DEAD Box RhlB RNA Helicase Physically Associates with Exoribonuclease PNPase to Degrade Double-stranded RNA Independent of the Degradosome-assembling Region of RNase E*

The Escherichia coli RNA degradosome is a multicomponent ribonucleolytic complex consisting of three major proteins that assemble on a scaffold provided by the C-terminal region of the endonuclease, RNase E. Using an E. coli two-hybrid system, together with BIAcore apparatus, we investigated the ability of three proteins, polynucleotide phosphorylase (PNPase), RhlB RNA helicase, and enolase, a glycolytic protein, to interact physically and functionally independently of RNase E. Here we report that Rh1B can physically bind to PNPase, both in vitro and in vivo, and can also form homodimers with itself. However, binding of RhlB or PNPase to enolase was not detected under the same conditions. BIAcore analysis revealed real-time, direct binding for bimolecular interactions between Rh1B units and for the RhlB interaction with PNPase. Furthermore, in the absence of RNase E, purified RhlB can carry out ATP-dependent unwinding of double-stranded RNA and consequently modulate degradation of double-stranded RNA together with the exonuclease activity of PNPase. These results provide evidence for the first time that both functional and physical interactions of individual degradosome protein components can occur in the absence of RNase E and raise the prospect that the RNase E-independent complexes of RhlB RNA helicase and PNPase, detected in vivo, may constitute mini-machines that assist in the degradation of duplex RNA in structures physically distinct from multicomponent RNA degradosomes.

Various approaches revealing protein-protein interactions in the degradosome indicate that the C-terminal region of RNase E serves as a scaffold that directly binds PNPase, RhlB RNA helicase, and enolase (24,25). No other interactions among these component proteins have been detected (25) or reported. Recently, a mini-degradosome complex (26) containing the scaffold region (without the N-terminal enzymatic region of RNase E), RhlB RNA helicase, and PNPase was reconstituted in vitro. These experiments revealed a functional interaction between RhlB RNA helicase and PNPase: RhlB helicase bound to the RNase E C-terminal region leads to subsequent ATP-activated degradation of a stem-loop segment intermediate RNA by PNPase. In the absence of the RNase E C-terminal region, RNA degradation does not occur. This led to the proposal that interaction between both RNase E-bound PNPase and RhlB helicase attacks the 3Ј-end of a structured mRNA (26). Recently, it has been shown that, in vivo, the steady state levels of individual component degradosome proteins differ and that degradosome protein components can exist both bound and unbound to RNase E (27). Thus the potential exists for the formation of protein complexes other than those formed on the RNase E scaffold to target RNAs for degradation. Here, we have addressed this question by investigating possible interactions among degradosome protein components independently of RNase E. We show, both in vivo and in vitro, that RhlB RNA helicase binds directly to PNPase and that the binding aids double-stranded RNA degradation by PNPase. These findings indicate that functionally important interactions between degradosome components can occur in the absence of the degradosome scaffold and suggest the existence of a dynamic network of degradosome components in vivo that targets various kinds of structured RNAs for rapid degradation by PNPase-RhlB complexes.
E. coli Two-hybrid Assay-E. coli two-hybrid assays were performed on MacConkey agar plates containing 1% maltose as described (29). Antibiotic concentrations were 100 g/ml ampicillin and 34 g/ml chloramphenicol. The plates were incubated at 30°C for ϳ2-3 days; photographs of individual plates were taken using the Kodak DC120 digital camera (Eastman Kodak) with KDS1D program (Eastman Kodak).
The strength of interactions between individual pairs of proteins was determined by assaying ␤-galactosidase activity on bacterial suspensions as previously described (32). For individual protein assays, logphase 30°C bacterial cultures at cell densities of ϳ2-5 ϫ 10 8 cells/ml (A 600 of ϳ0.3-0.7) were used. Under identical experimental condition, we also carried out ␤-galactosidase activity measurements, resulting from the interaction between T25-zip and T18-zip leucine zipper motifs (29) to normalize individual protein assays. For individual pairs of protein-protein interactions, triplicate samples or more were studied.
