Physical and functional interactions among RNase E, polynucleotide phosphorylase and the cold-shock protein, CsdA: evidence for a ‘cold shock degradosome’



Escherichia coli contains at least five ATP-dependent DEAD-box RNA helicases which may play important roles in macromolecular metabolism, especially in translation and mRNA decay. Here we demonstrate that one member of this family, CsdA, whose expression is induced by cold shock, interacts physically and functionally with RNase E. Three independent approaches show that after a shift of cultures to 15°C, CsdA co-purifies with RNase E and other components of the RNA degradosome. Moreover, functional assays using reconstituted minimal degradosomes prepared from purified components in vitro show that CsdA can fully replace the resident RNA helicase of the RNA degradosome, RhlB. In addition, under these conditions, CsdA displays RNA-dependent ATPase activity. Taken together, our data are consistent with a model in which CsdA accumulates during the early stages of cold acclimatization and subsequently assembles into degradosomes with RNase E synthesized in cold-adapted cultures. These findings show that the RNA degradosome is a flexible macromolecular machine capable of adapting to altered environmental conditions.


Escherichia coli contains at least five ATP-dependent DEAD-box RNA helicases, DbpA, SrmB, RhlB, RhlE and CsdA (also called DeaD) (Kalman et al., 1991). These proteins appear to play important roles in macromolecular metabolism, especially in translation and mRNA decay, although their full impact on cellular physiology remains to be elucidated. DbpA is perhaps the best-characterized family member and functions as an RNA-specific ATPase activated by a specific structure within domain V of 23S rRNA (Nicol and Fuller-Pace, 1995; Tsu and Uhlenbeck, 1998; Pugh et al., 1999). As DbpA is not activated by either 70S ribosomes or 50S subunits, but is activated by denatured ribosomes, it is proposed to play a specific role in 50S biogenesis rather than translation. SrmB was first identified as a gene copy-dependent suppressor of a temperature-sensitive mutation in rplX, encoding L24, which is required for 50S subunit assembly (Nishi et al., 1988). Recent work has shown that SrmB plays an essential role in the biogenesis of 50S subunits in wild-type cells at low temperature (Charollais et al., 2003). SrmB has also been implicated in mRNA turnover through its ability to stabilize a hybrid T7-lacZ mRNA and to interact with poly(A) polymerase (Iost and Dreyfus, 1994; Raynal and Carpousis, 1999). RhlB was initially identified as a member of the family of DEAD-box proteins (Kalman et al., 1991) and was later found to be a component of the RNA degradosome (Miczak et al., 1996; Py et al., 1996). Its activity is stimulated by both RNase E and RNA. It functions in vitro to unwind stable RNA secondary structures and co-ordinate attack by the 3′-exoribonuclease, polynucleotide phosphorylase (PNPase) (Py et al., 1996; Blum et al., 1999; Coburn et al., 1999). Knockout mutations in rhlB have little or no effect either on bacterial growth at various temperatures or on bulk mRNA decay, but severely impede the degradation of mRNAs containing REP sequences (Khemici and Carpousis, 2004). Thus, the RNA helicase activity provided to the RNA degradosome by RhlB may be unimportant for bulk mRNA decay in vivo, or redundant with another such function. RhlE is apparently non-essential (Ohmori, 1994), but can associate with poly(A) polymerase I (Raynal and Carpousis, 1999).

CsdA differs from the other four RNA helicases in that it bears a C-terminal extension (CTD) of ≈ 20 kDa unique to its subfamily. It unwinds double-stranded RNA in the absence of ATP and has been reported to lack ATPase activity (Jones et al., 1996; Lu et al., 1999). CsdA was originally identified as deaD, a gene dosage-dependent suppressor of a temperature-sensitive rpsB mutation, which reduces the amount of RpsA (S1) and RpsB (S2) proteins in ribosomes (Toone et al., 1991; Moll et al., 2002). Recently, CsdA has been shown to be involved in the biogenesis of 50S subunits (Charollais et al., 2004). CsdA may also be required for initiation of translation of mRNAs with extensive secondary structure (Lu et al., 1999). In addition, like SrmB, CsdA can interact with poly(A) polymerase and stabilize a T7-lacZ chimeric mRNA (Iost and Dreyfus, 1994; Raynal and Carpousis, 1999). Moreover, csdA mutants manifest a delay in the disappearance of cold-shock proteins possibly because of defects in mRNA decay at cold temperatures (Yamanaka and Inouye, 2001). Normally CsdA is present in small quantities in cells grown at 37°C but is induced significantly upon a shift to lower temperature (‘cold shock’) (Jones et al., 1996). Knockout mutations in csdA have little effect on cell growth at 37°C, but retard growth at 15°C (Jones et al., 1996). The unique CTD of CsdA is essential for cold temperature growth (C.S. Ramey, R.K. Beran and R.W. Simons, unpubl. data). The CsdA protein has been reported to localize exclusively to the ribosomal fraction of cold-shocked cells (Jones et al., 1996), but this result has been challenged (Moll et al., 2002).

Here, we show that CsdA associates with components of the RNA degradosome in vivo and in vitro, and that it can fully replace RhlB in a functional RNA degradosome assay in vitro. These observations argue that the RNA degradosome is a versatile and adaptable multicomponent molecular engine, with activities that accommodate growth at low temperatures.


Genetic interactions among RNase E and CsdA

As part of an ongoing investigation of the DEAD-box proteins of E. coli, we searched for suppressors of the cold-sensitive phenotype of a csdA mutant, knowing the potential for CsdA and other DEAD-box proteins to interact with other macromolecules (Toone et al., 1991; Jones et al., 1996; Yamanaka and Inouye, 2001), including components of the mRNA decay apparatus (Raynal and Carpousis, 1999). We constructed an in-frame deletion of the chromosomal copy of csdA (see Experimental procedures) and found that it displays a slow growth (‘cold-sensitive’) phenotype at 22°C (Fig. 1A, right hand sector). Growth at this temperature was partially restored by pRNE101, a complete copy of the rne gene on a low-copy vector (Fig. 1A, left hand sector). Overexpression of RNase E (pRNE101) also partially suppresses the cold-sensitive phenotype of the csdA mutant at 18°C (data not shown). The substitution of the wild-type chromosomal gene by its temperature-sensitive rne-1 allele (Ono and Kuwano, 1979; Mudd et al., 1990) exacerbated the growth defect of the csdA deletion at 22°C (Fig. 1B), although rne-1 itself exerts no phenotype at this temperature. These data demonstrate a genetic interaction between csdA and rne.

