RNase E is an essential endoribonuclease involved in RNA processing and mRNA degradation. The N-terminal half of the protein encompasses the catalytic domain; the C-terminal half is the scaffold for the assembly of the multienzyme RNA degradosome. Here we identify and characterize ‘segment-A’, an element in the beginning of the non-catalytic region of RNase E that is required for membrane binding. We demonstrate in vitro that an oligopeptide corresponding to segment-A has the propensity to form an amphipathic α-helix and that it avidly binds to protein-free phospholipid vesicles. We demonstrate in vitro and in vivo that disruption of segment-A in full-length RNase E abolishes membrane binding. Taken together, our results show that segment-A is necessary and sufficient for RNase E binding to membranes. Strains in which segment-A has been disrupted grow slowly. Since in vitro experiments show that phospholipid binding does not affect the ribonuclease activity of RNase E, the slow-growth phenotype might arise from a defect involving processes such as accessibility to substrates or interactions with other membrane-bound machinery. This is the first report demonstrating that RNase E is a membrane-binding protein and that its localization to the inner cytoplasmic membrane is important for normal cell growth.
In the model Gram-negative bacterium Escherichia coli, RNase E is an essential endoribonuclease involved in the maturation of stable RNA and the degradation of messenger RNA (Kushner, 2002; Carpousis, 2007). RNase E is a single-strand specific endonuclease, which is 5′-end-dependent, with a specificity for RNAs with 5′-monophosphate ends (Mackie, 1998; Jiang and Belasco, 2004). Recent X-ray crystallographic work has shown that the catalytic domain of RNase E contains a ‘pocket’ that engages the 5′-monophosphate group of an RNA substrate to trigger a conformational change that organizes the active site for cleavage (Callaghan et al., 2005; Koslover et al., 2008). In mRNA degradation, the rate-limiting step has been proposed to be either a ‘slow’ internal cleavage by RNase E or the conversion of the 5′-triphosphate end of the primary transcript to a 5′-monophosphate end (Celesnik et al., 2007). An E. coli enzyme that converts 5′-triphosphate ends to 5′-monophosphate ends has recently been identified and characterized (Deana et al., 2008). After the initial rate-limiting step, mRNA is rapidly degraded in a concerted reaction involving multiple cleavages by RNase E and the degradation of the RNA fragments to nucleotides by exoribonucleases.
The catalytic domain of RNase E, corresponding to residues 1–529, is a homotetramer that is organized as a dimer of dimers (Callaghan et al., 2005). RNase E has a large C-terminal non-catalytic extension (residues 530–1061) equal in length to the region that forms the catalytic domain. The non-catalytic region is at the centre of many physical and functional interactions involved in RNA processing and degradation. Most of the protein in the non-catalytic region is intrinsically unstructured (Callaghan et al., 2004). Residues 601–700 form an arginine-rich segment that binds RNA in vitro and that is believed to enhance the activity of RNase E in mRNA degradation in vivo (Lopez et al., 1999; Ow et al., 2000). Residues 701–1061 form a scaffold for interactions between RNase E and other enzymes including RNA helicase B (RhlB), enolase and polynucleotide phosphorylase (PNPase) (Kaberdin et al., 1998; Vanzo et al., 1998). This multienzyme complex is known as the RNA degradosome. The interaction of RNase E with PNPase, an exoribonuclease, brings together two enzymes important in the degradation of mRNA. RhlB, another component of the RNA degradosome, is a DEAD-box RNA helicase that facilitates mRNA degradation by RNase E and PNPase. The presence of RhlB in the RNA degradosome is necessary for RhlB activity in vitro and in vivo (Vanzo et al., 1998; Coburn et al., 1999; Khemici and Carpousis, 2004; Khemici et al., 2005). A different complex containing RNase E, Hfq and SgrS, a small regulatory RNA, is formed under conditions of phosphosugar stress (Morita et al., 2005). The formation of the complex with Hfq and SgrS requires the same region of RNase E that is necessary for the formation of the canonical RNA degradosome and evidence suggests that the degradosome is remodelled as a consequence of the new interaction. There is also evidence that RNase E can form a ‘cold shock’ RNA degradosome in which CsdA, another DEAD-box RNA helicase, is recruited to the complex (Khemici et al., 2004; Prud'homme-Genereux et al., 2004).
Previous work involving cell fractionation and electron microscopy has suggested that RNase E is associated with the inner cytoplasmic membrane (Miczak et al., 1991; Liou et al., 2001). More recent work employing a yeast two-hybrid analysis and fluorescence microscopy has confirmed and extended these results (Taghbalout and Rothfield, 2007). A protein–protein interaction between RNase E and MinD, a membrane-localized bacterial cytoskeletal protein, was demonstrated by two-hybrid analysis. RNase E tagged with yellow fluorescence protein (YFP) was detected in a filamentous structure that coils around the length of the cell. The other components of the RNA degradosome were also localized to the helical filament. Based on these results, it has been proposed that RNase E and the RNA degradosome are part of a bacterial cytoskeleton. However, although the localization of RNase E appears to be similar to that of the cytoskeletal proteins MinD and MreB, neither of these proteins is required for the localization of RNase E to the inner cytoplasmic membrane (Taghbalout and Rothfield, 2007). Since MinD is not required, the membrane localization of RNase E does not depend on the RNase E–MinD interaction. Here we elucidate the molecular basis for the localization of RNase E to the inner cytoplasmic membrane. In a reconstituted system, RNase E by itself binds to protein-free phospholipid vesicles. Thus, the interaction with the membrane does not require other protein factors. We show that ‘segment-A’, spanning residues 568–582 in the beginning of the non-catalytic region of RNase E, is necessary and sufficient for the binding of RNase E to the inner cytoplasmic membrane. Our results suggest that segment-A forms an amphipathic α-helix that is involved in membrane binding. These data are discussed in terms of the role of the membrane localization of RNase E in the structure and function of the RNA processing and degradation machinery of E. coli.
