Viable mutations affecting the 5′-phosphate sensor of RNase E, including R169Q or T170A, become lethal when combined with deletions removing part of the non-catalytic C-terminal domain of RNase E. The phosphate sensor is required for efficient autoregulation of RNase E synthesis as RNase E R169Q is strongly overexpressed with accumulation of proteolytic fragments. In addition, mutation of the phosphate sensor stabilizes the rpsT P1 mRNA as much as sixfold and slows the maturation of 16S rRNA. In contrast, the decay of other model mRNAs and the processing of several tRNA precursors are unaffected by mutations in the phosphate sensor. Our data point to the existence of overlapping mechanisms of substrate recognition by RNase E, which lead to a hierarchy of efficiencies with which its RNA targets are attacked.
RNase E is the principal intracellular RNase in Escherichia coli and related organisms. Although it was discovered through its role in processing 5S rRNA (Ghora and Apirion, 1978; Misra and Apirion, 1979), RNase E is now known to participate in most aspects of RNA metabolism in E. coli, including the processing of ribosomal, transfer and small regulatory RNAs as well as in the degradation of mRNAs (Coburn and Mackie, 1998a; Deutscher, 2006; Carpousis et al., 2009; Belasco, 2010). A number of endoribonucleases can play a role in initiating the decay of specific mRNAs, including RNase III, RNase G, RNase P and RNase Z. Nevertheless, a significant body of data implicates RNase E in determining the rate of decay of most mRNAs in E. coli (Carpousis et al., 2009; Belasco, 2010). However, compelling recent evidence shows that for some mRNAs, a step prior to the first cleavage is actually rate-limiting for turnover, namely the action of RppH, a pyrophosphatase (Celesnik et al., 2007; Deana et al., 2008). Elimination of RppH activity from E. coli does not compromise its viability but does lead to increased steady-state levels of a number of mRNAs. Among those examined directly, the half-lives increased 3- to 11-fold in an rppH mutant (Deana et al., 2008).
One of the hallmark properties of RNase E (and its paralogue, RNase G) is its strong preference for 5′-monophosphorylated RNA substrates (Mackie, 1998; Jiang et al., 2000; Tock et al., 2000). More recently, other structurally unrelated RNases have also been found to exhibit a 5′-monophosphate preference including RNases J1 and Y from B. subtilis (de la Sierra-Gallay et al., 2008; Yao and Bechhofer, 2010), Rat1 from S. pombe (Xiang et al., 2009) and Ago2 from A. fulgidus and H. sapiens (Ma et al., 2005; Frank et al., 2010). Although there are exceptional substrates (Kime et al., 2009) and debate in the literature as to whether the preference of RNase E or G for monophosphates reflects enhancements to Km or Vmax (or both), such substrates are generally turned over more efficiently by RNase E than their corresponding tri-phosphorylated forms (Jiang and Belasco, 2004; Jourdan and McDowall, 2008; Garrey et al., 2009). A ‘phosphate binding pocket’ was proposed as the structural basis for this preference (Mackie, 1998; Spickler et al., 2001). Subsequent determination of the structure of the catalytic domain of RNase E in complex with a substrate analogue verified this prediction and demonstrated the existence of a ‘sensor domain’ at one end of an RNA binding channel that could interact with the 5′-phosphate and first base of a substrate (Callaghan et al., 2005). In particular, R169 and T170 form H-bonds to the terminal phosphate while V124 stacks on the base of the first residue (Callaghan et al., 2005). Previous work showed that the mutation R169Q reduced the kcat of RNase E towards an oligonucleotide substrate ∼ 40-fold (Garrey et al., 2009). Nonetheless, a low copy plasmid carrying this mutation in a full-length copy of the rne gene could successfully complement a strain whose chromosomal rne gene was almost fully deleted. The complemented strain formed small colonies, grew more slowly than wild-type, and accumulated incompletely processed 5S rRNA precursors (Garrey et al., 2009). The viability of this strain was unanticipated and prompted us to investigate more closely the role of the 5′-sensor in RNA metabolism in E. coli. Our data point to the existence of overlapping mechanisms of substrate recognition by RNase E which lead to a hierarchy of efficiencies of RNA cleavage.
