RNase G complementation of rne null mutation identifies functional interrelationships with RNase E in Escherichia coli

Authors


Summary

The Escherichia coli endoribonucleases RNase E (Rne) and RNase G (Rng) have sequence similarity and broadly similar sequence specificity. Whereas the absence of Rne normally is lethal, we show here that E. coli bacteria that lack the rne gene can be made viable by overexpression of Rng. Rng-complemented cells accumulated precursors of 5S ribosomal RNA (rRNA) and the RNA component of RNase P (i.e. M1 RNA), indicating that normal processing of these Rne-cleaved RNAs was not restored by RNase G; additionally, neither 5S rRNA nor M1 RNA was generated from precursors by RNase G cleavage in vitro. Using DNA microarrays containing 4405 Escherichia coli open reading frames (ORFs), we identified mRNAs whose steady-state level was affected by Rne, Rng or the N-terminal catalytic domain of RNase E. Most transcript species affected by RNase E deficiency were also elevated in an rne deletion mutant complemented by Rng. However, approximately 100 mRNAs that accumulated in Rne-deficient cells were decreased by rng-complemention, thus identifying targets whose processing or degradation may be the basis for RNase E essentiality. Remarkably prominent in this group were mRNAs implicated in energy-generating pathways or in the synthesis or degradation of macromolecules.

Introduction

RNase E (Rne) is a multifunctional Escherichia coli ribonuclease that degrades bulk RNA (Ono and Kuwano, 1979), initiates decay of messenger and regulatory RNAs by cleaving them at highly specific sites (for recent reviews, see Coburn and Mackie, 1999; Steege, 2000), removes poly(A) tails from the 3′ ends of RNAs (Huang et al., 1998; Walsh et al., 2001), and assists in the maturation of a variety of catalytic RNAs, including 10Sa RNA (Lin-Chao et al., 1999), M1 RNA (Gurevitz et al., 1983), 5S rRNA (Ghora and Apirion, 1978) and 16S rRNA (Li et al., 1999; Wachi et al., 1999). CafA, which was initially identified by its role in chromosome segregation and cell division (Okada et al., 1994), was found subsequently to have 36% amino acid sequence identity with the N-terminal catalytic domain of RNase E (McDowall et al., 1993). Overproduction of CafA causes formation of chained cells and minicells (Okada et al., 1994) and, interestingly, this phenotype was also ob-served when RNase E was inactivated by temperature elevation in a strain carrying an rne temperature-sensitive (ts) mutation (i.e. the ams1 mutation) (Kuwano et al., 1977; Wang and Cohen, 1994). Recently, CafA has been shown to have endonuclease activity both in vivo (Li et al., 1999; Wachi et al., 1999; Umitsuki et al., 2001) and in vitro (Jiang et al., 2000; Tock et al., 2000), and consequently has been renamed RNase G (Rng). Its activity is 5′ end dependent (Jiang et al., 2000; Tock et al., 2000), like that of Rne (Mackie, 1998), and both ribonucleases attack substrates in A + U-rich regions. (Jiang et al., 2000; Tock et al., 2000). However, their precise cleavage sites are not identical (Li et al., 1999; Tock et al., 2000). Moreover, whereas cells lacking RNase E normally are non-viable (Apirion and Lassar, 1978; Ono and Kuwano, 1979), RNase G is dispensable for viability (Li et al., 1999; Wachi et al., 1999).

RNase G has been reported to partially suppress the temperature-sensitive growth phenotype of ams1 mutant bacteria when expressed from a multicopy plasmid (Wachi et al., 1997), but in another study did not allow growth of an RNase E-deficient strain when overproduced sixfold relative to wild-type RNase E levels (Jiang et al., 2000). We examined the functional relationship between these two enzymes and report here that RNase G expression can confer viability on an rne deletion mutant strain, indicating that Rng can carry out the essential action(s) of Rne. The ability of Rng to complement an rne deletion has enabled us to investigate the basis for RNase E essentiality and to identify specific RNAs attacked in E. coli cells by each of these enzymes.

Results

Complementation of rne deletion by overexpression of RNase G

To resolve the question of whether RNase G can substitute functionally for RNase E, we constructed an E. coli strain (KSL2000, Table 1) in which a chromosomal deletion in rne was complemented by a plasmid-borne rne gene under the control of an arabinose gene promoter (the Kmr plasmid pBAD-RNE, Fig. 1A and Table 1). Addition of 0.1% arabinose to cultures of KSL2000 allowed the synthesis of C-terminally hexahistidine-tagged RNase E at wild-type levels and consequently enabled survival and growth of this rne deletion mutant (see below). Introduction of an incompatible ampicillin resistance (Apr) plasmid expressing RNase G with a hexahistidine tag at the C-terminus under the control of the isoprogyl-thiogalacto-side (IPTG)-inducible lacUV5 promoter (pRNG3) into KSL2000 (Fig. 1A), and selection for the incoming plasmid by growing transformants containing both plasmids (pBAD-RNE and pRNG3) in the presence of ampicillin (50 μg ml–1) and 100–1000 μM IPTG for 40 generations, resulted in displacement of the resident Kmr plasmid by the Apr Rng-expressing construct – as indicated by both the antibiotic resistance phenotype and restriction enzyme analysis of plasmid DNA (data not shown here). Western blot analysis using anti-RNase E polyclonal antibody, which cross-reacted weakly with overproduced RNase G (lane 4), showed no detectable Rne protein in these cells (Fig. 1B). The presence of Rng at about twice the normal level of Rne, as judged from Western blot analysis using antibody to the histidine tag attached to each of these proteins (Fig. 1B), was sufficient for viability of the rne deletion mutant.