Protein Purifications and Western Blot Analyses-The FLAG-tagged PNPase, RhlB RNA helicase, and enolase were purified as previously described (3) except that cell debris was removed from cell lysate by centrifugation at 18,000 ϫ g (versus 10,000 ϫ g) for 15 min. To purify unbound protein components, the supernatant was precipitated with 60% (instead of 40%) saturation of ammonium sulfate, incubated on ice for 15-30 min and pelleted down by centrifugation at 12,000 ϫ g for 20 min. The pellet was washed with 60% saturated ammonium sulfate in buffer A (50 mM Tris-HCl, pH 7.5, 150 mM NaCl) and centrifuged at 12,000 ϫ g for 20 min. The pellet was dissolved in buffer A containing 20% glycerol, adjusted to 12.5 ml, and filtered with a 0.45-m filter. The protein was then purified on an anti-FLAG M2 affinity gel (Sigma) column according to the manufacturer's instruction by using 1 ml of solution of synthetic FLAG oligopeptide (1 mg/ml) (PAN Laboratory, Stanford University). Prior to elution, 1 ml of 0.1 mg/ml FLAG oligopeptide solution and 2 ml of buffer A containing 20% glycerol were used to remove/wash weakly interacting proteins. For BIAcore surface plasmon resonance analysis, the purified proteins were then dialyzed against 0.1 M phosphate buffer, pH 7.0, containing 50 mM NaCl at 4°C overnight. The enzymatic activities were monitored as described (see later section on activity assays).
BIAcore Surface Plasmon Resonance Analysis-Real time proteinprotein interactions were examined on a BIAcore instrument (BIAcore X). RhlB RNA helicase and bovine serum albumin were immobilized on different flow cells of a CM5 sensor chip using an amine-coupling kit (Amersham Biosciences). Briefly, the chip surface was first activated by injection of 35 l of 1:1 mixture of 0.4 M N-ethyl-NЈ-(dimethylaminopropyl) carbodiimide hydrochloride and 0.1 M N-hydroxysuccinimide. Native RhlB RNA helicase (70 l of 35 g/ml) in 100 mM phosphate buffer, 50 mM NaCl, pH 7.0, (optimal buffer components as determined by trial preconcentration experiments) was immobilized on one flow cell, and an additional flow cell was prepared as a blank background by immobilization of bovine serum albumin (50 g/ml) under the same buffer conditions. Remaining activated groups on each flow cell were blocked by injection of 35 l of 1 M ethanolamine HCl, pH 8.5. The chip was then washed twice with 10 l of 0.1 M NaOH to remove any non-covalently bound proteins. Finally, the system was primed with the running buffer (100 mM phosphate buffer, pH 7.0, 50 mM NaCl).
The interaction assays were performed with a constant (10, 20, or 30 l/min as indicated) flow rate at 25°C. Individually distinct concentrations, ranging from 5 to 50 nM (as shown in Fig. 3) of purified native RhlB RNA helicase, PNPase, or enolase were injected as analytes. The chemical binding groups were regenerated by sequentially washing the analytes with 10 l of injection of 0.1 M NaOH, 0.1 M glycine-HCl (pH 3.5) and 10% ethanol until a background level was attained.