Figure 1.

Suppression of the cold-sensitive phenotype of csdA deletions.
A. Partial suppression by overexpression of RNase E. Strain WJW45ΔcsdA was transformed with pACYC177-RNE 101 (left side) or with pACYC177 alone (right side). Plates were grown at 22°C.
B. Exacerbation of the phenotype by the rne-1 mutation. Derivatives of WJW45ΔcsdA carrying the indicated alleles of rne were grown at 22°C. WJW45rne-1 serves as the control.

Effect of CsdA on rne gene expression at low temperatures

To examine the possible involvement of CsdA in RNA degradosome function in vivo, we used an rne′–′lacZ fusion to quantify RNase E activity at normal (37°C) and low (22°C) temperatures. At the moderately low temperature of 22°C, CsdA is beneficial but not essential for cell growth, thereby enabling analysis under conditions of steady state, although slow, growth. Expression of the rne gene is negatively autoregulated post-transcriptionally, which requires an intact RNA degradosome, wherein the RNase E scaffold domain is involved (Jiang et al., 2000; Ow et al., 2000; Leroy et al., 2002). Accordingly, we tested the effects of the rne-131 deletion mutation, which removes the scaffold domain (Kido et al., 1996; Lopez et al., 1999), as well as the binding site for CsdA (A.J. Carpousis, pers. comm.), in cells that either were csdA+ or lacked CsdA by virtue of a csdA::kan substitution mutation (see Experimental procedures). As previously observed, rne′–′lacZ fusion expression increases about threefold in the rne-131 mutant grown at 37°C, consistent with loss of normal autoregulation (Table 1; cf. lines 1 and 3). Loss of CsdA increases fusion expression about 50% at the same temperature, whether the cell is rne+ or rne-131 (cf. lines 1 and 2, and 3 and 4). Although the basis of this effect of the csdA::kan mutation is unknown, it is apparently independent of the RNase E scaffold (see Discussion).

Table 1. Effects of the rne-131 and csdA::kan mutations on rne′′lacZ gene fusion expression in vivo.
 Relevant strain genotypea rne′′lacZ fusion expression (Miller units β-galactosidase activity)b
37°C csdA::kan/csdA+22°C csdA::kan/csdA+
  • a

    . Strains used were RS9074 (wild type), RS9195 (csdA::kan), RS9200 (rne-131) and RS9306 (rne-131 csdA::kan).

  • b

    . Each strain was grown in triplicate overnight in LB to saturation at the indicated temperatures, subcultured into fresh LB for growth to mid-log phase (≈ 0.4 OD) and then assayed for β-galactosidase activity as described (Simons et al., 1987). Controls showed that mutants were not selected during growth under these suboptimal conditions (data not shown). Data shown are averages of triplicates (variance was ≤10% of the means).

1Wild type2001.6×4101.9×
2 csdA::kan 3151.6×8001.9×
3 rne-131 6551.4×9100.8×
4 rne-131 csdA::kan 8901.4×7500.8×

At 22°C, both the rne-131 and csdA::kan mutations separately increase rne expression approximately twofold (Table 1, cf. lines 1–3). However, the csdA::kan allele exerts no effect in combination with the rne-131 mutation (line 4), suggesting its dependence on the RNase E scaffold domain. The apparent scaffold-dependent effect of CsdA is consistent with the RNase E-dependent involvement of CsdA in RNA degradosome function in vitro (see below). These results do not, however, distinguish effects of the CsdA mutation on mRNA decay versus potential effects on translation. Because the pattern of rne–lacZ mRNA decay is subtly different in these different strains (data not shown), it is not possible to directly compare their levels at steady state, thereby preventing an estimate of translational efficiency in the different cases.

Effect of CsdA on rne′–′lacZ fusion expression at the level of mRNA decay

To test whether CsdA controls the stability of the rne′–′lacZ mRNA directly, we examined the effects of csdA::kan and rne-131 mutations on its half-life at normal and low temperatures (Table 2). As seen previously, the rne-131 mutation increases rne mRNA stability at 37°C (Table 2; cf. lines 1 and 3). A similar increase in rne′–′lacZ mRNA stability also occurs at 22°C, suggesting that the RNA degradosome functions in vivo at lower temperatures. At 37°C, the csdA::kan mutation exerts little or no effect on rne′′lacZ mRNA decay, in either the rne+ or rne-131 background (Table 2, cf. lines 1–4, at 37°C). However, at 22°C, the csdA::kan mutation increases rne′′lacZ mRNA stability fivefold in the presence of normal RNase E (cf. lines 1 and 2). Likewise, the rne-131 mutation also increases rne′–′lacZ mRNA stability to a comparable level (Table 2; cf. lines 1 and 3), whereas a much smaller increase is observed when the csdA::kan mutation is introduced into a strain already bearing the rne-131 allele (cf. lines 3 and 4). These results at the level of rne′–′lacZ mRNA stability account for at least a part of the effects measured by gene fusion expression (Table 1).

Table 2. Effect of the rne-131 and csdA::kan mutations on the stability of the rne mRNA.
 Relevant strain genotypea rne′′lacZ mRNA half-life (min)b
37°C csdA::kan/csdA+22°C csdA::kan/csdA+
  • a

    . Strains are described in Table 1.

