Riboregulation by DsrA RNA: trans-actions for global economy




DsrA is an 87 nucleotide Escherichia coli RNA with extraordinary regulatory properties. The profound impact of its actions stems from DsrA regulating translation of two global transcription regulators, H-NS and RpoS (σs), by sequence-specific RNA–RNA interactions. H-NS is a major nucleoid-structuring and global repressor protein, and RpoS is the stationary phase and stress response sigma factor of RNA polymerase. DsrA changes its conformation to bind to these two different mRNA targets and thereby inhibits H-NS translation, while stimulating that of RpoS in a mechanistically distinct fashion. DsrA apparently binds both the start and the stop codons of hns mRNA and sharply decreases the mRNA half-life. DsrA also binds sequences in the 5′-untranslated leader region of rpoS mRNA, enhancing rpoS mRNA stability and RpoS translation. A cohort of genes, governed by H-NS repression and RpoS activation, are thus regulated. Low temperatures increase the levels of DsrA, with differential effects on H-NS and RpoS. Additionally, the RNA chaperone protein Hfq is involved with DsrA regulation, as well as with other small RNAs that also act on RpoS to co-ordinate stress responses. We address the possible functions of this genetic regulatory mechanism, as well as the advantages of using small RNAs as global regulators to orchestrate gene expression.

Small, stable RNA molecules provide an economical means for cells to achieve global regulation. A single small RNA specifying multiple target sequences by RNA–RNA interactions can do the work of many sequence-specific RNA-binding proteins (reviewed by Eddy, 1999). Small RNAs of Escherichia coli have been reviewed extensively (Inouye and Delihas, 1988; Delihas, 1995; Romeo, 1998; Eddy, 1999; Wassarman et al., 1999). Here, we will focus on DsrA, one such small, trans-acting stable RNA of E. coli.

DsrA changes its conformation to bind at least two different mRNAs, governing their translation and turnover (Majdalani et al., 1998; Lease et al., 1998; Lease and Belfort, 2000), with potential interactions at three more genes (Lease et al., 1998). As the global transcription regulators H-NS and RpoS (also called σs) are both targets of DsrA, there are dramatic downstream consequences of DsrA action (Fig. 1). To sustain this regulation, DsrA persists after its synthesis: the stability of DsrA (half-life 30–60 min; R. Lease and M. Belfort, unpublished; D. Sledjeski, personal communication) suggests that the transcriptional state of the cell can be maintained with high sequence specificity and minimal cost of maintenance.

Figure 1.

DsrA action at hns and rpoS, with downstream regulatory consequences. Solid arrows represent biochemically and genetically verified activities. To the left (–), DsrA blocks hns, relieving repression of downstream genes. To the right (+), DsrA activates rpoS, activating downstream genes. Large open and filled circles represent genes repressed and activated by hns and rpoS respectively; the shaded overlap region represents genes governed by antagonism of hns and rpoS. The dashed arrow, centre (±) indicates putative interactions suggested by bioinformatics (Lease et al., 1998).

Regulation by small RNAs would be particularly advantageous and flexible during stress responses, such as those governed by H-NS silencing (reviewed by Atlung and Ingmer, 1997; Williams and Rimsky, 1997) and RpoS activation (reviewed by Loewen and Hengge-Aronis, 1994; Hengge-Aronis, 1996), when the expenditure of energy must be tightly managed while multiple genes are co-ordinately regulated. Small RNAs are cost-effective under these circumstances, because the energy required to synthesize a small untranslated RNA that specifies multiple targets is minimal compared with the energetic cost of the synthesis and translation of multiple large mRNAs.

The dsrA gene was discovered in studies of DNA sequences in the downstream region of rcsA, a transcriptional activator of polysaccharide capsule synthesis genes (Sledjeski and Gottesman, 1995). Characterization of transcripts originating in this region led to the elucidation of the sequence, identification of the promoter and estimation of the size of the 87 nucleotide RNA. Additionally, a preliminary secondary structure based on minimum free energy was predicted. DsrA RNA acts indirectly to stimulate rcsA transcription by blocking H-NS (Fig. 1; Sledjeski and Gottesman, 1995), a global transcriptional ‘silencer’ and nucleoid-structuring protein (reviewed by Atlung and Ingmer, 1997; Williams and Rimsky, 1997).