Helicase Activity and PNPase Activity Assays-RNA oligonucleotides used for RNA helicase unwinding and the PNPase degradation assay were double-stranded substrates that resulted from the annealing of a short single-stranded oligo 5Ј-AGC GCA GUA CC and its complementary longer single stranded oligo 5Ј-ACA GUA UUU GGU ACU GCG CUC U (synthesized by Dharmacon Research Inc.). Underlined nucleotides indicate the complementary region with the short strand RNA oligo. The RNA sequences of long strand RNA are identical to the 5Ј-end sequences of RNAI, the antisense RNA of ColEI-type plasmids (18), and contain an RNase E cleavage site (33)(34)(35). Both strands were 5Ј-endlabeled with [␥-32 P]ATP as described previously (35). The labeled RNA strands were hybridized to form duplex RNA substrates made by combining "short" and "long" RNA in a 1:1.5 ratio. The excess of long RNA ensured that the majority of the labeled RNA formed duplex RNA substrates. RNA annealing was performed by adding equal volumes of 2ϫ hybridization buffer (40 mM HEPES-KOH, pH 7.6, 1 M NaCl, 0.2% SDS, 2 mM EDTA) to long and short RNA mixtures, in which samples were heated to 95°C for 5 min and then slowly cooled to room temperature. The duplex RNAs (25 fmol) were incubated with purified RhlB RNA helicase alone or with mixtures of PNPase and RhlB RNA helicase at 37°C in final reaction volumes of 20 l containing 20 mM HEPES-KOH, pH 7.5, 70 mM KCl, 2 mM dithiothreitol, 0.1 mg of bovine serum albumin, 1 mM magnesium acetate, and 3 mM ATP. Each reaction was terminated by adding a 2 l of stop solution (2.2% Ficoll 400, 0.11 M EDTA, 1.1% SDS, 0.275% bromphenol blue, 0.275% xylene cyanol FF). Samples were separated on a 16% polyacrylamide native gel (19:1 bisacrylamide) in 1ϫ TBE. Labeled RNAs were visualized by autoradiography and quantified by LAS-1000 plus (Fuji Film).

Degradosome Protein Components Are Able to Interact in
Vivo Independently of RNase E-General methods to identify interacting proteins or to study protein-protein interactions have been developed extensively. Among them, the two-hybrid systems currently represent the most powerful in vivo approaches. Here, we applied the E. coli two-hybrid approach (29) to study possible interactions among degradosome protein components. Individual degradosome component proteins were fused to two complementary fragments of the catalytic domain of Bordetella pertussis adenylate cyclase into plasmids pT25and pT18-based vectors, as described under "Experimental Procedures." As seen in Fig. 1A, the pink/red phenotypes displayed on MacConkey/maltose plates demonstrates that RhlB RNA helicase is able to interact not only with the RE12 fragment (residues 684 -784 of RNase E region are sufficient for interaction with the helicase) 1 but also with itself and the PNPase. However, under the same conditions, no interactions between the RhlB and enolase or enolase and PNPase (Fig. 1C) were detected. We quantified the ␤-galactosidase activity in individual transformants grown to log-phase culture to determine the efficiency of adenylate cyclase functional complementation, representative of the strength of functional interactions between two protein components (29). As shown in Fig. 1B, the strength of functional interactions is similar between RhlB RNA helicase-RE12 and RhlB-RhlB RNA helicases and both are stronger (about 4-fold) than that of the RhlB RNA helicase-PNPase interaction. Consistently, the ␤-galactosidase activity assay indicated a lack of any interaction between enolase and the two degradosome component proteins, RhlB and PNPase (Fig. 1, B and C).
Individually Specific Interactions between RhlB RNA Helicase, PNPase, and RE12 Are through the C-terminal Region of RhlB-To identify the regions of RhlB RNA helicase directly responsible for its protein-protein interactions, several trun-cated RhlB derivatives were constructed as described under "Experimental Procedures," based on the predicted conserved sequence and functional motifs of a DEAD box helicase family (36,37). As shown in Fig. 2A, RhlB1 consists of the ATPase motif. RhlB2 consists of the RhlB RNA helicase region including helicase and RNA interaction motifs. RhlB3 consists of the C-terminal region and lacks any conserved motifs. RhlB12, RhlB13, and RhlB23 consist of RhlB1 and -2, RhlB1 and -3, and RhlB2 and -3, respectively.
In the E. coli two-hybrid system, both color phenotypes on MacConkey/maltose plates and ␤-galactosidase activity assays showed that RhlB23 was sufficient to interact with RE12 and PNPase (Fig. 2, B-D, E-G, respectively). Interestingly, the region of RhlB sufficient for interaction with PNPase and RE12 fails to interact with itself. RhlB RNA helicase self-interaction requires full-length RhlB RNA helicase (Fig. 2, H and J).