RNase E binds to phospholipid vesicles and segment-A is necessary for binding
Although the sequence of the catalytic domain of RNase E is highly conserved throughout the proteobacteria, the non-catalytic region of RNase E, even among the closely related enterobacteria, is variable (Carpousis, 2002; Condon and Putzer, 2002). This sequence variation could reflect the intrinsically unstructured nature of the non-catalytic region (Callaghan et al., 2004; Marcaida et al., 2006). The non-catalytic region of E. coli RNase E, nevertheless, contains four isolated elements predicted to form limited secondary structure (Fig. 1A). For instance, ‘segment-C’ (residues 834–851), which is predicted to form a short α-helix, corresponds to the site in RNase E that binds enolase (Callaghan et al., 2004). A recent crystallographic study shows that the interaction between RNase E and enolase involves an intricate hydrogen-bonding network with a short α-helix formed by segment-C (Chandran and Luisi, 2006). Furthermore, among homologues of RNase E in the enterobacteria, segment-C corresponds to a conserved sequence motif suggesting that these proteins also interact with enolase.
Segment-A (residues 568–582) is also predicted to form a short α-helix (Callaghan et al., 2004), but until now no function has been attributed to this element. Comparison of strains expressing RNase E variants in which different regions of the protein have been deleted suggested that an element between residues 418 and 602 is required for localization to the periphery of the cell (Liou et al., 2001; Taghbalout and Rothfield, 2007). Inspection of the sequence of segment-A suggested that it could form an amphipathic α-helix. As depicted in the helical wheel diagram (Fig. 1B), the non-polar residues (yellow) cluster on one face of the helix (hydrophobic face), whereas the polar and charged residues are clustered on the opposite face (hydrophilic face). The hydrophobic face contains four leucine residues and three phenylalanine residues; amino acids with hydrophobic side-chains capable of making favourable interactions with the hydrophobic interior of a phospholipid bilayer. We therefore asked whether RNase E can bind to phospholipid vesicles in vitro. Multilamellar vesicles (MLVs) were prepared using protein-free phospholipids purified from E. coli. His-tagged versions of RNase E containing or lacking segment-A were overexpressed, highly purified under denaturing conditions and renatured as described (Experimental procedures). In parallel experiments using this protocol, we show that the renatured RNase E has endoribonuclease activity (see below). Membrane binding was measured by co-sedimentation of RNase E with the MLVs, which pellet during low-speed centrifugation. Protein in the pellet was analysed by SDS-PAGE. Visual inspection of the Coomassie-stained gel in Fig. 2A shows that wild-type RNase E binds quantitatively to the MLVs whereas no interaction is detected using the RNase E variant lacking segment-A (ΔA). In the absence of MLVs, a trace band is seen for RNase E, and this is likely due to a small amount of aggregated protein that pellets during low-speed centrifugation. From these results, we conclude that RNase E binds to phospholipid vesicles and that segment-A is necessary for binding.
We also tested two variants of RNase E with altered segment-A sequences. F/E is a variant of RNase E in which a phenylalanine residue 575 was changed to a glutamic acid (F575E). P+ is a variant in which a proline residue was inserted between residues 574 and 575. The F/E variant is predicted to conserve the putative α-helix, but destroy its amphipathic character, whereas the P+ variant is predicted to break the α-helix. Helical wheels of the F/E and P+ variants are shown in Fig. 1C and D respectively. The P+ variant is represented by helical wheel projections of the two halves of segment-A flanking the proline insertion. Both half-segments are predicted to have the propensity to form α-helices. The F/E and P+ variants bind to the MLVs as evidenced by co-sedimentation (Fig. 2B), but the amount of pelleted RNase E is less than observed with the wild-type protein.
In the co-sedimentation experiments (Fig. 2), 1 μg of RNase E in an 80 μl reaction corresponds to a nominal concentration of 27 nM based on the 464 kDa molecular weight of the RNase E tetramer. The Coomassie blue-stained gels in Fig. 2 were scanned to quantify the amount of RNase E that pellets with the MLVs. In the absence of MLVs, the small amount of RNase E that pellets corresponds to about 10% of the input protein. After subtracting this background, the amount of wild-type RNase E that pellets with the MLVs corresponds to about 80% of the input protein. This value was the same for both quantities of MLVs in Fig. 2A (32 and 60 μg) and comparable amounts of pelleted RNase E were measured over a range of 8–120 μg of MLVs (data not shown) demonstrating that the phospholipid vesicles are present in excess in these experiments. This quantification shows that wild-type RNase E binds to the MLVs with high affinity and that the complex is stable enough to pellet most of the RNase E during the centrifugation step. The variant of RNase E that is entirely lacking segment-A does not measurably interact with the MLVs after subtracting the background. Thus, even if the ΔA variant forms a weak equilibrium-binding complex, the complex is not stable enough to permit pelleting with the MLVs.