Table 1. E. coli strains and plasmids.
Strain or plasmidDescriptionSource of reference
  • a.

    See Experimental procedures for detailed description and construction scheme for plasmids.

  • Apr, ampicillin resistance; Kmr, kanamycin resistance; Tcr, tetracycline resistance; PBAD, BAD promoter; PlacUV5, lacUV5 promoter.

Strains
N3433 lacZ43, relA, spoT, thi-1 Goldblum and Apirion (1981)
KW153Same as N3433 but rne::cat [pMW135] Wang and Cohen (1994)
KSL1000Same as KW153 but recA::Tn10This study
KSL1004Same as N3433 but rngThis study
KSL2000Same as KSL1000 but [pBAD-RNE]This study
KSL2002Same as KSL2000 but [pNRNE5]This study
KSL2003Same as KSL2000 but [pLAC-RNE2]This study
KSL2004Same as KSL2000 but [pRNG3]This study
CS520 recA::Tn10Laboratory collection
Plasmids a
pSC101Tcr Cohen and Chang (1977)
pPM30pSC101 ori, Apr Meacock and Cohen (1980)
pBAD-RNEpSC101 ori, Kmr, rne under PBADThis study
pLAC-RNE2pSC101 ori, Apr, rne under PlacUV5 Lee and Cohen (2001)
pNRNE5pSC101 ori, Apr, N-rne under PlacUV5This study
pRNG3pSC101 ori, Apr, rng under PlacUV5This study
pTRP2ColE1 ori, TcrThis study
pET21aColE1 ori, AprNovagen
pET28aColE1 ori, KmrNovagen
pKD13Amplification of Kmr cassette Datsenko and Wanner (2000)
pM1Synthesis of pM1 RNAThis study
pM21Synthesis of GGG-RNAI RNA Helmer-Citterich et al. (1988)
Figure 1.

Functional complementation of RNase E by overproduction of RNase G.

A. Plasmid displacement procedure. Promoters for RNase genes (PBAD, PlacUV5), replication origin of plasmids (pSC101 ori), and antibiotic markers (Apr, Kmr) are indicated

B. Western blot analysis of strains obtained from the plasmid displacement experiments. 1, N3433 plus pPM30; 2, KSL2003; KSL2002; 4, KSL2004. The cultures were grown in the presence of 100 μM IPTG, except KSL2003 (10 μM), and harvested in the mid-log phase (OD600 = 0.6). The same membrane probed with anti-RNase E polyclonal antibody was stripped and reprobed with anti-His-tag polyclonal antibody.

C. Purified proteins from KSL2002–2004 were analysed on 7.5% SDS- polyacrylamide gel. One microgram of each His-tag affinity-purified protein was loaded in each lane and stained with Coomassie blue R-250. Smaller protein bands present in the RNase E lane were found by Western blot analysis to be degradation products of RNase E.

D. Two picomoles of 5′-end-labelled BR13 was incubated with no protein, 50 ng of RNase E, 100 ng of N-RNase E or 100 ng of RNase G in 40 μl of 1 × cleavage buffer at 37°C. A 10 μl aliquot of each sample was taken out at each time point indicated and mixed with an equal volume of formamide loading buffer. Samples were denatured at 70°C for 10 min and loaded onto a 15% polyacrylamide gel containing 8 M urea. The markers (10*) were a 1 nt ladder generated by limited polynucleotide phosphorylase (Sigma) decay of the substrate.

E. The percentage of uncleaved BR13 in the gel shown in D was quantitated using a Molecular Dynamics Phosphoimager and plotted.

In parallel experiments, the Apr plasmids pNRNE5 and pLAC-RNE2, which express N-Rne, the N-terminal catalytic domain of Rne and full-length Rne respectively, were similarly able to displace the resident Kmr pBAD-RNE plasmid when induced by IPTG. Plasmid pNRNE5 displaced the resident plasmid when N-Rne production was induced by 10, 100 or 1000 μM IPTG, whereas displacement by pLAC-RNE2 was not observed at an IPTG concentration >10 μM, consistent with the known toxicity of overproduction of RNase E (Diwa et al., 2000).

His-tagged N-Rne, Rne and Rng proteins were purified from the rne deletion mutant strains KSL2002–2004, respectively, by affinity column chromatography (Fig. 1C). The specific activity of the resulting Rne and N-Rne preparations on a single-stranded oligonucleotide substrate was approximately equal, as calibrated by their abilities to cleave the 13-base oligonucleotide BR13 (Fig. 1D and E) (McDowall et al., 1995). The specific activity of RNase G was slightly higher.

Processing of essential catalytic RNAs in an rne-deleted strain expressing N-Rne or Rng