Intragenic deletions that render R169Q lethal
Previous data showed that mutations in the phosphate sensor of RNase E retain catalytic activity and are viable although they retard growth (Garrey et al., 2009). The ∼ 500 residues of the C-terminal half of RNase E are not required for its catalytic activity, for viability, or for sensing 5′-monophosphates (Kido et al., 1996; Lopez et al., 1999; Jiang et al., 2000; Ow et al., 2000; Tock et al., 2000; Jiang and Belasco, 2004; Callaghan et al., 2005; Jourdan and McDowall, 2008; Garrey et al., 2009). Rather, these residues form several ‘microdomains’ that interact with the bacterial inner membrane, bacterial cytoskeleton or with other proteins to form the multi-enzyme RNA degradosome (Kido et al., 1996; Callaghan et al., 2004; Taghbalout and Rothfield, 2007; Khemici et al., 2008). In addition, the C-terminal domain has been reported to be required for the internal entry pathway of mRNA turnover (Joyce and Dreyfus, 1998), for recognition of some RNAs (Lopez et al., 1999; Kaberdin et al., 2000), and for interaction with Hfq and small RNAs (Morita et al., 2005). Accordingly, we tested whether otherwise viable deletions removing all or part of the C-terminal 500 residues of RNase E (see Fig. 1) retained viability when combined with mutations affecting the phosphate sensor (i.e. intragenic synthetic lethality). The appropriate plasmids conferring resistance to kanamycin were constructed so that they contained the native rne promoters, the full 5′-UTR, the complete native 3′-UTR and the rho-independent terminator (Garrey et al., 2009). Each was tested for its ability to displace a CmR plasmid containing the wild-type rne gene, pSBK1, from strain SK9714 (Ow et al., 2000). Table 1 summarizes the resultant data (cf. Garrey et al., 2009). In this procedure, pRne-SG1 (WT), pRne-SG4 (R169Q) or pRne-SG5 (T170A) are able to displace pSBK1 from strain SK9714 whereas pRne-SG6 (D303A) affecting the catalytic site is not. Deletion of residues C-terminal to position 529 in the absence of other mutations still permitted growth as shown by the viability of KmR/CmS transformants of SK9714 carrying pRne-SG11 (Δ530–1061), pRne-SG12 (Δ589–1061) or pRne-SG13 (Δ730–1061) as the sole source of RNase E (Table 1). All three strains, especially the first, exhibited smaller colony size and reduced growth rates in liquid media (data not shown). Additional shorter deletions spanning individual microdomains in the C-terminal scaffold region (see Fig. 1) including the Arg-rich region (608–644), the extended Arg-rich region with a putative coil–coli domain (589–723), the RhlB binding site (698–762), the enolase binding site (833–850) or the PNPase binding site (1021–1061) were also constructed and tested. All were able to displace pSBK1 efficiently (see pRne-SG14-18 in Table 1).
Table 1. Complementation of rneΔ1018 by derivatives of pRne-SG1.
All plasmids are based on pRne-SG1 as described in Experimental procedures and in Garrey et al. (2009) where it is named ‘pSG-Rne’.
Displacement was measured as described in Garrey et al. (2009) and refers to the fraction of colonies that have acquired resistance to kanamycin and lost resistance to chloramphenicol after > 20 doublings.
The small colony, slow growth phenotype of Δ530–1061 was further exacerbated by additional mutations in the 5′-sensor. In the most extreme case, Δ530–1061 was lethal in the presence of R169Q (pRne-SG 21 in Table 1) as no KmR/CmS transformants were recovered in two independent experiments. However, the combination of T170A and Δ530–1061 was viable as some KmR/CmS transformants, easily recognized by their small colony size, could be recovered and propagated (pRne-SG31; Table 1). Similar results were obtained with Δ589–1061, which was viable on its own or with T170A but lethal when combined with R169Q (pRne-SG22; Table 1). In contrast, a shorter C-terminal deletion (i.e. C-terminal to residue 730) was viable in the presence of R169Q (pRne-SG23; Table 1). These data imply that the 5′-phosphate sensor becomes essential when a function defined by residues between positions 530 and 729 is eliminated from RNase E. This region includes an Arg-rich segment as well as a coiled-coil domain (Callaghan et al., 2004); however, the viability of pSG24 (Table 1; R169Q/ Δ608–644) would eliminate a role for the Arg-rich region in complementing R169Q.