  • b

    . Total cellular RNA was analysed by Northern blotting and the rne′ lacZ species were quantified as described in Experimental procedures. Half-lives were determined in two completely independent analyses, each with independently isolated RNA samples, and the standard error between these experiments are shown for each genetic case.

  • c

    . Approximate values, as bands corresponding to the rne′ lacZ mRNA from rne+ cells grown at 37°C were faint, making half-life determination difficult.

1Wild type≈ 1.5c≤1.0× 1.7 ± 0.24.9×
2 csdA::kan ≈ 1.2c≤1.0× 8.3 ± 0.54.9×
3 rne-131 10 ± 0.5 0.6× 7.2 ± 0.51.6×
4 rne-131 csdA::kan 6.4 ± 0.3 0.6×11.9 ± 0.81.6×

Co-purification of CsdA with the RNA degradosome

Cultures of strain CF881 were grown to saturation either at 37°C or at 15°C and degradosomes purified from strain CF881 as described by Coburn and Mackie (1998) with the following modifications. Cultures were grown in 500 ml of Terrific broth (1.3% w/v granulated peptone, 2.7% yeast extract, 0.4% glycerol, 0.017 M KH2PO4 and 0.072 M K2HPO4) to saturation, either at 37°C or 15°C, the latter taking 3 days. Cells were harvested by centrifugation and washed twice in 10 mM Tris-HCl, pH 7.5, 10 mM Mg-acetate and 10 mM KCl. Cells were lysed and processed as described (Carpousis et al., 1994; Coburn and Mackie, 1998) including chromatography on SP-sepharose. The final step in purification used Biogel A 5M (Bio-Rad) size exclusion chromatography in 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 0.1 mM DTT, 5% glycerol and 0.5% Genapol X-080. The column was 1 cm × 50 cm and the flow rate was 0.5 ml min−1. Degradosomes eluted between 40 and 52 ml (37°C) or 32–44 ml (15°C); pooled fractions were concentrated to 0.4–0.8 mg ml−1 using centrifugal concentrators (Millipore). Figure 2 illustrates a separation of the proteins from the two preparations. The profile in Fig. 2, lane 2 (cells grown at 37°C) is similar to published separations (Py et al., 1996; Coburn and Mackie, 1998). In contrast, the preparation from cells grown at 15°C (Fig. 2, lane 3) contains a number of bands in addition to those also present in lane 2. Each of the bands in lane 3 was excised from the stained gel and identified by tryptic digestion and mass spectrometry. In addition to RNase E (27 peptides identified), PNPase (17 peptides), enolase (12 peptides) and RhlB (14 peptides), we also identified RNA polymerase β′-subunit (32 peptides), RNA polymerase β-subunit (14 peptides), RNA polymerase α-subunit (8 peptides), pyruvate dehydrogenase E1 subunit (32 peptides), pyruvate dehydrogenase E2 subunit (16 peptides) and CsdA (21 peptides). The co-purification of CsdA with the RNA degradosome was anticipated from the genetic suppression experiments (Fig. 1); however, the co-purification of RNA polymerase and pyruvate dehydrogenase was a surprise.

Figure 2.

Composition of purified degradosomes from cultures grown at different temperatures. Cultures of strain CF881 were grown to saturation either at 37°C (lane 2) or at 15°C (lane 3) and degradosomes purified from the cell pellets as described (Coburn and Mackie, 1998). Lane 1 shows markers (masses indicated in the left margin); the identity of bands in lanes 2 and 3 is given in the right margin (see the text).

To determine whether the RNA degradosome forms a physical complex with RNA polymerase and/or pyruvate dehydrogenase, we used an alternative purification using FLAG-tagged RNase E (Miczak et al., 1996). Cultures of BL21 (DE3)/pRE196 were grown at 37°C to mid-log phase, then shifted to 15°C for 2 h. FLAG-RNase E was induced at the time of down-shift. Cells were harvested and lysed, and the FLAG-RNase E and associated proteins were captured with a monoclonal anti-FLAG antibody linked to agarose beads. The proteins associated with RNase E were assessed by staining and by Western blotting and included PNPase, enolase, RhlB and CsdA in relative amounts comparable to those shown in Fig. 2, lane 3 (data not shown, but see also Fig. 4). No RNA polymerase subunits were recovered and only small quantities of pyruvate dehydrogenase E1 and E2 subunits were observed. We conclude that the purification of these two macromolecular complexes with the RNA degradosome after cold shock is largely coincidental reflecting the large sizes of these complexes rather than their direct physical interaction. Moreover, pyruvate dehydrogenase is also induced by cold shock (Jones and Inouye, 1996), increasing the likelihood of its detection in non-specific associations.

Figure 4.

Physical interactions among RNase E, Pnp and CsdA in vivo. Wild-type (lanes 1–4), pnp::Tn5 (lanes 5–8) or csdA::Tn5 (lanes 9–12) cells were grown at either 37°C (odd lanes) or 15°C (even lanes), while expressing either a FLAG epitope alone or FLAG-RNase E as shown above the figure and described in Experimental procedures. After pulse-labelling of proteins with [35S]-methionine, RNase E was immunoprecipitated using anti-FLAG M2 affinity gel (Sigma). Precipitated proteins were separated electrophoretically and visualized by phosphorimaging. The positions of molecular mass standards, and purified RNase E, PNPase and CsdA, determined by Coomassie Blue staining (not shown), are shown in the left and right margins respectively.

