Str, Sp Altier et al. (2000a)

RNA isolation

Total bacterial RNA was isolated from mid-log cultures by killing the bacteria with the addition of 0.15 volumes of 95% ethanol, 5% phenol, pH 4.3, pelleting the bacteria, resuspending the pellet in 10 mM Tris, 1 mM EDTA (TE) containing 0.5 mg ml−1 lysozyme, adding 1 ml of 10% SDS and incubating this suspension at 64°C for 2 min. After incubation, 11 ml of 1 M sodium acetate at pH 5.2 was added. An equal volume of phenol was then added, and the suspension was incubated at 64°C for 6 min with frequent mixing. Cultures were centrifuged at 7000 g for 10 min at 4°C. The aqueous layer was removed and mixed with an equal volume of chloroform and centrifuged at 7000 g for 5 min at 4°C. The aqueous layer was removed and a 1/10th volume of 3 M sodium acetate at pH 5.2 was added. Nucleic acid was precipitated with cold isopropanol and pelleted by centrifugation at 10 000 g for 25 min at 4°C. The pellet was washed with 80% ethanol and resuspended in 1 ml of nuclease-free water. To this suspension, 20 µl of 1 M Tris (pH 8.3) and 10 µl of 1 M magnesium chloride and a total of 500 U of RNase inhibitor and 250 U of RNase-free DNase were added, and the mixture was incubated at 37°C for 30 min. The RNA sample was then extracted once each with phenol and phenol–chloroform and twice with chloroform. A 1/10th volume of 3 M sodium acetate at pH 5.2 was added, and RNA was precipitated with isopropanol, washed with 80% ethanol and resuspended in nuclease-free water. The RNA concentration was measured with a spectrophotometer.

DNA microarrays

A total of 50 µg of RNA was transcribed to DNA and labelled with Cy3- or Cy5-conjugated dUTP using reverse transcriptase (Superscript II®; Invitrogen) and random hexamers as primers. The RNA was then hydrolysed by incubating labelled probes with 0.1 M sodium hydroxide final concentration at 65°C for 10 min. The sodium hydroxide was neutralized with the addition of hydrochloric acid to a final concentration of 0.1 M. Unincorporated nucleotides were removed using a polymerase chain reaction (PCR) purification kit (Qiagen) according to the manufacturer's instructions. Equal volumes of labelled probes from wild type and the csrA mutant strain were mixed with an equal volume of hybridization solution consisting of 50% formamide, 10× SSC and 0.2% SDS. Slides were prehybridized in 25% formamide, 5× SSC and 0.1% SDS at 42°C. Probes were hybridized simultaneously to a chip containing three replicate arrays spotted onto CMT-UltraGAPS® (Corning) slides. A second chip was hybridized with the dyes reversed to normalize for any differences in incorporation or fluorescence of each dye. Chips were scanned using a ScanArray 5000 laser scanner (GSI Lumonics), signals were recorded with scanarray 2.1 software and then quantified using quantarray 3.0 software (Packard BioScience). Ratios were calculated between the two conditions (i.e. mutant strain/wild type) (Eisen and Brown, 1999). Genes with signals less than two standard deviations (SD) above background controls in both conditions (experiment and control) were considered not detected.

Electron microscopy

A droplet of LB broth containing ≈ 108 bacteria ml−1 was placed onto a Formvar- and carbon-coated 2000 mesh copper grid, and liquid was removed after 3 min by wicking with filter paper. A droplet of 2% aqueous phosphotungstic acid (PTA) at pH 7.2 was placed on the grid before the surface dried completely. After 30 s, the PTA was completely removed by wicking with filter paper. The grid was examined with a transmission electron microscope.

Protein isolation

Outer membrane proteins were isolated as described previously (Kumar et al., 2001). Overnight cultures were centrifuged for 5 min at 7000 g. Pellets were washed once with 20 mM Tris, 10 mM EDTA, pH 8 (TE), then resuspended in the same buffer. Bacteria were disrupted by sonication for 1 min, followed by a 2 min rest, then an additional 1 min sonication. Samples were centrifuged for 5 min at 7000 g to remove debris, and the resulting supernatant was centrifuged for 1 h at 100 000 g. The pellet was resuspended in TE, and the protein concentration was estimated using Bradford reagent. A 1/10th volume of 10% Sarcosyl was added, and samples were incubated at 4°C for 1 h. Samples were centrifuged again for 1 h at 100 000 g. Pellets were resuspended in 1% SDS and boiled for 10 min. Proteins were separated using 10% SDS-PAGE and stained with Coomassie brilliant blue R-250. For sequencing, proteins were transferred to a polyvinylidene difluoride (PVDF) membrane, and amino-terminal sequence was determined by the Protein Sequencing Core Facility at the University of North Carolina.

β-Galactosidase assays

Triplicate or more overnight cultures of each bacterial strain were grown overnight without aeration at 37°C and assayed as described previously (Miller, 1992). Strains carrying eut-38::MudJ were grown in LB broth buffered to pH 8.0 with 100 mM Hepes as above but were also supplemented with 10 mM ethanolamine and 15 nM cobinamide dicyanide to induce expression of the eut operon.

Motility assays

Bacteria were grown as aerated overnight cultures, and 10 µl of each was spotted onto semi-solid (0.35%) LB agar. Plates were incubated at 37°C for 3–5 h in a humidified incubator, and strains were assessed for motility. The csrA mutant strain was assessed further until 48 h after inoculation.

Statistical analysis

For β-galactosidase assays, a one-way analysis of variance was used to determine whether the mean of at least one strain differed from that of any of the others. Then, multiple comparison tests (least square differences t-test at a P≤ 0.05) were used to determine which means differed (SAS System for Windows 8e). For the microarrays, analysis of variance (anova) based on mixed model analysis was used (SAS System for Windows 8e). Changes in gene expression were considered to be significant if the t-statistic of the log(2) of the fold change for a given gene had a P-value < 0.05 (Wolfinger et al., 2001).


We thank John Roth for generously providing strains, and Brian Ahmer and Kelly Hughes for sharing unpublished results. We acknowledge Michael Dykstra and the Laboratory for Advanced Electron and Light Optical Methods at the North Carolina State University College of Veterinary Medicine for assistance in specimen preparation and microscopy support. We also thank David Klapper of the UNC Protein Sequencing Core Facility for his help with protein sequencing. We thank Vygna Tran and Felisa Blackmer for technical assistance. This work was supported in part by NIH grants R01AI34829-13 and RO1AI43283-4 to M.M.