Loss of autoregulation in R169Q
We assessed the impact of R169Q and other rne alleles on the expression of RNase E by two methods. In the first, the relative amount of the Rne protein in each strain was assessed by Western blotting of boiled whole cells harvested at the same density. A representative blot is shown in Fig. 2A. As expected, at non-permissive temperature, there is an increase in the intensity of RNase E in lane 2 (rne-1) in agreement with past work (Mudd and Higgins, 1993). Strikingly, the amount of RNase E and its proteolytic breakdown products is strongly increased in R169Q at 37°C (lane 4) to a much greater extent than in the temperature-sensitive allele in lane 2. Otherwise, the amount of RNase E in other samples is only moderately greater than in lane 1. The same blot was stripped and reprobed for PNPase to control for loading (Fig. 2C). We have assumed that none of the deletions in RNase E will affect expression of PNPase significantly. The amount of RNase E in each lane of Fig. 2A was quantified and normalized to the amount in lane 1 (WT; strain MG1693) as shown in Fig. 2B. This analysis shows that the amount of RNase E in R169Q is 9.4-fold greater than in the WT strain (compare lane 4 with 1) and 5.5-fold higher than in lane 3 (pRne-SG1[WT]). It was not possible to quantify expression of RNase E carrying C-terminal deletions as the antiserum used is directed primarily at C-terminal epitopes (see also Leroy et al., 2002).
In the second method, plasmids of interest were introduced into strain CJ1828, which contains a chromosomal rne::lacZ fusion in an rne-1 background (Jain and Belasco, 1995). All plasmids tested complemented the temperature-sensitive phenotype of this strain. β-Galactosidase activity was assayed during exponential growth at 37°C (see Experimental procedures). The data in Table 2 show that WT RNase E expressed from pRne-SG1 is able to repress β-galactosidase activity almost ninefold. In contrast, RNase E R169Q from pRne-SG4 is less than half as efficient while RNase E T170A and the deletions encoded by pRne-SG12 and pRne-SG13 exhibit almost wild-type levels of repressor activity. Interestingly, RNase E 1–529 from pRne-SG11 is also a poor repressor in agreement with earlier findings that the C-terminal half of RNase E is required for its autoregulation (Mudd and Higgins, 1993; Jiang et al., 2000; Leroy et al., 2002). Taken together, the two methods demonstrate that a major consequence of the R169Q mutation is loss of autoregulation.
Table 2. β-Galactosidase activity expressed from a chromosomal rne::lacZ fusion in CJ1828.
β-Galactosidase activity was assayed as described in Experimental procedures and is the average of three independent determinations.
Repression is the inverse of ratio of activity in the mutant relative to the empty vector.
470 ± 120
530 ± 24
60 ± 3
140 ± 14
77 ± 4
160 ± 12
68 ± 2
65 ± 1
Impact of the 5′-sensor on mRNA stability
We initially employed the chromosomal rpsT mRNAs as endogenous reporters since they are substrates for RppH and are sensitive to the status of the 5′-phosphate in vitro and in vivo (Mackie, 1998; Mackie, 2000; Celesnik et al., 2007; Deana et al., 2008). We also employed circular rpsT mRNA generated in vivo from a permuted group I intron (Mackie, 2000) as a reporter for end-independent decay as a circular RNA must, by definition, enter an ‘internal entry’ pathway (Joyce and Dreyfus, 1998; Baker and Mackie, 2003). Representative Northern blots are shown in Fig. 3A–C while Table 3 summarizes a large set of measurements of half-lives with appropriate controls. The upper half of Table 3 (‘chromosomal alleles’) shows the effect of two well characterized rne alleles, rne-1 (Apirion and Lassar, 1978; McDowall et al., 1993) and rne-131 (Kido et al., 1996). The former is the more severe, increasing the half-life of the P1 and circular rpsT mRNAs up to sixfold and the P2 mRNA by ∼ 2.5-fold. The milder rne-131 allele causes a modest (∼ 1.3-fold) stabilization of the two chromosomal rpsT mRNAs, but a fivefold stabilization of circular rpsT mRNA. The pnp-7 and ΔpcnB mutations do not affect the initial rate of disappearance of the rpsT P1 or P2 mRNAs under the conditions employed here (cf. Coburn and Mackie, 1998b). These controls confirm that RNase E is the primary ribonuclease involved in initiating the decay of the rpsT mRNAs.