Quantification of the co-sedimentation experiment with the F/E and P+ variants indicated that 10% and 40% of the input protein pelleted with the MLVs respectively (see Fig. 2B). The amount of pelleted RNase E was the same at 32 and 60 μg of MLVs. The simplest interpretation of these results is that both variants form an equilibrium-binding complex with the MLVs, but these complexes are less stable than the complex with wild-type RNase E. The results with the F/E mutant shows that the change of a single residue in segment-A significantly weakens membrane binding in vitro and suggests that the hydrophobic phenylalanine at residue 575 is important for the binding to lipid membranes. The partial binding by the P+ variant suggests that the formation of an α-helix might not be necessary for membrane binding. Nevertheless, this result needs to be interpreted in light of the work presented in the next section.
Peptide-A has the propensity to form an α-helix
To evaluate the propensity of segment-A to form an α-helix, we measured the circular dichroism (CD) spectrum of a synthetic 21-residue oligopeptide corresponding to segment-A, which we will refer to as peptide-A. The sequence of peptide-A is shown in Fig. 1A. The underlined residues are predicted to have the propensity to form an α-helix. We also recorded spectra from a set of related peptides with amino acid changes that are expected to disrupt the α-helix or perturb its amphipathic character. The F/E and P+ variants have already been described in the previous section. In the FF/AA variant, two phenylalanines on the non-polar face of the helix were changed to alanine (Fig. 1C). This variant is predicted to still have the propensity to form an α-helix, but with reduced hydrophobicity of the non-polar face of the helix.
The CD spectrum of peptide-A has a profile characteristic of an α-helix and the addition of large unilamellar vesicles (LUVs) prepared from protein-free E. coli phospholipids increases the α-helical character (Fig. 3A). The F/E, P+ and FF/AA variants exhibit little if any α-helical character in the absence of LUVs. In Table 1, the CD spectra were decomposed to estimate the α-helical content. Peptide-A by itself is estimated to be 30% α-helical and the content increases to 75% in the presence of LUVs. The α-helical content of the F/E and FF/AA variants increases from 7% to 28% and 3% to 37%, respectively, in the presence of LUVs. The α-helical content of the P+ variant is essentially unchanged by the addition of membrane (7% versus 9% respectively). Although the decomposition may have limited accuracy, the cross-comparisons of the values provide an indication of the relative helical propensities of the different peptides. These data indicate that the wild-type peptide has the greatest helical propensity in isolation and that the wild-type, F/E and FF/AA variants undergo a partial coil to helix transition in the presence of LUVs. The P+ variant does not make this transition.
Peptide, 300 μM, in 6.25 mM NaKPO4, pH 7.4. Large unilamellar vesicles (LUVs) were added at a 17:1 lipid : peptide ratio. Spectra (180–250 nm) were decomposed with DICHROWEB using the K2D method.
wt & LUVs
F/E & LUVs
P+ & LUVs
FF/AA & LUVs
The melting temperature of peptide-A was also measured by CD spectra. At 100 μM, peptide-A has a reversible melting profile with a transition mid-point of roughly 42°C, but the melting profile shifted considerably at 1 mM, so that the mid-point becomes roughly 60°C, suggesting that the peptide might be self-associating (data not shown). The mid-point of the melting profile was also shifted to higher temperature in the presence of LUVs, suggesting that they interact; however, the transition is very broad (data not shown). Taken together, the CD data show that peptide-A has the propensity to form an amphipathic α-helix that is stabilized by interaction with the phospholipid bilayer.
Peptide-A avidly binds to phospholipid vesicles
The CD data show that there is an interaction between LUVs and wild-type peptide-A as well as the F/E and FF/AA variants at high concentrations of peptide (300 μM). To quantify these interactions we investigated the binding of peptide-A to LUVs by isothermal titration calorimetry (ITC). Several conditions were tested and measurable heat release could be observed for experiments in which either LUVs were titrated with a concentrated peptide stock, or the peptide was titrated with a concentrated LUV stock. When the peptide was diluted into buffer without LUVs, a strong endothermic reaction resulted, which suggests that the peptide is dissociating from an oligomeric state (data not shown). The ITC profile could be matched with a rough fit using a simple model for dissociation of a dimer, but the imperfection of this fit suggests that the peptide might be dissociating from a tetrameric or higher-order oligomeric form. The estimated heat change for dissociation at 20°C is 3.7 kcal mol−1 and the Kd is 3.9 × 10−4 M. Figure 3B shows the profile of an ITC experiment in which peptide-A was injected into buffer with LUVs. These data have been corrected for the endothermic process of peptide self-dissociation. The estimated heat change of the binding of peptide to LUVs is −5.8 kcal mol−1 and the Kd is 1.3 × 10−6 M. The stoichiometry of binding corresponds to roughly 40 phospholipid molecules for every peptide based on the average molecular mass of an E. coli phospholipid. Less heat is released at lower temperature, which suggests that the binding of peptide-A and membrane is associated with a negative heat capacity change – a hallmark of many macromolecular associations. None of the variants, F/E, FF/AA or P+, gave any measurable heat release under the conditions used for the wild-type peptide-A. Although it is possible that these peptides interact with the membrane in a process that is driven solely by entropy change, this seems unlikely. The ITC results therefore suggest that the interaction detected by CD with the F/E and FF/AA variants is very weak. Taken together with the CD measurements, these results provide strong evidence that the interaction with the phospholipid bilayer involves an amphipathic α-helix formed by segment-A.