Interactions among RNase E, Pnp and CsdA

We performed co-immunoprecipitation experiments with purified or partially purified proteins to assess whether RNase E and CsdA could associate directly using methods described previously (Miczak et al., 1996; Vanzo et al., 1998; Coburn et al., 1999). Partially purified RNase E expressed with an N-terminal FLAG epitope (Miczak et al., 1996) was mixed with purified Pnp and CsdA, both of which contained N-terminal hexahistidine tags (see Experimental procedures). The partially purified FLAG-RNase E protein associated with significant amounts of Pnp that were present in the extract (Fig. 3, lane 1), consistent with earlier findings (Carpousis et al., 1994; Miczak et al., 1996; Py et al., 1996; Coburn et al., 1999; R.E. Edge and G.A. Mackie, unpubl. results). The data in Fig. 3, lane 4 show that when all three proteins were mixed under conditions amenable to reconstitution, an anti-FLAG antibody was able to precipitate both Pnp and CsdA in addition to FLAG-RNase E. However, the anti-FLAG antibody did not precipitate purified Pnp or CsdA in the absence of FLAG-RNase E (Fig. 3, lanes 2 and 3). We also varied the ratio of CsdA to FLAG-RNase E and found the optimal ratio to be 10 µg CsdA per 100 µg of a partially fractionated extract containing FLAG-RNase E although a significant signal could be detected at somewhat lower ratios (data not shown).

Figure 3.

Interactions among RNase E, Pnp and CsdA in vitro. Partly purified FLAG-RNase E, purified HIS-Pnp and/or purified HIS-CsdA were mixed as shown above the Figure and incubated as described in Experimental procedures. Complexes were co-precipitated using anti-FLAG M2 affinity gel (Sigma) and individual components were detected by Western blotting (see Experimental procedures). The image is a composite of three blots, each probed with the antibody listed on the left. Each vertical lane represents one immunoprecipitation experiment.

To extend these findings, cultures expressing either the FLAG peptide alone or FLAG-RNase E were labelled with [35S]-methionine at 37°C or at 15°C for 2 h after the shift to lower temperature. Extracts were prepared and subjected to immunoprecipitation with anti-FLAG antibody and the precipitates resolved by electrophoresis (see Experimental procedures). Although there was considerable background, a comparison of lanes 3 and 4 in Fig. 4 showed that labelled proteins identical in mobility to RNase E, Pnp and CsdA were co-precipitated by the anti-FLAG antibody. These identifications were confirmed by Western blotting (data not shown). The band corresponding to CsdA in lane 4 was weak or missing in lane 3, consistent with the induction of CsdA by cold shock (Jones et al., 1996). The same experiment was repeated in a pnp::Tn5 mutant (Fig. 4, lanes 5–8). As expected, the band corresponding to Pnp was missing in lanes 7 and 8. The band corresponding to CsdA was very prominent among the proteins co-precipitated by the anti-FLAG antibody in lane 8 (labelling at 15°C). This band’s intensity was significantly increased relative to lane 7 (labelling at 37°C). As expected, this band disappeared when labelling was performed in a csdA::Tn5 strain (Fig. 4, lanes 9–12). In the latter case, a band corresponding to Pnp was clearly co-precipitated with FLAG-RNase E at either labelling temperature (Fig. 4, cf. lanes 11 and 12). We have not attempted to identify some of the other relatively intense bands in lanes 7, 8 and 11. Together, these data show that FLAG-RNase E interacts with Pnp and CsdA that are newly synthesized after cold shock and that the presence of one of these proteins is not required for the binding of the other.

Functional equivalence of RhlB and CsdA in vitro

Coburn et al. (1999) have shown that RNase E, Pnp and RhlB will associate with each other in vitro and display ATP- and phosphate-dependent degradation of a structured RNA substrate. The activity of the reconstituted ‘minimal degradosome’ is qualitatively identical to that of purified degradosomes. In view of the association of CsdA with RNase E in vivo and in vitro, we tested whether CsdA could functionally substitute for RhlB in a reconstituted minimal degradosome. The data in Fig. 5A and B show a time-course of disappearance of a ≈ 375 nt malEF RNA substrate dependent on RNase E, Pnp, CsdA, ATP and phosphate. The full-length substrate gradually disappeared whereas two intermediates accumulated, especially in the earliest time points (Fig. 5A, lanes 2 and 3). The shorter of these, labelled RSR in Fig. 5A, represents a product shortened by about 30–40 nt at its 3′-terminus. The yield of the RSR subsequently fell with continued incubation. In contrast, in the absence of ATP (Fig. 5B), the two intermediates persist and do not disappear. Similar data were obtained for assays in which RhlB and CsdA were interchanged (Fig. 5C). In all cases, the RSR reached a maximal yield within 5 min. In the presence of ATP (closed symbols in Fig. 5C), the RSR subsequently largely disappeared within 60 min. Its rate of disappearance did not differ significantly when CsdA was substituted for RhlB. In the absence of ATP (open symbols in Fig. 5C), the yield of RSR declined slowly in the presence of either CsdA or RhlB, most probably because of slow endonucleolytic cleavage rather than 3′ degradation (Coburn et al., 1999). The slow loss of the RSR was less pronounced at 30°C than at 37°C (data not shown). We did not attempt to determine whether the optimal concentrations of CsdA or RhlB differed significantly in this assay; both were in modest excess relative to RNase E.

Figure 5.

Functional equivalence of RhlB and CsdA in vitro.
A. Purified RNase E, Pnp and CsdA were mixed and assayed against in vitro transcribed, radiolabelled malEF RNA in the presence of phosphate and ATP as described in Experimental procedures. Aliquots were removed at the times (in min) after the addition of enzymes as shown above the panel. RNA intermediates were separated on a 6% denaturing polyacrylamide gel and were visualized using phosphorimaging. A schematic drawing of substrate and product is shown in the right margin.
B. As in (A), but in the absence of ATP.
C. Quantification of the RSR intermediate. Reconstitution of ‘mini-degradosomes’ was performed as described in Experimental procedures using RNase E, Pnp and RhlB (squares) or CsdA (circles). Assays included phosphate, with (▪, •) or without (□, ○) ATP. The yield of the ≈ 340 nt RSR intermediate is shown as a percentage of the signal of the intact malEF RNA at time 0.