RNase E was discovered initially by its ability to process 9S rRNA and generate 5S rRNA in E. coli cells, and the shift of rne ts bacteria to a non-permissive temperature leads to the accumulation of non-matured precursors of 5S rRNA in vivo (Ghora and Apirion, 1978; Gurevitz et al., 1983); pM1 RNA, the precursor of the RNA component of RNase P (Kole and Altman, 1979; Reed et al., 1982), also accumulates in these cells. We found that complementation of the rne deletion mutant by production of sufficient Rng to restore cell viability did not detectably assist in the processing of either of these essential catalytic RNA precursors (Figs 2B and 3B). The 5S rRNA precursor bands accumulating in rne-deleted cells complemented by Rng (i.e. the KSL2004 strain; Fig. 2B, lane 6) were identical in size to, and greater in quantity than, the species accumulating in rne-deleted cells in which synthesis of RNase E from the pBAD-RNE plasmid (i.e. the KSL2000 strain) was turned off by shift to media lacking arabinose (Fig. 2B, lane 2). In contrast, 9S rRNA was processed normally in the chromosomal rne deletion mutant complemented by N-Rne (KSL2002; Fig. 2B, lane 5). Together, these results argue against the notion that processing of 9S rRNA is the essential function provided by Rng complementation of rne null mutant bacteria; supporting this view is the failure of purified RNase G to cleave 9S rRNA transcripts in vitro (Fig. 2C), which independently has been observed by others (Tock et al., 2000). These results further imply that either 5S rRNA is not required for functional ribosome assembly in E. coli (as has been found recently for Bacillus subtilis; Condon et al., 2001) or, alternatively, that a pathway independent of both RNase E and RNase G can accomplish 5S rRNA maturation.

Figure 2.

Processing of 9S rRNA.

A. In vitro synthesized 9S rRNA transcript. RNase E cleavage sites are indicated as a, b and c and binding sites for probes used in B are shown.

B. Northern blot analysis of 5S rRNA precursors, which were identified by probing with 5S rRNA (probe I) and then retesting with 9S rRNA-specific probes that bind 5′ to residue 39 (probe II) and 3′ to residue 15 (probe III) of 9S rRNA. Lanes 1–6 contain 30 μg of total RNA prepared from: 1, N3433 plus pPM30; 2, KSL2000 (no arabinose); 3, KSL2000 (0.1% arabinose); 4, KSL2003 (10 μM IPTG); 5, KSL2002 (100 μM IPTG); 6, KSL2004 (100 μM IPTG). To prepare total RNA from KSL2000 (no arabinose), culture was grown to middle log phase in the presence of 0.1% arabinose, harvested, washed twice with plain Luria–Bertani (LB) medium and reinoculated into LB medium containing no arabinose (OD600 = 0.1). It was further incubated for 150 min (OD600 = 0.5) at 37°C and 250 rpm and harvested for total RNA preparation. Lane 0 contains 10 ng of 9S transcript used in C that was cleaved with purified RNase E as a reference.

C. 9S cleavage in vitro. One picomole of in vitro synthesized 9S rRNA transcript internally labelled with [32P]-UTP was cleaved in the same way described in Fig. 1D except that double the quantity of proteins was used (100 ng of RNase E, 200 ng of N-RNase E, or 200 ng of RNase G) and loaded onto a 9% polyacrylamide gel containing 8 M urea.

Figure 3.

Processing of pM1 RNA.

A. In vitro synthesized pM1 RNA transcript. RNase E cleavage sites are indicated as arrowheads and M1 probe binding site is shown.

B. Northern blot analysis of M1 RNA precursors. Lanes 1–6 contain the same samples described in Fig. 2B.

C. One picomole of pM1 RNA transcript, which was synthesized in vitro, internally labelled with [32P]-UTP, was cleaved in the same way described in Fig. 2C and loaded onto 6% polyacrylamide gel containing 8 M urea.

Analogous results were obtained for pM1 RNA in the Rng-complemented rne deletion strain; the precursor band (415 nt) accumulating in Rng-complemented cells was the same size as in Rne-depleted bacteria (Fig. 3B). These data support the previous suggestion that an RNase E-independent pathway exists for pM1 RNA processing (Kim et al., 1999). This putative pathway also does not seem to involve the action of RNase G, as the ratio of precursor to mature M1 RNA was greater (0.48) in the rne-deleted cells complemented by Rng (strain KSL2004) than was observed following the turn-off of Rne in the same parental strain (KSL2000) (0.13). Furthermore, unlike RNase E and N-RNase E, purified RNase G did not cleave pM1 RNA in vitro (Fig. 3C).

Limited cleavage of RNA I by RNase G in vivo

RNase E controls the copy number of ColE1 type replicons by cleaving RNAI, an antisense repressor of the replication primer, RNAII (Lin-Chao and Cohen, 1991). During the course of our experiments, we observed that Rng overexpression, either in the presence of a functional rne gene or in bacteria deleted for rne, did not affect the copy number of the ColE1-type plasmids (pTRP2 or pET28a, Table 2). As the copy number of such plasmids normally is modulated by RNA I cleavage, this obser-vation suggested that overproduction of RNase G to a level that confers cell viability does not inactivate RNA I in vivo. This interpretation was confirmed by Northern blot analysis, which showed the persistence in Rng-complemented cells of RNA I-hybridizable species that accumulated following repression of RNase E production (Fig. 4B, lanes 2 and 5). Moreover, notwithstanding a decrease in plasmid copy number in rne-deleted cells complemented by Rng, RNA I-hybridizable species were detected at a higher concentration in this strain (Fig. 4B, lane 5) than in wild-type controls (Fig. 4B, lanes 1 and 3). In contrast, complementation by N-Rne was associated with normal processing of RNAI (Fig. 4B, lane 4). In vitro, purified RNase G cleaved a derivative of 5′-triphosphorylated RNA I (GGG-RNA I) at one-fifteenth and one-fifth the rate, respectively, of RNase E and N-RNase E (Fig. 4C), as has been observed by others (Jiang et al., 2000).

Table 2. Effects of RNase G overexpression on copy number of ColE1 type plasmid A.
  rne + (DH5α) rne (KSL2002–2004)
pSC101
derivative
ColE1
plasmid
IPTG (mM)aCopy
numberb
ColE1
plasmid
IPTG (mM)Copy
numberb
  • a. IPTG was added at OD 600 = 0.1 and cultures were harvested for plasmid preparation 2 h after induction.