Table 3. Half-lives of the rpsT mRNAs in strain SK9714 complemented by mutant plasmids.
Half-lives are reported to two significant digits and are based on 2–5 determinations. Errors are standard deviations from the mean.
Circular rpsT mRNA was measured in strains transformed with a chloramphenicol-resistant derivative of pGM110 (Mackie, 2000).
Cultures were shifted to 39.5°C for 20 min prior to extraction of RNA.
nd, not determined.
64 ± 6
90 ± 6
460 ± 55
360 ± 34
200 ± 44
420 ± 27
120 ± 8
110 ± 14
260 ± 24
260 ± 18
2000 ± 250
130 ± 6
210 ± 25
1700 ± 140
100 ± 7
150 ± 6
330 ± 120
83 ± 8
140 ± 17
450 ± 29
120 ± 1
200 ± 7
210 ± 14
75 ± 11
110 ± 32
430 ± 54
71 ± 4
110 ± 14
320 ± 110
310 ± 54
320 ± 40
The lower half of Table 3 (‘Plasmid alleles in rneΔ1018’) shows the effect of a number of viable rne alleles on the stability of the rpsT mRNAs. Single mutations in the 5′-sensor exerted strong effects on the P1 mRNA species, in particular. Comparison of the data in Fig. 3C (R169Q) with those in Fig. 3A (WT) shows the relative intensification and enhanced persistence of the 447 nt P1 species due to loss of a functional 5′-sensor in R169Q (pRne-SG4). This resulted in a sixfold increase in the half-life of the P1 mRNA (from 64 to 360 s), comparable to the effect of rne-1 in strain SK5665 (Table 3). T170A (pRne-SG5 in Table 3) caused a less pronounced increase, but still resulted in a twofold stabilization of P1. Curiously, the P2 mRNA species was less affected by either mutation, being stabilized ≤ twofold. In contrast, the circular rpsT mRNA was insensitive to R169Q, as would be expected.
Viable deletions removing all or part of the C-terminal half of RNase E also affected the stability of the rpsT mRNAs although to a lesser extent than R169Q (see Table 3). The Δ530–1061 allele in pRne-SG11 provoked the largest effect, increasing the half-lives of the P1 mRNA ≥ fourfold. As in the case of the point mutations, the P2 mRNA was less severely affected (∼ twofold increase). Circular rpsT mRNA was also strongly stabilized, ∼ fivefold, by the Δ530–1061 allele in pRne-SG11. The other two C-terminal deletions, Δ589–1061 in pRne-SG12 and Δ731–1061 in pRne-SG13, also produced modest increases in the lifetimes of the P1 and P2 mRNAs, similar to those of the rne-131 allele in strain KBC1008. The stabilization of both species obtained with these three deletions suggests the existence of an additional, compensatory RNA recognition determinant in the C-terminus of RNase E, an idea consistent with earlier findings (Cormack et al., 1993; Lopez et al., 1999; Kaberdin et al., 2000; Leroy et al., 2002). Thus, we tested four shorter, internal deletions (pSG-Rne 14–16 and pSG-Rne 18) in an effort to identify the putative compensatory site. Only pSG-Rne 18 (Δ589–723) differed from its parent. It resulted in stabilization of both P1 and P2 rpsT mRNAs by approximately twofold (see Table 3). The measured rpsT half-lives in this strain are almost identical to those measured in pSG-Rne 12 (Δ589–1061) suggesting that the putative compensatory site maps to the Arg-rich and coiled-coil domains spanning residues 608–712 (Callaghan et al., 2004; Fig. 1). Curiously, circular rpsT mRNA was destabilized twofold in pSG-Rne 18 (Δ589–723), an anomaly that we are investigating.