The ␤-galactosidase activity assay indicates that the strength of interaction between RhlB23 and RE12 is a little weaker than that between full-length RhlB RNA helicase and RE12 (Fig. 2D). In contrast, the strength of interaction between RhlB23 and PNPase is similar to that between full-length RhlB RNA helicase and PNPase (Fig. 2G).

Real-time Detection of Protein-Protein Interaction by BIAcore Analyses Reveals Specific, Direct Binding of RhlB RNA Helicase to Itself and PNPase but Not to
Enolase-To examine direct interactions between components of the degradosome, we used purified component proteins and an in vitro real-time binding assay using BIAcore apparatus (38). The unbound and FLAG-tagged protein components PNP, RhlB RNA helicase, and enolase were purified from BZ99 (an E. coli mutant containing only the RNase E N-terminal region (amino acid residues 1-602)), and their enzymatic activities were confirmed, as described under "Experimental Procedures." Fig. 3A shows Coomassie Blue staining of high purity purified proteins. For the BIAcore surface plasmon resonance analyses (38), RhlB RNA helicase was first immobilized on a CM5 sensor chip, and different concentrations of three distinct native degradosome proteins, enolase, RhlB helicase, and PNPase (Fig. 3, B-D), were individually injected over the chip surface, and the sensorgrams recorded. In these analyses, the sensorgrams, in general, show an initial rapid increase in response units upon protein injection followed by the association of injected proteins with molecules coated on the chip and a slower association phase characteristic of reversible protein interaction, resulting in a plateau indicative of interacting saturation or equilibrium (Fig. 3). After a change of buffer, there is a rapid decrease in response unit/signal followed by the dissociation phase. As shown in Fig. 3B, the sensorgrams did not detect any RhlBenolase interaction, whereas significant interaction responses were detected for RhlB with itself and with PNPase (Fig. 3, C  and D, respectively). These results indicated that RhlB helicase was able to directly interact with PNPase and itself but not with enolase (Fig. 3B). The sensorgrams (Fig. 3, C-E) also indicated that the affinity of the RhlB RNA helicase interaction with itself was higher than that of its interaction with PNPase (only 12.5 nM is shown in Fig. 3E). These results are consistent with the results of our E. coli two-hybrid assay (Fig. 1B).
We further addressed whether the non-monomeric form of RhlB RNA helicase can directly interact with PNPase. As shown by sensorgrams, purified RhlB RNA helicase interacted with itself, after which these self-associated molecules of RhlB RNA helicase were still able to bind to PNPase (Fig. 4). Thus, these results illustrated that under the experimental conditions carried out, a dimeric (or multimeric) form of RhlB RNA helicase can physically bind to PNPase.
RhlB RNA Helicase Shows Functional Interaction with PNPase Independent of RNase E-In contrast to previously reported degradosome studies, which used a mini-degradosome complex containing the C-terminal region of RNase E, here we assessed the possibility of functional interaction between RhlB and PNPase in the absence of RNase E. The unwinding activity of RhlB RNA helicase was examined using classical gel electrophoresis assays (39), as described under "Experimental Procedures." Both single-stranded RNA oligonucleotides were labeled with 32 P and annealed to form a double-stranded RNA substrate (labeled as double-stranded RNA). As shown in Fig. 5 RhlB RNA helicase was able to unwind duplex RNA (Fig. 5,  lanes 2-7), and this effect was RhlB-concentration-dependent (compare lanes 6 and 7 versus 5). As a result of the unwinding activity of RhlB, a decrease in double-stranded RNA and an increase in single-stranded RNA were observed. In the absence of RhlB under the same reaction conditions PNPase quickly degraded single stranded RNA (labeled as long or short) but not FIG. 1. Degradosome protein-protein interactions as demonstrated by the E. coli two-hybrid system. The various chimeric degradosome protein-constructs used for individual complementation assays are shown. MacConkey/maltose agar plate assays and ␤-galactosidase activity assays were performed as described under "Experimental Procedures." A, C, chimeric degradosome protein-construct pairs T25RhlB-RE12T18, T25RhlB-RhlBT18, and T25RhlB-PNPT18, but not isolated individual