Subsequent work has shown that, in addition to antagonizing H-NS action, the small RNA stimulates translation of RpoS, the stationary phase and stress response sigma factor (Fig. 1) in log phase growth at low temperatures (20–30°C) but not at higher temperatures (42°C; Sledjeski et al., 1996). Low temperatures stimulate the synthesis of DsrA by a combination of enhanced transcription and increased termination efficiency (Sledjeski et al., 1996; F. Repoila, N. Majdalani and S. Gottesman, in preparation). Low temperatures enhance DsrA action on RpoS (Sledjeski et al., 1996) and H-NS (R. Lease and M. Belfort, unpublished). Even the low levels of DsrA normally produced from the chromosome influence RpoS (Sledjeski et al., 1996), whereas the higher DsrA levels produced from plasmids are required to produce effects on H-NS (Sledjeski et al., 1996; Lease et al., 1998; Majdalani et al., 1998). Thus, conditions that enhance DsrA synthesis, such as low temperature, may govern the balance of H-NS and RpoS regulation by DsrA.

DsrA contains blocks of sequence complementary to target genes

How does DsrA affect two targets in completely different ways? A partial answer to this tantalizing question was found by computer sequence analysis of DsrA, which revealed regions of complementarity within DsrA to five different genes (hns, rpoS, argR, ilvIH and rbsD) (Lease et al., 1998). The regions of complementarity cluster together in two different regions (Fig. 2A) near the 5′ end (rpoS and rbsD) and near the middle of DsrA (hns, argR and ilvIH). These complementary sequences and their location within DsrA suggested that RNA–RNA interactions between DsrA and target mRNAs might account for the different activities of DsrA. Indeed, independent bioinformatic and compensatory mutagenesis studies of DsrA with hns (Lease et al., 1998) and rpoS (Majdalani et al., 1998) revealed that DsrA acts in trans by RNA–RNA basepairing interactions of two different DsrA sequences with these two targets.

Figure 2.

DsrA conformational changes promote RNA–RNA interactions and regulation. The refined DsrA secondary structure is shown alone (A) and in complex with target RNAs (B and C).

A. The structure of DsrA determined by RNase footprinting and thermodynamic predictions. Numbers in circles delineate stem–loops. The rpoS-complementary sequences (rpoS′) are in a grey box; the two hns-complementary sequences (hns′) are in an open box and in white letters on a black background for the start and stop codon interaction regions respectively.

B. The DsrA–hns RNA interaction and melting of stem–loop 2, with coaxial stack interaction model below.

C. The DsrA–rpoS RNA interaction and melting of stem–loop 1. The rpoS translational operator (Brown and Elliott, 1997) is shown below, with the Shine–Dalgarno sequence in bold letters. Colons (:) represent G:U basepairs; the open circle (○) represents a non-canonical G:A basepair.

Further bioinformatic analysis revealed that precisely those three of the five DsrA-complementary mRNAs that were postulated to interact near the middle of DsrA, hns, argR and ilvIH, each contain a second, discrete DsrA-complementary sequence. Strikingly, although the first set of DsrA-complementary sequences are near the start codons of these three genes, the new set of sequences occurs near their stop codons. Within DsrA, the corresponding complementary sequences are either adjacent (for hns and argR) or overlapping (for ilvIH) and reside in the middle portion of DsrA (Fig. 2A and B, cf. boxed sequences and white letters on black background) (Lease and Belfort, 2000).

DsrA as a dynamic structure

The structure of DsrA, based on thermodynamic predictions, was proposed to consist of three stem–loops (Sledjeski and Gottesman, 1995). Further structural detail was obtained by RNase footprinting of DsrA with single- and double-strand specific nucleases in vitro (Lease and Belfort, 2000). This confirmed the proposed structure of the first and third stem–loops of DsrA, while suggesting a different structure for stem–loop 2 (Fig. 2A). This central stem–loop apparently remains intact, while DsrA stem–loop 1 melts out to basepair with rpoS mRNA (Fig. 2C). In contrast, stem–loop 1 is predicted to remain intact while DsrA is basepairing with hns mRNA by melting out stem–loop 2 (Fig. 2B) (Lease and Belfort, 2000). Indeed, genetic analyses suggest that stem–loops 1 and 2 can affect their different targets independently of one another (Lease et al., 1998; Majdalani et al., 1998). This provides a dynamic basis not only for the separation of activities but also for distinct modes of regulation.