Combining T170A with Δ530–1061 in SK9714/pSG31 produced consistent results as the half-life of the rpsT P1 mRNA is 310 s in the double mutation, higher than in T170A (120 s) or Δ530–1061 (260 s) (Table 3). The rpsT P2 mRNA exhibits the same trends. We were unable to recover a viable derivative of SK9714 containing both R169Q and Δ530–1061 and thus were unable to test whether the former allele would further exacerbate the effect of the latter.
We also examined three other well-characterized mRNAs, rpsO, lpp and ompA, for their response to the R169Q allele. These data are summarized in Table 4 while representative Northern blots for rpsO mRNAs are shown in Fig. 3 D-F. None of these RNAs is particularly sensitive to mutations in the 5′-sensor. The rpsO mRNA displays less than twofold increases in half-life in either R169Q or T170A; however, the differences are not significant. Moreover, the half-life of the rpsO mRNA did not increase significantly in strain SK9714/pSG31 (Δ530–1061/T170A) compared with the single mutants alone, unlike the rpsT mRNAs. Likewise, the lpp mRNA was unaffected by any of the rne alleles tested, including partial or full deletion of the C-terminal scaffold region. Although only a few of the ‘shuffled’ strains were tested, the ompA mRNA was unaffected by R169Q or T170A, consistent with the presence of a protective 5′-stem-loop on its 5′-end (Emory et al., 1992).
Table 4. Half-lives of selected mRNAs in strain SK9714 complemented by mutant plasmids
Effect of 5′-sensor mutations on stable RNA processing
We previously found that SK9714/pRne-SG4 (R169Q) exhibited a defect in the processing of 5S rRNA precursors that was qualitatively indistinguishable from that caused by rne-1 (Garrey et al., 2009; see below). We have now examined the processing of representative tRNA precursors, both polycistronic (argX and glyW) or monocistronic (pheU) that are known to depend on RNase E (Li and Deutscher, 2002; Ow and Kushner, 2002). Figure 4A shows the accumulation of hisR tRNA from the argX operon. A longer precursor, shown by the arrow in the left margin, is visible in SK5665 (rne-1) at non-permissive temperature (lane 2), but not in ‘shuffled’ derivatives of SK9714 carrying R169Q (pRne-SG4; lane 4) or rne Δ530–1061 (pRne-SG11; lane 10) as the sole source of RNase E activity. Shorter deletions (pRne-SG 14–18) were also tested but did not accumulate longer precursors (data not shown). Similar behaviour was observed for pheU tRNA from the pheU operon (Fig. 4B; compare lanes 1, 4 and 10 with lane 2), cysT tRNA from the glyW operon (Fig. 4C); argX tRNA from the argX operon, and leuZ tRNA (not shown). We conclude that none of the tRNAs examined requires the 5′-sensor or any part of the C-terminal scaffold domain for correct processing.
To assess the impact of the 5′-sensor on rRNA maturation, we measured accumulation of immature 16S and 5S rRNA (i.e. with all or part of the 5′-spacer intact) by primer extension. A schematic diagram illustrating the 5′-spacer and known cleavage sites for 16S rRNA is shown in Fig. 5A. The results of the primer extension analysis are shown in Fig. 5C and quantified in Fig. 5D. The two major products detected in this experiment are 136 and 202 nt, corresponding, respectively, to the 5′-termini of mature 16S rRNA and its immediate precursor generated by RNase E cleavage between residues −67 and −66. Longer products corresponding to cleavage by RNase III or to the primary transcript were readily detected with longer exposures (not shown). These endpoints were confirmed by separating the products on a sequencing gel with an appropriate sequence ladder (data not shown). The pattern of cDNAs was qualitatively similar in each of the strains examined. The 202 nt cDNA was, however, noticeably more intense in SK9714/pRne-SG4 (R169Q) (Fig. 5B, lane 3). Figure 5C summarizes quantitative measurements of the intensity of the 202 and 136 nt bands. In the wild-type strain, SK9714/pRne-SG1, immature 16S rRNA constitutes only 5% of the mature (i.e. the ratio of 202 nt to 136 nt products is 0.049). This ratio increases modestly to 0.065 in SK9714/pRne-SG5 (T170A) but quite significantly to 0.28 in SK9714/pRne-SG4 (R169Q) (Fig. 5B, lane 3). Deletion of the C-terminus of RNase E in SK9714/pRne-SG11 also increases the fraction of immature 16S rRNA. These data show that the sites of processing of pre-16S rRNA by RNase III and RNase E are normal in the mutant strains; however, the