Cellular localization of RNase E variants with defective segment-A
In order to construct strains for fluorescence microscopy, we used KSL2000, which has a chromosomal disruption of the rne gene that is complemented by the rne gene on a low-copy-number plasmid (Lee and Cohen, 2003). In our control strain, a plasmid expressing RNase E fused to yellow fluorescence protein (YFP), under the control of its own expression signals, was introduced into KSL2000 using a protocol that displaces the resident plasmid with the incoming plasmid. Visualization by fluorescence microscopy showed that cells expressing wild-type RNase E–YFP have a ghost-like appearance due to the intense signal at the periphery and the void in the interior (Fig. 4A). The images for wild-type RNase E in Fig. 4A are typical for the strain that we constructed. We only rarely see cells with the helical filamentous structure previously described by Taghbalout and Rothfield (2007). They report detecting filaments with constructs using either the endogenous rne expression signals or a lac promoter. They also detected helical filaments by indirect immunofluorescence using RNase E with an HA epitope tag. Although we cannot explain why we only rarely see a helical filament, it is important to emphasize that in the work presented here we can easily distinguish the difference between the peripheral localization of wild-type RNase E–YFP and the localization of the variants affecting the membrane binding activity of segment-A.
Figure 4A also shows images of strains in which wild-type RNase E was replaced by variants with defective segment-A function. Surprisingly, the ΔA variant is localized in two or three bright foci. These foci are often located at the poles of the cell. DAPI staining showed that RNase E in the foci appears to be distinct from the DNA in the nucleoid (data not shown). The pattern of fluorescence with the F582E variant, which disrupts the amphipathic character of the α-helix (Fig. 1C), is comparable to the ΔA variant. Its complete delocalization from the periphery demonstrates the importance of the highly conserved phenylalanine at this position (see sequence alignments below). Other variants have an intermediate phenotype: P+ and F575E exhibit a partial delocalization from the periphery that correlates with a mixture of bright foci and a mottled distribution at the periphery of the cell. The A+, 2A+, 3A+ and 4A+ variants are a series in which one, two, three or four alanines were inserted between residues 574 and 575 (same position as P+). Only the 2A+ variant, which is predicted to place half of the hydrophobic residues on the opposite face of the helix (Fig. 1D), delocalizes RNase E from the inner cytoplasmic membrane. The pattern of fluorescence of the 2A+ variant is similar to that of the F575E and P+ variants.
The level of expression of the RNase E variants was measured by Western blotting using an antibody against the YFP tag (Fig. 4B). The ΔA variant is twofold overexpressed compared with wild type based on the serial dilutions in the Western blot. This is also the case for the F575E and F582E variants whereas the other proteins have a level of expression similar to wild type. Thus, although the micrographs in Fig. 4A might give the impression that certain RNase E variants are highly overexpressed, this is not the case. The intensity of the foci reflects the high concentration of RNase E in a relatively small region within the cell. The expression of RNase E is autoregulated by a post-transcriptional mechanism that responds to the level of RNase E activity in the cell (Jain and Belasco, 1995). The overexpression of RNase E variants delocalized from the inner cytoplasmic membrane could reflect either a deficit of ribonuclease activity or deregulation of the system-controlling expression.
Growth of mutant strains expressing RNase E variants with defective segment-A
The strain expressing wild-type RNase E fused to YFP grows normally whereas the strain expressing RNase E lacking segment-A grows slowly. This phenotype is readily visible during growth on LB agar (Fig. 4C) and the defect was verified by the determination of growth rates in liquid cultures. The slower growth of the ΔA variant was also observed in constructs expressing FLAG-tagged RNase E and thus does not depend on the presence of the YFP tag (data not shown). The doubling time for the strain expressing wild-type RNase E was about 34 min as compared with about 55 min for the strain expressing the ΔA variant (Table 2). The growth rates of other strains expressing variants affecting segment-A function were also slower (Fig. 4C and Table 2). The results presented in Fig. 4 and Table 2 show that there is a correlation between the extent of delocalization of RNase E from the inner cytoplasmic membrane, the formation of foci and slow growth. Since the enzymatic activity of RNase E is essential for cell growth, these results suggest two possibilities that are not mutually exclusive: either the localization to the inner cytoplasmic membrane is necessary for normal enzymatic activity of RNase E or the sequestration of RNase E in the foci interferes with its activity.
Average doubling time and standard deviation from three independent determinations. Growth was in LB media at 37°C.