  • b. Plasmid copy number was calculated relative to a concurrently present pSC101 derivative, whose replication is independent of Rne ( Lin-Chao and Cohen, 1991), by setting the molar ratio of pSC101 derivative plasmid (pPM30) to ColE1 plasmid (pTRP2 or pET28a) to 1. Densitometric measurements of bands corresponding to each plasmid were converted to actual ratios after normalizing the values according to the size of pSC101 derivative plasmids.

  • c.

    The parental strain of KSL2002–2004, N3433, was used as a wild-type control as rne is an essential gene.

pPM30pTRP211pET28a0.11c
pLAC-RNE2pTRP217.5 ± 0.3pET28a0.011.0 ± 0.3
pNRNE5pTRP214.6 ± 1.1pET28a0.10.7 ± 0.0
pRNG3pTRP211.2 ± 0.4pET28a0.10.5 ± 0.1
Figure 4.

Cleavage of RNA I.

A. In vitro synthesized GGG-RNA I transcript. The RNase E cleavage site is indicated as an arrowhead and the RNA I probe binding site is shown.

B. Northern blot analysis of RNA I species. Lanes 1–5 contain 30 μg of total RNA prepared from: 1, N3433 plus pPM30 and pET28a; 2, KSL2000 plus pET21a (no arabinose); 3, KSL2003 plus pET28a (10 μM IPTG); 4, KSL20002 plus pET28a (100 μM IPTG); 5, KSL2004 plus pET28a (100 μM IPTG). Lane 0 contains 10 ng of GGG-RNA I that was cleaved with purified RNase E as a reference. GGG-RNA I transcript was not gel purified to make a use of 3′ truncated GGG-RNA I species as a size marker. *Band corresponding to the size of uncleaved pppRNA I.

C. One picomole of GGG-RNA I transcript, which was synthesized in vitro, internally labeled with [32P]-UTP, was cleaved in the same way described in Fig. 2. Samples were loaded onto 10% polyacrylamide gel containing 8 M urea.

Action of RNase E and RNase G on mRNA decay

The experiments shown in Figs 2 and 3 indicate that RNase G complementation of an rne deletion mutation does not stem from any ability of RNase G to process precursors of 5S rRNA and M1 RNA. To identify RNAs whose cellular levels are affected by RNase G complementation of RNase E deficiency, we used DNA microarrays containing 4405 E. coli ORFs to assess globally the effects of these ribonucleases on mRNA abundance. These experiments compared abundance of cDNAs differentially labelled with fluorescence dyes and hybridized to DNA templates arrayed on glass slides. cDNAs were prepared from RNA samples isolated from isogenic N3433-derived strains that contained deletions in rne or rng (Fig. 5A) and harboured pPM30, pBAD-RNE, pNRNE5 or pRNG3; the level of ribonuclease expression was determined in the same cultures used for RNA isolation (Fig. 5B). Interestingly, the amount of RNase G protein produced from plasmid pRNG3 in an rne+ background (N3433) was about one-quarter the level observed in the isogenic rne deletion mutant (N3433rne) under the same culture conditions, suggesting that RNase E may downregulate the expression of RNase G by stimulating decay of the rng transcript. This notion was supported by our micro-array data, which showed a threefold increase in steady-state level of the endogenous rng transcript in RNase E-deficient cells (i.e. N3433rne plus pBAD-RNE cultured in the absence of arabinose) (Table 3). The RNase E cleavage site(s) that we postulate is present in rng transcripts may reside in the protein-coding region of rng, as pRNG3 does not include the 5′ or 3′ untranslated regions of this gene.

Figure 5.

Strains used in DNA microarray analysis.

A. Construction of rng knock-out strain. The procedure is described in the Experimental procedures section and the primers used to amplify the rng region from chromosome are rng-5′-UTR (5′−GTGGATTTTCTTGCTGATGC) and rng-3′ (5′−ATTACGACGTCAAACTGCTC). Diagrammatic representation of polymerase chain reaction (PCR)-amplified regions encompassing rng in the chromosome of N3433, N3433rng::Kmr after insertion of a kanamycin marker, and N3433rng after its subsequent removal are shown on the left. Corresponding PCR products analyzed by electrophoresis in 1% agarose gel and stained by ethidium bromide are shown on the right. The markers were bands in an 1 kbp DNA ladder purchased from Invitrogen. FRT stands for FLP recognition target.

B. Western blot analysis of strains used in DNA microarray analysis. Strains used were: 1, N3433rne plus pBAD-RNE (no arabinose); 2, N3433 plus pPM30 (no arabinose); 3, N3433 plus pPM30 (100 μM IPTG); 4, N3433 plus pRNG3 (100 μM IPTG); 5, N3433rne plus pRNG3 (100 μM IPTG); 6, N3433rne plus pNRNE5 (100 μM IPTG); 7, N3433rng plus pPM30 (100 μM IPTG). N3433rne plus pBAD-RNE (no arabinose) was grown as described in Fig. 2B. The membrane probed with anti-RNase E monoclonal antibody was stripped and reprobed with anti-His-tag polyclonal antibody and subsequently with anti-S1 polyclonal antibody to provide an internal standard to demonstrate that different lanes contain the same amount of cell extract.

Table 3. Steady-state level of mRNAs encoding degradosome components and RNase G.
StrainPlasmidRnea rng dnaKenomopApnpppkrhlB
  • a.