chimeric proteins, revealed a positive interaction as shown by the red colony phenotype. The positive control T25Zip-ZipT18's interaction is also shown. No interactions between construct pairs T25RhlB-EnoT18 and T25Eno-PNPT18 were detected. A reversed chimeric construct of the Eno-PNP construct pair (i.e. T25PNP-EnoT18) confirmed a lack of protein-protein interaction between enolase and PNPase. B, ␤-galactosidase assay used to study the relative strength of protein-protein interactions. The chimeric Zipdimer was used as positive control, (reaction activity was normalized to 100% and compared with other chimeric pair interactions) as shown. The results represent the mean value from at least three independent cultures: RE12, fragment of RNase E, which is sufficient for an interaction with RhlB RNA helicase; PNP, PNPase; RhlB, RhlB RNA helicase; and ENO, enolase. All plasmid constructs encoding individually chimeric proteins are described under "Experimental Procedures." double-stranded RNA (Fig. 5, lanes 8 -10 versus 11). Collectively, these experiments indicate that the unwinding ability of RhlB enables PNPase to rapidly degrade duplex doublestranded RNA. DISCUSSION Here we report that RNA degradosome protein component RhlB interacts with PNPase independently of RNase E, which up to this point has been believed to be the necessary building scaffold for interactions among degradosome proteins. RhlB and PNPase interact through the C-terminal region of RhlB RNA helicase enabling PNPase to degrade double-stranded RNA substrates. We also show that monomers of RhlB interact among themselves, although at a site distinct from the one that interacts with PNPase. It is interesting that Rh1B binds more strongly to itself and to an RNase E fragment than it does to PNPase. There are a number of reasons, both biochemical and cellular, that may explain this result. First, the interaction surface (between the two proteins) may be more favorable for RhlB binding to RNase E than for the RhlB-PNPase interaction. Second, from a cellular perspective, RNase E is a mem-brane-associated protein of relatively low abundance compared with the highly abundant cytoplasmic PNPase; PNPase is present in E. coli in excess of RNase E and has been detected in cells unattached to the RNase E scaffold (27). If the RhlB-PNPase interaction is stronger than the RhlB-RNase E interaction, this may affect the availability of RhlB to bind to RNase E and form degradosome complexes. This in turn could compromise RNA degradation via a membrane-associated degradosome response.
Our results indicate that to degrade double-stranded RNA, PNPase would need to recruit Rh1B helicase to open the double-stranded RNA secondary structure, thereby allowing the exonuclease action of PNPase to degrade the RNA, lending support to the hypothesis that DEAD box RhlB helicase function is necessary for degradosome-mediated RNA decay. Our findings also suggest that multimers of Rh1B RNA helicase may form and bind directly to molecules of PNPase, independently of RNase E, to form mini-degradation engines in E. coli capable of removing RNA degradation intermediates (products) that have secondary structures.

FIG. 2. Specific interaction between
RhlB RNA helicase and PNPase is mediated via the C-terminal region of RhlB, a region not adequate by itself for RhlB-RhlB self-interaction. A, a schematic diagram shows the conserved regions of the DEAD box helicase family as regards E. coli RhlB RNA helicase and its truncated derivatives RhlB1, RhlB2, RhlB3, RhlB12, RhlB13, and RhlB23. Amino acid residues of individual derivatives are indicated. B-J, the E. coli twohybrid assay was performed to study chimeric protein interactions as described in Fig. 1 in which bacteria were grown on either MacConkey/maltose plates (B, E, H) or were cultured in the presence of isopropyl ␤-D-thiogalactoside plus appropriate antibiotics (D, G, J) (Karimova et al.,Ref. 29). The chimeric protein interaction pairs are also shown (C, F, I). ␤-galactosidase assays were used to study the strength of protein-protein interactions. The positive control (T25Zip-ZipT18) and the method used for comparison of protein-protein interaction strength are described in the legend to Fig. 1. The results showed there is no self-interaction detected between various RhlB-truncated derivatives except for a full-length RhlB-RhlB interaction; the interaction between RhlB and PNPase is associated with the C-terminal region of RhlB.