We also tested whether purified CsdA would manifest any activity independent of RNase E. CsdA exhibited readily detectable ATPase activity that was stimulated substantially in the presence of RNA (Fig. 6). DNA would not substitute (data not shown). This activity was not stimulated further by RNase E (or by PNPase) in the absence or presence of RNA. Under the same conditions, the ATPase activity of a comparable amount of RhlB in purified degradosomes was less than 10% of that of CsdA. We also assayed the concentration dependence of the ATPase activity on RNA. The extent of ATP hydrolysis in a fixed time reached a plateau at approximately one molecule of RNA per molecule of CsdA (data not shown).

Figure 6.

ATPase activity of CsdA and degradosomes. Purified HIS-CsdA was incubated with [γ-32P]-ATP at 30°C as described in Experimental procedures, alone (□), in the presence of unlabelled rpsT mRNA (▪) or with both rpsT mRNA and purified RNase E ( inline image ). Purified degradosomes were assayed similarly in the presence (•) or absence (○) of rpsT RNA. Aliquots were removed at the times indicated and spotted on a PEI-cellulose plate which was developed in 0.375 M KH2PO4, pH 3.5. The released Pi was quantified using phosphorimaging. The results represent the average value of at least five assays.


Altered degradosomes after cold shock

Many functions have previously been ascribed to the DEAD-box RNA helicase, CsdA: ribosomal assembly (Toone et al., 1991; Moll et al., 2002; Charollais et al., 2004; R.K. Beran, C.S. Ramey and R.W. Simons, unpubl. obs.), translational initiation (Lu et al., 1999), stabilization of mRNA (Iost and Dreyfus, 1994), interaction with poly(A) polymerase (Raynal and Carpousis, 1999), degradation of cold-shock mRNAs (Yamanaka and Inouye, 2001) and survival at low temperature (Jones et al., 1996). Our data add an additional function to this list, namely, association with the RNA degradosome after cold shock. Several lines of evidence support this assertion. First, CsdA co-purifies with degradosomes isolated from cultures grown at 15°C in significant yields, approaching one copy per degradosome. Moreover, this association is detected in two independent methods of purification, biochemical and immunological. Second, pulse-labelling shows that CsdA associates with RNase E synthesized after adaptation to low temperature, suggesting de novo assembly of a CsdA-containing degradosome, although we cannot rule out association of CsdA with pre-existing degradosomes accumulated before cold shock. Third, purified CsdA can associate directly with RNase E in vitro, but not with PNPase. Fourth, CsdA appears to play a role in mRNA decay at low temperature in vivo, in a manner dependent on the degradosome-assembly domain of RNase E (i.e. its scaffold). Finally and most dramatically, CsdA can fully replace RhlB in an in vitro assay for degradosome function. Taken together, these data show that an alternative form of the degradosome assembles in vivo after cold shock and that, by implication, it can perform all the functions of the native (37°C) degradosome.

It has been reported previously that CsdA unwinds a 29 bp duplex in the absence of ATP (Jones et al., 1996). Jones et al. used a great excess of CsdA over RNA, raising the possibility that the observed unwinding activity was non-specific. In our hands, the unwinding activity of CsdA depends on RNase E when it is coupled to the degradation of the malEF RNA by PNPase and occurs when substrate is in excess. And in our hands, the unwinding activity of CsdA depends on RNase E and occurs when substrate is in excess. It has also been reported that CsdA lacks ATPase activity even in the presence of polynucleotides (Lu et al., 1999). In our experience, the observed ATPase activity of CsdA is very dependent on the buffer used and this could explain the discrepancy between our data and those reported earlier.

Although it is tempting to suggest that CsdA fully replaces RhlB in the degradosome at cold temperatures, the data should be interpreted cautiously in this regard. RhlB can still be detected in degradosomes purified from cold-adapted cultures. Moreover, pulse-labelling shows that RhlB is synthesized during the recovery phase of cold shock and is co-immunoprecipitated with RNase E. Two models are consistent with these observations. First (the ‘two helicase’ model), both RhlB and CsdA associate simultaneously with RNase E, but bind to separate sites, as shown by Khemici et al. (2004). In this model, cold-adapted degradosomes contain an additional component, CsdA. Alternatively, a heterogeneous population of degradosomes exists in cold-shocked cells, some containing RhlB, some containing CsdA and some containing both helicases. For technical reasons, we have not been able to distinguish between these models using immunological methods, as the titre of the available anti-RhlB antibodies is too low to determine whether CsdA co-immunopurifies with RhlB via RNase E. Moreover, free CsdA is in considerable excess over RNase E. Nonetheless, we favour the latter model as it is possible that CsdA would be in considerable excess over RhlB during the early stages of cold acclimatized growth (see below). Interestingly, RNase E preparations from Rhodobacter capsulatus also contain two helicases, one identified as RhlB and the other almost certainly as CsdA (Jager et al., 2001).

Functional consequences of a ‘cold shock degradosome’

Although the discovery of the degradosome was very satisfying, until recently its functional significance has been unclear and, at times, controversial (Kido et al., 1996; Coburn and Mackie, 1998; Steege, 2000; Beran et al., 2003). Recent evidence shows that RhlB associated with RNase E is required for the degradation of REP sequences, a secondary structure element found in many E. coli mRNAs (Khemici and Carpousis, 2004). Lower temperatures are expected to stabilize RNA secondary structures. Indeed, many cold-shock proteins have been implicated in RNA metabolism or translation (Jones and Inouye, 1996). Strains deleted for CsdA are cold sensitive, showing that when RhlB is expressed at normal levels it cannot replace CsdA (Jones et al., 1996). Conversely, our data show that CsdA can fully replace RhlB in vitro to promote the degradation of the malEF RNA, which contains a duplicated REP sequence. This is particularly noteworthy in view of the fact that RhlB and CsdA appear to bind to spatially separated sites in RNase E (Khemici et al., 2004). Nonetheless, they are functionally equivalent as each permits coupled unwinding and phosphorolysis of structured RNA substrates.