    Rne concentration was estimated from Western blot analysis shown in Fig. 5. (N) indicates N-terminal catalytic domain of RNase E.

  • All other values are from DNA microarray data shown in Fig. 6.

N3433rnepBAD-RNE0.053.1 ± 0.71.8 ± 0.21.3 ± 0.11.5 ± 0.32.2 ± 0.32.8 ± 0.72.2 ± 0.3
N3433rnepNRNE520 (N)1.2 ± 0.20.4 ± 0.00.3 ± 0.00.4 ± 0.10.9 ± 0.01.1 ± 0.10.9 ± 0.1
N3433rnepRNG3047.3 ± 0.21.3 ± 0.10.9 ± 0.11.0 ± 0.01.9 ± 0.11.5 ± 0.02.0 ± 0.0
N3433rngpPM3010.2 ± 0.11.5 ± 0.01.9 ± 0.11.8 ± 0.21.2 ± 0.11.1 ± 0.11.1 ± 0.1
N3433pRNG3113.5 ± 2.90.6 ± 0.10.6 ± 0.10.8 ± 0.21.2 ± 0.11.0 ± 0.11.0 ± 0.0

As shown in Fig. 6A, a modal distribution of mRNA abundance was shifted 0.4 to 0.6-fold (x-axis log2 ratio) in rne-deleted bacteria made viable by overexpression of Rng, indicating that the steady-state level of more than 40% of all mRNAs increased at least 1.5-fold; a similar shift was observed in cells complemented by N-Rne. This implies that neither the N-terminal catalytic domain of Rne nor Rng restored RNA decay to normal in rne-deleted bacteria (Fig. 6A, graphs 1–3), a conclusion made earlier for N-Rne (Lopez et al., 1999; Ow et al., 2000). Deletion of rng in rne+ bacteria increased the abundance of 18 cellular mRNAs by more than 1.5-fold (Table 4), identifying these mRNAs as possible targets of RNase G. Among the transcripts most prominently affected by rng deletion were adhE, which is known to be induced by anaerobic growth, pgi, glk, tpiA and eno, which encode enzymes involved in glycolysis, and mopA and clpB, which specify heat shock proteins. Eleven of the mRNAs that were stabilized in the N3433rng mutant (adhE, pgi, clpB, tpiA, eno, mopA, ndh, mgsA, panB, ahpF and B1983) were decreased in abundance when RNase G expression was elevated in rne+ cells (N3433 plus pRNG3; Table 4), supporting this interpretation. Some mRNAs (i.e. rpoE, murA, ybqJ and B2326) that were decreased in abundance in the rng deletion mutant, N3433rng, were also less abundant in N3433, which contained pRNG3, implying that their decrease in N3433rng is not a direct consequence of rng deletion.

Figure 6.

Analysis of microarray data.

A. Distributions of log-transformed fluorescence ratios obtained from individual arrays. The x-axis shows frequency of occurrence and the y-axis log2 ratio (red/green).

B. Hierarchical clustering of the 105 genes. A total of 1113 genes were selected whose relative mRNA levels were 1.5-fold or more changed in at least 5 out of 10 slides. Of these1113 genes,105 whose mRNA levels were decreased by overproduction of N-RNase E and RNase G were further clustered. See Experimental Procedures for detailed microarray procedures. Cy-5-labelled cDNAs were synthesized with RNA samples taken from strains: 1, N3433rne plus pBAD-RNE (no arabinose); 2, N3433rne plus pNRNE5 (100 μM IPTG); 3, N3433rne plus pRNG3 (100 μM IPTG); 4, N3433rng plus pPM30 (100 μM IPTG); 5, N3433 plus pRNG3 (100 μM IPTG). RNA samples from isogenic wild-type strain (N3433) harbouring pPM30 were labelled with Cy-3 and comparisons between paired mutant and wild-type cultures were done directly. cDNA ratios are represented in tabular form according to the colour scale shown at the top; rows correspond to individual genes and columns correspond to different strains. Red shades represent an increase and green shades represent a decrease in the level of the transcript in wild-type cells. Black indicates no detectable change in transcript level and grey represents the absence of data. The amounts of each RNase present in the strains used are indicated as a threshold of cellular concentration of each protein in wild-type strain (N3433). The amount of each protein was estimated from Western blot analysis (Fig. 5B) and microarray data (Table 2).

Table 4. Genes whose mRNA level was changed by deletion of rng.
GeneaDescriptionLevel of mRNA N3433rngbN3433 (pRNG3)b
  • a.

    Forty-two genes whose mRNA level changed 1.5-fold in three out of four slides prepared with RNA samples from N3433rng were initially chosen. Twenty-three of the 42 genes whose standard error was equal to or less than 20% of the average level for that mRNA were further selected and listed.

  • b.