A 3Ј-5Ј exonuclease complex, termed the "exosome," has been identified in yeast cells (41,42) with exosome protein component equivalents also discovered in mammalian cells (43). Recent modeling experiments (44) suggest that exosomes may use the 3Ј-5Ј exonuclease PNPase at their core. Although, E. coli degradosome models have considered RNase E to be at the core of this bacterial ribonucleolytic complex (25), our discovery of PNPase-based complexes that do not include RNase E, provides a molecular basis for alternative pathways of E. coli RNA decay independent of RNase E-mediated endonucleolytic cleavages. Recent evidence suggesting that functional inactivation of chloramphenicol acetyltransferase mRNA in E. coli cell extracts occurs by a mechanism other than endonucleolytic cleavage (45), and earlier findings showing that mutations of the RNase E-encoding rne gene affects chemical half-life but not functional decay (13) are consistent with this view.
It has been suggested that enolase is the ␤ subunit of ␣ 3 ␤ 2 form PNPase (PNPase is the ␣ subunit) (4), although no experimental proof has been forthcoming. However, our results indicate that RhlB RNA helicase, not enolase, is a native component of the PNPase complex. We detected no binding between enolase and PNPase (a similar conclusion also has been discussed by Kü hnel and Luisi (Ref. 46; see "Discussion")). The yeast two-hybrid system also failed to detect any interaction between the two proteins (25). Furthermore, we have shown in this study that RhlB contains as ATP-dependent RNA unwinding ability (not demonstrated prior to this study) and that this ATP-activated unwinding activity enables PNPase to degrade double-stranded RNA; whether the binding of RhlB to PNPase regulates PNPase catalytic activity or vise versa remains to be clarified.
Far Western and yeast two-hybrid analyses of interactions between RNase E truncated polypeptides and RhlB have shown that RhlB binds directly to both the central (500 -752) and C-terminal (734 -1061) regions of RNase E, suggesting that important determinants for the binding are located in the small overlap region (734 -752) of RNase E segments; other interactions among degradosome protein components were not detected (25). Using E. coli two hybrid and BIAcore biosensor FIG. 3. Kinetic analyses of degradosome protein component interactions. A, Coomassie Blue staining of purified proteins PNPase, RhlB RNA helicase, and enolase separated by 8% SDS-PAGE. Protein purification is described under "Experimental Procedures." B-D, kinetic data set collected for individual analytes (i.e. degradosome protein components) binding to an RhlB-surface chip. Degradosome component proteins at various concentrations, as indicated, were injected over the RhlB surface. Injections were carried out as described under "Experimental Procedures." Normalized binding responses for enolase (B), RhlB RNA helicase (C), and PNPase (D) were reproducible for individual injections (data not shown). Individual injections were performed at least four to five times. E, comparison of the kinetic data for RhlB RNA helicase, enolase, and PNPase, respectively, injected over the RhlB surface chip. No response was detected for enolase-RhlB, however, a "strong" response was detected for PNPase-RhlB, and an even stronger response for RhlB-RhlB interaction reaction. A detailed account for the basic shape of the graph can be found under "Experimental Procedures" and "Results." FIG. 4. Self-interacted form of RhlB-RhlB RNA helicase binds to PNPase. The distinct protein analytes and their concentrations are indicated on the sensorgram. RhlB RNA helicase was injected over the RhlB chip surface; after the dissociation phase, injection of the second analyte (i.e. the PNPase) over the RhlB-RhlB surface was carried out to generate the data presented in this sensorgram. analyses, we detected RhlB self-interaction and direct binding to PNPase, in addition to its interaction with the RNase E (684 -784). These results indicate that the protein conformation of various fusion products of RhlB and PNPase used for yeast and E. coli two-hybrid analyses differ and thus that it is appropriate to explore multiple approaches for the study of protein-protein interactions. Use of the BIAcore biosensor to analyze real time interactions among degradosome protein components and between protein component and RNA substrates to determine their kinetic constants may help to elucidate the mechanism by which the dynamic processes of degradosome assembly and RNA degradation interrelate.