Our in vivo data now show that CsdA is required for efficient decay of an rne′–′lacZ fusion mRNA at low temperature, in a manner dependent on the C-terminal scaffold domain of RNase E. By implication, CsdA is required for normal RNase E activity at cold temperatures. The most direct interpretation of these data is that CsdA must interact with RNase E directly, consistent with the formation of a specialized cold-shock degradosome. This view, coupled with the observation that a third DEAD-box helicase, RhlE, can also substitute for RhlB in vitro (Khemici et al., 2004), argues that the degradosome is a versatile ‘machine’, able to adapt to changing growth conditions.

It remains possible that CsdA affects in vivo mRNA stability indirectly at low temperatures, for example, by limiting the supply of ribosomes. However, our data show that such an effect would have to depend on RNase E, yet the available evidence suggests that CsdA participates in ribosome assembly and translation independently of RNase E. Nonetheless, we do suggest that the requirement for CsdA in processes such as ribosome assembly at low temperature is the source of the scaffold-independent effects of CsdA seen in the data of Table 1.

Model for assembly of a cold-shock degradosome

Immediately upon cold shock, many mRNAs are stabilized substantially, suggesting that the mRNA decay machinery is temporarily inactivated (Zangrossi et al., 2000; Beran and Simons, 2001; Mathy et al., 2001; Yamanaka and Inouye, 2001). Although the mechanism of inactivation is unknown, it is possible that the rigidification of the cytoplasmic membrane, which accompanies cold shock (Janoff et al., 1979), results in the trapping of RNase E and the degradosome in a non-functional state until membrane fluidity is restored. The synthesis of CsdA and other cold-shock proteins is induced early after the shift to lower temperature (Jones et al., 1996), whereas the expression of ‘non-cold shock’ proteins is repressed. As a consequence, a significant pool of CsdA accumulates during the ≈ 2 h period after cold shock and before recovery. Thus, once the synthesis of RNase E reinitiates after cells become acclimatized to the new ambient temperature, CsdA associates with the available RNase E. The fraction of cold-adapted degradosomes containing both CsdA and RhlB would depend on the time when RhlB synthesis resumes and its rate of accumulation relative to RNase E and CsdA. Our data suggest that excess CsdA is required to saturate RNase E in vitro, implying a lower affinity for RNase E compared with RhlB. This could explain why degradosomes containing RhlB are assembled at low temperature even when CsdA is so abundant and also why little or no CsdA is found in degradosomes isolated from cultures grown at 37°C, where CsdA is expressed at a basal level. In addition to binding to RNase E, CsdA can (and does) play other functions independent of RNase E during immediate adaptation to cold and the subsequent recovery.

If there are two distinct populations of degradosomes (those with and those without CsdA), there is one important consequence: any function of RhlB-containing degradosomes at 37°C can also be performed by degradosomes assembled during cold shock in the presence of CsdA. To date, the only established function of the degradosome is the degradation of REP sequences and probably other highly structured RNAs (Lopez et al., 1999; Leroy et al., 2002; Khemici and Carpousis, 2004). Our data show that this occurs efficiently in vitro when CsdA substitutes for RhlB, and that this substitution has likely functional consequences in vivo. In view of its other reported roles in ribosomal assembly and translation, CsdA is truly a multifunctional DEAD-box helicase.

Experimental procedures

Bacterial strains and plasmids

BL21(DE3) was obtained from Novagen. Plasmids pflagLRC and pRE196, encoding inducible FLAG epitope and FLAG-RNase E constructs, respectively, were obtained from Dr Sue Lin-Chao (Miczak et al., 1996). Plasmid pEP′18 (Py et al., 1996) was obtained from Dr C.F. Higgins. Plasmid pECSRA201, encoding the csdA sequence, was obtained from Dr Pamela Jones (University of Georgia, Athens). The csdA gene was removed from this plasmid by digestion with NdeI and HindIII and the resultant ≈2 kb fragment was ligated into NdeI- and HindIII-digested pET15b (Novagen) to generate pRS3486. To express the gene products, BL21(DE3) was transformed by the corresponding plasmid.

The pnp::Tn5 allele was transduced from strain JC357 (obtained from Dr Claude Portier) into BL21(DE3) by P1 phage and selection for kanamycin resistance. The absence of Pnp protein in this new strain, RS9215, was verified by Western blotting, and by growth defects at 15°C. Likewise, the csdA::kan allele from JC7623 csdA::kan (obtained from Dr Pamela Jones) was transduced into BL21(DE3), thereby creating strain RS9216. Disruption of the csdA gene was confirmed by DNA sequencing of the amplified csdA::kan construct, and by a lack of signal on an anti-CsdA Western blot. The in-frame deletion of csdA in strain WJW45 has been described (Charollais et al., 2004).

RS9157 [P90C rne131 zce-726::Tn10 (TcR)] was constructed by co-transducing rne-131 and zce-726::Tn10 from BZ31 (kindly provided by M. Dreyfus) into P90C (Miller, 1992), selecting for TcR. Presence of the co-transduced rne-131 mutation was confirmed by Western blot identification of truncated RNase E and genomic sequencing. RS9052 and RS9171 were constructed by transducing the csdA::kan allele from JC7623 csdA::kan into P90C and RS9157, respectively, selecting for KmR. RS9074, RS9195, RS9200 and RS9306 were constructed by lysogenizing P90C, RS9052, RS9157 and RS9171, respectively, with λez1 (Jain and Belasco, 1994), which bears the rne′–′lacZ fusion and was kindly provided by Dr Joel Belasco. The csdA::kan mutation was confirmed by the absence of CsdA in Western blot analysis and genomic sequencing.


RNase E protein was purified by preparative gel electrophoresis followed by electroelution as described previously ( Cormack et al., 1993; Coburn et al., 1999; Mackie et al., 2001). Pnp and RhlB were purified by chromatographic techniques as reported (Coburn and Mackie, 1998; Coburn et al., 1999; A. Prud’homme-Généreux, unpubl. data). Degradosomes were purified from strain CF881 as previously described (Coburn and Mackie, 1998).