Stem loop 3 is the transcription terminator for DsrA (Sledjeski and Gottesman, 1995) and is predicted to remain intact when DsrA is in both repressing and activating mode. Substitution of stem–loop 3 with a heterologous terminator sequence permits DsrA function (Majdalani et al., 1998). Termination of DsrA, thought to be a function of temperature and stem–loop 3 stability, is critical for DsrA function. Mutations that disrupt the structure of stem–loop 3 reduce DsrA activity, as readthrough transcripts are non-functional (Sledjeski and Gottesman, 1995).

Nuclease footprinting of DsrA in the presence and absence of hns RNA in vitro supported the second region of potential complementarity between DsrA and hns RNA also predicted by computer analysis (Fig. 2A and B, white letters on a black background). The two hns-complementary regions are adjacent within DsrA, so that this pair of interactions between hns and DsrA has the potential to form a coaxial stack that loops out the central portion of hns RNA (Fig. 2B). Analogous coaxial stacks can also be modelled for the potential DsrA–argR and DsrA–ilvIH RNA–RNA interactions (Lease and Belfort, 2000), suggesting a common theme of regulation by binding both start and stop codon regions.

DsrA action at hns

DsrA reduces H-NS protein levels (Lease et al., 1998). Autoregulation of hns transcription (reviewed by Williams and Rimsky, 1997) dictates that decreases in H-NS protein levels should lead to increases in hns RNA levels. As steady-state levels of hns RNA are unaffected by DsrA (Lease et al., 1998), DsrA was predicted to decrease hns RNA stability to balance the increased hns RNA synthesis. Indeed, when de novo synthesis of hns RNA was blocked with rifamycin, DsrA strongly enhanced the turnover of hns RNA, decreasing the half-life of hns RNA > eightfold (Lease and Belfort, 2000). Consistent with these analyses, DsrA decreased the steady-state RNA levels of an hns deletion mutant in which autoregulation is absent (Lease et al., 1998). In summary, DsrA blocks H-NS synthesis through enhanced turnover of hns mRNA via RNA–RNA interactions (Fig. 2B). DsrA could act by exposing or creating an RNase-sensitive site within hns RNA that enhances turnover (Lease and Belfort, 2000) or by blocking translation to enhance turnover (reviewed by Kushner, 1996).

Oddly, another function was found for DsrA, namely increased induction of λ prophage in recA mutants (Rozanov et al., 1998). This could be related to DsrA effects on hns, the mutation of which has been found to modulate supercoiling of chromosomal DNA and enhance phage mu transposition (reviewed by Williams and Rimsky, 1997). Alternatively, the effect could be mediated by DsrA action on other targets.

DsrA action at rpoS

The translational stimulation of RpoS was found to occur by DsrA binding to the 5′ untranslated region of rpoS RNA (Majdalani et al., 1998). In this model, a cis-acting translational operator, comprising an inhibitory stem–loop that sequesters the Shine–Dalgarno sequence of rpoS (Brown and Elliott, 1997), binds DsrA and opens, permitting access of ribosomes and enhancing translation (Fig. 2C; Lease et al., 1998; Majdalani et al., 1998; Lease and Belfort, 2000). RNA turnover studies with rifamycin also show that DsrA enhances the stability of rpoS RNA about threefold (Lease and Belfort, 2000). This mRNA stabilization effect is presumably the indirect result of enhanced translation (reviewed by Kushner, 1996).

Other small RNAs in related pathways

Several small RNAs apparently share the property of regulation at rpoS. OxyS, another such small RNA, is proposed to act both at rpoS and at fhlA, an activator of the oxidative damage response, when OxyS is induced by oxidative stress (Altuvia et al., 1997; 1998; Zhang et al., 1998). A third small RNA, RprA, complements the dsrA deletion mutant phenotype at rpoS, although the mechanism of action is unclear (N. Majdalani and S. Gottesman, manuscript in preparation). Thus, several small RNAs influence the specific expression of rpoS, again highlighting the involvement and utility of these small RNAs in stress responses.

Regulation of DsrA action and synthesis

Hfq (HF-I), an RNA chaperone (Schuppli et al., 1997; Tsui et al., 1997) and global regulatory protein (Muffler et al., 1997), may be a potentiator of DsrA action and synthesis. The proposed mechanism of DsrA action at rpoS is similar to that used by Hfq, which apparently destabilizes the rpoS translational operator (Brown and Elliott, 1997). In some hfq strains, DsrA activity at hns and rpoS is compromised (D. Sledjeski and C. Whitman, submitted). It is thus possible that Hfq and DsrA work in concert, but this remains to be defined further. Another potential level of Hfq involvement stems from the finding that certain hfq mutant strains produce only low levels of DsrA (R. Lease and M. Belfort, unpublished). As this protein is both an RNA chaperone and a global regulator, this effect on DsrA may result from both decreased transcription (R. Lease and M. Belfort, unpublished) and enhanced turnover of DsrA RNA in hfq-null mutants (D. Sledjeski, personal communication).