    Level of mRNA was calculated from at least two independent experimental sets and shown as average ± standard error.

adhE Alcohol dehydrogenase3.7 ± 0.20.7 ± 0.1
pgi Glucosephosphate isomerase 12.8 ± 0.50.7 ± 0.1
glk Glucokinase 12.3 ± 0.11.0 ± 0.1
clpB Heat shock protein, ClpB2.3 ± 0.50.6 ± 0.1
tpiA Triosephosphate isomerase 12.0 ± 0.10.8 ± 0.0
eno Enolase1.9 ± 0.10.6 ± 0.1
mopA Heat shock protein, GroEL1.8 ± 0.20.8 ± 0.2
ndh Respiratory NADH dehydrogenase1.8 ± 0.30.9 ± 0.2
yhgI Hypothetical protein1.8 ± 0.21.7 ± 0.0
yfhO Putative aminotransferase 11.8 ± 0.21.6 ± 0.1
yeiP Putative elongation factor1.7 ± 0.11.0 ± 0.1
mgsA Methylglyoxal synthase1.7 ± 0.10.7 ± 0.0
panB 3-Methyl-2-oxobutanoate hydroxymethyltransferase1.7 ± 0.20.8 ± 0.0
nagB Glucosamine-6-phosphate deaminase1.7 ± 0.21.0 ± 0.1
ahpF Alkyl hydroperoxide reductase, F52a subunit1.6 ± 0.10.8 ± 0.0
srlA PTS system, glucitol/sorbitol-specific IIC component1.6 ± 0.11.4 ± 0.1
B1983 Hypothetical protein1.6 ± 0.10.8 ± 0.1
acs Acetyl-CoA synthetase1.6 ± 0.11.0 ± 0.2
rpoE RNA polymerase, sigma-E factor0.7 ± 0.10.8 ± 0.1
murA UDP-N-glucosamine 1-carboxyvinyltransferase0.6 ± 0.10.6 ± 0.0
ybqJ Putative carboxylase0.6 ± 0.10.7 ± 0.2
B2326 Putative transporting ATPase0.5 ± 0.00.6 ± 0.0
rng Cytoplasmic axial filament, RNase G0.2 ± 0.113.5 ± 2.9

We hypothesized that the population of transcripts that increased in abundance during RNase E deficiency but decreased during RNase G overexpression in the mutant cells would include mRNAs whose cleavage by RNase G enables complementation. A total of 105 mRNAs, including transcripts that encode three protein components (DnaK, enolase and GroEL) of the E. coli degradosome complex (Carpousis et al., 1994; Py et al., 1994; Miczak et al., 1996; Py et al., 1996), satisfied these criteria (Fig. 6B). The effects of mutation or overexpression of RNase E and RNase G on the abundance of transcripts encoding these and other degradosome components is shown in Table 3. Transcript levels of polynucleotide phosphorylase, polyphosphate kinase and RNA helicase were increased in abundance following depletion of RNase E and returned to normal following complementation by N-Rne, suggesting that their levels are regulated by RNase E. However, the steady-state level of these transcripts was not affected by Rng expression, indicating that they are not substrates for RNase G in vivo.

Discussion

The experiments we report indicate that overexpression of RNase G allows E. coli cells to survive and reproduce in the absence of the rne gene and RNase E protein, indicating that RNase G has the inherent ability to carry out the functions that make RNase E normally essential. The level of RNase G that accomplished complementation was twice the normal cellular concentration of RNase E (Fig. 1B) and more than 50 times the normal amount of RNase G (Table 2; see Experimental procedures). Although expression of Rng previously was found not to complement Rne deficiency (Jiang et al., 2000), the earlier results may reflect the toxicity (Okada et al., 1994) of the very high cellular concentrations of Rng used in those investigations. Alternatively, the epitope-tagged RNase G variant used by Jiang et al. (2000), which contains 31 extra amino acids at the amino terminus, may have been functionally impaired in vivo although it showed in vitro ribonucleolytic activity comparable to N-Rne.

Characteristic decay intermediates of 9S rRNA and the precursor of M1 RNA accumulated in vivo in rne-deleted cells complemented by RNase G. These in vivo observations, together with in vitro evidence that purified RNase G does not cleave 9S and pM1 RNA in vitro (Tock et al., 2000; our results), argue that RNase G complementation of rne-deleted bacteria is not accomplished through the processing of these essential catalytic RNA precursors. While we do not exclude the possibility that appropriate processing of these RNAs by RNase G occurs in vivo at an undetectable level or that maturation of 9S rRNA into 5S rRNA is not required by E. coli, the simplest inter-pretation of our results is that a pathway that does not involve either RNase E or RNase G can carry out 9S and pM1 RNA processing in E. coli cells. Consistent with this view is recent evidence for the existence of an RNase E-independent pathway in the maturation of M1 RNA (Kim et al., 1999).

adh mRNA and 16S rRNA precursor, both of which have been shown to be natural substrates for RNase G in vivo, are cleaved at their 5′ ends by RNase III, leaving 5′-monophosphorylated termini, prior to attack by RNase G (Aristarkhov et al., 1996; Li et al., 1999; Wachi et al., 1999; Umitsuki et al., 2001). In Rng-deficient cells, the level of adh transcript is significantly higher than in wild-type cells (Umitsuki et al., 2001), suggesting that the RNase III-processed adh mRNA accumulates. Moreover, translation of adh mRNA, which requires cleavage in the 5′ untranslated region by RNase III (Aristarkhov et al., 1996), is increased. In our investigations, Rng over-expression, unlike overexpression of Rne and N-Rne, did not affect the copy number of ColE1-type plasmids, and Northern blotting showed accumulation of RNA I-hybridizing transcripts. These in vivo findings are consistent with evidence that purified RNase G cleaves 5′-triphosphorylated RNA I in vitro less efficiently than either RNase E or N-RNase E (Jiang et al., 2000; also our results). As RNase G prefers 5′-monophosphorylated substrates (Tock et al., 2000), it normally may act in vivo primarily or entirely on substrates generated by other endonucleases, as has been shown for adh mRNA and 16S rRNA.