Partially purified FLAG-RNase E.  Cultures of BL21(DE3)/pRE196 (Miczak et al., 1996) were grown with vigorous shaking at 37°C in 150 ml of LB broth and 50 mg ml−1 ampicillin to an OD600 = 0.4. FLAG-RNase E was induced by adding IPTG to 0.5 mM together with 150 ml of fresh medium. Induction was allowed to proceed for 4 h at 30°C. Cells were collected by centrifugation, resuspended in buffer A [50 mM Tris-HCl (pH 7.6), 10 mM MgCl2, 60 mM NH4Cl, 0.5 mM EDTA] supplemented with 7.5% glycerol, 100 mM DTT, 2 µg ml−1 aprotinin, 0.8 µg ml−1 leupeptin and 0.8 µg ml−1 pepstatin A, and lysed by two passages through a French pressure cell (8000 psi). The lysate was treated with 20 µg DNase I (Sigma) and 1 mM phenylmethanesulphonyl fluoride (PMSF) for 10 min on ice, and centrifuged at 30 000 g for 45 min at 4°C. The supernatant was made to 26% (w/v) with ammonium sulphate, and the precipitated FLAG-RNase E was recovered by centrifugation at 17 000 g for 20 min at 4°C. The pellet was resuspended in buffer A containing 1 mM DTT, 7.5% (v/v) glycerol, 200 µM PMSF, 2 µg ml−1 aprotinin, 0.8 µg ml−1 leupeptin and 0.8 µg ml−1 pepstatin A. The protein concentration was determined using the Bradford Assay (Bio-Rad).

HIS-Pnp.  Cultures of BL21(DE3)/pEP′18 were grown as described above, but at 30°C. After induction with 1 mM IPTG for 2 h, cells were collected by centrifugation, suspended in 5 ml of buffer W (50 mM Tris-HCl, pH 8.0, 500 mM NaCl) supplemented with 1 mM PMSF and lysed in a French pressure cell (10 000 psi). The lysate was treated with 20 µg DNase I on ice for 10 min, and the extract was centrifuged at 30 000 g for 45 min at 4°C. The supernatant was passed over a TALON Metal Affinity column (Clontech). The column was washed with buffer W containing 10 mM imidazole, and the bound proteins were eluted with buffer W supplemented with 50 mM imidazole. The fractions containing Pnp were collected, concentrated using Millipore centrifugal filters to 1.2 ml, and subjected to size exclusion chromatography on a Bio-Gel A-0.5m column (1.5 cm × 35 cm, Bio-Rad) in buffer C [25 mM Tris-HCl (pH 7.7), 100 mM NaCl, 5% glycerol, 1 mM DTT, 0.1 mM EDTA]. Fractions containing Pnp were pooled and concentrated using Centricon centrifugal filter device (Millipore).

HIS-CsdA.  Cultures of BL21(DE3)/pRS3486 were grown at 37°C as above and induced for 3 h with 1 mM IPTG. Cells were recovered by centrifugation, resuspended in buffer B [20 mM Tris-HCl (pH 8.0), 500 mM NaCl, 5 mM MgCl2] supplemented with 1.7 µg ml−1 lysozyme, 0.05 units ml−1 DNase I and 10 mM imidazole, and left on ice for 30 min. They were then lysed by sonication with eight 10 s pulses. The extract was centrifuged at 12 000 g for 30 min at 4°C. The supernatant was passed twice through a Ni++ NTA agarose column (Qiagen) equilibrated with buffer B + 10 mM imidazole. The column was washed with 2.5 column volumes of buffer B + 20 mM imidazole, and the bound proteins were eluted in 1 ml of buffer B + 1 M imidazole. The eluate was concentrated using Centricon centrifugal filter device to a final volume of 500 µl, and supplemented with an equal volume of glycerol. This protein preparation was stored at −20°C.

In vitro co-immunoprecipitation

Reconstitutions were performed by mixing partially purified FLAG-RNase E (100 µg of total protein from an AS26 fraction), 20 µg of purified HIS-Pnp, and 20 µg of purified HIS-CsdA in 200 µl of buffer A containing 7.5% (v/v) glycerol, 50 mM DTT, 2 µg ml−1 aprotinin, 0.8 µg ml−1 leupeptin and 0.8 µg ml−1 pepstatin A. This mixture was incubated for 20 min at 30°C. To pull down FLAG-RNase E, 40 µl of anti-FLAG M2-agarose affinity gel (Sigma) was added to the mixture, and incubated with gentle agitation at 4°C for 2 h. The beads were washed three times in 500 µl of TBS buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl), and then incubated for 30 min at 4°C with 100 µl of a 150 µg ml−1 FLAG-epitope (Sigma) solution to elute bound proteins. This solution (10 µl) was loaded on a 7.5% SDS-PAGE gel, and subsequently blotted to a nitrocellulose membrane. This blot was probed using anti-RNase E, anti-Pnp or anti-CsdA rabbit polyclonal anti-sera (used at 1:10 000).