generation of the mature 16S 5′-end is retarded. Comparable data for 5S rRNA processing were also obtained (Fig. 5E). The major products detected by primer extension were 61 and 64 nt, corresponding to mature 5S rRNA and its immediate precursor, the product of RNase E cleavage at the ‘a’ site respectively (see the diagram in Fig. 5B). Among the minor products, a 145 nt cDNA was readily detected in total RNA from a temperature-sensitive mutant (rne-1) (Fig. 5E, lane 2) or from R169Q (lane 5). This corresponds to the 5′-end created by the action of RNase III. The accumulation of this product in R169Q shows that it is normally recognized by the 5′-sensor domain of RNase E. A ∼ 129 nt cDNA detected in all samples and a faint 99 nt band in lane 5 correspond to cleavages ∼+16 and ∼+45 residues 3′ to the RNase III site respectively. Occurrence of the +16 cleavage in R169Q implies that this site is 5′-end independent.
A hierarchy of substrate recognition by RNase E
This work was prompted by the unexpected finding that the rne R169Q mutation is viable although it severely affects RNase E activity in vitro by inactivating the enzyme's phosphate sensing mechanism (Garrey et al., 2009). Our data shed new light on the functions of this sensor and the diverse ways in which RNase E can recognize substrates. In particular, we have identified a hierarchy by which different classes of substrate are initially bound. The first hierarchy encompasses tRNA precursors including transcripts from the argX, pheU, glyW and leuZ operons as well as the 30S precursor to rRNAs (see below). These RNAs are processed without accumulation of intermediates in strains lacking the phosphate sensor, the C-terminal domain (residues 530–1061) or both the phosphate sensor and the Arg-rich region (residues 608–644). We presume that such substrates bind with high affinity to the catalytic domain of RNase E by direct interaction with the RNA-binding channel (Callaghan et al., 2005; Kime et al., 2009) without requiring the participation of other parts of the enzyme. The lpp mRNA may also fall into this class, but there is as yet no evidence that its decay depends on RNase E (O'Hara et al., 1995). A second hierarchy is exemplified by circular rpsT mRNA, which is insensitive to the status of the phosphate sensor but requires residues in the C-terminal domain of RNase E for the initial step in its degradation. Other mRNAs, which are also degraded by a direct entry (or ‘internal entry’) pathway, have also been reported to require the C-terminus of RNase E as their degradation is slowed by the rne-131 allele (Joyce and Dreyfus, 1998; Lopez et al., 1999; Baker and Mackie, 2003). The C-terminal domain and/or the proteins bound to it may provide direct contacts to the RNA or may localize RNase E to an environment that favours end-independent substrate recognition. In principle, there is a third hierarchy of RNAs, which depend on only the phosphate sensor, but not on the C-terminal domain, for the initial step of their degradation. We have not identified such RNAs in this survey although they likely exist. Short oligonucleotides, for example, fit this criterion in vitro (Jiang and Belasco, 2004; Jourdan and McDowall, 2008; Garrey et al., 2009). RNAs whose decay is initiated by RNase G would also fall into a similar group. The final hierarchy includes RNAs, which require the phosphate sensor and at least part of the C-terminal domain of RNase E. The rpsT P1 mRNA is an excellent example of this class of substrate as it is stabilized more strongly by the double mutation T170A/Δ530–1061 than by either single mutation alone. The rpsO mRNA behaves similarly, but its dependence on the phosphate sensor or the C-terminal domain is much weaker. This class of substrates may also include the processing of the 9S precursor to 5S rRNA. This substrate accumulates in rne R169Q (Garrey et al., 2009) and is reported to require part of the C-terminal domain for efficient cleavage in vitro, especially at the ‘b’ site (Kaberdin et al., 2000). In both the third and fourth hierarchies, there is an implied requirement for a prior step that generates a 5′-monophosphate to enable the engagement of RNase E (or RNase G). In the case of rpsT mRNA, this step is catalysed by RppH, a pyrophosphohydrolase (Celesnik et al., 2007; Deana et al., 2008). However, this requirement can be fulfilled by other enzymes, including RNase III (e.g. to generate the 9S precursor to 5S rRNA).