Our global investigations of mRNA steady-state levels using DNA microarrays containing 4405 E. coli ORFs identified mRNA species affected by depletion of RNase E, deletion of the rng gene or overproduction of RNase G. We found only 11 mRNAs (adhE, pgi, clpB, tpiA, eno, mopA, ndh, mgsA, panB, ahpF and B1983) whose overall abundance was controlled by the cellular concentration of RNase G. The most prominently affected transcripts encode translation products that function in the utilization of energy sources (adhE, pgi, ndh, tpiA and eno), suggesting that RNase G may be required for RNA metabolism under specialized growth conditions. Interestingly, the transcripts affected by RNase G also became more abundant when RNase E was depleted.

A shift in the modal distribution of mRNA abundance by 0.4- to 0.6-fold (x-axis log2 ratio) occurred in rne-deleted strains made viable by overexpression of either N-Rne or Rng in Rne-depleted cells (Fig. 6A, graphs 1–3). This indicates that restoration of viability by the N-terminal catalytic domain of Rne or Rng was not accompanied by return of mRNA degradation to entirely normal; never-theless, the elevated steady-state level observed for 105 mRNAs after depletion of RNase E was completely or partially reversed. These mRNAs included the 11 that were increased in abundance by deletion of rng in a strain that contains an intact and functional rne gne (see above), confirming the effect of Rng on their decay. We infer that the remaining transcripts in this group are degraded by both RNase E and RNase G. Remarkably, about 90% of the 105 mRNAs encode proteins involved in energy-generating pathways (glycolysis, aerobic respiration and the tricarboxylic acid cycle), or the metabolism of macromolecules and intermediary compounds.

The finding that the cellular level of metabolic gene transcripts is affected by both rne deletion and RNase G complementation suggests that the lethality normally associated with the absence of RNase E may result from the perturbation of cell metabolism. While the native cellular level of RNase G is insufficient to deal with the accumulation of metabolic gene transcripts in cells deficient in RNase E, as we observed by microarray analysis, overexpression of RNase G reversed this accumulation and enabled cell viability.

Experimental procedures

Strains and plasmids

Strains and plasmids used in this study are listed in Table 1. The isogenic E. coli K-12 strains KW153(rne::cat) (Wang and Cohen, 1994), KSL1000 (rne::cat, recA::Tn10) and KSL1004 (rng) are N3433 (lacZ43, relA, spoT, thi-1) (Goldblum and Apirion, 1981) derivatives. KSL1000 was constructed from KW153 by phage P1-mediated transduction using CS520 (recA::Tn10) as a donor strain. KSL1004 was constructed by deleting the open reading frame (ORF) of rng using the procedure described by Datsenko and Wanner (2000). PCR primers used were 5′−rng-KO (5′−GTGAGAAAAGGGA TAAACATGACGGCTGAATTGTTAGTAAACGTAACGGTG TAGGCTGGAGCTGCTTC; the 16 codons of the 5′-rng ORF are in bold type) and 3′−rng-KO (5′−TTACATCATTACGAC GTCAAACTGCTCCTGGTTATAGAGCGGTTCAATATTC CGGGGATCCGTCGACC; the complementary sequence of the 16 codons of 3′-rng ORF is in bold type) and pKD13 (Datsenko and Wanner, 2000) was used as a template. Insertion of a kanamycin resistance marker in rng and its subsequent deletion were confirmed by amplifying the chromosomal region encompassing the rng locus (Fig. 5A), and the absence of rng transcripts in these strains was confirmed by microarray analysis (see Results).

Plasmid pBAD-RNE is a derivative of pSC101 (Cohen and Chang, 1973; 1977) that directs the conditional synthesis of a full-length carboxy-terminally tagged form of E. coli RNase E under control of the BAD promoter (Fig. 1A). A hexahistidine affinity tag was inserted right before the stop codon of rne (PVEHHHHHH•, wild-type C-terminus is underlined and stop codon is indicated as •) and a stronger ribosome binding sequence (RBS) was incorporated (5′-GCGGCCGCAGGAG GTTACGATG; the RBS is underlined and the start codon is in bold type). pLAC-RNE2 is a derivative of pPM30 (Meacock and Cohen, 1980) that contains the same RNase E gene of pBAD-RNE under control of lacUV5 promoter. pNRNE5 is a derivative of pLAC-RNE2 which encodes a carboxy-terminally tagged, truncated form of RNase E lacking the C-terminal 562 residues (N-RNase E). It was constructed by replacing the NotI and SpeI fragment (entire rne coding region) of pLAC-RNE2 with PCR DNA containing N-terminal 499 residues of RNase E and a hexahistidine affinity tag. The C-terminal amino acid sequence of N-RNase E is TLSHHH HHH• (the wild-type sequence of residues 497–499 is underlined and the stop codon is indicated as •). Plasmid pRNG3 encodes E. coli RNase G with a hexahistidine affinity tag at the C-terminus and was constructed by replacing the NotI and SpeI fragment of pNRNE5 with PCR DNA containing all 495 residues of RNase G and a C-terminus tag. A stronger RBS and a more common start codon (ATG instead of GTG) were incorporated for a better expression of RNase G in this construct (5′-GCGGCCGCTTTAAGAAGGAGATATACAT ATG; the RBS is underlined and the start codon is in bold type), and these modifications resulted in a twofold increase in Rng production compared with that expressed from a similar construct containing intact 5′-UTR of rng (data not shown here). pTRP2 was constructed by cloning a mutant tryptophan promoter (Hui and de Boer, 1987) into the EcoRI and PstI sites of pBR322 and the promoter cassette was prepared by annealing two oligonucleotides, Ptrpc top (5′-AATTCGAGCTGTTGACAATTAATCATCGAACTAGTTTA ATGTGTGGAAGATCTGCAG) and Ptrpc bottom (5′-GATCC TGCAGATCTTCCACACATTAAACTAGTTCGATGATTA ATTGTCAACAGCTCG), followed by digestion with PstI. pM1 plasmid was constructed by amplifying the region of pM1 RNA from N3433 chromosome using two PCR primers, M1–5′ (5′-ATCCGCGGATCCATTTAGGTGACACTATAGAA GCTGACCAGACAGTCGC; SP6 promoter is underlined and 5′ region of pM1 RNA is in bold type) and M1–3′ (5′-AT GAATTCAAGCTTTAAAGCAAAAACCCGCCGAAG; the 3′ region of pM1 RNA is in bold type) and cloning into the BamHI and HindIII sites of pBluescriptSK(+) (Stratagene).