Several other proteins, namely H-NS, StpA and LeuO, all affect DsrA levels. H-NS regulates LeuO, which directly or indirectly represses DsrA (Klauck et al., 1997). A further level of complexity derives from StpA, a paralogue of H-NS that regulates genes together with H-NS (Sonden and Uhlin, 1996; Zhang et al., 1996). DsrA levels increased slightly in both stpA and hns, but not stpA hns strains (R. Lease and M. Belfort, unpublished), so it is possible that StpA and H-NS repress DsrA by acting together.

Open questions about complex regulatory interplays

What role in nature is accomplished by co-ordinate regulation of the target genes of DsrA? Two verified targets of DsrA, hns and rpoS, and a putative target, argR, are all global transcriptional regulators. Thus, DsrA is expected to orchestrate far-reaching effects on the transcriptional state of the cell, with the co-ordination of H-NS and RpoS regulation having a global impact on gene expression (Fig. 1). Interestingly, RpoS and H-NS are considered antagonists at a variety of stress response genes. These include genes involved in low-temperature adaptation, osmotic shock, starvation and stationary phase survival, as well as induction of virulence factors such as curli fimbriae. However, H-NS represses many genes not activated by RpoS, including several virulence factors as well as genes induced by anaerobiosis, low osmolarity, high temperature and low pH (reviewed by Atlung and Ingmer, 1997). Given that low DsrA levels enhance RpoS translation without affecting H-NS, whereas high DsrA levels enhance RpoS translation and block H-NS (Sledjeski et al., 1996; Lease et al., 1998), control of DsrA levels could differentially and globally govern regulation of genes by RpoS with or without downregulating H-NS. The role of DsrA is therefore likely to change in response to the environmental context and physiological status of the cell.

H-NS, StpA and Hfq are all proteins with nucleoid structuring and global regulatory activities (Azam and Ishihama, 1999; Azam et al., 1999 and references therein) and are intimately intertwined with DsrA in regulatory loops. H-NS and StpA cross-regulate by repressing each others' transcription (Sonden and Uhlin, 1996; Zhang et al., 1996), so that StpA levels rise when DsrA blocks H-NS synthesis (Lease et al., 1998). DsrA repression of H-NS therefore enhances StpA levels. Furthermore, StpA, like Hfq, is also an RNA chaperone (Zhang et al., 1995; Cusick and Belfort, 1998; Clodi et al., 1999), suggesting a possible regulatory role because RNA chaperones resolve misfolded RNAs (Herschlag, 1995), a problem that increases with low temperature.

Small RNA as regulator

The RNA world hypothesis predicates RNA as both catalyst and information system that predates current, predominantly DNA-specified biology (Gesteland et al., 1999). Thus, RNA–RNA interactions, either alone or in conjunction with proteins, would suffice for regulation in the absence of DNA. Although small RNAs have probably arisen de novo rather than having evolved from earlier, purely RNA-based systems (Eddy, 1999), the existence of these RNA–RNA and ribonucleoprotein (RNP) elements suggests RNA-based regulation in an RNA or RNP world. Nevertheless, the economy of small RNAs in terms of energy cost, coupled with their structural dynamics and propensity for multiple sequence-specific regulatory roles, probably accounts for their prevalence in the world of DNA and proteins. Here, with their lengthy half-lives, they are particularly well-suited to conduct sequence-specific regulation in demanding environments, in which energy must be conserved. It is therefore not surprising that DsrA and other small RNAs play critical roles in co-ordinating genetic responses, particularly in times of stress. DsrA, with its ability to undergo conformational changes, is thus well equipped as an orchestrator of global regulation by RNA–RNA interactions, co-ordinating RNA expression and turnover with exquisite specificity and economy.


We thank Susan Gottesman, Nadim Majdalani, Darren Sledjeski, Gigi Storz and Aixia Zhang for contributions of unpublished data and for reading the manuscript. The work in our laboratory was funded by NIH grants GM39422 and GM44844 to M.B and GM18542 to R.L.


  1. Present address: Department of Biophysics, Jenkins Hall, Johns Hopkins University, 3400 N. Charles St., Baltimore, MD 21218-2685, USA.