Role of the phosphate sensor in autoregulation of RNase E
Autoregulation of RNase E is believed to occur in at least three steps. First, RNase E binds to a hairpin structure (hp2) in its long 5′-untranslated sequence (Diwa et al., 2000; Shuck et al., 2009). Subsequently, the rne mRNA undergoes an initial cleavage (Mudd and Higgins, 1993). This cleavage then promotes the ‘bulk’ degradation of the rne mRNA, presumably catalysed by RNase E in its 5′-end-dependent mode. Our data show that regulated synthesis of RNase E is largely abrogated in rne R169Q and is characterized by overexpression of RNase E accompanied by accumulation of partially degraded fragments. It is known that the initial binding step as assessed by UV-cross-linking is impaired by the R169A mutation (Shuck et al., 2009). It is possible that R169Q acts similarly. Nonetheless, we believe that by analogy to the stabilization of the rpsT P1 mRNA, R169Q would greatly retard later events in autoregulation. Failure to recognize the first cleavage product efficiently would result in persistence of the ‘clipped’rne mRNA and accumulation of RNase E and its proteolytic fragments. Although RNase E is well known to be highly sensitive to partial proteolysis following cell lysis (Mudd and Higgins, 1993; Coburn and Mackie, 1998a), accumulation of proteolytic fragments has not previously been associated with its autoregulation. We postulate that the binding of an RNA via the phosphate sensor which is believed to induce a conformational change in RNase E (‘the mousetrap model’; Koslover et al., 2008) stabilizes the enzyme. In the absence of 5′-end-dependent RNA binding, the mutant RNase E would become susceptible to proteolysis. This raises the very interesting possibility that RNase E is also regulated post-translationally by the availability of its substrates.
Maturation of ribosomes
An unexpected finding of this investigation is a requirement of both the phosphate sensor and the C-terminal domain of RNase E for efficient maturation of 16S rRNA. The final step in the maturation of the 5′-end of 16S rRNA is normally catalysed by RNase G (Li et al., 1999). Since the −66 cleavage product (see Fig. 5A) accumulates normally in the wild-type but is enhanced in the mutant strains tested, and the expression of RNase G is unaffected by R169Q (data not shown), we believe that the explanation is indirect. We suggest that the action of RppH on mRNAs as well as the 5′-end-independent cleavage of various RNAs by RNase E and other enzymes, including RNase III, generate a pool of 5′-monophosphorylated RNAs. In the absence of a functional 5′-phosphate sensor on RNase E, the pool of potential substrates for RNase G is greatly expanded. Consequently, this pool competes with this enzyme's normal substrate, the immediate precursor to 16S rRNA. This hypothesis will explain the observed accumulation of immature 16S rRNA and is consistent with reports that RNase G can complement a deficiency of RNase E when the former is highly overexpressed (Lee et al., 2002; Chung et al., 2010).
Strains and plasmids
Escherichia coli K12 strains MG1693 (thyA715, rph-1), SK5665 (thyA715, rph-1, rne1), SK5691 (thyA715, rph-1, pnp7), SK7988 (thyA715, rph-1, pcnB::KmR) and SK9714 (thyA715, rph-1, rneΔ1018::bla, recA56, srlD300::Tn10 TcR/pSBK1 [rne+, CmR]) were obtained from Dr Sidney R. Kushner, University of Georgia. KBC1008 (thyA715, rph-1, rne131) was constructed by Dr Kristian Baker from MG1693 by P1 transduction. CJ1828 (araD39, Δ(ara, leu)7697, ΔlacX74, galU, galK, hsr, strA, rne-1; zce::Tn10, λez1 [rne::lacZ]) was obtained from Dr Joel Belasco (Jain and Belasco, 1995). Plasmids encoding rne are based on pRne-SG1 (originally called pSG-Rne in Garrey et al., 2009). This plasmid contains the entire rne 5′-UTR, coding sequence and 3′-UTR (3881 bp) inserted into the unique SalI site of pWSK129, a 6–8 copy plasmid expressing KmR. Point mutations and deletions were constructed using overlap PCR as described in the supplementary material to Garrey et al. (2009) and maintained in DH5α (obtained from Invitrogen). The sequences of oligonucleotides used for these constructions are available upon request. Mutations were verified by sequencing prior to reconstruction and again in the full-length plasmid. Cultures were grown in Luria–Bertani (LB) medium supplemented with thymidine (25 µg ml−1), and carbenicillin (100 µg ml−1, tetracycline (15 µg ml−1) and kanamycin or chloramphenicol (20 µg ml−1) as required. Displacement of pSBK1 from strain SK9714 by derivatives of pRne-SG1 (‘plasmid shuffling’) was performed as described (Garrey et al., 2009).