Protein work and Western blotting

N-Rne, Rne and Rng were purified from KSL2002, 2003 and 2004, respectively, using the HisBind purification kit (Novagen). The cultures were grown to mid-log phase, induced with 1 mM IPTG for 1.5 h and harvested. The enzyme was eluted from columns using 100 mM imidazole, concentrated and stored as described previously (McDowall et al., 1995). The protein concentration was calculated using Coomassie brilliant blue G250 as described by Sedmak and Grossberg (1977) and using bovine serum albumin as a standard. The proteins were run on a 7.5% sodium dodecyl sulphate (SDS) polyacrylamide gel, and gels were electroblotted to a nitrocellulose membrane and probed as described previously (Lee and Cohen, 2001).

Northern blot analysis

Total RNA was isolated as described by Lin-Chao and Cohen (1991). Thirty micrograms of total RNA sample was denatured at 70°C for 10 min in an equal volume of formamide- loading buffer and loaded onto an 8% polyacrylamide gel containing 8 M urea. The gels were electroblotted onto a nylon membrane (Zeta-Probe® blotting membrane, Bio-Rad) in 0.5 × TBE at 80 V for 6 h. The RNA samples were fixed by baking the membrane at 80°C for 30 min and hybridized at 55°C with 32P-end-labelled oligonucleotides according to the manufacturer’s instruction manual. Oligonucleotides used for studying 9S rRNA processing are probe I (5′-ACTACCATC GGCGCTACGGC-3′), probe II (5′-CAGGCTGAAAATCTTCT CTC-3′) and probe III (5′-TTTCGACTGAGCCTTTCGTT-3′). M1 probe (5′-GCTCTCTGTTGCACTGGTCG-3′) and RNA I probe (5′-GGATCAAGAGCTACCAACTC-3′) were used for probing M1 RNA and RNA I respectively. When blots were reprobed, they were stripped twice at 95°C for 20 min in buffer containing 0.1 × SSC (standard saline citrate) and 0.5% SDS.

Synthesis of RNA and in vitro cleavage assay

RNA I transcript universally labelled with [α-32P]-UTP was synthesized from PCR DNA using the MEGAshortscript™ T7 kit (Ambion) according to the manufacturer’s instructions. The PCR DNA was amplified from pM21 (Helmer-Citterich et al., 1988) using two primers, T7 (5′-TAATACGACTCACTATAGG) and 3′−RNA I (5′-AACAAAAAAACCACCGCTACCAGCG). The pM1 RNA transcript universally labeled with [α-32P]-UTP was synthesized from DraI-digested pM1 plasmid using the MEGAscript™ SP6 kit (Ambion). This transcript contains exactly the same sequence found in pM1 RNA in vivo. 9S ribosomal RNA was synthesized as previously described (Lee and Cohen, 2001). All RNA transcripts were purified from 6% polyacrylamide gel containing 8 M urea. RNase E cleavage assays were performed as described previously (McDowall et al., 1995).

Microarray procedures

Relative mRNA levels were determined by parallel two-colour hybridization to DNA microarrays (Schena et al., 1995) on glass slides containing 4405 known and predicted ORFs (Blattner et al., 1997). Comparative measurements of transcript abundance were performed by directly determining the abundance of each gene’s transcripts relative to the wild-type sample (N3433 plus pPM30). RNA samples taken from N3433 plus pRNG3, N3433rng plus pPM30, and N3433rne plus pBAD-RNE, pNRNE5, or pRNG3 were labelled with Cy-5, and RNA from N3433 plus pPM30 was labelled with Cy-3. Comparisons between paired mutant and wild-type cultures were done directly. Two kinds of wild-type cultures were used: one was grown in the same way as N3433rne plus pBAD-RNE (no arabinose) for comparison with N3433rne plus pBAD-RNE (no arabinose); the other was grown in the presence of 100 μM IPTG for comparison with N3433 plus pRNG3, N3433rng plus pPM30, and N3433rne plus pNRNE5, or pRNG3. Preparation of RNA samples, cDNA synthesis, hybridization and analysis of spots were performed as described by Khodursky et al. (2000).

Analysis of data was performed with the software avail-able at http://genome www4.stanford.edu/MicroArray/SMD/ restech.html and http://rana.lbl.gov. We measured relative mRNA abundance under appropriate conditions in five replicate comparisons, selecting 1113 genes whose relative mRNA levels were changed 1.5-fold or more in at least 5 of 10 slides. A total of 105 of 1113 genes whose mRNA levels were decreased by overproduction of N-RNase E and RNase G were further clustered by the method of Eisen et al. (1998).

Acknowledgements

These studies were supported by NIH grants AI08619 and GM54158 to S. N. C. We thank Dr Björn Sohlberg for helpful discussions and comments on the manuscript.

Ancillary