33.7 ± 0.2
54.8 ± 1.6
44.9 ± 1.2
47.6 ± 0.7
47.2 ± 1.4
Foci formation by RNase E with defective segment-A requires the non-catalytic region
In order to further explore the effect of the delocalization of RNase E from the periphery of the cell on the formation of the foci and cell growth, we examined two additional variants of RNase E–YFP; one encompassing the catalytic domain plus segment-A (residues 1–592) and the other encompassing the catalytic domain alone (residues 1–529). As expected, RNase E (1–592) is localized to the membrane whereas RNase E (1–529) is cytoplasmic (Fig. 5A). Note, however, that RNase E (1–529) is distributed uniformly throughout the cell: it does not form foci at the poles. Since RNase E (1–529) is effectively a ΔA variant lacking the C-terminal half of RNase E, the uniform distribution is strong evidence that the formation of the foci at the poles of the cell requires the non-catalytic region of RNase E. The growth of the strain expressing RNase E (1–592) is comparable to wild type and the growth of the strain expressing RNase E (1–529) is only slightly slower (Fig. 5B). The growth defect of RNase E (1–529) is thus milder than the ΔA, F582E and F575E variants of full-length RNase E (compare Figs 4C and 5B). This result suggests that the growth defect of the strains expressing the full-length RNase E variants could be at least partly due to the formation of the foci at the poles of the cell. The level of expression of RNase E (1–592) is normal whereas the RNase E (1–529) is clearly overexpressed (Fig. 4B). This suggests that localization to the inner cytoplasmic membrane might be necessary for normal levels of RNase E activity.
The endoribonuclease activity of RNase E is not affected by binding to phospholipid vesicles
To test the effect of phospholipid vesicles on the enzymatic activity of RNase E, we performed in vitro assays in the absence or presence of MLVs. As described for the binding experiments in Fig. 2, RNase E was highly purified under denaturing conditions and then renatured. This approach was used since natively purified RNase E co-purifies with the protein components of the RNA degradosome and tightly bound RNA (Carpousis et al., 1994; Miczak et al., 1996; Bessarab et al., 1998). To the best of our knowledge, nobody has devised a protocol to separate RNase E from the associated protein and RNA under native conditions. It has been demonstrated by others that enzymatic activity can be restored by renaturation of highly purified RNase E (Cormack et al., 1993; McDowall and Cohen, 1996; Mackie et al., 2001). The activity of the renatured RNase E was assayed using the 9Sa RNA substrate, a fragment of the 9S precursor of 5S rRNA. The 9Sa substrate and the assay conditions used here have been previously described in detail (Carpousis et al., 1994; 2001). In the assay, equal amounts of enzyme were assayed in the absence or presence of MLVs, or in the presence of Triton X-100, a non-ionic detergent. Triton X-100 was included in these assays since it has been used routinely in the purification and assay of native RNase E. Visual inspection of the image in Fig. 6 reveals no significant difference in ribonuclease activity under the three conditions and this was confirmed by scanning the gel and quantifying the results (data not shown). From these results, we conclude that the binding of RNase E to the phospholipid vesicles has no effect on the digestion of the 9Sa substrate under conditions that have routinely been used to assay the endoribonuclease activity of RNase E.
Conservation of motif-A in RNase E homologues of the β- and γ-proteobacteria
A sequence alignment of a selection of RNase E homologues from the enterobacteria is presented in Fig. 7A. Although the entire RNase E protein sequence was used to make the alignment, we present only the part of the alignment corresponding to segment-A. The conserved 15-residue sequence, motif-A, has a characteristic set of evenly spaced hydrophobic residues (highlighted in blue). Although internal prolines break α-helices, the conserved N-terminal proline of the enterobacterial motif-A is a permitted residue that might help to stabilize the structure by capping the terminus. In the enterobacteria, the beginning of motif-A is located 30–50 residues from the end of the catalytic domain. Motif-A is flanked on both sides by proline-rich sequences. The proline-rich regions of the E. coli homologue are predicted to be intrinsically unstructured (Callaghan et al., 2004). Figure 7B shows a sequence alignment of a selection of RNase E homologues from the γ-proteobacteria, a large family of Gram-negative bacteria that includes the enterobacteria. Although the conservation of sequence is less striking than for the enterobacteria, the characteristic evenly spaced hydrophobic residues and the flanking proline-rich sequences are clearly conserved.
We have also detected a related sequence, motif-A′, in RNase E homologues of the β-proteobacteria (Fig. 7C). Although only nine residues in length, it is predicted to form an amphipathic α-helix and it is flanked by proline-rich sequences. Helical wheel projections of a selection of motif-A and motif-A′ sequences are presented in Fig. 7D. The sequence of Hahella chejuensis is presented as one of the rare exceptions where the hydrophobic face of the α-helix appears to be interrupted by a charged residue (position 8, arginine). An arginine at this position could nonetheless have the role of a hydrophobic residue, in which its aliphatic portion is buried in the hydrophobic interior and the charged guanidinium moiety ‘snorkels’ to the surface where it interacts with the polar lipid head groups. In general, motif-A contains seven hydrophobic residues whereas the shorter motif-A′ sequences contains only six hydrophobic residues. Both motifs have a propensity for amino acids with small side-chains, such as glycine and serine, on the hydrophilic face of the α-helix. Glycine in particular is predicted to destabilize the α-helix (Chou and Fasman, 1978) and the conservation of small residues suggests that the hydrophilic face might have another, as yet unidentified, function. In summary, motif-A and the related motif-A′ sequence are ubiquitous in RNase E homologues of the β- and γ-proteobacteria. These motifs are predicted to form an autonomous amphipathic α-helix that could anchor RNase E to the inner cytoplasmic membrane in these bacteria.