Preparation and analysis of RNA
Cultures were grown at 37°C in supplemented LB medium to mid-log phase (A600 ≤ 0.5). Rifampicin (dissolved in CH3OH) was added to 200 µg ml−1 to arrest transcription. Samples of 400 µl were withdrawn immediately before and after this addition and boiled in 200 µl of 1.5% (w/v) SDS, 300 mM Na-acetate, 30 mM EDTA for 60 s, then chilled on wet ice. Subsequent samples were taken at timed intervals and processed similarly. Samples were subsequently extracted serially with water-saturated phenol (pH 4.3), phenol-chloroform-isoamyl alcohol (25:24:1) and 2-butanol prior to precipitation with 2.5 volumes ethanol. RNA was recovered by centrifugation, dissolved in buffered 2 M NH4-acetate and reprecipitated with 2.5 volumes ethanol. After recovery and washing with 80% ethanol, RNA was dissolved in sterile H2O and quantified spectrophotometrically. For Northern blotting, samples containing 5 µg of total RNA were precipitated by ethanol in the presence of carrier yeast RNA then dissolved with buffered 90% formamide, boiled for 75 s, and separated on 5–8% polyacrylamide gels containing 8 M urea depending on the RNA to be detected. Separated RNAs were electrophoretically transferred to Hybond NX (GE Healthcare) and fixed with UV. Annealing was performed at 45.5°C in the presence of 50% formamide or at 37°C in its absence with the following probes: rpsT, 32P-49E cRNA (Mackie, 1986); rpsO, 5′-32P-AGTGCTACCTGAACTTCGG; lpp, 5′-32P-AGTAGAACCCAGGATTACCG; ompA, 5′-32P-CCCAGACGAGTGTAGATGTC; cysT, 5′-32P-GGAGTCGAACCGGACTAGACGG; hisR, 5′-32P-GGATTCGAACCCACGACAACTG; pheU, 5′-32P-GACTCGGAATTGAACCAAGGACAC.
Analysis of proteins
Cultures of interest were grown at 37°C to an A600 ≤ 0.5 when 1 ml was harvested by centrifugation, suspended in 200 µl sample buffer [120 mM Tris-HCl, pH 6.8, 3% (w/v) SDS, 0.1 M DTT, 10% (w/v) glycerol and 5 mM EDTA] and boiled for 5 min. Portions of each extract were separated by electrophoresis, blotted to nitrocellulose and detected with antisera against RNase E (1:20 000) or PNPase (1:10 000) as described (Rouleau et al., 1994) except that the secondary antibody was HRP-conjugated mouse anti-rabbit serum and detection used a Licor Odyssey.
Cultures of CJ1828 (Jain and Belasco, 1995) were transformed with pRne-SG1 or its derivatives. Cultures were grown at 37°C in LB supplemented with tetracycline and kanamycin. Samples were taken during mid-exponential growth and assayed as described by Miller (1972). Experiments were repeated three times.
We thank Dr Sidney Kushner, University of Georgia, and Dr Joel Belasco, New York University, for their gifts of strains, Dr Murray Deutscher for advice, and Drs George Jones and Ken McDowall for commenting on the manuscript. Operating support was provided by the Canadian Institutes of Health Research (MOP-5396) and the University of British Columbia.