In vivo co-immunoprecipitation

Strains containing pflagLRC or pRE196 expressing the FLAG epitope or N-terminally FLAG-tagged RNase E, respectively, were grown in M9 minimal media supplemented with all the amino acids except methionine and cysteine, in the presence of 150 µg ml−1 ampicillin at 37°C to an A600 = 0.4. A 5 ml portion was pulse-labelled with 50 µCi of [35S]-methionine (NEN, 1000 Ci mmol−1) and 1 mM IPTG for 15 min at 37°C, then washed by adding 5 ml of ice-cold minimal media containing 0.5% sodium azide and 0.3% methionine to stop incorporation of the label. The cells were quick frozen in an ethanol/dry-ice bath. The remainder of the culture was grown at 15°C for 2 h, at which point 5 ml was removed and pulse-labelled for 30 min at 15°C as described above. Labelled cells were suspended and extracts were prepared as described by Carpousis et al. (1994), with some modifications. Briefly, cells were lysed in 250 µl of 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 3 mM EDTA, 5% (v/v) glycerol, 1 mM DTT, 1.5 mg ml−1 lysozyme (Sigma), 1 mM PMSF, 2 mg ml−1 aprotinin, 0.8 mg ml−1 leupeptin, 0.8 mg ml−1 pepstatin A, and mixed on ice periodically over the course of 1 h. This extract was supplemented with 125 µl of 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 30 mM magnesium acetate, 5% (v/v) glycerol, 1 mM DTT, 3% (v/v) Triton X-100, 1 mM PMSF, 2 mg ml−1 aprotinin, 0.8 mg ml−1 leupeptin, 0.8 mg ml−1 pepstatin A, 20 mg ml−1 DNase I and four units of RNase T1 (Pharmacia), and left on ice for an additional 30 min. NH4Cl was then added to 1 M, with continuous stirring on ice over the course of 30 min. Cell debris was pelleted by centrifugation for 15 min at 16 000 g. The supernatant was precipitated with (NH4)2SO4 to 40% of saturation, recovered by centrifugation for 10 min at 16 000 g, and suspended in 500 µl of TBS. Immunoprecipitation with 40 µl of anti-FLAG M2-agarose affinity gel was performed as described above. Proteins in the immunoprecipitate were resolved by electrophoresis and detected by phosphorimaging.

Exonuclease assay

The MalEF RNA was synthesized in vitro in the presence of [32P]-CTP from the plasmid pCH77 linearized with EcoRI (McLaren et al., 1991; Mackie et al., 2001). Assays were assembled in a 40 µl volume as described previously (Py et al., 1996; Mackie et al., 2001) with slight modifications. Briefly, 1 mg ml−1 purified RNase E protein, 2 mg ml−1 Pnp and 1 mg ml−1 RhlB or HIS-CsdA were incubated together in 20 mM Tris-HCl (pH 7.5), 1.5 mM DTT, 1 mM MgCl2, 20 mM KCl for 15 min at 37°C to regenerate a ‘minimal degradosome’ (Coburn et al., 1999). The enzymes were then mixed with 1.6 nM MalEF RNA, 10 mM sodium phosphate, 0.1 mg ml−1 BSA, 0.05% (v/v) Genapol X-080, in the presence or absence of 3 mM ATP. The reaction was incubated at 37°C. Samples (4 µl) were removed from the assay at various times, and quenched with three volumes of loading buffer containing 90% formamide, 22 mM Tris, 22 mM boric acid, 0.5 mM EDTA, 0.1% xylene cyanol FF, 0.1% bromophenol blue and 0.1% SDS. Portions of these samples were separated on a 6% polyacrylamide gel containing 8 M urea, and visualized by phosphorimaging.

ATPase assay

The ATP hydrolytic activity of DEAD-box helicases was assayed by measuring the release of [32P]-Pi from [32P]-ATP. Reactions were assembled in a 10 µl volume, and contained ≈ 10 nM helicase (750 ng ml−1 CsdA or 4.8 µg ml−1 degradosome) in 20 mM Tris-HCl (pH 7.5), 1.5 mM DTT, 1 mM MgCl2, 20 mM KCl, 10 mM Na-phosphate (pH 7.5), 0.05% Genapol X-080, 25 mg ml−1 BSA, 0.1 mM ATP, 2.5 mCi [32P]-ATP (3000 Ci mM−1) and 20 nM in vitro transcribed rpsT mRNA (Mackie and Genereaux, 1993). Some assays also contained 150 pg double-stranded DNA, and up to 7.5 mg ml−1 RNase E protein. Assays were assembled on ice, and the reactions were initiated by the addition of helicase. The mixture was incubated at 15°C or 30°C. At various times, a 1 µl aliquot was removed and spotted on a poly(ethylenimine) (PEI)-cellulose thin layer chromatography (TLC) plate. This plate was developed in 0.375 M KH2PO4 at pH 3.5, dried and visualized by phosphorimaging.

Northern blot analysis

Strains (Table 1) were grown in LB at the indicated temperatures until saturated, subcultured 25-fold into fresh LB and incubated again until mid-exponential growth (A600≈ 0.5) was reached, whereupon rifampicin (400 µg ml−1) was added. Aliquots were then collected at 10 s and 1, 2, 4 and 8 min intervals, quick-frozen in a dry ice-ethanol bath, and total cellular RNA was isolated using the RNAwizTM method (Ambion) and quantified by absorbance. Northern blot analysis was performed essentially as previously described (Ausubel et al., 1992). Total cellular RNA (10 µg per lane) was electrophoresed on a 1.2% agarose-formaldehyde gel, with MOPS buffer [0.04 M MOPS (pH 7.0), 0.05 M sodium acetate, 0.001 M EDTA]. Gels were rinsed several times in RNase-free water and soaked in 500 ml 10× SSC (1.5 M NaCl, 0.15 M trisodium citrate, pH 7.0) for 45 min with rocking. RNA was then transferred to a neutral nylon membrane (Amersham Hybond-N) by capillary action with 20× SSC. RNA was cross-linked to the membrane by UV irradiation (Stratagene UV Stratalinker 1800) and probed overnight at 50°C with a 1274 bp 32P-internally labelled probe made by polymerase chain reaction (PCR) of the rne′′lacZ fusion with primers corresponding to regions in the rne (5′-TGTTGATT ACGGCGCTGAACG-3′) and lacZ (5′-TAGAGATTCGGGAT TTCGGCGCT-3′) portions. Blots were visualized and quantified on a Molecular Dynamics phosphorimager, with ImageQuant software.


This work was supported by grants from CIHR to G.A.M., from ARC (Number 4633) and MENRT (ACI Dynamique et Réactivité des Assemblages Biologiques) to M. Dreyfus, and from NIH and NSF to R.W.S. I.I. thanks M. Dreyfus for his support. R.K.B. was an Edwin Pauley Fellow and C.S.R. was a Beckman Scholar. We thank Ashley Hagiya for assistance with the Northern blot analysis, and all members or our laboratories for useful discussions.