The principal finding of this work is that RNase E binds to protein-free phospholipid vesicles. Like other proteins associated with the inner cytoplasmic membrane of E. coli, RNase E contains an autonomous element that serves as the membrane anchor. Previously characterized examples of E. coli proteins that bind to the phospholipid bilayer via a membrane anchor include FtsA, FtsY and MinD (Zhou and Lutkenhaus, 2003; Pichoff and Lutkenhaus, 2005; Parlitz et al., 2007). These proteins contain an amphipathic α-helix that is believed to bind to the lipid bilayer by inserting residues on the hydrophobic face of the α-helix into the hydrophobic interior of the membrane. Additional interactions may involve positively charged lysine and arginine residues on the hydrophilic face of the α-helix and the negatively charged head groups of the polar lipids. Previous work suggested that the peripheral localization of RNase E involved the N-terminal catalytic region of RNase E (Liou et al., 2001). Here we show that segment-A, the membrane anchor, is structurally distinct from the catalytic domain of RNase E. The catalytic domain spans residues 1–529 of RNase E; segment-A is located between residues 565 and 582. The proline-rich unstructured regions flanking segment-A tether the N-terminal catalytic domain and the C-terminal non-catalytic region to the membrane. This organization is conserved throughout the β- and γ-proteobacteria. RNase E homologues found in species outside of the β- and γ-proteobacteria have a different structure. For instance, we do not detect an autonomous amphipathic α-helix downstream of the catalytic domain of RNase E homologues of the α-proteobacteria.
We have shown that full-length RNase E binds to phospholipid vesicles in vitro. The deletion of segment-A (ΔA) abolishes this binding. The substitution of the phenylalanine at position 575 by a glutamic acid (F575E) or the insertion of a proline between residues 574 and 575 (P+) has a milder effect, leading to the partial loss of binding in vitro. We have demonstrated that segment-A has the propensity to form an α-helix by CD measurements with peptide-A, a 21-residue oligopeptide encompassing segment-A. Peptide-A is partially α-helical by itself in solution and its helical content increases in the presence of phospholipid vesicles. Direct evidence for a physical interaction between peptide-A and the lipid bilayer was obtained by ITC. The interaction with phospholipid vesicles results in a strong exothermic reaction, and the binding profile could be fitted with a dissociation constant of approximately 1 μM for the interaction. Three variants of peptide-A were analysed by CD and ITC. The P+, F/E and FF/AA variants have little if any propensity to form an α-helix in solution and an interaction with phospholipid vesicles could not be detected by ITC. Our results show that peptide-A avidly binds to phospholipid bilayers and that the membrane-bound form is in an α-helical conformation. Taken together, these results show that segment-A is necessary and sufficient for anchoring RNase E to the phospholipid bilayer.
Our in vitro binding results with full-length RNase E agree remarkably well with the cellular localization of the corresponding proteins in vivo as determined by fluorescence microscopy. The ΔA variant is fully delocalized from the periphery of the cell whereas the F575E and P+ variants are only partially delocalized. The correlation between the in vitro and in vivo results provides strong support for the conclusion that the peripheral localization of RNase E is due to binding to the phospholipid bilayer. These results suggest an explanation for a puzzling observation in the work by Taghbalout and Rothfield (2007). After presenting yeast two-hybrid evidence for a protein–protein interaction between RNase E and MinD, a membrane-localized bacterial cytoskeletal protein, and proposing that RNase E is part of a bacterial cytoskeleton, they showed that RNase E was still localized to the periphery of the cell in strains where the genes encoding MinD or MreB, another cytoskeletal protein, were disrupted. Our results show that the principal determinant for the peripheral localization of RNase E is the interaction with the phospholipid bilayer by segment-A. If there is a protein–protein interaction between RNase E and MinD in E. coli, then our results suggest that the membrane anchor of RNase E is necessary for such an interaction. The requirement could be indirect or direct. In the first case, membrane binding might be required for protein–protein interactions with components of the bacterial cytoskeleton. In the second case, segment-A might have a dual function as a membrane anchor and as a structural element for protein–protein interactions with other membrane proteins.
Our in vivo results demonstrate that there is a correlation between delocalization of RNase E from the inner cytoplasmic membrane, the formation of foci at the poles of the cell and slow growth. The slow growth suggests a deficiency of RNase E activity in the cell. This could involve, for example, a mechanism in which the optimal structure of the catalytic domain of RNase E depends on membrane binding. However, in an in vitro assay, there was no difference in the activity of RNase E in the absence or presence of phospholipid vesicles. It is thus possible that the slow-growth phenotype arises from a defect involving processes such as accessibility of RNase E to its substrates or interactions of RNase E with other membrane-bound machinery. Our results with the RNase E (1–592) and RNase E (1–529) variants strongly suggest that the non-catalytic region of RNase E is required for the formation of foci at the poles of the cell by RNase E variants that have defective segment-A function. In the strains where foci are detected, RNase E is associated with the other components of the degradosome as evidenced by co-immunoprecipitation results (our unpublished results). The formation of the foci at the poles of the cell thus appears to involve the RNA degradosome. One hypothesis is that the cytoplasmic RNA degradosome is somehow targeted to the pole. For instance, if the synthesis and assembly of the RNA degradosome occurs at the pole, then the cytoplasmic RNA degradosome might accumulate at the pole due to its inability to bind to the membrane. To test this hypothesis, we are currently looking for determinants in the non-catalytic region of RNase E that might be involved in localization to the pole. Another explanation for the foci might be that the cytoplasmic RNA degradosome aggregates into a very large complex that is excluded from the nucleoid.
Although our work establishes that RNase E associates directly with the inner cytoplasmic membrane, the consequence of this localization on RNA processing and degradation is less clear. Since the association of RNase E with the inner cytoplasmic membrane might be reversible, it is possible that the membrane is a storehouse and that the active form of the enzyme is the small proportion of molecules that are cytoplasmic. The association of RNase E to the membrane might be regulated by interactions with regulatory proteins or covalent modification. Storehousing RNase E would create a reserve of RNA processing and degrading activity that could be mobilized without the need to synthesize new enzyme. We raise this possibility since previous work has suggested that there is an excess of RNase E relative to the amount required to sustain the growth of E. coli (Sousa et al., 2001; Jain et al., 2002). On the other hand, the original membrane fractionation experiments examining the cellular localization of RNase E suggested that RNase III and RNase P are also membrane associated (Miczak et al., 1991). More recent work by fluorescence microscopy has shown that poly(A) polymerase (PAP I) is at least partially associated with the cytoplasmic membrane (Jasiecki and Wegrzyn, 2005). Taken together, these observations suggest that RNase E, RNase III, RNase P, PNPase, RhlB and PAP I are at least partly localized to the inner cytoplasmic membrane. This set of enzymes comprises the core of the RNA processing and degradation machinery in E. coli. Studies in Bacillus subtilis using fluorescent tagged RNA polymerase and ribosomes have shown that the RNA polymerase is associated principally with the nucleoid and the ribosomes are localized almost exclusively to the cytoplasmic space outside of the nucleoid (Lewis et al., 2000). These results have been interpreted as evidence for a compartmentalization of transcription and translation despite the lack of a nuclear membrane in bacteria. The localization of RNase E and other enzymes to the inner cytoplasmic membrane suggests that RNA processing and degradation is also compartmentalized and that a class of transcripts, which remains to be identified, is processed or degraded on the inner cytoplasmic membrane.
Strains and plasmids
The KSL2000 E. coli strain (lacZ43, relA, spoT, thi-1, rne::cat, recA::Tn10[pBAD-RNE]) was kindly provided by S. Cohen (Lee and Cohen, 2003). The resident pBAD-RNE plasmid was displaced by the plasmids constructed in this study. The plasmid expressing RNase E fused to YFP is a derivative of pAM-rne (Vanzo et al., 1998). Insertion of the YFP coding sequence was made by a three-step cross-over PCR procedure using the following pairs of oligonucleotides: 5′-gtacgtccgcaagatgtacaggtt/5′-ggtgaacagctcctcgcccttgctctcaacaggttgcggacgcgcagg, 5′-agcaagggcgaggagctgttcacc/5′-ttacttgtacagctcgtccatgcc and 5′-ggcatggacgagctgtacaagtaataattagctcaaagtaatcaagcc/5′-cccctgcagacggcaaacgcgattttt. The yfp sequences are indicated in bold type. The yfp DNA was amplified using pFX234 (Gordon et al., 2004) and the rne DNA using pAM-rne. The final product of the cross-over PCR was digested by EcoO1091 and AsiSI and ligated into pAM-rne at the same sites to construct the plasmid pVK207. Plasmids expressing variants of RNase E–YFP were derived from pVK207. Deletions were constructed by inverse PCR of pVK207. Point mutations and insertions were introduced using a QuikChange II kit (Stratagene). These plasmids are described in Table 3.
Table 3. Plasmids.
Chromosomal PstI fragment containing rne gene cloned into pAM238 (pSC101ori, spectinomycin resistance)
T7 expression of RNase E with a C-terminal His-tag
Cells were grown in LB at 37°C to late logarithmic phase (OD600 = 0.6), washed once in M9 minimal media and visualized with a fluorescence microscope (Leica). Images were acquired using a MetaVue imaging system (Universal Imaging).
Sedimentation assay for RNase E binding to lipid vesicles
The N305D substitution was introduced in the catalytic domain of RNase E as described (Callaghan et al., 2005) using the plasmid pET15b-rne to create the plasmid pLP46 (Table 3). Segment-A encoded in pLP46 was deleted by replacement of a PmlI–MluI restriction fragment with the comparable fragment from pVK221 to create the pLP50. Mutant polypeptides were expressed in BL21(DE3) and purified on Ni-NTA spin columns (Qiagen) following the procedure of the manufacturer for denaturing conditions in urea except that two additional washes were made in 0.8 M urea, 20 mM imidazole pH 7, 800 mM NaCl, 10% glycerol, 0.1 mM EDTA, 0.1 mM DTT supplemented with EDTA-free Protease Inhibitor Cocktail (Roche). Elution was performed with the same buffer except that imidazole was increased to 300 mM. In this procedure, the RNase E is renatured on the Ni-NTA resin by decreasing the concentration of urea to 0.8 M.
Escherichia coli phospholipids were purchased from Avanti Polar Lipids. MLVs were prepared essentially as described (Zhou and Lutkenhaus, 2003). Briefly, E. coli phospholipids (4 mg) were re-suspended on ice in 100 μl of 25 mM Tris HCl, pH 7.5, 40% glycerol with vigorous vortexing. Lipids were diluted rapidly with 50 volumes (5 ml) of 25 mM Tris HCl, pH 7.5, 50 mM NaCl (TN buffer) and incubated for 30 min on ice. The lipids were centrifuged 20 min at 55 000 r.p.m. in a Beckman TLA 100.3 rotor at 4°C. The pellet was suspended in 2 ml of TN buffer supplemented with EDTA-free Protease Inhibitor Cocktail and incubated at 65°C for 2 h with occasional vortexing.
Approximately 1 μg of RNase E (10 μl of eluate) was incubated with 70 μl of MLVs (32 or 60 μg in the TN buffer containing protease inhibitors) at 30°C for 20 min. The MLVs were pelleted by centrifugation at 15 000 g for 2 min at room temperature. The protein in the pellet was separated by SDS-PAGE (7.5%) and visualized by Coomassie blue staining (Safe Stain, Invitrogen).
CD spectroscopy and isothermal titration calorimetry
The wild-type peptide-A was prepared by Clonestar. Variants of peptide-A were synthesized and HPLC purified by the Protein and Nucleic Acids Chemistry Facility, Department of Biochemistry, University of Cambridge. Peptides were first gel filtered into buffer using PD-10 (G25) columns (GE Healthcare). Peptide concentrations were determined by amino acid composition analyses. LUVs were prepared from E. coli lipids extracted by acetone and ether as described (Ames, 1968). The lipids were dissolved in ITC binding buffer (50 mM Tris HCl pH 7.5, 50 mM NaCl, 50 mM KCl) containing 1 mM dithiothreitol and extruded using a 400 nm polycarbonate filter (Avanti Polar Lipids).
Circular dichroism spectra were recorded using an Aviv model 215 spectrometer. The signal from the instrument was calibrated with a standard of camphorsulphonic acid. Twenty spectra were recorded and averaged in the wavelength range from 250 to 180 nm for peptide solutions in the 70–100 μM concentration range with 12.5 mM NaKPO4 pH 7.4 buffer at 25°C using 400 μl, 1 mm path-length quartz cuvettes. For comparisons, 10 spectra were also recorded and averaged from 250 to 190 nm in 50 mM NaKPO4 buffer pH 7.4. Reference spectra of the corresponding buffer were subtracted. Spectra in the presence of lipids were recorded for each peptide at the same concentration in the presence of 1.1 mM LUVs from 250 to 196 nm. The lipid mixture samples were made by dilution from a 21 mM stock of lipids in 50 mM Tris HCl, pH 7.5, 50 mM NaCl, 50 mM KCl, 1 mM DTT. Secondary structure analysis was performed with DICHROWEB (Lobley et al., 2002; Whitmore and Wallace, 2004) using the ContinLL program and the SP175 reference set (Lees et al., 2006). Thermal denaturation profiles were recorded by fixing the wavelength at 222 nm, which is a minima for helical signature, and varying the temperature from 4°C to 85°C.
Isothermal titration calorimetry measurements were made using a Microcal ITC-VP instrument. Lipids and peptides were degassed and gel filtered using PD-10 (G25) columns (GE Healthcare). A variety of conditions was explored, but the best results were obtained by injecting solutions of peptide in the 0.5 mM concentration range into solutions of LUVs (0.8 mM) at 35°C.
Total-cell extracts were separated on 7.5% SDS-PAGE as described (Leroy et al., 2002) and then transferred to Hybond-C-Extra (GE Healthcare). The YFP-tagged proteins were detected using anti-GFP antibodies (Invitrogen) and an ECL kit (GE Healthcare) following the recommendations of the manufacturers. The membrane was analysed using a luminescence imager (LAS-4000, FUJI).
RNase E enzymatic activity
The coding region of the wild-type rne gene was PCR amplified and cloned into pET21b to create pET21b-rne, which encodes RNase E with a C-terminal His-tag (Table 3). RNase E was expressed in BL21 (DE3) and purified on Ni-NTA spin columns (Qiagen) as described above (Sedimentation assay). The enzyme was stored on ice since pilot experiments suggested that freezing resulted in a loss of activity. RNase E activity was assayed using the 9Sa RNA substrate as described (Carpousis et al., 1994; Carpousis et al., 2001) except that Triton X-100 was omitted unless indicated otherwise. From the activity test in Fig. 6, we estimate that the specific activity of the renatured enzyme is 16 000 U mg−1 (see Carpousis et al., 2001 for definition of a unit and comparisons to natively purified RNase E).
In Cambridge, we thank Mike Weldon for peptide syntheses and Peter Sharratt for amino acid composition analyses, Joe Maman and Martin Moncrieffe for helpful advice about CD spectra, Barbara Woebking and Rik van Veen for preparation of E. coli lipid vesicles. In Toulouse, we thank Dave Lane for helpful advice on fluorescence microscopy and critical comments on the manuscript, and Béatrice Clouet-d'Orval for critical comments on the manuscript. B.F.L. is supported by the Wellcome Trust. The group in Toulouse is supported by the Centre National de la Recherche Scientifique (CNRS) with additional funding from the Agence Nationale de la Recherche (ANR) to A.J.C. (Grant NT05_1-44659). Funding for international scientific exchange was provided by the Franco-British Alliance Programme (B.F.L. and A.J.C.).